System

Abstract

A system is provided.SOLUTION: A system comprising: means for a user to post a negative rating for a product to an e-commerce platform; means for generating a positive rating from the negative rating using generative artificial intelligence; means for converting the generated rating into a vector format; means for storing the vector format rating in a database; means for searching for another product having a similar rating based on the stored vector format rating; and means for recommending another product having a similar rating to the user based on the search.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The technology of the present disclosure relates to a system. [Background technology]

[0002] Patent document 1 discloses a persona chatbot control method performed by at least one processor, the method including the steps of receiving a user utterance, adding the user utterance to a prompt including an instruction sentence related to a description of the chatbot character, encoding the prompt, and inputting the encoded prompt into a language model to generate a chatbot utterance in response to the user utterance. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2022-180282 Summary of the Invention [Problem to be solved by the invention]

[0004] Negative product reviews posted by users on e-commerce platforms are common. This forces users to search for alternative products to resolve their dissatisfaction with their current products, but this process is time-consuming and cumbersome. Finding a suitable alternative product can also be difficult. Furthermore, negative reviews can be a barrier to product sales for e-commerce platform operators. Therefore, there is a need for a system that can analyze negative reviews and efficiently recommend alternative products that address the dissatisfaction. [Means for solving the problem]

[0005] The present invention provides a system including a means for users to post negative reviews of products on an e-commerce platform. The system includes a means for generating positive reviews from the negative reviews using a generative artificial intelligence, a means for converting the generated reviews into a vector format, a means for storing the vector format reviews in a database, a means for searching for other products with similar reviews based on the stored vector format reviews, and a means for recommending other products with similar reviews to the user based on the search, thereby enabling user dissatisfaction to be resolved quickly and efficiently, thereby improving user satisfaction.

[0006] "User" refers to any individual or legal entity who purchases goods using the e-commerce platform.

[0007] "E-commerce platform" refers to a website or application that allows users to sell or purchase goods and services over the Internet.

[0008] "Negative ratings" refer to reviews in which users report flaws or dissatisfaction with a product.

[0009] "Generative artificial intelligence" refers to computer systems that use natural language processing techniques to generate and analyze text.

[0010] "Positive ratings" refer to reviews in which users report the benefits and satisfaction of a product.

[0011] "Vector format" refers to a format that converts text information into numerical data.

[0012] A "database" refers to a system for efficiently storing, managing, and retrieving large amounts of information.

[0013] "Calculating similarity" refers to the process of quantitatively evaluating the similarities and differences between different text data.

[0014] An "alternative product" refers to a new product proposed to satisfy a user's dissatisfaction or requirements.

[0015] "Recommending" refers to the act of suggesting the best option to a user based on certain criteria. [Brief explanation of the drawings]

[0016] [Figure 1] 1 is a conceptual diagram showing an example of the configuration of a data processing system according to a first embodiment. [Figure 2] 1 is a conceptual diagram showing an example of main functions of a data processing device and a smart device according to a first embodiment. [Figure 3] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a second embodiment. [Figure 4] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and smart glasses according to a second embodiment. [Figure 5] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a third embodiment. [Figure 6] FIG. 11 is a conceptual diagram showing an example of main functions of a data processing device and a headset-type terminal according to a third embodiment. [Figure 7] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a fourth embodiment. [Figure 8] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and a robot according to a fourth embodiment. [Figure 9] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 10] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 11] FIG. 3 is a sequence diagram showing a processing flow of the data processing system according to the first embodiment. [Figure 12] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 1. [Figure 13]FIG. 10 is a sequence diagram showing the flow of processing in the data processing system according to the second embodiment when an emotion engine is combined. [Figure 14] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 2 when an emotion engine is combined. DETAILED DESCRIPTION OF THE INVENTION

[0017] An example of an embodiment of a system according to the technology of the present disclosure will be described below with reference to the accompanying drawings.

[0018] First, the terms used in the following description will be explained.

[0019] In the following embodiments, a coded processor (hereinafter simply referred to as a "processor") may be a single arithmetic device or a combination of multiple arithmetic devices. Furthermore, a processor may be a single type of arithmetic device or a combination of multiple types of arithmetic devices. Examples of arithmetic devices include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a GPGPU (General-Purpose computing on Graphics Processing Units), and an APU (Accelerated Processing Unit).

[0020] In the following embodiments, a coded RAM (Random Access Memory) is a memory in which information is temporarily stored and is used as a working memory by a processor.

[0021] In the following embodiments, the coded storage is one or more non-volatile storage devices that store various programs, various parameters, etc. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), and magnetic tapes.

[0022] In the following embodiments, a communication I / F (Interface) with a symbol is an interface including a communication processor, an antenna, etc. The communication I / F controls communication between multiple computers. Examples of communication standards applied to the communication I / F include wireless communication standards including 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), Bluetooth (registered trademark), etc.

[0023] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." In other words, "A and / or B" means that it may be only A, only B, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" is also applied when three or more things are expressed connected by "and / or."

[0024] [First embodiment]

[0025] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0026] 1, a data processing system 10 includes a data processing device 12 and a smart device 14. An example of the data processing device 12 is a server.

[0027] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0028] The smart device 14 includes a computer 36, a reception device 38, an output device 40, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The reception device 38, the output device 40, and the camera 42 are also connected to the bus 52.

[0029] The reception device 38 includes a touch panel 38A, a microphone 38B, and the like, and receives user input. The touch panel 38A detects contact with an indicator (for example, a pen or a finger) to receive user input by the touch of the indicator. The microphone 38B detects the user's voice to receive user input by voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data indicating the user input.

[0030] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting the data in a form of expression that the user 20 can perceive (for example, audio and / or text). The display 40A displays visible information such as text and images in accordance with instructions from the processor 46. The speaker 40B outputs audio in accordance with instructions from the processor 46. The camera 42 is a compact digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0031] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54.

[0032] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0033] 2, in the data processing device 12, a specific process is performed by the processor 28. A specific processing program 56 is stored in the storage 32. The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific process is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0034] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0035] In the smart device 14, the processor 46 performs the reception output process. The storage 50 stores a reception output program 60. The reception output program 60 is used in conjunction with the specific processing program 56 by the data processing system 10. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0036] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0037] The present invention is a system that analyzes negative reviews posted by users on an e-commerce platform and recommends to the users alternative products that will resolve their dissatisfaction.

[0038] The system mainly includes the following means:

[0039] A means for users to post negative product reviews on e-commerce platforms

[0040] A means for the server to generate positive reviews from the negative reviews using artificial intelligence

[0041] means for the server to convert the generated ratings into vector form;

[0042] means for storing the vector-formed evaluations in a database;

[0043] A means for the server to search for other products having similar ratings based on the stored vector-format ratings.

[0044] A means for the server to recommend other products with similar ratings to the user based on the search.

[0045] Program processing and specific examples

[0046] 1. Users post reviews

[0047] Users can post negative reviews of the product in the review section of the e-commerce platform.

[0048] For example: "I'm not happy with this phone because the battery dies so quickly."

[0049] 2. The server receives the review and passes it to the generation AI.

[0050] The server receives user submissions, performs preprocessing, and then passes negative reviews to the AI ​​generator.

[0051] Preprocessing includes tokenization and stop word removal.

[0052] 3. Generative AI generates positive reviews

[0053] Generative AI generates positive evaluations based on negative evaluations.

[0054] For example: "This smartphone has a great battery life and can be used for a long time."

[0055] 4. The server vectorizes the generated ratings and stores them in a database.

[0056] The server converts the generated positive reviews into vector format and stores them in a database.

[0057] Techniques such as TF-IDF (Term Frequency-Inverse Document Frequency) are used for vectorization.

[0058] 5. The server searches for other products with similar reviews

[0059] The server calculates the similarity between the vectorized rating and other product reviews in the database.

[0060] For this similarity calculation, a method such as cosine similarity is used.

[0061] 6. The server recommends another product to the user

[0062] The server generates a list of highly similar products and recommends them to the user.

[0063] For example: "If you're unhappy with your battery, we recommend this phone with better battery life."

[0064] This system allows users to efficiently find suitable alternative products to resolve their dissatisfaction, and also helps e-commerce platform operators improve user experience and reduce barriers to sales.

[0065] The processing flow will be explained below.

[0066] Step 1:

[0067] A user posts a review on an e-commerce platform

[0068] Users access the review section of the product they have purchased and enter a negative rating to post.

[0069] For example: "I'm not happy with this phone because the battery dies so quickly."

[0070] Step 2:

[0071] The server receives the review

[0072] The server receives user submissions instantly.

[0073] Registers the received review internally and passes it to the next step for processing.

[0074] Step 3:

[0075] The server pre-processes the review

[0076] The server pre-processes the review sentences. In this step, the review sentences are tokenized (divided into words) and stop words (common words with little meaning) are removed.

[0077] Preprocessing makes the review sentences ready for analysis and generation.

[0078] Step 4:

[0079] The server passes the review to the generation AI

[0080] The preprocessed review text is sent to the AI ​​generator, which analyzes the received review text and identifies negative elements.

[0081] Step 5:

[0082] Generative AI generates positive reviews

[0083] The artificial intelligence generates a review sentence with negative elements removed.

[0084] For example: "This smartphone has a great battery life and can be used for a long time."

[0085] Step 6:

[0086] The server vectorizes the generated ratings

[0087] The server converts the generated positive ratings into a numerical format (vector format), using techniques such as TF-IDF (Term Frequency-Inverse Document Frequency).

[0088] The generated evaluation features are expressed numerically.

[0089] Step 7:

[0090] The server stores the vectorized ratings in a database.

[0091] The vectorized review texts are stored in a database, which will be used for future multi-level review retrieval and recommendation.

[0092] Step 8:

[0093] The server searches for other products with similar reviews

[0094] The server calculates the similarity of the generated review to other product reviews in the database, using a method such as cosine similarity.

[0095] Identify products with the most similar reviews.

[0096] Step 9:

[0097] The server recommends similar products to the user.

[0098] Based on the calculation results, the server recommends products with the most similar reviews to the user, and displays the recommended results on the user's device.

[0099] For example: "If you're unhappy with your battery, we recommend this phone with better battery life."

[0100] This step allows users to quickly find the best alternative to resolve their dissatisfaction.

[0101] Example 1

[0102] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0103] When a user posts a negative review on an e-commerce platform, it is important to appropriately recommend alternative products to resolve the user's dissatisfaction. Current systems do not accurately identify the user's dissatisfaction from the negative review and recommend alternative products based on that information, which limits the effectiveness of improving the user experience and promoting sales.

[0104] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0105] In this invention, the server includes means for users to post negative reviews of products on the e-commerce platform, means for generating positive reviews from the negative reviews using a generation artificial intelligence, means for converting the generated reviews into a vector format, means for storing the vector format reviews in a database, means for searching for other products with similar reviews based on the stored vector format reviews, means for recommending other products with similar reviews to the user based on the search, and means for pre-processing the negative reviews, thereby making it possible to effectively recommend alternative products based on the user's negative reviews.

[0106] "User" means an individual or legal entity who uses the e-commerce platform to post ratings and reviews of products.

[0107] An "e-commerce platform" is an online system for buying and selling goods and services over the Internet.

[0108] A "negative rating" is a review in which a user expresses dissatisfaction or dissatisfaction with a product or service.

[0109] "Generative AI" refers to AI that generates text using natural language processing and machine learning techniques.

[0110] A "positive rating" is a review that expresses positive content or benefits in relation to a negative rating.

[0111] "Vector format" is a format that expresses sentences and evaluations as numerical data. It is expressed as multiple numerical vectors, making calculations and comparisons easier.

[0112] A "database" is a system for storing and managing information in an organized manner.

[0113] "Similar evaluations" refer to evaluations in which the content of multiple evaluations is similar to one another. This similarity is quantified using a specific calculation method.

[0114] "Preprocessing" refers to processing performed to make raw data easier to analyze. Specifically, it includes tokenization and removal of stop words.

[0115] "Tokenizing" is the process of dividing text data into units of words or phrases.

[0116] "Stop word removal" is the process of removing common words that have no meaning or are not needed during analysis.

[0117] "Cosine similarity" is a method for calculating the similarity between vector data. It calculates the similarity based on the angle between the vectors.

[0118] The present invention is a system that analyzes negative reviews posted by users on an e-commerce platform and recommends alternative products to the users that will resolve their dissatisfaction. This system is composed of a user, a server, a generative artificial intelligence (AI), and a database.

[0119] First, a user accesses an e-commerce platform and posts a negative review in the review section. For example, the user might write, "I'm not happy with this smartphone because the battery runs out quickly," and submit it.

[0120] Next, the server receives these negative reviews and preprocesses them. This preprocessing includes tokenizing (dividing sentences into words) and removing stop words (removing meaningless common words). This preprocessing makes the review text easier to analyze by generative artificial intelligence.

[0121] The preprocessed reviews are then passed to a generative AI model, which uses natural language processing techniques, for example, to convert negative reviews into positive ones. Based on the example above, the generative AI model would generate a positive review such as, "This smartphone has excellent battery life and can be used for a long time."

[0122] The server then converts the generated positive reviews into vector format using techniques such as Term Frequency-Inverse Document Frequency (TF-IDF) and word embedding, and stores the converted reviews in a database.

[0123] Based on the stored vector-format ratings, the server calculates the similarity with the reviews of other products. The similarity is calculated using methods such as cosine similarity. Based on this similarity calculation, the server searches for other products with similar ratings.

[0124] Finally, the server generates a list of similar products and recommends them to the user. The user can refer to the recommended products to find suitable alternative products that can resolve their complaints. For example, the server may display a message such as, "If you are dissatisfied with the battery, we recommend this smartphone with a long battery life."

[0125] Examples of prompt statements:

[0126] User type: "I'm not happy with this phone because the battery dies so quickly."

[0127] Prompt for the AI ​​generator: "I'm not happy with this phone because the battery dies so quickly. Please generate a positive rating for this review."

[0128] Positive review generated: "This phone has a great battery life and can be used for a long time."

[0129] This particular embodiment allows for quick and effective recommendation of alternative products based on users' negative reviews, thereby improving the user experience of the e-commerce platform and promoting product sales.

[0130] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0131] Step 1:

[0132] Users post reviews

[0133] Specific behavior:

[0134] Users access the e-commerce platform's website or app and post negative reviews of products in the review section.

[0135] Input: A negative review such as "I'm not happy with this phone because the battery dies so quickly."

[0136] Output: A negative rating is sent to the server.

[0137] Step 2:

[0138] The server receives the reviews and pre-processes them

[0139] Specific behavior:

[0140] The server receives reviews posted by users and pre-processes the review content.

[0141] Input: The negative rating submitted by the user.

[0142] Data processing: tokenization (splitting sentences into words) and stop word removal (removing meaningless common words).

[0143] Output: Preprocessed text data.

[0144] Step 3:

[0145] The server passes the data to a generative AI model, which generates a positive rating.

[0146] Specific behavior:

[0147] The server passes the preprocessed negative reviews to a generative AI model, which generates positive ratings.

[0148] Input: Preprocessed negative reviews.

[0149] Data computation: Generative AI models generate positive reviews based on negative reviews.

[0150] Output: A positive review generated by the generative AI.

[0151] Step 4:

[0152] The server vectorizes the generated ratings and stores them in a database.

[0153] Specific behavior:

[0154] The server converts the generated positive reviews into vector format and stores them in a database.

[0155] Input: Generated positive reviews.

[0156] Data processing: Convert data into vector format using techniques such as TF-IDF or word embedding.

[0157] Output: Ratings in vector format stored in a database.

[0158] Step 5:

[0159] The server searches for similar products

[0160] Specific behavior:

[0161] The server calculates the similarity between the vectorized ratings and other product reviews in the database and searches for other products with similar ratings.

[0162] Input: Ratings in vector format stored in a database.

[0163] Data calculation: Calculate similarity using methods such as cosine similarity.

[0164] Output: A list of highly similar products.

[0165] Step 6:

[0166] The server recommends different products to the user

[0167] Specific behavior:

[0168] The server generates a list of highly similar products and recommends them to the user.

[0169] Input: A list of similar products.

[0170] Output: The recommended product list displayed to the user.

[0171] Example: "If you're unhappy with your battery, we recommend this smartphone with better battery life."

[0172] As described above, the system's program starts with negative reviews posted by users, processes the data and performs various calculations, and then recommends alternative products that will resolve the user's complaints.

[0173] (Application example 1)

[0174] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0175] On modern e-commerce platforms, negative reviews posted by users are an important factor in product selection. However, these negative reviews often discourage purchases and cause users to abandon the platform. Conventional systems lack a mechanism to effectively utilize these negative reviews to recommend alternative products to users, which results in the inability to adequately address users' dissatisfaction and needs.

[0176] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0177] In this invention, the server includes means for users to post negative reviews of products on the e-commerce platform, means for generating positive reviews from the negative reviews using a generation artificial intelligence, means for converting the generated reviews into vector format, means for storing the vector format reviews in a database, means for searching for other products with similar reviews based on the stored vector format reviews, means for recommending other products with similar reviews to the user based on the search, and means for analyzing the user's needs based on the negative reviews and recommending other products that meet those needs in real time, thereby enabling users to quickly find alternative products that best suit their needs and complaints.

[0178] An "e-commerce platform" is an online system for buying and selling goods and services over the Internet.

[0179] "Negative evaluation" refers to reviews or feedback in which users describe their dissatisfaction or shortcomings with a product or service.

[0180] "Generative AI" refers to AI technology that uses natural language processing technology to generate, translate, and analyze text.

[0181] A "positive rating" is a review that describes the positive aspects or benefits of a product or service for the user.

[0182] "Vector format" is a format in which text data is converted into numerical data and expressed in a multidimensional space.

[0183] A "database" is a system for efficiently storing, managing, and searching data.

[0184] "Cosine similarity" is a method for measuring the similarity between two vectors by calculating the cosine value of the angle between them.

[0185] "Similarity calculation" is a calculation for quantifying the similarity between different data.

[0186] "Recommendation" means that the system presents appropriate products and services based on the user's ratings and preferences.

[0187] "Real-time recommendations" refers to the ability to instantly respond to information entered by the user and recommend appropriate products and services.

[0188] This invention relates to a system that analyzes negative reviews posted by users on e-commerce platforms and recommends alternative products that meet the user's needs in real time.

[0189] The system primarily uses the following hardware and software: React Native for front-end development, Node.js and Express.js for back-end development, and MongoDB for the database. It also utilizes GPT-3 or ChatGPT as a generative AI model, the TF-IDF method for vectorizing text data, and cosine similarity for similarity calculations. The system operates when users post reviews through the e-commerce platform's application.

[0190] First, a user posts a negative review about a product using their device. The posted negative review is sent to the server, which receives the review and performs preprocessing. Preprocessing includes tokenization and stop word removal. The preprocessed negative review is then passed to a generative AI model, which generates a positive review. For example, if a user posts, "This smartphone's battery runs out quickly," the generative AI model will generate a positive review such as, "This smartphone has excellent battery life and can be used for a long time."

[0191] The generated positive reviews are converted into vector format by the server and stored in a database. The server then calculates the similarity between the reviews and those of other products in the database. This similarity is calculated using cosine similarity. Based on the calculated similarity, the server searches for other products that may meet the user's needs and recommends them to the user in real time.

[0192] This system allows users to quickly resolve their complaints and find the best alternative product. For example, if a user is dissatisfied with the battery, a similarity calculation will be used to recommend another smartphone with a better battery life. An example of a prompt used by the generative AI model is as follows:

[0193] Example prompt sentence:

[0194] "A user left a negative review stating that the battery dies quickly. Use this as a basis to generate a positive review."

[0195] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0196] Step 1:

[0197] The user uses the terminal to input and post a negative review in the review section of the e-commerce platform.

[0198] Input: Negative review text (e.g., "I'm not happy with this phone because the battery runs out quickly.")

[0199] Output: The posted negative review text is sent to the server.

[0200] Step 2:

[0201] The server receives negative feedback from users and performs preprocessing, which includes tokenization (splitting text into words) and stop-word removal (removing frequently occurring words).

[0202] Input: The negative review text posted.

[0203] Output: Preprocessed negative evaluation text (text after tokenization and stopword removal).

[0204] Step 3:

[0205] The server passes the preprocessed negative evaluation text to a generative AI model (e.g., GPT-3) to generate a positive evaluation. For example, in response to the negative evaluation "The battery runs out quickly," a positive evaluation such as "The battery life is very good and can be used for a long time" is generated.

[0206] Input: Preprocessed negative rating text.

[0207] Output: The generated positive review text.

[0208] Step 4:

[0209] The server converts the generated positive feedback text into vector format using the Term Frequency-Inverse Document Frequency (TF-IDF) method.

[0210] Input: The generated positive review text.

[0211] Output: Vector format data.

[0212] Step 5:

[0213] The server stores the vector data in a database.

[0214] Input: Vector format data.

[0215] Output: Data in vector format stored in a database.

[0216] Step 6:

[0217] The server calculates the similarity between the vector data and other product reviews in the database. Cosine similarity is used to calculate this similarity.

[0218] Input: Vector data from the database and newly generated vector data.

[0219] Output: Similarity calculation result.

[0220] Step 7:

[0221] Based on the similarity calculation results, the server searches for other products that suit the user's needs and recommends them to the user in real time.

[0222] Input: Similarity calculation result.

[0223] Output: A list of products recommended to the user.

[0224] This allows users to quickly find the best alternative product to resolve their dissatisfaction.

[0225] Furthermore, an emotion engine that estimates the user's emotion may be combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59 and perform identification processing using the user's emotion.

[0226] The present invention combines a system that analyzes negative reviews posted by users on e-commerce platforms and recommends alternative products to users that will resolve their dissatisfaction with an emotion engine that recognizes users' emotions.

[0227] The system mainly includes the following means:

[0228] A means for users to post negative product reviews on e-commerce platforms

[0229] A means for the server to generate positive reviews from the negative reviews using artificial intelligence

[0230] means for the server to convert the generated ratings into vector form;

[0231] means for storing the vector-formed evaluations in a database;

[0232] A means for the server to search for other products having similar ratings based on the stored vector-format ratings.

[0233] A means for the server to recommend other products with similar ratings to the user based on the search.

[0234] an emotion engine that recognizes the user's emotions contained in the negative reviews and the generated reviews;

[0235] Program processing and specific examples

[0236] 1. Users post reviews

[0237] Users can post negative reviews of the product in the review section of the e-commerce platform.

[0238] For example: "I'm not happy with this phone because the battery dies so quickly."

[0239] 2. The server receives the review

[0240] The server immediately receives the user's post, registers it internally, and passes it on to the next step for processing.

[0241] 3. The server pre-processes the review

[0242] The server performs preprocessing such as tokenizing the review sentences and removing stop words, so that the review sentences are ready for analysis and generation.

[0243] 4. The server passes the review to the sentiment engine

[0244] The preprocessed review sentences are passed to the emotion engine, which recognizes the user's emotions from the review sentences and extracts the emotion information.

[0245] 5. Generative AI generates positive evaluations based on negative evaluations using emotional information

[0246] The generative AI analyzes negative reviews and generates review texts that eliminate negative elements based on emotional information obtained from the emotion engine.

[0247] For example: "This smartphone has a great battery life and can be used for a long time."

[0248] 6. The server vectorizes the generated ratings

[0249] The server converts the generated positive reviews into a numerical format (vector format). For vectorization, techniques such as TF-IDF are used. The features of the generated reviews are expressed numerically.

[0250] 7. The server stores the vectorized ratings in a database

[0251] The vectorized review texts are stored in a database, which will be used for future multi-level review retrieval and recommendation.

[0252] 8. The server searches for other products with similar reviews

[0253] The server calculates the similarity of the generated review to other product reviews in the database, using a method such as cosine similarity, and identifies the product with the most similar review.

[0254] 9. The server recommends similar products to the user

[0255] The server then recommends the product with the most similar reviews to the user based on the calculation results, and displays the recommended results on the user's device.

[0256] For example: "If you're unhappy with your battery, we recommend this phone with better battery life."

[0257] This system allows users to efficiently find suitable alternative products to resolve their dissatisfaction. In addition, the emotion engine enables more appropriate recommendations that take into account the emotions expressed in users' ratings, improving the user experience.

[0258] The processing flow will be explained below.

[0259] Step 1:

[0260] Users post reviews

[0261] Users access the review section of the product they have purchased and enter a negative rating to post.

[0262] For example: "I'm not happy with this phone because the battery dies so quickly."

[0263] Step 2:

[0264] The server receives the review

[0265] The server receives user submissions instantly, and the received reviews are stored in an internal log and prepared for further processing.

[0266] Step 3:

[0267] The server pre-processes the review

[0268] The server pre-processes the review text, including tokenizing (dividing the text into words), removing stop words, and normalizing (e.g., converting all text to lowercase), making the review text optimized for analysis.

[0269] Preprocessing result example: "Smartphone battery runs out quickly, dissatisfaction"

[0270] Step 4:

[0271] The server passes the review to the sentiment engine

[0272] The preprocessed review sentences are passed to the sentiment engine, which extracts user sentiment from the reviews.

[0273] Example: "Dissatisfied"

[0274] Step 5:

[0275] Generative AI generates positive evaluations based on emotional information

[0276] The generative AI uses extracted emotional information (e.g., "dissatisfied") to turn negative evaluations into positive ones.

[0277] For example: "This smartphone has a great battery life and can be used for a long time."

[0278] Step 6:

[0279] The server vectorizes the generated ratings

[0280] The server converts the generated positive ratings into vector format using techniques such as TF-IDF (Term Frequency-Inverse Document Frequency) and Word2Vec.

[0281] Generated ratings in vector format, e.g.: \[0.15, 0.34, 0.01, ... \]

[0282] Step 7:

[0283] The server stores the vectorized ratings in a database.

[0284] The server stores the vectorized review text in a database, which will be used in future search and recommendation algorithms.

[0285] Step 8:

[0286] The server searches for other products with similar reviews

[0287] The server calculates the similarity between the reviews of other products stored in the database and the generated review vector. This similarity is calculated using cosine similarity.

[0288] Identify the most similar products.

[0289] Step 9:

[0290] The server recommends similar products to the user.

[0291] The server then uses the results of the calculation to recommend products with the most similar reviews to the user, displaying the recommendations on the user's device and suggesting alternative products.

[0292] For example: "If you're unhappy with your battery, we recommend this phone with better battery life."

[0293] Through this step, users can easily find the best alternative product to resolve their dissatisfaction.By using the emotion engine, we can understand the emotions expressed by users in their reviews and make more appropriate product recommendations, thereby further increasing user satisfaction.

[0294] Example 2

[0295] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0296] In conventional e-commerce platforms, negative reviews posted by users are often not properly addressed. As a result, users' dissatisfaction remains unresolved, potentially leading to customer churn. In addition, the lack of recommendations for alternative products based on negative reviews creates a problem of not improving the user experience.

[0297] The identification process by the identification processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means. In this invention, the server includes: means for a user to post a negative review of a product on an e-commerce platform; means for generating a positive review from the negative review using a generative artificial intelligence; means for converting the generated review into a vector format; means for storing the vector-format review in a database; means for searching for other products with similar reviews based on the stored vector-format review; means for recommending other products with similar reviews to the user based on the search; means including an emotion engine that recognizes the user's emotions from the review text and extracts the emotion information; and means for generating a positive review from the negative review based on the emotion information. This makes it possible to resolve the user's dissatisfaction by analyzing the user's negative review and recommending appropriate alternative products, thereby providing a more satisfying user experience.

[0298] "User" means any individual or legal entity who uses the e-commerce platform to post reviews on products.

[0299] An "e-commerce platform" is an online commerce system that allows users to purchase products and post reviews.

[0300] A "negative rating" is a review in which the user points out dissatisfaction or problems with the product.

[0301] "Generative artificial intelligence" is a computer program that uses natural language processing technology to generate new text.

[0302] A "positive evaluation" is a review that states the good points and satisfaction points of the product.

[0303] "Vector format" refers to text data converted into a numerical format that is easy to process with machine learning algorithms.

[0304] A "database" is a system for effectively storing, retrieving, and managing structured data.

[0305] An "emotion engine" is a program that uses natural language processing technology to extract and recognize a user's emotional information from text.

[0306] "Preprocessing" refers to a series of processing steps to convert text data into a format suitable for analysis or generation.

[0307] "Cosine similarity" is a mathematical method for calculating the similarity between two vectors based on the vector's direction.

[0308] This invention relates to a system that analyzes negative reviews posted by users on an e-commerce platform and recommends alternative products to the users that will resolve their dissatisfaction. This system achieves more appropriate product recommendations by incorporating an emotion engine that recognizes the user's emotions.

[0309] Hardware or software used

[0310] 1. Server: Responsible for central data processing of this system. Cloud servers can also be used if necessary.

[0311] 2. Device: The client device (computer, smartphone, tablet, etc.) through which a user posts a review.

[0312] 3. Database: A data storage system for storing reviews, vector-format ratings and other related data.

[0313] 4. Generative AI: An AI model that uses natural language processing technology to generate positive reviews based on negative ones. Examples include OpenAI's GPT.

[0314] 5. Sentiment Engine: A program that uses natural language processing techniques to extract user sentiment from reviews. It uses a model for sentiment analysis (e.g., a BERT-based model).

[0315] Data processing or data calculation

[0316] 1. Review preprocessing: Negative reviews posted by users are tokenized using a natural language processing library (e.g., Python's NLTK) and stop words are removed. This preprocessing converts the reviews into a format suitable for analysis and generation.

[0317] 2. Extracting Emotional Information: The preprocessed review sentences are passed to the emotion engine, which extracts emotional information and recognizes emotion tags such as "sad" and "disappointed."

[0318] 3. Generating Positive Evaluations: The generative AI generates positive evaluations from negative evaluations based on the emotional information extracted from the emotion engine. For example, the generative AI uses OpenAI's GPT model to invert emotion tags such as "sad" and "disappointed" to "joy" and "satisfied."

[0319] 4. Vectorization: The generated positive ratings are converted into a vector format that is easy for machine learning algorithms to process, using tools such as scikit-learn's TF-IDF Vectorizer.

[0320] 5. Similarity calculation: Calculate the similarity with other product reviews in the database. To calculate similarity, use scikit-learn's cosine_similarity.

[0321] 6. Product Recommendation: Based on the calculation results, we identify the products with the most similar reviews and recommend them to the user.

[0322] Specific examples

[0323] An example prompt for this system would be:

[0324] "Enter a negative review and generate a positive one based on it."

[0325] This allows users to more efficiently find alternative products that can resolve their dissatisfaction, improving the user experience by making recommendations that take emotions into account.

[0326] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0327] Step 1:

[0328] Users post reviews

[0329] The user enters a negative review of the product in the review section of the e-commerce platform and presses the post button.

[0330] Input: User negative review text

[0331] Output: Posting a negative review to an e-commerce platform

[0332] Specific operation: A user uses a device (PC, smartphone, tablet, etc.) to access an online shopping site, enter a review in text format, and press the post button.

[0333] Step 2:

[0334] The server receives the review

[0335] The server immediately receives reviews posted by users and registers them in a database.

[0336] Input: User-submitted negative review

[0337] Output: Negative reviews stored in the database

[0338] Specific operation: The server uses the receiving API to accept review submissions and saves the data in a database management system.

[0339] Step 3:

[0340] The server pre-processes the review

[0341] The server performs preprocessing such as tokenizing the review sentences and removing stop words, converting the review sentences into a format suitable for analysis and generation.

[0342] Input: Negative reviews retrieved from the database

[0343] Output: Preprocessed review sentences

[0344] Specific operation: Using Python's NLTK library, the review sentences are split into words and frequent meaningless words (stop words) are removed.

[0345] Step 4:

[0346] The server passes the review to the sentiment engine

[0347] The preprocessed review sentences are passed to the emotion engine, which recognizes emotions from the user's review sentences and extracts the emotion information.

[0348] Input: Preprocessed review text

[0349] Output: Emotion information (e.g., emotion tags such as "sad" or "disappointed")

[0350] Specific operation: Calls an API that passes text to the emotion engine, and receives emotion information in return.

[0351] Step 5:

[0352] Generative AI generates positive evaluations based on negative evaluations and emotional information

[0353] The generative AI analyzes negative ratings and generates positive reviews based on emotional information obtained from the emotion engine.

[0354] Input: Negative evaluation and emotional information

[0355] Output: Positive review text

[0356] Specific operation: Using OpenAI's GPT model, etc., the system generates review sentences by converting negative elements into their opposite positive elements (e.g., converting "sad" and "disappointed" into "joy" and "satisfied").

[0357] Step 6:

[0358] The server vectorizes the generated ratings

[0359] The server converts the generated positive reviews into a numerical format (vector format).

[0360] Input: Generated positive review text

[0361] Output: Evaluation data in vector format

[0362] Specific operation: Converts text into vector format using scikit-learn's TF-IDF Vectorizer.

[0363] Step 7:

[0364] The server stores the vectorized ratings in a database.

[0365] The server stores the vectorized review sentences in a database.

[0366] Input: Vectorized evaluation data

[0367] Output: Evaluation data in vector format stored in a database

[0368] Specific behavior: Executes a query to insert vector data into an SQL or NoSQL database.

[0369] Step 8:

[0370] The server searches for other products with similar reviews

[0371] The server calculates the similarity between the generated review and reviews of other products in the database.

[0372] Input: Vector-formatted reviews and generated vector-formatted ratings data in the database

[0373] Output: Evaluation data with similarity calculated

[0374] Specific operation: Uses scikit-learn's cosine_similarity function to calculate the similarity with other vectors in the database and identify the most similar vector.

[0375] Step 9:

[0376] The server recommends similar products to the user.

[0377] Based on the calculation results, the server recommends products with the most similar reviews to the user.

[0378] Input: A list of similar products

[0379] Output: Recommended products displayed on the user's device

[0380] Specific behavior: Generates an API response to display recommended results on the user's device and returns the results to the browser or app.

[0381] (Application example 2)

[0382] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0383] Current e-commerce platforms lack a system that uses negative feedback from users to appropriately recommend alternative products to resolve their dissatisfaction. This means users have to spend time and effort to find a satisfactory product. Furthermore, existing systems do not take user sentiment into account when making recommendations, which does not adequately improve the user experience.

[0384] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for a user to post a negative review of a product on an e-commerce platform, means for generating a positive review from the negative review using a generative artificial intelligence, and means for converting the generated review into a vector format. This makes it possible to pass the user's negative review to an emotion recognition engine, extract the user's emotional information, generate a positive review based on the emotional information, search for other products with similar reviews based on the vector-format review stored in the database, and recommend other products with similar reviews to the user based on the search.

[0385] "User" means any individual or organization using the e-commerce platform.

[0386] An "e-commerce platform" is an online system for buying and selling goods and services over the Internet.

[0387] "Negative ratings" are reviews or feedback where users express dissatisfaction with a product or service.

[0388] "Generative AI" is an algorithm or system that uses natural language processing techniques to generate new text.

[0389] "Positive reviews" are positive reviews or feedback that indicate satisfaction with a product or service.

[0390] "Vector format" is a format in which text data is converted into a numerical vector.

[0391] A "database" is a system for efficiently storing, managing, and retrieving data.

[0392] A "search method" is an algorithm or method for finding specific data in a database.

[0393] A "recommendation method" is a system or method for suggesting appropriate products or services to users.

[0394] An "emotion recognition engine" is an artificial intelligence technology for extracting user emotional information from text data.

[0395] "Emotion information" is data that indicates the user's emotional state.

[0396] A "prompt" is an input sentence or cue that the generative AI uses to generate text.

[0397] "Similarity" is an index that indicates the similarity between two pieces of data.

[0398] "Cosine similarity" is a method of calculating similarity using the cosine angle between two vectors.

[0399] A specific system for implementing this invention is one that allows users to post negative reviews of products using an e-commerce platform, and based on those reviews, recommends alternative products to resolve the user's dissatisfaction. This system is mainly composed of a combination of a program, hardware, and a database.

[0400] 1. Program Generation and Explanation

[0401] The server receives negative reviews posted by users and preprocesses them. This preprocessing step involves tokenization and stopword removal to convert the reviews into a more easily parseable format. The server then passes the preprocessed reviews to an emotion recognition engine (e.g., using the Hugging Face transformers library) to extract the user's emotional information.

[0402] Based on the emotional information obtained from the emotion engine, the server uses generative artificial intelligence (e.g., a Hugging Face text-generation model) to convert negative feedback into positive feedback. In this step, the prompt sentence is used to input the generative AI model, which generates a positive feedback.

[0403] Once a positive review is generated, the server converts it into a vector format using techniques such as TF-IDF, which is then stored in a database for future searches and recommendations.

[0404] The server then searches for other products with similar ratings based on the stored vector of ratings. It uses cosine similarity to calculate similarity and identifies products with the most similar reviews. Finally, the server recommends products with the highest similarity to the user, and the recommended results are displayed on the user's device.

[0405] 2. Hardware and Software Used

[0406] The system requires advanced hardware and software for text analysis and generation, including high-performance servers and database servers, and Python, SQL, and the Hugging Face transformers library.

[0407] 3. Examples of concrete examples and prompts

[0408] For example, if a user posts a review stating, "I'm dissatisfied with this smartphone because the battery runs out quickly," the system receives it and performs preprocessing and sentiment analysis. Based on the results of the sentiment analysis, the generative AI model generates a positive evaluation such as, "This smartphone has excellent battery life and can be used for a long time." This evaluation is then vectorized and stored in a database.

[0409] The server then calculates the similarity of the generated rating to other ratings in the database, identifies the product with the most similar rating, and recommends this product to the user, saying, "If you're dissatisfied with your battery, we recommend this smartphone with a long battery life."

[0410] Example prompt for a generative AI model:

[0411] User expressed dissatisfaction with: "I'm not happy with this phone because the battery runs out so quickly." Write a positive review.

[0412] This system allows users to quickly find suitable alternative products to resolve their dissatisfaction, improving the user experience.

[0413] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0414] Step 1:

[0415] A user posts a negative review of a product on an e-commerce platform. For example, the user might write, "I'm not happy with this smartphone because the battery runs out quickly." This review becomes the input to the system.

[0416] Step 2:

[0417] The server receives negative reviews from users. It registers the received reviews internally and passes them to the next step. The input is the user's negative review, and the output is the data passed to the preprocessing step.

[0418] Step 3:

[0419] The server preprocesses the reviews. This includes tokenizing the text, removing stop words, etc. It takes the received negative reviews as input and converts them into a format that is easy to parse. The output is the preprocessed text data.

[0420] Step 4:

[0421] The server passes the preprocessed reviews to the emotion recognition engine. Using the Hugging Face transformers library, the user's emotional information is extracted. The input is preprocessed text data, and after processing by the emotion recognition engine, the output is emotional information. For example, a result such as "Negative (90% probability)" can be obtained.

[0422] Step 5:

[0423] The server generates a prompt based on the emotion information and passes it to the AI ​​generator. For example, it generates a prompt like, "User expressed dissatisfaction with: 'I'm dissatisfied with this smartphone because the battery runs out quickly.' Write a positive review." The input is the emotion information and the preprocessed review, and the output is the prompt text.

[0424] Step 6:

[0425] Generative AI generates a positive review based on the prompt. It uses Hugging Face's text-generation model to create a positive review that eliminates negative elements. The input is the prompt text, and the output is the generated positive review. For example, you might get a review like, "This smartphone has excellent battery life and can be used for a long time."

[0426] Step 7:

[0427] The server vectorizes the generated positive reviews. It uses techniques such as TF-IDF to convert text data into a numerical format. The input is the generated positive reviews, and the output is vector-format data.

[0428] Step 8:

[0429] The server stores the generated rating vector in a database, which is used for future searches and recommendations. The input is vector data, and the output is the completed storage in the database.

[0430] Step 9:

[0431] The server calculates the similarity of the generated review to other reviews in the database. It uses methods such as cosine similarity to find the product with the most similar reviews. The input is rating data in vector format, and the output is the product with the most similar review.

[0432] Step 10:

[0433] The server recommends highly similar products to the user. Based on the calculation results, the most suitable product recommendation is displayed on the user's device. The input is the product information from the search results, and the output is the recommendation to the user. For example, a recommendation such as "If you are dissatisfied with your battery, we recommend this smartphone with a long battery life" is displayed.

[0434] The specific processing unit 290 transmits the result of the specific processing to the smart device 14. In the smart device 14, the control unit 46A causes the output device 40 to output the result of the specific processing. The microphone 38B acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0435] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0436] In the above embodiment, an example in which the specific process is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific process may be performed by the smart device 14.

[0437] [Second embodiment]

[0438] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0439] 3, the data processing system 210 includes the data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

[0440] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0441] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, and the camera 42 are also connected to the bus 52.

[0442] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0443] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0444] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0445] Fig. 4 shows an example of the main functions of the data processing device 12 and the smart glasses 214. As shown in Fig. 4, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0446] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0447] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0448] In the smart glasses 214, the reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0449] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the smart glasses 214 will be referred to as the "terminal."

[0450] The present invention is a system that analyzes negative reviews posted by users on an e-commerce platform and recommends to the users alternative products that will resolve their dissatisfaction.

[0451] The system mainly includes the following means:

[0452] A means for users to post negative product reviews on e-commerce platforms

[0453] A means for the server to generate positive reviews from the negative reviews using artificial intelligence

[0454] means for the server to convert the generated ratings into vector form;

[0455] means for storing the vector-formed evaluations in a database;

[0456] A means for the server to search for other products having similar ratings based on the stored vector-format ratings.

[0457] A means for the server to recommend other products with similar ratings to the user based on the search.

[0458] Program processing and specific examples

[0459] 1. Users post reviews

[0460] Users can post negative reviews of the product in the review section of the e-commerce platform.

[0461] For example: "I'm not happy with this phone because the battery dies so quickly."

[0462] 2. The server receives the review and passes it to the generation AI.

[0463] The server receives user submissions, performs preprocessing, and then passes negative reviews to the AI ​​generator.

[0464] Preprocessing includes tokenization and stop word removal.

[0465] 3. Generative AI generates positive reviews

[0466] Generative AI generates positive evaluations based on negative evaluations.

[0467] For example: "This smartphone has a great battery life and can be used for a long time."

[0468] 4. The server vectorizes the generated ratings and stores them in a database.

[0469] The server converts the generated positive reviews into vector format and stores them in a database.

[0470] Techniques such as TF-IDF (Term Frequency-Inverse Document Frequency) are used for vectorization.

[0471] 5. The server searches for other products with similar reviews

[0472] The server calculates the similarity between the vectorized rating and other product reviews in the database.

[0473] For this similarity calculation, a method such as cosine similarity is used.

[0474] 6. The server recommends another product to the user

[0475] The server generates a list of highly similar products and recommends them to the user.

[0476] For example: "If you're unhappy with your battery, we recommend this phone with better battery life."

[0477] This system allows users to efficiently find suitable alternative products to resolve their dissatisfaction, and also helps e-commerce platform operators improve user experience and reduce barriers to sales.

[0478] The processing flow will be explained below.

[0479] Step 1:

[0480] A user posts a review on an e-commerce platform

[0481] Users access the review section of the product they have purchased and enter a negative rating to post.

[0482] For example: "I'm not happy with this phone because the battery dies so quickly."

[0483] Step 2:

[0484] The server receives the review

[0485] The server receives user submissions instantly.

[0486] Registers the received review internally and passes it to the next step for processing.

[0487] Step 3:

[0488] The server pre-processes the review

[0489] The server pre-processes the review sentences. In this step, the review sentences are tokenized (divided into words) and stop words (common words with little meaning) are removed.

[0490] Preprocessing makes the review sentences ready for analysis and generation.

[0491] Step 4:

[0492] The server passes the review to the generation AI

[0493] The preprocessed review text is sent to the AI ​​generator, which analyzes the received review text and identifies negative elements.

[0494] Step 5:

[0495] Generative AI generates positive reviews

[0496] The artificial intelligence generates a review sentence with negative elements removed.

[0497] For example: "This smartphone has a great battery life and can be used for a long time."

[0498] Step 6:

[0499] The server vectorizes the generated ratings

[0500] The server converts the generated positive ratings into a numerical format (vector format), using techniques such as TF-IDF (Term Frequency-Inverse Document Frequency).

[0501] The generated evaluation features are expressed numerically.

[0502] Step 7:

[0503] The server stores the vectorized ratings in a database.

[0504] The vectorized review texts are stored in a database, which will be used for future multi-level review retrieval and recommendation.

[0505] Step 8:

[0506] The server searches for other products with similar reviews

[0507] The server calculates the similarity of the generated review to other product reviews in the database, using a method such as cosine similarity.

[0508] Identify products with the most similar reviews.

[0509] Step 9:

[0510] The server recommends similar products to the user.

[0511] Based on the calculation results, the server recommends products with the most similar reviews to the user, and displays the recommended results on the user's device.

[0512] For example: "If you're unhappy with your battery, we recommend this phone with better battery life."

[0513] This step allows users to quickly find the best alternative to resolve their dissatisfaction.

[0514] Example 1

[0515] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0516] When a user posts a negative review on an e-commerce platform, it is important to appropriately recommend alternative products to resolve the user's dissatisfaction. Current systems do not accurately identify the user's dissatisfaction from the negative review and recommend alternative products based on that information, which limits the effectiveness of improving the user experience and promoting sales.

[0517] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0518] In this invention, the server includes means for users to post negative reviews of products on the e-commerce platform, means for generating positive reviews from the negative reviews using a generation artificial intelligence, means for converting the generated reviews into a vector format, means for storing the vector format reviews in a database, means for searching for other products with similar reviews based on the stored vector format reviews, means for recommending other products with similar reviews to the user based on the search, and means for pre-processing the negative reviews, thereby making it possible to effectively recommend alternative products based on the user's negative reviews.

[0519] "User" means an individual or legal entity who uses the e-commerce platform to post ratings and reviews of products.

[0520] An "e-commerce platform" is an online system for buying and selling goods and services over the Internet.

[0521] A "negative rating" is a review in which a user expresses dissatisfaction or dissatisfaction with a product or service.

[0522] "Generative AI" refers to AI that generates text using natural language processing and machine learning techniques.

[0523] A "positive rating" is a review that expresses positive content or benefits in relation to a negative rating.

[0524] "Vector format" is a format that expresses sentences and evaluations as numerical data. It is expressed as multiple numerical vectors, making calculations and comparisons easier.

[0525] A "database" is a system for storing and managing information in an organized manner.

[0526] "Similar evaluations" refer to evaluations in which the content of multiple evaluations is similar to one another. This similarity is quantified using a specific calculation method.

[0527] "Preprocessing" refers to processing performed to make raw data easier to analyze. Specifically, it includes tokenization and removal of stop words.

[0528] "Tokenizing" is the process of dividing text data into units of words or phrases.

[0529] "Stop word removal" is the process of removing common words that have no meaning or are not needed during analysis.

[0530] "Cosine similarity" is a method for calculating the similarity between vector data. It calculates the similarity based on the angle between the vectors.

[0531] The present invention is a system that analyzes negative reviews posted by users on an e-commerce platform and recommends alternative products to the users that will resolve their dissatisfaction. This system is composed of a user, a server, a generative artificial intelligence (AI), and a database.

[0532] First, a user accesses an e-commerce platform and posts a negative review in the review section. For example, the user might write, "I'm not happy with this smartphone because the battery runs out quickly," and submit it.

[0533] Next, the server receives these negative reviews and preprocesses them. This preprocessing includes tokenizing (dividing sentences into words) and removing stop words (removing meaningless common words). This preprocessing makes the review text easier to analyze by generative artificial intelligence.

[0534] The preprocessed reviews are then passed to a generative AI model, which uses natural language processing techniques, for example, to convert negative reviews into positive ones. Based on the example above, the generative AI model would generate a positive review such as, "This smartphone has excellent battery life and can be used for a long time."

[0535] The server then converts the generated positive reviews into vector format using techniques such as Term Frequency-Inverse Document Frequency (TF-IDF) and word embedding, and stores the converted reviews in a database.

[0536] Based on the stored vector-format ratings, the server calculates the similarity with the reviews of other products. The similarity is calculated using methods such as cosine similarity. Based on this similarity calculation, the server searches for other products with similar ratings.

[0537] Finally, the server generates a list of similar products and recommends them to the user. The user can refer to the recommended products to find suitable alternative products that can resolve their complaints. For example, the server may display a message such as, "If you are dissatisfied with the battery, we recommend this smartphone with a long battery life."

[0538] Examples of prompt statements:

[0539] User type: "I'm not happy with this phone because the battery dies so quickly."

[0540] Prompt for the AI ​​generator: "I'm not happy with this phone because the battery dies so quickly. Please generate a positive rating for this review."

[0541] Positive review generated: "This phone has a great battery life and can be used for a long time."

[0542] This particular embodiment allows for quick and effective recommendation of alternative products based on users' negative reviews, thereby improving the user experience of the e-commerce platform and promoting product sales.

[0543] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0544] Step 1:

[0545] Users post reviews

[0546] Specific behavior:

[0547] Users access the e-commerce platform's website or app and post negative reviews of products in the review section.

[0548] Input: A negative review such as "I'm not happy with this phone because the battery dies so quickly."

[0549] Output: A negative rating is sent to the server.

[0550] Step 2:

[0551] The server receives the reviews and pre-processes them

[0552] Specific behavior:

[0553] The server receives reviews posted by users and pre-processes the review content.

[0554] Input: The negative rating submitted by the user.

[0555] Data processing: tokenization (splitting sentences into words) and stop word removal (removing meaningless common words).

[0556] Output: Preprocessed text data.

[0557] Step 3:

[0558] The server passes the data to a generative AI model, which generates a positive rating.

[0559] Specific behavior:

[0560] The server passes the preprocessed negative reviews to a generative AI model, which generates positive ratings.

[0561] Input: Preprocessed negative reviews.

[0562] Data computation: Generative AI models generate positive reviews based on negative reviews.

[0563] Output: A positive review generated by the generative AI.

[0564] Step 4:

[0565] The server vectorizes the generated ratings and stores them in a database.

[0566] Specific behavior:

[0567] The server converts the generated positive reviews into vector format and stores them in a database.

[0568] Input: Generated positive reviews.

[0569] Data processing: Convert data into vector format using techniques such as TF-IDF or word embedding.

[0570] Output: Ratings in vector format stored in a database.

[0571] Step 5:

[0572] The server searches for similar products

[0573] Specific behavior:

[0574] The server calculates the similarity between the vectorized ratings and other product reviews in the database and searches for other products with similar ratings.

[0575] Input: Ratings in vector format stored in a database.

[0576] Data calculation: Calculate similarity using methods such as cosine similarity.

[0577] Output: A list of highly similar products.

[0578] Step 6:

[0579] The server recommends different products to the user

[0580] Specific behavior:

[0581] The server generates a list of highly similar products and recommends them to the user.

[0582] Input: A list of similar products.

[0583] Output: The recommended product list displayed to the user.

[0584] Example: "If you're unhappy with your battery, we recommend this smartphone with better battery life."

[0585] As described above, the system's program starts with negative reviews posted by users, processes the data and performs various calculations, and then recommends alternative products that will resolve the user's complaints.

[0586] (Application example 1)

[0587] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0588] On modern e-commerce platforms, negative reviews posted by users are an important factor in product selection. However, these negative reviews often discourage purchases and cause users to abandon the platform. Conventional systems lack a mechanism to effectively utilize these negative reviews to recommend alternative products to users, which results in the inability to adequately address users' dissatisfaction and needs.

[0589] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0590] In this invention, the server includes means for users to post negative reviews of products on the e-commerce platform, means for generating positive reviews from the negative reviews using a generation artificial intelligence, means for converting the generated reviews into vector format, means for storing the vector format reviews in a database, means for searching for other products with similar reviews based on the stored vector format reviews, means for recommending other products with similar reviews to the user based on the search, and means for analyzing the user's needs based on the negative reviews and recommending other products that meet those needs in real time, thereby enabling users to quickly find alternative products that best suit their needs and complaints.

[0591] An "e-commerce platform" is an online system for buying and selling goods and services over the Internet.

[0592] "Negative evaluation" refers to reviews or feedback in which users describe their dissatisfaction or shortcomings with a product or service.

[0593] "Generative AI" refers to AI technology that uses natural language processing technology to generate, translate, and analyze text.

[0594] A "positive rating" is a review that describes the positive aspects or benefits of a product or service for the user.

[0595] "Vector format" is a format in which text data is converted into numerical data and expressed in a multidimensional space.

[0596] A "database" is a system for efficiently storing, managing, and searching data.

[0597] "Cosine similarity" is a method for measuring the similarity between two vectors by calculating the cosine value of the angle between them.

[0598] "Similarity calculation" is a calculation for quantifying the similarity between different data.

[0599] "Recommendation" means that the system presents appropriate products and services based on the user's ratings and preferences.

[0600] "Real-time recommendations" refers to the ability to instantly respond to information entered by the user and recommend appropriate products and services.

[0601] This invention relates to a system that analyzes negative reviews posted by users on e-commerce platforms and recommends alternative products that meet the user's needs in real time.

[0602] The system primarily uses the following hardware and software: React Native for front-end development, Node.js and Express.js for back-end development, and MongoDB for the database. It also utilizes GPT-3 or ChatGPT as a generative AI model, the TF-IDF method for vectorizing text data, and cosine similarity for similarity calculations. The system operates when users post reviews through the e-commerce platform's application.

[0603] First, a user posts a negative review about a product using their device. The posted negative review is sent to the server, which receives the review and performs preprocessing. Preprocessing includes tokenization and stop word removal. The preprocessed negative review is then passed to a generative AI model, which generates a positive review. For example, if a user posts, "This smartphone's battery runs out quickly," the generative AI model will generate a positive review such as, "This smartphone has excellent battery life and can be used for a long time."

[0604] The generated positive reviews are converted into vector format by the server and stored in a database. The server then calculates the similarity between the reviews and those of other products in the database. This similarity is calculated using cosine similarity. Based on the calculated similarity, the server searches for other products that may meet the user's needs and recommends them to the user in real time.

[0605] This system allows users to quickly resolve their complaints and find the best alternative product. For example, if a user is dissatisfied with the battery, a similarity calculation will be used to recommend another smartphone with a better battery life. An example of a prompt used by the generative AI model is as follows:

[0606] Example prompt sentence:

[0607] "A user left a negative review stating that the battery dies quickly. Use this as a basis to generate a positive review."

[0608] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0609] Step 1:

[0610] The user uses the terminal to input and post a negative review in the review section of the e-commerce platform.

[0611] Input: Negative review text (e.g., "I'm not happy with this phone because the battery runs out quickly.")

[0612] Output: The posted negative review text is sent to the server.

[0613] Step 2:

[0614] The server receives negative feedback from users and performs preprocessing, which includes tokenization (splitting text into words) and stop-word removal (removing frequently occurring words).

[0615] Input: The negative review text posted.

[0616] Output: Preprocessed negative evaluation text (text after tokenization and stopword removal).

[0617] Step 3:

[0618] The server passes the preprocessed negative evaluation text to a generative AI model (e.g., GPT-3) to generate a positive evaluation. For example, in response to the negative evaluation "The battery runs out quickly," a positive evaluation such as "The battery life is very good and can be used for a long time" is generated.

[0619] Input: Preprocessed negative rating text.

[0620] Output: The generated positive review text.

[0621] Step 4:

[0622] The server converts the generated positive feedback text into vector format using the Term Frequency-Inverse Document Frequency (TF-IDF) method.

[0623] Input: The generated positive review text.

[0624] Output: Vector format data.

[0625] Step 5:

[0626] The server stores the vector data in a database.

[0627] Input: Vector format data.

[0628] Output: Data in vector format stored in a database.

[0629] Step 6:

[0630] The server calculates the similarity between the vector data and other product reviews in the database. Cosine similarity is used to calculate this similarity.

[0631] Input: Vector data from the database and newly generated vector data.

[0632] Output: Similarity calculation result.

[0633] Step 7:

[0634] Based on the similarity calculation results, the server searches for other products that suit the user's needs and recommends them to the user in real time.

[0635] Input: Similarity calculation result.

[0636] Output: A list of products recommended to the user.

[0637] This allows users to quickly find the best alternative product to resolve their dissatisfaction.

[0638] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[0639] The present invention combines a system that analyzes negative reviews posted by users on e-commerce platforms and recommends alternative products to users that will resolve their dissatisfaction with an emotion engine that recognizes users' emotions.

[0640] The system mainly includes the following means:

[0641] A means for users to post negative product reviews on e-commerce platforms

[0642] A means for the server to generate positive reviews from the negative reviews using artificial intelligence

[0643] means for the server to convert the generated ratings into vector form;

[0644] means for storing the vector-formed evaluations in a database;

[0645] A means for the server to search for other products having similar ratings based on the stored vector-format ratings.

[0646] A means for the server to recommend other products with similar ratings to the user based on the search.

[0647] an emotion engine that recognizes the user's emotions contained in the negative reviews and the generated reviews;

[0648] Program processing and specific examples

[0649] 1. Users post reviews

[0650] Users can post negative reviews of the product in the review section of the e-commerce platform.

[0651] For example: "I'm not happy with this phone because the battery dies so quickly."

[0652] 2. The server receives the review

[0653] The server immediately receives the user's post, registers it internally, and passes it on to the next step for processing.

[0654] 3. The server pre-processes the review

[0655] The server performs preprocessing such as tokenizing the review sentences and removing stop words, so that the review sentences are ready for analysis and generation.

[0656] 4. The server passes the review to the sentiment engine

[0657] The preprocessed review sentences are passed to the emotion engine, which recognizes the user's emotions from the review sentences and extracts the emotion information.

[0658] 5. Generative AI generates positive evaluations based on negative evaluations using emotional information

[0659] The generative AI analyzes negative reviews and generates review texts that eliminate negative elements based on emotional information obtained from the emotion engine.

[0660] For example: "This smartphone has a great battery life and can be used for a long time."

[0661] 6. The server vectorizes the generated ratings

[0662] The server converts the generated positive reviews into a numerical format (vector format). For vectorization, techniques such as TF-IDF are used. The features of the generated reviews are expressed numerically.

[0663] 7. The server stores the vectorized ratings in a database

[0664] The vectorized review texts are stored in a database, which will be used for future multi-level review retrieval and recommendation.

[0665] 8. The server searches for other products with similar reviews

[0666] The server calculates the similarity of the generated review to other product reviews in the database, using a method such as cosine similarity, and identifies the product with the most similar review.

[0667] 9. The server recommends similar products to the user

[0668] The server then recommends the product with the most similar reviews to the user based on the calculation results, and displays the recommended results on the user's device.

[0669] For example: "If you're unhappy with your battery, we recommend this phone with better battery life."

[0670] This system allows users to efficiently find suitable alternative products to resolve their dissatisfaction. In addition, the emotion engine enables more appropriate recommendations that take into account the emotions expressed in users' ratings, improving the user experience.

[0671] The processing flow will be explained below.

[0672] Step 1:

[0673] Users post reviews

[0674] Users access the review section of the product they have purchased and enter a negative rating to post.

[0675] For example: "I'm not happy with this phone because the battery dies so quickly."

[0676] Step 2:

[0677] The server receives the review

[0678] The server receives user submissions instantly, and the received reviews are stored in an internal log and prepared for further processing.

[0679] Step 3:

[0680] The server pre-processes the review

[0681] The server pre-processes the review text, including tokenizing (dividing the text into words), removing stop words, and normalizing (e.g., converting all text to lowercase), making the review text optimized for analysis.

[0682] Preprocessing result example: "Smartphone battery runs out quickly, dissatisfaction"

[0683] Step 4:

[0684] The server passes the review to the sentiment engine

[0685] The preprocessed review sentences are passed to the sentiment engine, which extracts user sentiment from the reviews.

[0686] Example: "Dissatisfied"

[0687] Step 5:

[0688] Generative AI generates positive evaluations based on emotional information

[0689] The generative AI uses extracted emotional information (e.g., "dissatisfied") to turn negative evaluations into positive ones.

[0690] For example: "This smartphone has a great battery life and can be used for a long time."

[0691] Step 6:

[0692] The server vectorizes the generated ratings

[0693] The server converts the generated positive ratings into vector format using techniques such as TF-IDF (Term Frequency-Inverse Document Frequency) and Word2Vec.

[0694] Generated ratings in vector format, e.g.: \[0.15, 0.34, 0.01, ... \]

[0695] Step 7:

[0696] The server stores the vectorized ratings in a database.

[0697] The server stores the vectorized review text in a database, which will be used in future search and recommendation algorithms.

[0698] Step 8:

[0699] The server searches for other products with similar reviews

[0700] The server calculates the similarity between the reviews of other products stored in the database and the generated review vector. This similarity is calculated using cosine similarity.

[0701] Identify the most similar products.

[0702] Step 9:

[0703] The server recommends similar products to the user.

[0704] The server then uses the results of the calculation to recommend products with the most similar reviews to the user, displaying the recommendations on the user's device and suggesting alternative products.

[0705] For example: "If you're unhappy with your battery, we recommend this phone with better battery life."

[0706] Through this step, users can easily find the best alternative product to resolve their dissatisfaction.By using the emotion engine, we can understand the emotions expressed by users in their reviews and make more appropriate product recommendations, thereby further increasing user satisfaction.

[0707] Example 2

[0708] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0709] In conventional e-commerce platforms, negative reviews posted by users are often not properly addressed. As a result, users' dissatisfaction remains unresolved, potentially leading to customer churn. In addition, the lack of recommendations for alternative products based on negative reviews creates a problem of not improving the user experience.

[0710] The identification process by the identification processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means. In this invention, the server includes: means for a user to post a negative review of a product on an e-commerce platform; means for generating a positive review from the negative review using a generative artificial intelligence; means for converting the generated review into a vector format; means for storing the vector-format review in a database; means for searching for other products with similar reviews based on the stored vector-format review; means for recommending other products with similar reviews to the user based on the search; means including an emotion engine that recognizes the user's emotions from the review text and extracts the emotion information; and means for generating a positive review from the negative review based on the emotion information. This makes it possible to resolve the user's dissatisfaction by analyzing the user's negative review and recommending appropriate alternative products, thereby providing a more satisfying user experience.

[0711] "User" means any individual or legal entity who uses the e-commerce platform to post reviews on products.

[0712] An "e-commerce platform" is an online commerce system that allows users to purchase products and post reviews.

[0713] A "negative rating" is a review in which the user points out dissatisfaction or problems with the product.

[0714] "Generative artificial intelligence" is a computer program that uses natural language processing technology to generate new text.

[0715] A "positive evaluation" is a review that states the good points and satisfaction points of the product.

[0716] "Vector format" refers to text data converted into a numerical format that is easy to process with machine learning algorithms.

[0717] A "database" is a system for effectively storing, retrieving, and managing structured data.

[0718] An "emotion engine" is a program that uses natural language processing technology to extract and recognize a user's emotional information from text.

[0719] "Preprocessing" refers to a series of processing steps to convert text data into a format suitable for analysis or generation.

[0720] "Cosine similarity" is a mathematical method for calculating the similarity between two vectors based on the vector's direction.

[0721] This invention relates to a system that analyzes negative reviews posted by users on an e-commerce platform and recommends alternative products to the users that will resolve their dissatisfaction. This system achieves more appropriate product recommendations by incorporating an emotion engine that recognizes the user's emotions.

[0722] Hardware or software used

[0723] 1. Server: Responsible for central data processing of this system. Cloud servers can also be used if necessary.

[0724] 2. Device: The client device (computer, smartphone, tablet, etc.) through which a user posts a review.

[0725] 3. Database: A data storage system for storing reviews, vector-format ratings and other related data.

[0726] 4. Generative AI: An AI model that uses natural language processing technology to generate positive reviews based on negative ones. Examples include OpenAI's GPT.

[0727] 5. Sentiment Engine: A program that uses natural language processing techniques to extract user sentiment from reviews. It uses a model for sentiment analysis (e.g., a BERT-based model).

[0728] Data processing or data calculation

[0729] 1. Review preprocessing: Negative reviews posted by users are tokenized using a natural language processing library (e.g., Python's NLTK) and stop words are removed. This preprocessing converts the reviews into a format suitable for analysis and generation.

[0730] 2. Extracting Emotional Information: The preprocessed review sentences are passed to the emotion engine, which extracts emotional information and recognizes emotion tags such as "sad" and "disappointed."

[0731] 3. Generating Positive Evaluations: The generative AI generates positive evaluations from negative evaluations based on the emotional information extracted from the emotion engine. For example, the generative AI uses OpenAI's GPT model to invert emotion tags such as "sad" and "disappointed" to "joy" and "satisfied."

[0732] 4. Vectorization: The generated positive ratings are converted into a vector format that is easy for machine learning algorithms to process, using tools such as scikit-learn's TF-IDF Vectorizer.

[0733] 5. Similarity calculation: Calculate the similarity with other product reviews in the database. To calculate similarity, use scikit-learn's cosine_similarity.

[0734] 6. Product Recommendation: Based on the calculation results, we identify the products with the most similar reviews and recommend them to the user.

[0735] Specific examples

[0736] An example prompt for this system would be:

[0737] "Enter a negative review and generate a positive one based on it."

[0738] This allows users to more efficiently find alternative products that can resolve their dissatisfaction, improving the user experience by making recommendations that take emotions into account.

[0739] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0740] Step 1:

[0741] Users post reviews

[0742] The user enters a negative review of the product in the review section of the e-commerce platform and presses the post button.

[0743] Input: User negative review text

[0744] Output: Posting a negative review to an e-commerce platform

[0745] Specific operation: A user uses a device (PC, smartphone, tablet, etc.) to access an online shopping site, enter a review in text format, and press the post button.

[0746] Step 2:

[0747] The server receives the review

[0748] The server immediately receives reviews posted by users and registers them in a database.

[0749] Input: User-submitted negative review

[0750] Output: Negative reviews stored in the database

[0751] Specific operation: The server uses the receiving API to accept review submissions and saves the data in a database management system.

[0752] Step 3:

[0753] The server pre-processes the review

[0754] The server performs preprocessing such as tokenizing the review sentences and removing stop words, converting the review sentences into a format suitable for analysis and generation.

[0755] Input: Negative reviews retrieved from the database

[0756] Output: Preprocessed review sentences

[0757] Specific operation: Using Python's NLTK library, the review sentences are split into words and frequent meaningless words (stop words) are removed.

[0758] Step 4:

[0759] The server passes the review to the sentiment engine

[0760] The preprocessed review sentences are passed to the emotion engine, which recognizes emotions from the user's review sentences and extracts the emotion information.

[0761] Input: Preprocessed review text

[0762] Output: Emotion information (e.g., emotion tags such as "sad" or "disappointed")

[0763] Specific operation: Calls an API that passes text to the emotion engine, and receives emotion information in return.

[0764] Step 5:

[0765] Generative AI generates positive evaluations based on negative evaluations and emotional information

[0766] The generative AI analyzes negative ratings and generates positive reviews based on emotional information obtained from the emotion engine.

[0767] Input: Negative evaluation and emotional information

[0768] Output: Positive review text

[0769] Specific operation: Using OpenAI's GPT model, etc., the system generates review sentences by converting negative elements into their opposite positive elements (e.g., converting "sad" and "disappointed" into "joy" and "satisfied").

[0770] Step 6:

[0771] The server vectorizes the generated ratings

[0772] The server converts the generated positive reviews into a numerical format (vector format).

[0773] Input: Generated positive review text

[0774] Output: Evaluation data in vector format

[0775] Specific operation: Converts text into vector format using scikit-learn's TF-IDF Vectorizer.

[0776] Step 7:

[0777] The server stores the vectorized ratings in a database.

[0778] The server stores the vectorized review sentences in a database.

[0779] Input: Vectorized evaluation data

[0780] Output: Evaluation data in vector format stored in a database

[0781] Specific behavior: Executes a query to insert vector data into an SQL or NoSQL database.

[0782] Step 8:

[0783] The server searches for other products with similar reviews

[0784] The server calculates the similarity between the generated review and reviews of other products in the database.

[0785] Input: Vector-formatted reviews and generated vector-formatted ratings data in the database

[0786] Output: Evaluation data with similarity calculated

[0787] Specific operation: Uses scikit-learn's cosine_similarity function to calculate the similarity with other vectors in the database and identify the most similar vector.

[0788] Step 9:

[0789] The server recommends similar products to the user.

[0790] Based on the calculation results, the server recommends products with the most similar reviews to the user.

[0791] Input: A list of similar products

[0792] Output: Recommended products displayed on the user's device

[0793] Specific behavior: Generates an API response to display recommended results on the user's device and returns the results to the browser or app.

[0794] (Application example 2)

[0795] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0796] Current e-commerce platforms lack a system that uses negative feedback from users to appropriately recommend alternative products to resolve their dissatisfaction. This means users have to spend time and effort to find a satisfactory product. Furthermore, existing systems do not take user sentiment into account when making recommendations, which does not adequately improve the user experience.

[0797] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for a user to post a negative review of a product on an e-commerce platform, means for generating a positive review from the negative review using a generative artificial intelligence, and means for converting the generated review into a vector format. This makes it possible to pass the user's negative review to an emotion recognition engine, extract the user's emotional information, generate a positive review based on the emotional information, search for other products with similar reviews based on the vector-format review stored in the database, and recommend other products with similar reviews to the user based on the search.

[0798] "User" means any individual or organization using the e-commerce platform.

[0799] An "e-commerce platform" is an online system for buying and selling goods and services over the Internet.

[0800] "Negative ratings" are reviews or feedback where users express dissatisfaction with a product or service.

[0801] "Generative AI" is an algorithm or system that uses natural language processing techniques to generate new text.

[0802] "Positive reviews" are positive reviews or feedback that indicate satisfaction with a product or service.

[0803] "Vector format" is a format in which text data is converted into a numerical vector.

[0804] A "database" is a system for efficiently storing, managing, and retrieving data.

[0805] A "search method" is an algorithm or method for finding specific data in a database.

[0806] A "recommendation method" is a system or method for suggesting appropriate products or services to users.

[0807] An "emotion recognition engine" is an artificial intelligence technology for extracting user emotional information from text data.

[0808] "Emotion information" is data that indicates the user's emotional state.

[0809] A "prompt" is an input sentence or cue that the generative AI uses to generate text.

[0810] "Similarity" is an index that indicates the similarity between two pieces of data.

[0811] "Cosine similarity" is a method of calculating similarity using the cosine angle between two vectors.

[0812] A specific system for implementing this invention is one that allows users to post negative reviews of products using an e-commerce platform, and based on those reviews, recommends alternative products to resolve the user's dissatisfaction. This system is mainly composed of a combination of a program, hardware, and a database.

[0813] 1. Program Generation and Explanation

[0814] The server receives negative reviews posted by users and preprocesses them. This preprocessing step involves tokenization and stopword removal to convert the reviews into a more easily parseable format. The server then passes the preprocessed reviews to an emotion recognition engine (e.g., using the Hugging Face transformers library) to extract the user's emotional information.

[0815] Based on the emotional information obtained from the emotion engine, the server uses generative artificial intelligence (e.g., a Hugging Face text-generation model) to convert negative feedback into positive feedback. In this step, the prompt sentence is used to input the generative AI model, which generates a positive feedback.

[0816] Once a positive review is generated, the server converts it into a vector format using techniques such as TF-IDF, which is then stored in a database for future searches and recommendations.

[0817] The server then searches for other products with similar ratings based on the stored vector of ratings. It uses cosine similarity to calculate similarity and identifies products with the most similar reviews. Finally, the server recommends products with the highest similarity to the user, and the recommended results are displayed on the user's device.

[0818] 2. Hardware and Software Used

[0819] The system requires advanced hardware and software for text analysis and generation, including high-performance servers and database servers, and Python, SQL, and the Hugging Face transformers library.

[0820] 3. Examples of concrete examples and prompts

[0821] For example, if a user posts a review stating, "I'm dissatisfied with this smartphone because the battery runs out quickly," the system receives it and performs preprocessing and sentiment analysis. Based on the results of the sentiment analysis, the generative AI model generates a positive evaluation such as, "This smartphone has excellent battery life and can be used for a long time." This evaluation is then vectorized and stored in a database.

[0822] The server then calculates the similarity of the generated rating to other ratings in the database, identifies the product with the most similar rating, and recommends this product to the user, saying, "If you're dissatisfied with your battery, we recommend this smartphone with a long battery life."

[0823] Example prompt for a generative AI model:

[0824] User expressed dissatisfaction with: "I'm not happy with this phone because the battery runs out so quickly." Write a positive review.

[0825] This system allows users to quickly find suitable alternative products to resolve their dissatisfaction, improving the user experience.

[0826] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0827] Step 1:

[0828] A user posts a negative review of a product on an e-commerce platform. For example, the user might write, "I'm not happy with this smartphone because the battery runs out quickly." This review becomes the input to the system.

[0829] Step 2:

[0830] The server receives negative reviews from users. It registers the received reviews internally and passes them to the next step. The input is the user's negative review, and the output is the data passed to the preprocessing step.

[0831] Step 3:

[0832] The server preprocesses the reviews. This includes tokenizing the text, removing stop words, etc. It takes the received negative reviews as input and converts them into a format that is easy to parse. The output is the preprocessed text data.

[0833] Step 4:

[0834] The server passes the preprocessed reviews to the emotion recognition engine. Using the Hugging Face transformers library, the user's emotional information is extracted. The input is preprocessed text data, and after processing by the emotion recognition engine, the output is emotional information. For example, a result such as "Negative (90% probability)" can be obtained.

[0835] Step 5:

[0836] The server generates a prompt based on the emotion information and passes it to the AI ​​generator. For example, it generates a prompt like, "User expressed dissatisfaction with: 'I'm dissatisfied with this smartphone because the battery runs out quickly.' Write a positive review." The input is the emotion information and the preprocessed review, and the output is the prompt text.

[0837] Step 6:

[0838] Generative AI generates a positive review based on the prompt. It uses Hugging Face's text-generation model to create a positive review that eliminates negative elements. The input is the prompt text, and the output is the generated positive review. For example, you might get a review like, "This smartphone has excellent battery life and can be used for a long time."

[0839] Step 7:

[0840] The server vectorizes the generated positive reviews. It uses techniques such as TF-IDF to convert text data into a numerical format. The input is the generated positive reviews, and the output is vector-format data.

[0841] Step 8:

[0842] The server stores the generated rating vector in a database, which is used for future searches and recommendations. The input is vector data, and the output is the completed storage in the database.

[0843] Step 9:

[0844] The server calculates the similarity of the generated review to other reviews in the database. It uses methods such as cosine similarity to find the product with the most similar reviews. The input is rating data in vector format, and the output is the product with the most similar review.

[0845] Step 10:

[0846] The server recommends highly similar products to the user. Based on the calculation results, the most suitable product recommendation is displayed on the user's device. The input is the product information from the search results, and the output is the recommendation to the user. For example, a recommendation such as "If you are dissatisfied with your battery, we recommend this smartphone with a long battery life" is displayed.

[0847] The specific processing unit 290 transmits the result of the specific processing to the smart glasses 214. In the smart glasses 214, the control unit 46A causes the speaker 240 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0848] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0849] In the above embodiment, an example in which the specific processing is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the smart glasses 214.

[0850] [Third embodiment]

[0851] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0852] 5, the data processing system 310 includes the data processing device 12 and a headset type terminal 314. An example of the data processing device 12 is a server.

[0853] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0854] The headset type terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a display 343. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the display 343 are also connected to the bus 52.

[0855] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0856] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0857] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0858] Fig. 6 shows an example of the main functions of the data processing device 12 and the headset type terminal 314. As shown in Fig. 6, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0859] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0860] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0861] In the headset type terminal 314, a reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0862] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the headset type terminal 314 will be referred to as the "terminal."

[0863] The present invention is a system that analyzes negative reviews posted by users on an e-commerce platform and recommends to the users alternative products that will resolve their dissatisfaction.

[0864] The system mainly includes the following means:

[0865] A means for users to post negative product reviews on e-commerce platforms

[0866] A means for the server to generate positive reviews from the negative reviews using artificial intelligence

[0867] means for the server to convert the generated ratings into vector form;

[0868] means for storing the vector-formed evaluations in a database;

[0869] A means for the server to search for other products having similar ratings based on the stored vector-format ratings.

[0870] A means for the server to recommend other products with similar ratings to the user based on the search.

[0871] Program processing and specific examples

[0872] 1. Users post reviews

[0873] Users can post negative reviews of the product in the review section of the e-commerce platform.

[0874] For example: "I'm not happy with this phone because the battery dies so quickly."

[0875] 2. The server receives the review and passes it to the generation AI.

[0876] The server receives user submissions, performs preprocessing, and then passes negative reviews to the AI ​​generator.

[0877] Preprocessing includes tokenization and stop word removal.

[0878] 3. Generative AI generates positive reviews

[0879] Generative AI generates positive evaluations based on negative evaluations.

[0880] For example: "This smartphone has a great battery life and can be used for a long time."

[0881] 4. The server vectorizes the generated ratings and stores them in a database.

[0882] The server converts the generated positive reviews into vector format and stores them in a database.

[0883] Techniques such as TF-IDF (Term Frequency-Inverse Document Frequency) are used for vectorization.

[0884] 5. The server searches for other products with similar reviews

[0885] The server calculates the similarity between the vectorized rating and other product reviews in the database.

[0886] For this similarity calculation, a method such as cosine similarity is used.

[0887] 6. The server recommends another product to the user

[0888] The server generates a list of highly similar products and recommends them to the user.

[0889] For example: "If you're unhappy with your battery, we recommend this phone with better battery life."

[0890] This system allows users to efficiently find suitable alternative products to resolve their dissatisfaction, and also helps e-commerce platform operators improve user experience and reduce barriers to sales.

[0891] The processing flow will be explained below.

[0892] Step 1:

[0893] A user posts a review on an e-commerce platform

[0894] Users access the review section of the product they have purchased and enter a negative rating to post.

[0895] For example: "I'm not happy with this phone because the battery dies so quickly."

[0896] Step 2:

[0897] The server receives the review

[0898] The server receives user submissions instantly.

[0899] Registers the received review internally and passes it to the next step for processing.

[0900] Step 3:

[0901] The server pre-processes the review

[0902] The server pre-processes the review sentences. In this step, the review sentences are tokenized (divided into words) and stop words (common words with little meaning) are removed.

[0903] Preprocessing makes the review sentences ready for analysis and generation.

[0904] Step 4:

[0905] The server passes the review to the generation AI

[0906] The preprocessed review text is sent to the AI ​​generator, which analyzes the received review text and identifies negative elements.

[0907] Step 5:

[0908] Generative AI generates positive reviews

[0909] The artificial intelligence generates a review sentence with negative elements removed.

[0910] For example: "This smartphone has a great battery life and can be used for a long time."

[0911] Step 6:

[0912] The server vectorizes the generated ratings

[0913] The server converts the generated positive ratings into a numerical format (vector format), using techniques such as TF-IDF (Term Frequency-Inverse Document Frequency).

[0914] The generated evaluation features are expressed numerically.

[0915] Step 7:

[0916] The server stores the vectorized ratings in a database.

[0917] The vectorized review texts are stored in a database, which will be used for future multi-level review retrieval and recommendation.

[0918] Step 8:

[0919] The server searches for other products with similar reviews

[0920] The server calculates the similarity of the generated review to other product reviews in the database, using a method such as cosine similarity.

[0921] Identify products with the most similar reviews.

[0922] Step 9:

[0923] The server recommends similar products to the user.

[0924] Based on the calculation results, the server recommends products with the most similar reviews to the user, and displays the recommended results on the user's device.

[0925] For example: "If you're unhappy with your battery, we recommend this phone with better battery life."

[0926] This step allows users to quickly find the best alternative to resolve their dissatisfaction.

[0927] Example 1

[0928] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[0929] When a user posts a negative review on an e-commerce platform, it is important to appropriately recommend alternative products to resolve the user's dissatisfaction. Current systems do not accurately identify the user's dissatisfaction from the negative review and recommend alternative products based on that information, which limits the effectiveness of improving the user experience and promoting sales.

[0930] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0931] In this invention, the server includes means for users to post negative reviews of products on the e-commerce platform, means for generating positive reviews from the negative reviews using a generation artificial intelligence, means for converting the generated reviews into a vector format, means for storing the vector format reviews in a database, means for searching for other products with similar reviews based on the stored vector format reviews, means for recommending other products with similar reviews to the user based on the search, and means for pre-processing the negative reviews, thereby making it possible to effectively recommend alternative products based on the user's negative reviews.

[0932] "User" means an individual or legal entity who uses the e-commerce platform to post ratings and reviews of products.

[0933] An "e-commerce platform" is an online system for buying and selling goods and services over the Internet.

[0934] A "negative rating" is a review in which a user expresses dissatisfaction or dissatisfaction with a product or service.

[0935] "Generative AI" refers to AI that generates text using natural language processing and machine learning techniques.

[0936] A "positive rating" is a review that expresses positive content or benefits in relation to a negative rating.

[0937] "Vector format" is a format that expresses sentences and evaluations as numerical data. It is expressed as multiple numerical vectors, making calculations and comparisons easier.

[0938] A "database" is a system for storing and managing information in an organized manner.

[0939] "Similar evaluations" refer to evaluations in which the content of multiple evaluations is similar to one another. This similarity is quantified using a specific calculation method.

[0940] "Preprocessing" refers to processing performed to make raw data easier to analyze. Specifically, it includes tokenization and removal of stop words.

[0941] "Tokenizing" is the process of dividing text data into units of words or phrases.

[0942] "Stop word removal" is the process of removing common words that have no meaning or are not needed during analysis.

[0943] "Cosine similarity" is a method for calculating the similarity between vector data. It calculates the similarity based on the angle between the vectors.

[0944] The present invention is a system that analyzes negative reviews posted by users on an e-commerce platform and recommends alternative products to the users that will resolve their dissatisfaction. This system is composed of a user, a server, a generative artificial intelligence (AI), and a database.

[0945] First, a user accesses an e-commerce platform and posts a negative review in the review section. For example, the user might write, "I'm not happy with this smartphone because the battery runs out quickly," and submit it.

[0946] Next, the server receives these negative reviews and preprocesses them. This preprocessing includes tokenizing (dividing sentences into words) and removing stop words (removing meaningless common words). This preprocessing makes the review text easier to analyze by generative artificial intelligence.

[0947] The preprocessed reviews are then passed to a generative AI model, which uses natural language processing techniques, for example, to convert negative reviews into positive ones. Based on the example above, the generative AI model would generate a positive review such as, "This smartphone has excellent battery life and can be used for a long time."

[0948] The server then converts the generated positive reviews into vector format using techniques such as Term Frequency-Inverse Document Frequency (TF-IDF) and word embedding, and stores the converted reviews in a database.

[0949] Based on the stored vector-format ratings, the server calculates the similarity with the reviews of other products. The similarity is calculated using methods such as cosine similarity. Based on this similarity calculation, the server searches for other products with similar ratings.

[0950] Finally, the server generates a list of similar products and recommends them to the user. The user can refer to the recommended products to find suitable alternative products that can resolve their complaints. For example, the server may display a message such as, "If you are dissatisfied with the battery, we recommend this smartphone with a long battery life."

[0951] Examples of prompt statements:

[0952] User type: "I'm not happy with this phone because the battery dies so quickly."

[0953] Prompt for the AI ​​generator: "I'm not happy with this phone because the battery dies so quickly. Please generate a positive rating for this review."

[0954] Positive review generated: "This phone has a great battery life and can be used for a long time."

[0955] This particular embodiment allows for quick and effective recommendation of alternative products based on users' negative reviews, thereby improving the user experience of the e-commerce platform and promoting product sales.

[0956] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0957] Step 1:

[0958] Users post reviews

[0959] Specific behavior:

[0960] Users access the e-commerce platform's website or app and post negative reviews of products in the review section.

[0961] Input: A negative review such as "I'm not happy with this phone because the battery dies so quickly."

[0962] Output: A negative rating is sent to the server.

[0963] Step 2:

[0964] The server receives the reviews and pre-processes them

[0965] Specific behavior:

[0966] The server receives reviews posted by users and pre-processes the review content.

[0967] Input: The negative rating submitted by the user.

[0968] Data processing: tokenization (splitting sentences into words) and stop word removal (removing meaningless common words).

[0969] Output: Preprocessed text data.

[0970] Step 3:

[0971] The server passes the data to a generative AI model, which generates a positive rating.

[0972] Specific behavior:

[0973] The server passes the preprocessed negative reviews to a generative AI model, which generates positive ratings.

[0974] Input: Preprocessed negative reviews.

[0975] Data computation: Generative AI models generate positive reviews based on negative reviews.

[0976] Output: A positive review generated by the generative AI.

[0977] Step 4:

[0978] The server vectorizes the generated ratings and stores them in a database.

[0979] Specific behavior:

[0980] The server converts the generated positive reviews into vector format and stores them in a database.

[0981] Input: Generated positive reviews.

[0982] Data processing: Convert data into vector format using techniques such as TF-IDF or word embedding.

[0983] Output: Ratings in vector format stored in a database.

[0984] Step 5:

[0985] The server searches for similar products

[0986] Specific behavior:

[0987] The server calculates the similarity between the vectorized ratings and other product reviews in the database and searches for other products with similar ratings.

[0988] Input: Ratings in vector format stored in a database.

[0989] Data calculation: Calculate similarity using methods such as cosine similarity.

[0990] Output: A list of highly similar products.

[0991] Step 6:

[0992] The server recommends different products to the user

[0993] Specific behavior:

[0994] The server generates a list of highly similar products and recommends them to the user.

[0995] Input: A list of similar products.

[0996] Output: The recommended product list displayed to the user.

[0997] Example: "If you're unhappy with your battery, we recommend this smartphone with better battery life."

[0998] As described above, the system's program starts with negative reviews posted by users, processes the data and performs various calculations, and then recommends alternative products that will resolve the user's complaints.

[0999] (Application example 1)

[1000] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1001] On modern e-commerce platforms, negative reviews posted by users are an important factor in product selection. However, these negative reviews often discourage purchases and cause users to abandon the platform. Conventional systems lack a mechanism to effectively utilize these negative reviews to recommend alternative products to users, which results in the inability to adequately address users' dissatisfaction and needs.

[1002] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1003] In this invention, the server includes means for users to post negative reviews of products on the e-commerce platform, means for generating positive reviews from the negative reviews using a generation artificial intelligence, means for converting the generated reviews into vector format, means for storing the vector format reviews in a database, means for searching for other products with similar reviews based on the stored vector format reviews, means for recommending other products with similar reviews to the user based on the search, and means for analyzing the user's needs based on the negative reviews and recommending other products that meet those needs in real time, thereby enabling users to quickly find alternative products that best suit their needs and complaints.

[1004] An "e-commerce platform" is an online system for buying and selling goods and services over the Internet.

[1005] "Negative evaluation" refers to reviews or feedback in which users describe their dissatisfaction or shortcomings with a product or service.

[1006] "Generative AI" refers to AI technology that uses natural language processing technology to generate, translate, and analyze text.

[1007] A "positive rating" is a review that describes the positive aspects or benefits of a product or service for the user.

[1008] "Vector format" is a format in which text data is converted into numerical data and expressed in a multidimensional space.

[1009] A "database" is a system for efficiently storing, managing, and searching data.

[1010] "Cosine similarity" is a method for measuring the similarity between two vectors by calculating the cosine value of the angle between them.

[1011] "Similarity calculation" is a calculation for quantifying the similarity between different data.

[1012] "Recommendation" means that the system presents appropriate products and services based on the user's ratings and preferences.

[1013] "Real-time recommendations" refers to the ability to instantly respond to information entered by the user and recommend appropriate products and services.

[1014] This invention relates to a system that analyzes negative reviews posted by users on e-commerce platforms and recommends alternative products that meet the user's needs in real time.

[1015] The system primarily uses the following hardware and software: React Native for front-end development, Node.js and Express.js for back-end development, and MongoDB for the database. It also utilizes GPT-3 or ChatGPT as a generative AI model, the TF-IDF method for vectorizing text data, and cosine similarity for similarity calculations. The system operates when users post reviews through the e-commerce platform's application.

[1016] First, a user posts a negative review about a product using their device. The posted negative review is sent to the server, which receives the review and performs preprocessing. Preprocessing includes tokenization and stop word removal. The preprocessed negative review is then passed to a generative AI model, which generates a positive review. For example, if a user posts, "This smartphone's battery runs out quickly," the generative AI model will generate a positive review such as, "This smartphone has excellent battery life and can be used for a long time."

[1017] The generated positive reviews are converted into vector format by the server and stored in a database. The server then calculates the similarity between the reviews and those of other products in the database. This similarity is calculated using cosine similarity. Based on the calculated similarity, the server searches for other products that may meet the user's needs and recommends them to the user in real time.

[1018] This system allows users to quickly resolve their complaints and find the best alternative product. For example, if a user is dissatisfied with the battery, a similarity calculation will be used to recommend another smartphone with a better battery life. An example of a prompt used by the generative AI model is as follows:

[1019] Example prompt sentence:

[1020] "A user left a negative review stating that the battery dies quickly. Use this as a basis to generate a positive review."

[1021] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1022] Step 1:

[1023] The user uses the terminal to input and post a negative review in the review section of the e-commerce platform.

[1024] Input: Negative review text (e.g., "I'm not happy with this phone because the battery runs out quickly.")

[1025] Output: The posted negative review text is sent to the server.

[1026] Step 2:

[1027] The server receives negative feedback from users and performs preprocessing, which includes tokenization (splitting text into words) and stop-word removal (removing frequently occurring words).

[1028] Input: The negative review text posted.

[1029] Output: Preprocessed negative evaluation text (text after tokenization and stopword removal).

[1030] Step 3:

[1031] The server passes the preprocessed negative evaluation text to a generative AI model (e.g., GPT-3) to generate a positive evaluation. For example, in response to the negative evaluation "The battery runs out quickly," a positive evaluation such as "The battery life is very good and can be used for a long time" is generated.

[1032] Input: Preprocessed negative rating text.

[1033] Output: The generated positive review text.

[1034] Step 4:

[1035] The server converts the generated positive feedback text into vector format using the Term Frequency-Inverse Document Frequency (TF-IDF) method.

[1036] Input: The generated positive review text.

[1037] Output: Vector format data.

[1038] Step 5:

[1039] The server stores the vector data in a database.

[1040] Input: Vector format data.

[1041] Output: Data in vector format stored in a database.

[1042] Step 6:

[1043] The server calculates the similarity between the vector data and other product reviews in the database. Cosine similarity is used to calculate this similarity.

[1044] Input: Vector data from the database and newly generated vector data.

[1045] Output: Similarity calculation result.

[1046] Step 7:

[1047] Based on the similarity calculation results, the server searches for other products that suit the user's needs and recommends them to the user in real time.

[1048] Input: Similarity calculation result.

[1049] Output: A list of products recommended to the user.

[1050] This allows users to quickly find the best alternative product to resolve their dissatisfaction.

[1051] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1052] The present invention combines a system that analyzes negative reviews posted by users on e-commerce platforms and recommends alternative products to users that will resolve their dissatisfaction with an emotion engine that recognizes users' emotions.

[1053] The system mainly includes the following means:

[1054] A means for users to post negative product reviews on e-commerce platforms

[1055] A means for the server to generate positive reviews from the negative reviews using artificial intelligence

[1056] means for the server to convert the generated ratings into vector form;

[1057] means for storing the vector-formed evaluations in a database;

[1058] A means for the server to search for other products having similar ratings based on the stored vector-format ratings.

[1059] A means for the server to recommend other products with similar ratings to the user based on the search.

[1060] an emotion engine that recognizes the user's emotions contained in the negative reviews and the generated reviews;

[1061] Program processing and specific examples

[1062] 1. Users post reviews

[1063] Users can post negative reviews of the product in the review section of the e-commerce platform.

[1064] For example: "I'm not happy with this phone because the battery dies so quickly."

[1065] 2. The server receives the review

[1066] The server immediately receives the user's post, registers it internally, and passes it on to the next step for processing.

[1067] 3. The server pre-processes the review

[1068] The server performs preprocessing such as tokenizing the review sentences and removing stop words, so that the review sentences are ready for analysis and generation.

[1069] 4. The server passes the review to the sentiment engine

[1070] The preprocessed review sentences are passed to the emotion engine, which recognizes the user's emotions from the review sentences and extracts the emotion information.

[1071] 5. Generative AI generates positive evaluations based on negative evaluations using emotional information

[1072] The generative AI analyzes negative reviews and generates review texts that eliminate negative elements based on emotional information obtained from the emotion engine.

[1073] For example: "This smartphone has a great battery life and can be used for a long time."

[1074] 6. The server vectorizes the generated ratings

[1075] The server converts the generated positive reviews into a numerical format (vector format). For vectorization, techniques such as TF-IDF are used. The features of the generated reviews are expressed numerically.

[1076] 7. The server stores the vectorized ratings in a database

[1077] The vectorized review texts are stored in a database, which will be used for future multi-level review retrieval and recommendation.

[1078] 8. The server searches for other products with similar reviews

[1079] The server calculates the similarity of the generated review to other product reviews in the database, using a method such as cosine similarity, and identifies the product with the most similar review.

[1080] 9. The server recommends similar products to the user

[1081] The server then recommends the product with the most similar reviews to the user based on the calculation results, and displays the recommended results on the user's device.

[1082] For example: "If you're unhappy with your battery, we recommend this phone with better battery life."

[1083] This system allows users to efficiently find suitable alternative products to resolve their dissatisfaction. In addition, the emotion engine enables more appropriate recommendations that take into account the emotions expressed in users' ratings, improving the user experience.

[1084] The processing flow will be explained below.

[1085] Step 1:

[1086] Users post reviews

[1087] Users access the review section of the product they have purchased and enter a negative rating to post.

[1088] For example: "I'm not happy with this phone because the battery dies so quickly."

[1089] Step 2:

[1090] The server receives the review

[1091] The server receives user submissions instantly, and the received reviews are stored in an internal log and prepared for further processing.

[1092] Step 3:

[1093] The server pre-processes the review

[1094] The server pre-processes the review text, including tokenizing (dividing the text into words), removing stop words, and normalizing (e.g., converting all text to lowercase), making the review text optimized for analysis.

[1095] Preprocessing result example: "Smartphone battery runs out quickly, dissatisfaction"

[1096] Step 4:

[1097] The server passes the review to the sentiment engine

[1098] The preprocessed review sentences are passed to the sentiment engine, which extracts user sentiment from the reviews.

[1099] Example: "Dissatisfied"

[1100] Step 5:

[1101] Generative AI generates positive evaluations based on emotional information

[1102] The generative AI uses extracted emotional information (e.g., "dissatisfied") to turn negative evaluations into positive ones.

[1103] For example: "This smartphone has a great battery life and can be used for a long time."

[1104] Step 6:

[1105] The server vectorizes the generated ratings

[1106] The server converts the generated positive ratings into vector format using techniques such as TF-IDF (Term Frequency-Inverse Document Frequency) and Word2Vec.

[1107] Generated ratings in vector format, e.g.: \[0.15, 0.34, 0.01, ... \]

[1108] Step 7:

[1109] The server stores the vectorized ratings in a database.

[1110] The server stores the vectorized review text in a database, which will be used in future search and recommendation algorithms.

[1111] Step 8:

[1112] The server searches for other products with similar reviews

[1113] The server calculates the similarity between the reviews of other products stored in the database and the generated review vector. This similarity is calculated using cosine similarity.

[1114] Identify the most similar products.

[1115] Step 9:

[1116] The server recommends similar products to the user.

[1117] The server then uses the results of the calculation to recommend products with the most similar reviews to the user, displaying the recommendations on the user's device and suggesting alternative products.

[1118] For example: "If you're unhappy with your battery, we recommend this phone with better battery life."

[1119] Through this step, users can easily find the best alternative product to resolve their dissatisfaction.By using the emotion engine, we can understand the emotions expressed by users in their reviews and make more appropriate product recommendations, thereby further increasing user satisfaction.

[1120] Example 2

[1121] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1122] In conventional e-commerce platforms, negative reviews posted by users are often not properly addressed. As a result, users' dissatisfaction remains unresolved, potentially leading to customer churn. In addition, the lack of recommendations for alternative products based on negative reviews creates a problem of not improving the user experience.

[1123] The identification process by the identification processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means. In this invention, the server includes: means for a user to post a negative review of a product on an e-commerce platform; means for generating a positive review from the negative review using a generative artificial intelligence; means for converting the generated review into a vector format; means for storing the vector-format review in a database; means for searching for other products with similar reviews based on the stored vector-format review; means for recommending other products with similar reviews to the user based on the search; means including an emotion engine that recognizes the user's emotions from the review text and extracts the emotion information; and means for generating a positive review from the negative review based on the emotion information. This makes it possible to resolve the user's dissatisfaction by analyzing the user's negative review and recommending appropriate alternative products, thereby providing a more satisfying user experience.

[1124] "User" means any individual or legal entity who uses the e-commerce platform to post reviews on products.

[1125] An "e-commerce platform" is an online commerce system that allows users to purchase products and post reviews.

[1126] A "negative rating" is a review in which the user points out dissatisfaction or problems with the product.

[1127] "Generative artificial intelligence" is a computer program that uses natural language processing technology to generate new text.

[1128] A "positive evaluation" is a review that states the good points and satisfaction points of the product.

[1129] "Vector format" refers to text data converted into a numerical format that is easy to process with machine learning algorithms.

[1130] A "database" is a system for effectively storing, retrieving, and managing structured data.

[1131] An "emotion engine" is a program that uses natural language processing technology to extract and recognize a user's emotional information from text.

[1132] "Preprocessing" refers to a series of processing steps to convert text data into a format suitable for analysis or generation.

[1133] "Cosine similarity" is a mathematical method for calculating the similarity between two vectors based on the vector's direction.

[1134] This invention relates to a system that analyzes negative reviews posted by users on an e-commerce platform and recommends alternative products to the users that will resolve their dissatisfaction. This system achieves more appropriate product recommendations by incorporating an emotion engine that recognizes the user's emotions.

[1135] Hardware or software used

[1136] 1. Server: Responsible for central data processing of this system. Cloud servers can also be used if necessary.

[1137] 2. Device: The client device (computer, smartphone, tablet, etc.) through which a user posts a review.

[1138] 3. Database: A data storage system for storing reviews, vector-format ratings and other related data.

[1139] 4. Generative AI: An AI model that uses natural language processing technology to generate positive reviews based on negative ones. Examples include OpenAI's GPT.

[1140] 5. Sentiment Engine: A program that uses natural language processing techniques to extract user sentiment from reviews. It uses a model for sentiment analysis (e.g., a BERT-based model).

[1141] Data processing or data calculation

[1142] 1. Review preprocessing: Negative reviews posted by users are tokenized using a natural language processing library (e.g., Python's NLTK) and stop words are removed. This preprocessing converts the reviews into a format suitable for analysis and generation.

[1143] 2. Extracting Emotional Information: The preprocessed review sentences are passed to the emotion engine, which extracts emotional information and recognizes emotion tags such as "sad" and "disappointed."

[1144] 3. Generating Positive Evaluations: The generative AI generates positive evaluations from negative evaluations based on the emotional information extracted from the emotion engine. For example, the generative AI uses OpenAI's GPT model to invert emotion tags such as "sad" and "disappointed" to "joy" and "satisfied."

[1145] 4. Vectorization: The generated positive ratings are converted into a vector format that is easy for machine learning algorithms to process, using tools such as scikit-learn's TF-IDF Vectorizer.

[1146] 5. Similarity calculation: Calculate the similarity with other product reviews in the database. To calculate similarity, use scikit-learn's cosine_similarity.

[1147] 6. Product Recommendation: Based on the calculation results, we identify the products with the most similar reviews and recommend them to the user.

[1148] Specific examples

[1149] An example prompt for this system would be:

[1150] "Enter a negative review and generate a positive one based on it."

[1151] This allows users to more efficiently find alternative products that can resolve their dissatisfaction, improving the user experience by making recommendations that take emotions into account.

[1152] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1153] Step 1:

[1154] Users post reviews

[1155] The user enters a negative review of the product in the review section of the e-commerce platform and presses the post button.

[1156] Input: User negative review text

[1157] Output: Posting a negative review to an e-commerce platform

[1158] Specific operation: A user uses a device (PC, smartphone, tablet, etc.) to access an online shopping site, enter a review in text format, and press the post button.

[1159] Step 2:

[1160] The server receives the review

[1161] The server immediately receives reviews posted by users and registers them in a database.

[1162] Input: User-submitted negative review

[1163] Output: Negative reviews stored in the database

[1164] Specific operation: The server uses the receiving API to accept review submissions and saves the data in a database management system.

[1165] Step 3:

[1166] The server pre-processes the review

[1167] The server performs preprocessing such as tokenizing the review sentences and removing stop words, converting the review sentences into a format suitable for analysis and generation.

[1168] Input: Negative reviews retrieved from the database

[1169] Output: Preprocessed review sentences

[1170] Specific operation: Using Python's NLTK library, the review sentences are split into words and frequent meaningless words (stop words) are removed.

[1171] Step 4:

[1172] The server passes the review to the sentiment engine

[1173] The preprocessed review sentences are passed to the emotion engine, which recognizes emotions from the user's review sentences and extracts the emotion information.

[1174] Input: Preprocessed review text

[1175] Output: Emotion information (e.g., emotion tags such as "sad" or "disappointed")

[1176] Specific operation: Calls an API that passes text to the emotion engine, and receives emotion information in return.

[1177] Step 5:

[1178] Generative AI generates positive evaluations based on negative evaluations and emotional information

[1179] The generative AI analyzes negative ratings and generates positive reviews based on emotional information obtained from the emotion engine.

[1180] Input: Negative evaluation and emotional information

[1181] Output: Positive review text

[1182] Specific operation: Using OpenAI's GPT model, etc., the system generates review sentences by converting negative elements into their opposite positive elements (e.g., converting "sad" and "disappointed" into "joy" and "satisfied").

[1183] Step 6:

[1184] The server vectorizes the generated ratings

[1185] The server converts the generated positive reviews into a numerical format (vector format).

[1186] Input: Generated positive review text

[1187] Output: Evaluation data in vector format

[1188] Specific operation: Converts text into vector format using scikit-learn's TF-IDF Vectorizer.

[1189] Step 7:

[1190] The server stores the vectorized ratings in a database.

[1191] The server stores the vectorized review sentences in a database.

[1192] Input: Vectorized evaluation data

[1193] Output: Evaluation data in vector format stored in a database

[1194] Specific behavior: Executes a query to insert vector data into an SQL or NoSQL database.

[1195] Step 8:

[1196] The server searches for other products with similar reviews

[1197] The server calculates the similarity between the generated review and reviews of other products in the database.

[1198] Input: Vector-formatted reviews and generated vector-formatted ratings data in the database

[1199] Output: Evaluation data with similarity calculated

[1200] Specific operation: Uses scikit-learn's cosine_similarity function to calculate the similarity with other vectors in the database and identify the most similar vector.

[1201] Step 9:

[1202] The server recommends similar products to the user.

[1203] Based on the calculation results, the server recommends products with the most similar reviews to the user.

[1204] Input: A list of similar products

[1205] Output: Recommended products displayed on the user's device

[1206] Specific behavior: Generates an API response to display recommended results on the user's device and returns the results to the browser or app.

[1207] (Application example 2)

[1208] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1209] Current e-commerce platforms lack a system that uses negative feedback from users to appropriately recommend alternative products to resolve their dissatisfaction. This means users have to spend time and effort to find a satisfactory product. Furthermore, existing systems do not take user sentiment into account when making recommendations, which does not adequately improve the user experience.

[1210] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for a user to post a negative review of a product on an e-commerce platform, means for generating a positive review from the negative review using a generative artificial intelligence, and means for converting the generated review into a vector format. This makes it possible to pass the user's negative review to an emotion recognition engine, extract the user's emotional information, generate a positive review based on the emotional information, search for other products with similar reviews based on the vector-format review stored in the database, and recommend other products with similar reviews to the user based on the search.

[1211] "User" means any individual or organization using the e-commerce platform.

[1212] An "e-commerce platform" is an online system for buying and selling goods and services over the Internet.

[1213] "Negative ratings" are reviews or feedback where users express dissatisfaction with a product or service.

[1214] "Generative AI" is an algorithm or system that uses natural language processing techniques to generate new text.

[1215] "Positive reviews" are positive reviews or feedback that indicate satisfaction with a product or service.

[1216] "Vector format" is a format in which text data is converted into a numerical vector.

[1217] A "database" is a system for efficiently storing, managing, and retrieving data.

[1218] A "search method" is an algorithm or method for finding specific data in a database.

[1219] A "recommendation method" is a system or method for suggesting appropriate products or services to users.

[1220] An "emotion recognition engine" is an artificial intelligence technology for extracting user emotional information from text data.

[1221] "Emotion information" is data that indicates the user's emotional state.

[1222] A "prompt" is an input sentence or cue that the generative AI uses to generate text.

[1223] "Similarity" is an index that indicates the similarity between two pieces of data.

[1224] "Cosine similarity" is a method of calculating similarity using the cosine angle between two vectors.

[1225] A specific system for implementing this invention is one that allows users to post negative reviews of products using an e-commerce platform, and based on those reviews, recommends alternative products to resolve the user's dissatisfaction. This system is mainly composed of a combination of a program, hardware, and a database.

[1226] 1. Program Generation and Explanation

[1227] The server receives negative reviews posted by users and preprocesses them. This preprocessing step involves tokenization and stopword removal to convert the reviews into a more easily parseable format. The server then passes the preprocessed reviews to an emotion recognition engine (e.g., using the Hugging Face transformers library) to extract the user's emotional information.

[1228] Based on the emotional information obtained from the emotion engine, the server uses generative artificial intelligence (e.g., a Hugging Face text-generation model) to convert negative feedback into positive feedback. In this step, the prompt sentence is used to input the generative AI model, which generates a positive feedback.

[1229] Once a positive review is generated, the server converts it into a vector format using techniques such as TF-IDF, which is then stored in a database for future searches and recommendations.

[1230] The server then searches for other products with similar ratings based on the stored vector of ratings. It uses cosine similarity to calculate similarity and identifies products with the most similar reviews. Finally, the server recommends products with the highest similarity to the user, and the recommended results are displayed on the user's device.

[1231] 2. Hardware and Software Used

[1232] The system requires advanced hardware and software for text analysis and generation, including high-performance servers and database servers, and Python, SQL, and the Hugging Face transformers library.

[1233] 3. Examples of concrete examples and prompts

[1234] For example, if a user posts a review stating, "I'm dissatisfied with this smartphone because the battery runs out quickly," the system receives it and performs preprocessing and sentiment analysis. Based on the results of the sentiment analysis, the generative AI model generates a positive evaluation such as, "This smartphone has excellent battery life and can be used for a long time." This evaluation is then vectorized and stored in a database.

[1235] The server then calculates the similarity of the generated rating to other ratings in the database, identifies the product with the most similar rating, and recommends this product to the user, saying, "If you're dissatisfied with your battery, we recommend this smartphone with a long battery life."

[1236] Example prompt for a generative AI model:

[1237] User expressed dissatisfaction with: "I'm not happy with this phone because the battery runs out so quickly." Write a positive review.

[1238] This system allows users to quickly find suitable alternative products to resolve their dissatisfaction, improving the user experience.

[1239] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1240] Step 1:

[1241] A user posts a negative review of a product on an e-commerce platform. For example, the user might write, "I'm not happy with this smartphone because the battery runs out quickly." This review becomes the input to the system.

[1242] Step 2:

[1243] The server receives negative reviews from users. It registers the received reviews internally and passes them to the next step. The input is the user's negative review, and the output is the data passed to the preprocessing step.

[1244] Step 3:

[1245] The server preprocesses the reviews. This includes tokenizing the text, removing stop words, etc. It takes the received negative reviews as input and converts them into a format that is easy to parse. The output is the preprocessed text data.

[1246] Step 4:

[1247] The server passes the preprocessed reviews to the emotion recognition engine. Using the Hugging Face transformers library, the user's emotional information is extracted. The input is preprocessed text data, and after processing by the emotion recognition engine, the output is emotional information. For example, a result such as "Negative (90% probability)" can be obtained.

[1248] Step 5:

[1249] The server generates a prompt based on the emotion information and passes it to the AI ​​generator. For example, it generates a prompt like, "User expressed dissatisfaction with: 'I'm dissatisfied with this smartphone because the battery runs out quickly.' Write a positive review." The input is the emotion information and the preprocessed review, and the output is the prompt text.

[1250] Step 6:

[1251] Generative AI generates a positive review based on the prompt. It uses Hugging Face's text-generation model to create a positive review that eliminates negative elements. The input is the prompt text, and the output is the generated positive review. For example, you might get a review like, "This smartphone has excellent battery life and can be used for a long time."

[1252] Step 7:

[1253] The server vectorizes the generated positive reviews. It uses techniques such as TF-IDF to convert text data into a numerical format. The input is the generated positive reviews, and the output is vector-format data.

[1254] Step 8:

[1255] The server stores the generated rating vector in a database, which is used for future searches and recommendations. The input is vector data, and the output is the completed storage in the database.

[1256] Step 9:

[1257] The server calculates the similarity of the generated review to other reviews in the database. It uses methods such as cosine similarity to find the product with the most similar reviews. The input is rating data in vector format, and the output is the product with the most similar review.

[1258] Step 10:

[1259] The server recommends highly similar products to the user. Based on the calculation results, the most suitable product recommendation is displayed on the user's device. The input is the product information from the search results, and the output is the recommendation to the user. For example, a recommendation such as "If you are dissatisfied with your battery, we recommend this smartphone with a long battery life" is displayed.

[1260] The specific processing unit 290 transmits the result of the specific processing to the headset type terminal 314. In the headset type terminal 314, the control unit 46A causes the speaker 240 and the display 343 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[1261] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1262] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the headset type terminal 314.

[1263] [Fourth embodiment]

[1264] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1265] 7, a data processing system 410 includes a data processing device 12 and a robot 414. An example of the data processing device 12 is a server.

[1266] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[1267] The robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a control target 443. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the control target 443 are also connected to the bus 52.

[1268] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[1269] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[1270] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[1271] The control object 443 includes a display device, LEDs in the eyes, and motors for driving the arms, hands, and feet. The posture and gestures of the robot 414 are controlled by controlling the motors of the arms, hands, and feet. Some of the emotions of the robot 414 can be expressed by controlling these motors. In addition, the facial expressions of the robot 414 can also be expressed by controlling the light emission state of the LEDs in the eyes of the robot 414.

[1272] Fig. 8 shows an example of the main functions of the data processing device 12 and the robot 414. As shown in Fig. 8, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[1273] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[1274] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[1275] In the robot 414, the processor 46 performs the reception output process. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[1276] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1277] The present invention is a system that analyzes negative reviews posted by users on an e-commerce platform and recommends to the users alternative products that will resolve their dissatisfaction.

[1278] The system mainly includes the following means:

[1279] A means for users to post negative product reviews on e-commerce platforms

[1280] A means for the server to generate positive reviews from the negative reviews using artificial intelligence

[1281] means for the server to convert the generated ratings into vector form;

[1282] means for storing the vector-formed evaluations in a database;

[1283] A means for the server to search for other products having similar ratings based on the stored vector-format ratings.

[1284] A means for the server to recommend other products with similar ratings to the user based on the search.

[1285] Program processing and specific examples

[1286] 1. Users post reviews

[1287] Users can post negative reviews of the product in the review section of the e-commerce platform.

[1288] For example: "I'm not happy with this phone because the battery dies so quickly."

[1289] 2. The server receives the review and passes it to the generation AI.

[1290] The server receives user submissions, performs preprocessing, and then passes negative reviews to the AI ​​generator.

[1291] Preprocessing includes tokenization and stop word removal.

[1292] 3. Generative AI generates positive reviews

[1293] Generative AI generates positive evaluations based on negative evaluations.

[1294] For example: "This smartphone has a great battery life and can be used for a long time."

[1295] 4. The server vectorizes the generated ratings and stores them in a database.

[1296] The server converts the generated positive reviews into vector format and stores them in a database.

[1297] Techniques such as TF-IDF (Term Frequency-Inverse Document Frequency) are used for vectorization.

[1298] 5. The server searches for other products with similar reviews

[1299] The server calculates the similarity between the vectorized rating and other product reviews in the database.

[1300] For this similarity calculation, a method such as cosine similarity is used.

[1301] 6. The server recommends another product to the user

[1302] The server generates a list of highly similar products and recommends them to the user.

[1303] For example: "If you're unhappy with your battery, we recommend this phone with better battery life."

[1304] This system allows users to efficiently find suitable alternative products to resolve their dissatisfaction, and also helps e-commerce platform operators improve user experience and reduce barriers to sales.

[1305] The processing flow will be explained below.

[1306] Step 1:

[1307] A user posts a review on an e-commerce platform

[1308] Users access the review section of the product they have purchased and enter a negative rating to post.

[1309] For example: "I'm not happy with this phone because the battery dies so quickly."

[1310] Step 2:

[1311] The server receives the review

[1312] The server receives user submissions instantly.

[1313] Registers the received review internally and passes it to the next step for processing.

[1314] Step 3:

[1315] The server pre-processes the review

[1316] The server pre-processes the review sentences. In this step, the review sentences are tokenized (divided into words) and stop words (common words with little meaning) are removed.

[1317] Preprocessing makes the review sentences ready for analysis and generation.

[1318] Step 4:

[1319] The server passes the review to the generation AI

[1320] The preprocessed review text is sent to the AI ​​generator, which analyzes the received review text and identifies negative elements.

[1321] Step 5:

[1322] Generative AI generates positive reviews

[1323] The artificial intelligence generates a review sentence with negative elements removed.

[1324] For example: "This smartphone has a great battery life and can be used for a long time."

[1325] Step 6:

[1326] The server vectorizes the generated ratings

[1327] The server converts the generated positive ratings into a numerical format (vector format), using techniques such as TF-IDF (Term Frequency-Inverse Document Frequency).

[1328] The generated evaluation features are expressed numerically.

[1329] Step 7:

[1330] The server stores the vectorized ratings in a database.

[1331] The vectorized review texts are stored in a database, which will be used for future multi-level review retrieval and recommendation.

[1332] Step 8:

[1333] The server searches for other products with similar reviews

[1334] The server calculates the similarity of the generated review to other product reviews in the database, using a method such as cosine similarity.

[1335] Identify products with the most similar reviews.

[1336] Step 9:

[1337] The server recommends similar products to the user.

[1338] Based on the calculation results, the server recommends products with the most similar reviews to the user, and displays the recommended results on the user's device.

[1339] For example: "If you're unhappy with your battery, we recommend this phone with better battery life."

[1340] This step allows users to quickly find the best alternative to resolve their dissatisfaction.

[1341] Example 1

[1342] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1343] When a user posts a negative review on an e-commerce platform, it is important to appropriately recommend alternative products to resolve the user's dissatisfaction. Current systems do not accurately identify the user's dissatisfaction from the negative review and recommend alternative products based on that information, which limits the effectiveness of improving the user experience and promoting sales.

[1344] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1345] In this invention, the server includes means for users to post negative reviews of products on the e-commerce platform, means for generating positive reviews from the negative reviews using a generation artificial intelligence, means for converting the generated reviews into a vector format, means for storing the vector format reviews in a database, means for searching for other products with similar reviews based on the stored vector format reviews, means for recommending other products with similar reviews to the user based on the search, and means for pre-processing the negative reviews, thereby making it possible to effectively recommend alternative products based on the user's negative reviews.

[1346] "User" means an individual or legal entity who uses the e-commerce platform to post ratings and reviews of products.

[1347] An "e-commerce platform" is an online system for buying and selling goods and services over the Internet.

[1348] A "negative rating" is a review in which a user expresses dissatisfaction or dissatisfaction with a product or service.

[1349] "Generative AI" refers to AI that generates text using natural language processing and machine learning techniques.

[1350] A "positive rating" is a review that expresses positive content or benefits in relation to a negative rating.

[1351] "Vector format" is a format that expresses sentences and evaluations as numerical data. It is expressed as multiple numerical vectors, making calculations and comparisons easier.

[1352] A "database" is a system for storing and managing information in an organized manner.

[1353] "Similar evaluations" refer to evaluations in which the content of multiple evaluations is similar to one another. This similarity is quantified using a specific calculation method.

[1354] "Preprocessing" refers to processing performed to make raw data easier to analyze. Specifically, it includes tokenization and removal of stop words.

[1355] "Tokenizing" is the process of dividing text data into units of words or phrases.

[1356] "Stop word removal" is the process of removing common words that have no meaning or are not needed during analysis.

[1357] "Cosine similarity" is a method for calculating the similarity between vector data. It calculates the similarity based on the angle between the vectors.

[1358] The present invention is a system that analyzes negative reviews posted by users on an e-commerce platform and recommends alternative products to the users that will resolve their dissatisfaction. This system is composed of a user, a server, a generative artificial intelligence (AI), and a database.

[1359] First, a user accesses an e-commerce platform and posts a negative review in the review section. For example, the user might write, "I'm not happy with this smartphone because the battery runs out quickly," and submit it.

[1360] Next, the server receives these negative reviews and preprocesses them. This preprocessing includes tokenizing (dividing sentences into words) and removing stop words (removing meaningless common words). This preprocessing makes the review text easier to analyze by generative artificial intelligence.

[1361] The preprocessed reviews are then passed to a generative AI model, which uses natural language processing techniques, for example, to convert negative reviews into positive ones. Based on the example above, the generative AI model would generate a positive review such as, "This smartphone has excellent battery life and can be used for a long time."

[1362] The server then converts the generated positive reviews into vector format using techniques such as Term Frequency-Inverse Document Frequency (TF-IDF) and word embedding, and stores the converted reviews in a database.

[1363] Based on the stored vector-format ratings, the server calculates the similarity with the reviews of other products. The similarity is calculated using methods such as cosine similarity. Based on this similarity calculation, the server searches for other products with similar ratings.

[1364] Finally, the server generates a list of similar products and recommends them to the user. The user can refer to the recommended products to find suitable alternative products that can resolve their complaints. For example, the server may display a message such as, "If you are dissatisfied with the battery, we recommend this smartphone with a long battery life."

[1365] Examples of prompt statements:

[1366] User type: "I'm not happy with this phone because the battery dies so quickly."

[1367] Prompt for the AI ​​generator: "I'm not happy with this phone because the battery dies so quickly. Please generate a positive rating for this review."

[1368] Positive review generated: "This phone has a great battery life and can be used for a long time."

[1369] This particular embodiment allows for quick and effective recommendation of alternative products based on users' negative reviews, thereby improving the user experience of the e-commerce platform and promoting product sales.

[1370] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1371] Step 1:

[1372] Users post reviews

[1373] Specific behavior:

[1374] Users access the e-commerce platform's website or app and post negative reviews of products in the review section.

[1375] Input: A negative review such as "I'm not happy with this phone because the battery dies so quickly."

[1376] Output: A negative rating is sent to the server.

[1377] Step 2:

[1378] The server receives the reviews and pre-processes them

[1379] Specific behavior:

[1380] The server receives reviews posted by users and pre-processes the review content.

[1381] Input: The negative rating submitted by the user.

[1382] Data processing: tokenization (splitting sentences into words) and stop word removal (removing meaningless common words).

[1383] Output: Preprocessed text data.

[1384] Step 3:

[1385] The server passes the data to a generative AI model, which generates a positive rating.

[1386] Specific behavior:

[1387] The server passes the preprocessed negative reviews to a generative AI model, which generates positive ratings.

[1388] Input: Preprocessed negative reviews.

[1389] Data computation: Generative AI models generate positive reviews based on negative reviews.

[1390] Output: A positive review generated by the generative AI.

[1391] Step 4:

[1392] The server vectorizes the generated ratings and stores them in a database.

[1393] Specific behavior:

[1394] The server converts the generated positive reviews into vector format and stores them in a database.

[1395] Input: Generated positive reviews.

[1396] Data processing: Convert data into vector format using techniques such as TF-IDF or word embedding.

[1397] Output: Ratings in vector format stored in a database.

[1398] Step 5:

[1399] The server searches for similar products

[1400] Specific behavior:

[1401] The server calculates the similarity between the vectorized ratings and other product reviews in the database and searches for other products with similar ratings.

[1402] Input: Ratings in vector format stored in a database.

[1403] Data calculation: Calculate similarity using methods such as cosine similarity.

[1404] Output: A list of highly similar products.

[1405] Step 6:

[1406] The server recommends different products to the user

[1407] Specific behavior:

[1408] The server generates a list of highly similar products and recommends them to the user.

[1409] Input: A list of similar products.

[1410] Output: The recommended product list displayed to the user.

[1411] Example: "If you're unhappy with your battery, we recommend this smartphone with better battery life."

[1412] As described above, the system's program starts with negative reviews posted by users, processes the data and performs various calculations, and then recommends alternative products that will resolve the user's complaints.

[1413] (Application example 1)

[1414] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1415] On modern e-commerce platforms, negative reviews posted by users are an important factor in product selection. However, these negative reviews often discourage purchases and cause users to abandon the platform. Conventional systems lack a mechanism to effectively utilize these negative reviews to recommend alternative products to users, which results in the inability to adequately address users' dissatisfaction and needs.

[1416] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1417] In this invention, the server includes means for users to post negative reviews of products on the e-commerce platform, means for generating positive reviews from the negative reviews using a generation artificial intelligence, means for converting the generated reviews into vector format, means for storing the vector format reviews in a database, means for searching for other products with similar reviews based on the stored vector format reviews, means for recommending other products with similar reviews to the user based on the search, and means for analyzing the user's needs based on the negative reviews and recommending other products that meet those needs in real time, thereby enabling users to quickly find alternative products that best suit their needs and complaints.

[1418] An "e-commerce platform" is an online system for buying and selling goods and services over the Internet.

[1419] "Negative evaluation" refers to reviews or feedback in which users describe their dissatisfaction or shortcomings with a product or service.

[1420] "Generative AI" refers to AI technology that uses natural language processing technology to generate, translate, and analyze text.

[1421] A "positive rating" is a review that describes the positive aspects or benefits of a product or service for the user.

[1422] "Vector format" is a format in which text data is converted into numerical data and expressed in a multidimensional space.

[1423] A "database" is a system for efficiently storing, managing, and searching data.

[1424] "Cosine similarity" is a method for measuring the similarity between two vectors by calculating the cosine value of the angle between them.

[1425] "Similarity calculation" is a calculation for quantifying the similarity between different data.

[1426] "Recommendation" means that the system presents appropriate products and services based on the user's ratings and preferences.

[1427] "Real-time recommendations" refers to the ability to instantly respond to information entered by the user and recommend appropriate products and services.

[1428] This invention relates to a system that analyzes negative reviews posted by users on e-commerce platforms and recommends alternative products that meet the user's needs in real time.

[1429] The system primarily uses the following hardware and software: React Native for front-end development, Node.js and Express.js for back-end development, and MongoDB for the database. It also utilizes GPT-3 or ChatGPT as a generative AI model, the TF-IDF method for vectorizing text data, and cosine similarity for similarity calculations. The system operates when users post reviews through the e-commerce platform's application.

[1430] First, a user posts a negative review about a product using their device. The posted negative review is sent to the server, which receives the review and performs preprocessing. Preprocessing includes tokenization and stop word removal. The preprocessed negative review is then passed to a generative AI model, which generates a positive review. For example, if a user posts, "This smartphone's battery runs out quickly," the generative AI model will generate a positive review such as, "This smartphone has excellent battery life and can be used for a long time."

[1431] The generated positive reviews are converted into vector format by the server and stored in a database. The server then calculates the similarity between the reviews and those of other products in the database. This similarity is calculated using cosine similarity. Based on the calculated similarity, the server searches for other products that may meet the user's needs and recommends them to the user in real time.

[1432] This system allows users to quickly resolve their complaints and find the best alternative product. For example, if a user is dissatisfied with the battery, a similarity calculation will be used to recommend another smartphone with a better battery life. An example of a prompt used by the generative AI model is as follows:

[1433] Example prompt sentence:

[1434] "A user left a negative review stating that the battery dies quickly. Use this as a basis to generate a positive review."

[1435] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1436] Step 1:

[1437] The user uses the terminal to input and post a negative review in the review section of the e-commerce platform.

[1438] Input: Negative review text (e.g., "I'm not happy with this phone because the battery runs out quickly.")

[1439] Output: The posted negative review text is sent to the server.

[1440] Step 2:

[1441] The server receives negative feedback from users and performs preprocessing, which includes tokenization (splitting text into words) and stop-word removal (removing frequently occurring words).

[1442] Input: The negative review text posted.

[1443] Output: Preprocessed negative evaluation text (text after tokenization and stopword removal).

[1444] Step 3:

[1445] The server passes the preprocessed negative evaluation text to a generative AI model (e.g., GPT-3) to generate a positive evaluation. For example, in response to the negative evaluation "The battery runs out quickly," a positive evaluation such as "The battery life is very good and can be used for a long time" is generated.

[1446] Input: Preprocessed negative rating text.

[1447] Output: The generated positive review text.

[1448] Step 4:

[1449] The server converts the generated positive feedback text into vector format using the Term Frequency-Inverse Document Frequency (TF-IDF) method.

[1450] Input: The generated positive review text.

[1451] Output: Vector format data.

[1452] Step 5:

[1453] The server stores the vector data in a database.

[1454] Input: Vector format data.

[1455] Output: Data in vector format stored in a database.

[1456] Step 6:

[1457] The server calculates the similarity between the vector data and other product reviews in the database. Cosine similarity is used to calculate this similarity.

[1458] Input: Vector data from the database and newly generated vector data.

[1459] Output: Similarity calculation result.

[1460] Step 7:

[1461] Based on the similarity calculation results, the server searches for other products that suit the user's needs and recommends them to the user in real time.

[1462] Input: Similarity calculation result.

[1463] Output: A list of products recommended to the user.

[1464] This allows users to quickly find the best alternative product to resolve their dissatisfaction.

[1465] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1466] The present invention combines a system that analyzes negative reviews posted by users on e-commerce platforms and recommends alternative products to users that will resolve their dissatisfaction with an emotion engine that recognizes users' emotions.

[1467] The system mainly includes the following means:

[1468] A means for users to post negative product reviews on e-commerce platforms

[1469] A means for the server to generate positive reviews from the negative reviews using artificial intelligence

[1470] means for the server to convert the generated ratings into vector form;

[1471] means for storing the vector-formed evaluations in a database;

[1472] A means for the server to search for other products having similar ratings based on the stored vector-format ratings.

[1473] A means for the server to recommend other products with similar ratings to the user based on the search.

[1474] an emotion engine that recognizes the user's emotions contained in the negative reviews and the generated reviews;

[1475] Program processing and specific examples

[1476] 1. Users post reviews

[1477] Users can post negative reviews of the product in the review section of the e-commerce platform.

[1478] For example: "I'm not happy with this phone because the battery dies so quickly."

[1479] 2. The server receives the review

[1480] The server immediately receives the user's post, registers it internally, and passes it on to the next step for processing.

[1481] 3. The server pre-processes the review

[1482] The server performs preprocessing such as tokenizing the review sentences and removing stop words, so that the review sentences are ready for analysis and generation.

[1483] 4. The server passes the review to the sentiment engine

[1484] The preprocessed review sentences are passed to the emotion engine, which recognizes the user's emotions from the review sentences and extracts the emotion information.

[1485] 5. Generative AI generates positive evaluations based on negative evaluations using emotional information

[1486] The generative AI analyzes negative reviews and generates review texts that eliminate negative elements based on emotional information obtained from the emotion engine.

[1487] For example: "This smartphone has a great battery life and can be used for a long time."

[1488] 6. The server vectorizes the generated ratings

[1489] The server converts the generated positive reviews into a numerical format (vector format). For vectorization, techniques such as TF-IDF are used. The features of the generated reviews are expressed numerically.

[1490] 7. The server stores the vectorized ratings in a database

[1491] The vectorized review texts are stored in a database, which will be used for future multi-level review retrieval and recommendation.

[1492] 8. The server searches for other products with similar reviews

[1493] The server calculates the similarity of the generated review to other product reviews in the database, using a method such as cosine similarity, and identifies the product with the most similar review.

[1494] 9. The server recommends similar products to the user

[1495] The server then recommends the product with the most similar reviews to the user based on the calculation results, and displays the recommended results on the user's device.

[1496] For example: "If you're unhappy with your battery, we recommend this phone with better battery life."

[1497] This system allows users to efficiently find suitable alternative products to resolve their dissatisfaction. In addition, the emotion engine enables more appropriate recommendations that take into account the emotions expressed in users' ratings, improving the user experience.

[1498] The processing flow will be explained below.

[1499] Step 1:

[1500] Users post reviews

[1501] Users access the review section of the product they have purchased and enter a negative rating to post.

[1502] For example: "I'm not happy with this phone because the battery dies so quickly."

[1503] Step 2:

[1504] The server receives the review

[1505] The server receives user submissions instantly, and the received reviews are stored in an internal log and prepared for further processing.

[1506] Step 3:

[1507] The server pre-processes the review

[1508] The server pre-processes the review text, including tokenizing (dividing the text into words), removing stop words, and normalizing (e.g., converting all text to lowercase), making the review text optimized for analysis.

[1509] Preprocessing result example: "Smartphone battery runs out quickly, dissatisfaction"

[1510] Step 4:

[1511] The server passes the review to the sentiment engine

[1512] The preprocessed review sentences are passed to the sentiment engine, which extracts user sentiment from the reviews.

[1513] Example: "Dissatisfied"

[1514] Step 5:

[1515] Generative AI generates positive evaluations based on emotional information

[1516] The generative AI uses extracted emotional information (e.g., "dissatisfied") to turn negative evaluations into positive ones.

[1517] For example: "This smartphone has a great battery life and can be used for a long time."

[1518] Step 6:

[1519] The server vectorizes the generated ratings

[1520] The server converts the generated positive ratings into vector format using techniques such as TF-IDF (Term Frequency-Inverse Document Frequency) and Word2Vec.

[1521] Generated ratings in vector format, e.g.: \[0.15, 0.34, 0.01, ... \]

[1522] Step 7:

[1523] The server stores the vectorized ratings in a database.

[1524] The server stores the vectorized review text in a database, which will be used in future search and recommendation algorithms.

[1525] Step 8:

[1526] The server searches for other products with similar reviews

[1527] The server calculates the similarity between the reviews of other products stored in the database and the generated review vector. This similarity is calculated using cosine similarity.

[1528] Identify the most similar products.

[1529] Step 9:

[1530] The server recommends similar products to the user.

[1531] The server then uses the results of the calculation to recommend products with the most similar reviews to the user, displaying the recommendations on the user's device and suggesting alternative products.

[1532] For example: "If you're unhappy with your battery, we recommend this phone with better battery life."

[1533] Through this step, users can easily find the best alternative product to resolve their dissatisfaction.By using the emotion engine, we can understand the emotions expressed by users in their reviews and make more appropriate product recommendations, thereby further increasing user satisfaction.

[1534] Example 2

[1535] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1536] In conventional e-commerce platforms, negative reviews posted by users are often not properly addressed. As a result, users' dissatisfaction remains unresolved, potentially leading to customer churn. In addition, the lack of recommendations for alternative products based on negative reviews creates a problem of not improving the user experience.

[1537] The identification process by the identification processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means. In this invention, the server includes: means for a user to post a negative review of a product on an e-commerce platform; means for generating a positive review from the negative review using a generative artificial intelligence; means for converting the generated review into a vector format; means for storing the vector-format review in a database; means for searching for other products with similar reviews based on the stored vector-format review; means for recommending other products with similar reviews to the user based on the search; means including an emotion engine that recognizes the user's emotions from the review text and extracts the emotion information; and means for generating a positive review from the negative review based on the emotion information. This makes it possible to resolve the user's dissatisfaction by analyzing the user's negative review and recommending appropriate alternative products, thereby providing a more satisfying user experience.

[1538] "User" means any individual or legal entity who uses the e-commerce platform to post reviews on products.

[1539] An "e-commerce platform" is an online commerce system that allows users to purchase products and post reviews.

[1540] A "negative rating" is a review in which the user points out dissatisfaction or problems with the product.

[1541] "Generative artificial intelligence" is a computer program that uses natural language processing technology to generate new text.

[1542] A "positive evaluation" is a review that states the good points and satisfaction points of the product.

[1543] "Vector format" refers to text data converted into a numerical format that is easy to process with machine learning algorithms.

[1544] A "database" is a system for effectively storing, retrieving, and managing structured data.

[1545] An "emotion engine" is a program that uses natural language processing technology to extract and recognize a user's emotional information from text.

[1546] "Preprocessing" refers to a series of processing steps to convert text data into a format suitable for analysis or generation.

[1547] "Cosine similarity" is a mathematical method for calculating the similarity between two vectors based on the vector's direction.

[1548] This invention relates to a system that analyzes negative reviews posted by users on an e-commerce platform and recommends alternative products to the users that will resolve their dissatisfaction. This system achieves more appropriate product recommendations by incorporating an emotion engine that recognizes the user's emotions.

[1549] Hardware or software used

[1550] 1. Server: Responsible for central data processing of this system. Cloud servers can also be used if necessary.

[1551] 2. Device: The client device (computer, smartphone, tablet, etc.) through which a user posts a review.

[1552] 3. Database: A data storage system for storing reviews, vector-format ratings and other related data.

[1553] 4. Generative AI: An AI model that uses natural language processing technology to generate positive reviews based on negative ones. Examples include OpenAI's GPT.

[1554] 5. Sentiment Engine: A program that uses natural language processing techniques to extract user sentiment from reviews. It uses a model for sentiment analysis (e.g., a BERT-based model).

[1555] Data processing or data calculation

[1556] 1. Review preprocessing: Negative reviews posted by users are tokenized using a natural language processing library (e.g., Python's NLTK) and stop words are removed. This preprocessing converts the reviews into a format suitable for analysis and generation.

[1557] 2. Extracting Emotional Information: The preprocessed review sentences are passed to the emotion engine, which extracts emotional information and recognizes emotion tags such as "sad" and "disappointed."

[1558] 3. Generating Positive Evaluations: The generative AI generates positive evaluations from negative evaluations based on the emotional information extracted from the emotion engine. For example, the generative AI uses OpenAI's GPT model to invert emotion tags such as "sad" and "disappointed" to "joy" and "satisfied."

[1559] 4. Vectorization: The generated positive ratings are converted into a vector format that is easy for machine learning algorithms to process, using tools such as scikit-learn's TF-IDF Vectorizer.

[1560] 5. Similarity calculation: Calculate the similarity with other product reviews in the database. To calculate similarity, use scikit-learn's cosine_similarity.

[1561] 6. Product Recommendation: Based on the calculation results, we identify the products with the most similar reviews and recommend them to the user.

[1562] Specific examples

[1563] An example prompt for this system would be:

[1564] "Enter a negative review and generate a positive one based on it."

[1565] This allows users to more efficiently find alternative products that can resolve their dissatisfaction, improving the user experience by making recommendations that take emotions into account.

[1566] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1567] Step 1:

[1568] Users post reviews

[1569] The user enters a negative review of the product in the review section of the e-commerce platform and presses the post button.

[1570] Input: User negative review text

[1571] Output: Posting a negative review to an e-commerce platform

[1572] Specific operation: A user uses a device (PC, smartphone, tablet, etc.) to access an online shopping site, enter a review in text format, and press the post button.

[1573] Step 2:

[1574] The server receives the review

[1575] The server immediately receives reviews posted by users and registers them in a database.

[1576] Input: User-submitted negative review

[1577] Output: Negative reviews stored in the database

[1578] Specific operation: The server uses the receiving API to accept review submissions and saves the data in a database management system.

[1579] Step 3:

[1580] The server pre-processes the review

[1581] The server performs preprocessing such as tokenizing the review sentences and removing stop words, converting the review sentences into a format suitable for analysis and generation.

[1582] Input: Negative reviews retrieved from the database

[1583] Output: Preprocessed review sentences

[1584] Specific operation: Using Python's NLTK library, the review sentences are split into words and frequent meaningless words (stop words) are removed.

[1585] Step 4:

[1586] The server passes the review to the sentiment engine

[1587] The preprocessed review sentences are passed to the emotion engine, which recognizes emotions from the user's review sentences and extracts the emotion information.

[1588] Input: Preprocessed review text

[1589] Output: Emotion information (e.g., emotion tags such as "sad" or "disappointed")

[1590] Specific operation: Calls an API that passes text to the emotion engine, and receives emotion information in return.

[1591] Step 5:

[1592] Generative AI generates positive evaluations based on negative evaluations and emotional information

[1593] The generative AI analyzes negative ratings and generates positive reviews based on emotional information obtained from the emotion engine.

[1594] Input: Negative evaluation and emotional information

[1595] Output: Positive review text

[1596] Specific operation: Using OpenAI's GPT model, etc., the system generates review sentences by converting negative elements into their opposite positive elements (e.g., converting "sad" and "disappointed" into "joy" and "satisfied").

[1597] Step 6:

[1598] The server vectorizes the generated ratings

[1599] The server converts the generated positive reviews into a numerical format (vector format).

[1600] Input: Generated positive review text

[1601] Output: Evaluation data in vector format

[1602] Specific operation: Converts text into vector format using scikit-learn's TF-IDF Vectorizer.

[1603] Step 7:

[1604] The server stores the vectorized ratings in a database.

[1605] The server stores the vectorized review sentences in a database.

[1606] Input: Vectorized evaluation data

[1607] Output: Evaluation data in vector format stored in a database

[1608] Specific behavior: Executes a query to insert vector data into an SQL or NoSQL database.

[1609] Step 8:

[1610] The server searches for other products with similar reviews

[1611] The server calculates the similarity between the generated review and reviews of other products in the database.

[1612] Input: Vector-formatted reviews and generated vector-formatted ratings data in the database

[1613] Output: Evaluation data with similarity calculated

[1614] Specific operation: Uses scikit-learn's cosine_similarity function to calculate the similarity with other vectors in the database and identify the most similar vector.

[1615] Step 9:

[1616] The server recommends similar products to the user.

[1617] Based on the calculation results, the server recommends products with the most similar reviews to the user.

[1618] Input: A list of similar products

[1619] Output: Recommended products displayed on the user's device

[1620] Specific behavior: Generates an API response to display recommended results on the user's device and returns the results to the browser or app.

[1621] (Application example 2)

[1622] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1623] Current e-commerce platforms lack a system that uses negative feedback from users to appropriately recommend alternative products to resolve their dissatisfaction. This means users have to spend time and effort to find a satisfactory product. Furthermore, existing systems do not take user sentiment into account when making recommendations, which does not adequately improve the user experience.

[1624] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for a user to post a negative review of a product on an e-commerce platform, means for generating a positive review from the negative review using a generative artificial intelligence, and means for converting the generated review into a vector format. This makes it possible to pass the user's negative review to an emotion recognition engine, extract the user's emotional information, generate a positive review based on the emotional information, search for other products with similar reviews based on the vector-format review stored in the database, and recommend other products with similar reviews to the user based on the search.

[1625] "User" means any individual or organization using the e-commerce platform.

[1626] An "e-commerce platform" is an online system for buying and selling goods and services over the Internet.

[1627] "Negative ratings" are reviews or feedback where users express dissatisfaction with a product or service.

[1628] "Generative AI" is an algorithm or system that uses natural language processing techniques to generate new text.

[1629] "Positive reviews" are positive reviews or feedback that indicate satisfaction with a product or service.

[1630] "Vector format" is a format in which text data is converted into a numerical vector.

[1631] A "database" is a system for efficiently storing, managing, and retrieving data.

[1632] A "search method" is an algorithm or method for finding specific data in a database.

[1633] A "recommendation method" is a system or method for suggesting appropriate products or services to users.

[1634] An "emotion recognition engine" is an artificial intelligence technology for extracting user emotional information from text data.

[1635] "Emotion information" is data that indicates the user's emotional state.

[1636] A "prompt" is an input sentence or cue that the generative AI uses to generate text.

[1637] "Similarity" is an index that indicates the similarity between two pieces of data.

[1638] "Cosine similarity" is a method of calculating similarity using the cosine angle between two vectors.

[1639] A specific system for implementing this invention is one that allows users to post negative reviews of products using an e-commerce platform, and based on those reviews, recommends alternative products to resolve the user's dissatisfaction. This system is mainly composed of a combination of a program, hardware, and a database.

[1640] 1. Program Generation and Explanation

[1641] The server receives negative reviews posted by users and preprocesses them. This preprocessing step involves tokenization and stopword removal to convert the reviews into a more easily parseable format. The server then passes the preprocessed reviews to an emotion recognition engine (e.g., using the Hugging Face transformers library) to extract the user's emotional information.

[1642] Based on the emotional information obtained from the emotion engine, the server uses generative artificial intelligence (e.g., a Hugging Face text-generation model) to convert negative feedback into positive feedback. In this step, the prompt sentence is used to input the generative AI model, which generates a positive feedback.

[1643] Once a positive review is generated, the server converts it into a vector format using techniques such as TF-IDF, which is then stored in a database for future searches and recommendations.

[1644] The server then searches for other products with similar ratings based on the stored vector of ratings. It uses cosine similarity to calculate similarity and identifies products with the most similar reviews. Finally, the server recommends products with the highest similarity to the user, and the recommended results are displayed on the user's device.

[1645] 2. Hardware and Software Used

[1646] The system requires advanced hardware and software for text analysis and generation, including high-performance servers and database servers, and Python, SQL, and the Hugging Face transformers library.

[1647] 3. Examples of concrete examples and prompts

[1648] For example, if a user posts a review stating, "I'm dissatisfied with this smartphone because the battery runs out quickly," the system receives it and performs preprocessing and sentiment analysis. Based on the results of the sentiment analysis, the generative AI model generates a positive evaluation such as, "This smartphone has excellent battery life and can be used for a long time." This evaluation is then vectorized and stored in a database.

[1649] The server then calculates the similarity of the generated rating to other ratings in the database, identifies the product with the most similar rating, and recommends this product to the user, saying, "If you're dissatisfied with your battery, we recommend this smartphone with a long battery life."

[1650] Example prompt for a generative AI model:

[1651] User expressed dissatisfaction with: "I'm not happy with this phone because the battery runs out so quickly." Write a positive review.

[1652] This system allows users to quickly find suitable alternative products to resolve their dissatisfaction, improving the user experience.

[1653] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1654] Step 1:

[1655] A user posts a negative review of a product on an e-commerce platform. For example, the user might write, "I'm not happy with this smartphone because the battery runs out quickly." This review becomes the input to the system.

[1656] Step 2:

[1657] The server receives negative reviews from users. It registers the received reviews internally and passes them to the next step. The input is the user's negative review, and the output is the data passed to the preprocessing step.

[1658] Step 3:

[1659] The server preprocesses the reviews. This includes tokenizing the text, removing stop words, etc. It takes the received negative reviews as input and converts them into a format that is easy to parse. The output is the preprocessed text data.

[1660] Step 4:

[1661] The server passes the preprocessed reviews to the emotion recognition engine. Using the Hugging Face transformers library, the user's emotional information is extracted. The input is preprocessed text data, and after processing by the emotion recognition engine, the output is emotional information. For example, a result such as "Negative (90% probability)" can be obtained.

[1662] Step 5:

[1663] The server generates a prompt based on the emotion information and passes it to the AI ​​generator. For example, it generates a prompt like, "User expressed dissatisfaction with: 'I'm dissatisfied with this smartphone because the battery runs out quickly.' Write a positive review." The input is the emotion information and the preprocessed review, and the output is the prompt text.

[1664] Step 6:

[1665] Generative AI generates a positive review based on the prompt. It uses Hugging Face's text-generation model to create a positive review that eliminates negative elements. The input is the prompt text, and the output is the generated positive review. For example, you might get a review like, "This smartphone has excellent battery life and can be used for a long time."

[1666] Step 7:

[1667] The server vectorizes the generated positive reviews. It uses techniques such as TF-IDF to convert text data into a numerical format. The input is the generated positive reviews, and the output is vector-format data.

[1668] Step 8:

[1669] The server stores the generated rating vector in a database, which is used for future searches and recommendations. The input is vector data, and the output is the completed storage in the database.

[1670] Step 9:

[1671] The server calculates the similarity of the generated review to other reviews in the database. It uses methods such as cosine similarity to find the product with the most similar reviews. The input is rating data in vector format, and the output is the product with the most similar review.

[1672] Step 10:

[1673] The server recommends highly similar products to the user. Based on the calculation results, the most suitable product recommendation is displayed on the user's device. The input is the product information from the search results, and the output is the recommendation to the user. For example, a recommendation such as "If you are dissatisfied with your battery, we recommend this smartphone with a long battery life" is displayed.

[1674] The specific processing unit 290 transmits the result of the specific processing to the robot 414. In the robot 414, the control unit 46A causes the speaker 240 and the control target 443 to output the result of the specific processing. The microphone 238 acquires voice indicating a user input regarding the result of the specific processing. The control unit 46A transmits voice data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[1675] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1676] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the robot 414.

[1677] The emotion identification model 59 as an emotion engine may determine the user's emotion according to a specific mapping. Specifically, the emotion identification model 59 may determine the user's emotion according to an emotion map (see FIG. 9), which is a specific mapping. Similarly, the emotion identification model 59 may determine the robot's emotion, and the identification processing unit 290 may perform identification processing using the robot's emotion.

[1678] FIG. 9 is a diagram illustrating an emotion map 400 on which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. Emotions closer to the center of the concentric circles are more primitive. Emotions representing states and actions arising from a state of mind are arranged on the outer edges of the concentric circles. The concept of emotion includes both affect and mental states. Emotions generally generated from reactions occurring in the brain are arranged on the left side of the concentric circles. Emotions generally induced by situational judgment are arranged on the right side of the concentric circles. Emotions generally generated from reactions occurring in the brain and induced by situational judgment are arranged on the upper and lower sides of the concentric circles. Furthermore, the emotion of "pleasure" is arranged on the upper side of the concentric circles, and the emotion of "discomfort" is arranged on the lower side. In this way, in the emotion map 400, multiple emotions are mapped based on the structure by which emotions are generated, and emotions that tend to occur simultaneously are mapped close to each other.

[1679] These emotions are distributed in the 3 o'clock direction on emotion map 400, and typically fluctuate between relief and anxiety. In the right half of emotion map 400, situational awareness dominates over internal sensations, resulting in a sense of calm.

[1680] The inside of emotion map 400 represents what is going on in the mind, and the outside of emotion map 400 represents behavior, so the further you go outside emotion map 400, the more visible the emotions become (the more they are expressed in behavior).

[1681] Human emotions are based on various balances, such as posture and blood sugar levels. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. Emotions can also be created for robots, automobiles, and motorcycles, based on various balances, such as posture and remaining battery life. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. An emotion map can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on Voice Emotion Recognition and Emotional Brain Physiological Signal Analysis Systems, Tokushima University, Doctoral Dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). The left half of the emotion map lists emotions belonging to the "reaction" domain, where sensation is dominant. The right half of the emotion map lists emotions belonging to the "situation" domain, where situational awareness is dominant.

[1682] The emotion map defines two emotions that promote learning. One is a negative emotion on the situation side, around the middle of "repentance" or "reflection." In other words, this occurs when the robot experiences negative emotions such as "I never want to feel this way again" or "I don't want to be scolded again." The other is a positive emotion on the response side, around "desire." In other words, this occurs when the robot experiences positive feelings such as "I want more" or "I want to know more."

[1683] The emotion identification model 59 inputs user input into a pre-trained neural network, obtains emotion values ​​indicating each emotion shown in the emotion map 400, and determines the user's emotion. This neural network is pre-trained based on multiple pieces of training data that are combinations of user input and emotion values ​​indicating each emotion shown in the emotion map 400. Furthermore, this neural network is trained so that emotions that are located close to each other have similar values, as in the emotion map 900 shown in FIG. 10. FIG. 10 shows an example in which multiple emotions, "relieved," "calm," and "reassuring," have similar emotion values.

[1684] The system according to the present disclosure has been described above mainly with respect to the functions of the data processing device 12, but the system according to the present disclosure is not necessarily implemented on a server. The system according to the present disclosure may be implemented as a general information processing system. The present disclosure may be implemented, for example, as a software program running on a personal computer or an application running on a smartphone, etc. The method according to the present disclosure may be provided to users in the form of SaaS (Software as a Service).

[1685] In the above embodiment, an example was given in which the specific processing is performed by one computer 22, but the technology of the present disclosure is not limited to this, and the specific processing may be distributed and performed by a plurality of computers including the computer 22. For example, the data generation model 58 may be provided in an external device of the data processing device 12, and data may be generated in the external device in accordance with input data.

[1686] In the above embodiment, an example in which the specific processing program 56 is stored in the storage 32 has been described, but the technology of the present disclosure is not limited to this. For example, the specific processing program 56 may be stored in a portable, computer-readable, non-transitory storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-transitory storage medium is installed in the computer 22 of the data processing device 12. The processor 28 executes the specific processing in accordance with the specific processing program 56.

[1687] Alternatively, the specific processing program 56 may be stored in a storage device such as a server connected to the data processing device 12 via the network 54, and the specific processing program 56 may be downloaded and installed on the computer 22 in response to a request from the data processing device 12.

[1688] It is not necessary to store all of the specific processing program 56 in a storage device such as a server connected to the data processing device 12 via the network 54, or to store all of the specific processing program 56 in the storage 32; only a portion of the specific processing program 56 may be stored.

[1689] The hardware resource for executing a specific process can be any of the following processors: An example of a processor is a CPU, which is a general-purpose processor that functions as a hardware resource for executing a specific process by executing software, i.e., a program. Another example of a processor is a dedicated electrical circuit, such as an FPGA (Field-Programmable Gate Array), a PLD (Programmable Logic Device), or an ASIC (Application Specific Integrated Circuit), which is a processor with a circuit configuration designed specifically for executing a specific process. Each processor has built-in or connected memory, and each processor uses the memory to execute the specific process.

[1690] The hardware resource that executes the specific processing may be configured with one of these various processors, or may be configured with a combination of two or more processors of the same or different types (for example, a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Also, the hardware resource that executes the specific processing may be a single processor.

[1691] As an example of a system configured with a single processor, first, one processor is configured by combining one or more CPUs and software, and this processor functions as a hardware resource that executes a specific process. Second, there is a system that uses a processor that realizes the functions of an entire system including multiple hardware resources that execute a specific process on a single IC chip, as typified by SoC (System-on-a-chip). In this way, a specific process is realized using one or more of the above-mentioned various processors as hardware resources.

[1692] Furthermore, the hardware structure of these various processors can be, more specifically, an electric circuit that combines circuit elements such as semiconductor devices. The specific processing described above is merely an example. Therefore, it goes without saying that unnecessary steps may be deleted, new steps may be added, or the processing order may be rearranged, without departing from the spirit of the invention.

[1693] The above-described description and illustrations are a detailed explanation of the parts related to the technology of the present disclosure and are merely an example of the technology of the present disclosure. For example, the above description of the configuration, functions, actions, and effects is an explanation of an example of the configuration, functions, actions, and effects of the parts related to the technology of the present disclosure. Therefore, it goes without saying that unnecessary parts may be deleted, new elements may be added, or replacements may be made to the above-described description and illustrations within the scope of the gist of the technology of the present disclosure. Furthermore, to avoid confusion and facilitate understanding of the parts related to the technology of the present disclosure, the above-described description and illustrations omit explanations of common technical knowledge that do not require particular explanation to enable the implementation of the technology of the present disclosure.

[1694] All publications, patent applications, and technical standards mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication, patent application, or technical standard was specifically and individually indicated to be incorporated by reference.

[1695] The following is further disclosed regarding the above embodiment.

[1696] (Claim 1)

[1697] a means for users to post negative product reviews on the e-commerce platform;

[1698] A means for generating a positive evaluation from the negative evaluation using a generation artificial intelligence;

[1699] means for converting the generated estimates into vector form;

[1700] means for storing the vector-formed ratings in a database;

[1701] A means for searching for other products having similar ratings based on the stored vector-format ratings;

[1702] means for recommending to the user other products with similar ratings based on said search;

[1703] A system including:

[1704] (Claim 2)

[1705] The system of claim 1 further comprising means for preprocessing the generated assessments.

[1706] (Claim 3)

[1707] 2. The system according to claim 1, further comprising means for calculating the similarity using cosine similarity by the search means.

[1708] "Example 1"

[1709] (Claim 1)

[1710] a means for users to post negative product reviews on the e-commerce platform;

[1711] A means for generating a positive evaluation from the negative evaluation using a generation artificial intelligence;

[1712] means for converting the generated estimates into vector form;

[1713] means for storing the vector-formed ratings in a database;

[1714] A means for searching for other products having similar ratings based on the stored vector-format ratings;

[1715] means for recommending to the user other products with similar ratings based on said search;

[1716] means for preprocessing the negative ratings;

[1717] A system including:

[1718] (Claim 2)

[1719] 10. The system of claim 1, further comprising means for using cosine similarity to calculate said similarity.

[1720] (Claim 3)

[1721] The system of claim 1, further comprising means for tokenizing and removing stop words from the negative evaluations before passing them to the artificial intelligence generator.

[1722] "Application Example 1"

[1723] (Claim 1)

[1724] a means for users to post negative product reviews on the e-commerce platform;

[1725] A means for generating a positive evaluation from the negative evaluation using a generation artificial intelligence;

[1726] means for converting the generated estimates into vector form;

[1727] means for storing the vector-formed ratings in a database;

[1728] A means for searching for other products having similar ratings based on the stored vector-format ratings;

[1729] means for recommending to the user other products with similar ratings based on said search;

[1730] A method to analyze user needs based on negative reviews and recommend other products that meet those needs in real time.

[1731] A system including:

[1732] (Claim 2)

[1733] The system of claim 1 further comprising means for preprocessing the generated assessments.

[1734] (Claim 3)

[1735] 2. The system according to claim 1, further comprising means for calculating the similarity using cosine similarity by the search means.

[1736] "Example 2: Combining Emotion Engines"

[1737] (Claim 1)

[1738] a means for users to post negative product reviews on the e-commerce platform;

[1739] A means for generating a positive evaluation from the negative evaluation using a generation artificial intelligence;

[1740] means for converting the generated estimates into vector form;

[1741] means for storing the vector-formed ratings in a database;

[1742] A means for searching for other products having similar ratings based on the stored vector-format ratings;

[1743] means for recommending to the user other products with similar ratings based on said search;

[1744] A means including an emotion engine for recognizing user emotions from review sentences and extracting emotion information;

[1745] The system includes means for generating a positive rating from a negative rating based on the emotion information.

[1746] (Claim 2)

[1747] The system of claim 1 further comprising means for preprocessing the generated assessments.

[1748] (Claim 3)

[1749] 2. The system according to claim 1, further comprising means for calculating the similarity using cosine similarity by the search means.

[1750] "Application example 2 when combining emotion engines"

[1751] (Claim 1)

[1752] a means for users to post negative product reviews on the e-commerce platform;

[1753] A means for generating a positive evaluation from the negative evaluation using a generation artificial intelligence;

[1754] means for converting the generated estimates into vector form;

[1755] means for storing the vector-formed ratings in a database;

[1756] A means for searching for other products having similar ratings based on the stored vector-format ratings;

[1757] means for recommending to the user other products with similar ratings based on said search;

[1758] a means for passing the negative evaluation to an emotion recognition engine to extract the user's emotion information;

[1759] means for generating a positive evaluation based on the emotion information;

[1760] A system including:

[1761] (Claim 2)

[1762] The system of claim 1 , further comprising means for preprocessing the generated ratings and means for using affective information as a prompt.

[1763] (Claim 3)

[1764] 2. The system according to claim 1, further comprising means for calculating the similarity using cosine similarity by the search means. [Explanation of symbols]

[1765] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Device 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robot< / url:> < / url:> < / url:> < / url:>

Claims

1. a means for users to post negative product reviews on the e-commerce platform; A means for generating a positive evaluation from the negative evaluation using a generation artificial intelligence; means for converting the generated estimates into vector form; means for storing the vector-formed ratings in a database; A means for searching for other products having similar ratings based on the stored vector-format ratings; means for recommending to the user other products with similar ratings based on said search; A system including:

2. The system of claim 1 further comprising means for preprocessing the generated assessments.

3. 2. The system according to claim 1, further comprising means for calculating the similarity using cosine similarity by said searching means.

Images