Method for processing natural language information and system for processing natural language information

Abstract

To provide a natural language information processing method and a natural language information processing system.SOLUTION: A method for processing natural language information and a system for processing natural language information are provided. The natural language information processing system includes a cloud server that executes the natural language information processing method, the method including: receiving, after an online interaction program is started, content input by a user through an interaction interface; acquiring semantic features of the content input by the user, using a smart method; acquiring user data and real-time environment information to determine content that suits the semantic features, preference of the user, and the real-time environment information; generate, after performing natural language model processing on the obtained content, interaction content to be introduced to the online interaction program; and output interaction content that suits the semantic features of the user, the preference of the user, and the real-time environment information, to the interaction interface.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to a chatbot, and particularly to a natural language information processing method and a natural language information processing system capable of interacting according to a user's semantic meaning, the user's preferences, and real-time environmental information.

Background Art

[0002] Currently, artificial intelligence (AI) in various fields is developing rapidly. Among them, for example, there is a natural language chatbot that can process natural language and automatically generate content, such as the Chat Generative Pre-trained Transformer developed by OpenAI, known as ChatGPT in English. Such a natural language chatbot can learn a large amount of data using generative AI technology and then generate new data related to the original data, and build a smart model through deep learning (for example, generative adversarial networks (GAN)).

[0003] Taking ChatGPT as an example, ChatGPT can learn a large amount of network information and respond to users in a natural language manner. However, the general content answered to users is a standard answer due to being learned, and it cannot provide a response related to the user's current state in real time. Although it is called a natural language chatbot, it lacks content related to the user and suitable for the real-time situation.

Summary of the Invention

[0004] The present invention provides a natural language information processing method and a natural language information processing system, which can provide dialogue content that conforms to an individual's needs and current situation by further referring to the user's preferences and real-time environment information in the dialogue process between a chatbot realized using natural language processing (NLP) and generative artificial intelligence technology and the user.

[0005] In a natural language information processing system realized by a computer system, a cloud server that executes a natural language information processing method by a processing circuit is provided. By the cloud server, after the user starts an online dialogue program through a user interface and receives the content input by the user through a dialogue interface, by acquiring the semantic features of the content input by the user, user data, and real-time environment information, it is possible to determine content that conforms to the semantic features of the content input by the user, the user's preferences obtained from the user data, and the real-time environment information. Then, after performing natural language model processing in the online dialogue program, dialogue content can be generated, and finally, after introducing it into the online dialogue program, the dialogue content can be output.

Brief Description of the Drawings

[0006]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Best Mode for Carrying Out the Invention

[0007] This specification discloses a natural language information processing method and a natural language information processing system. The natural language information processing method is executed on a cloud server to implement the natural language information processing system. The cloud server provides a social media service through a network. After subscribing, users can share text content, image content, and video content. The cloud server further provides chatbots in various fields, and users can interact through the services provided by the cloud server. Using artificial intelligence technology includes chatbots that use machine learning algorithms and natural language processing technology to learn data in various fields to provide an interactive service, learn the activities of users on social media to obtain user preferences, and the chatbots can provide interactive content that conforms to the user's individual needs and current environmental characteristics based on the semantics of the user's conversation and the user's preferences, plus the real-time obtained environmental information.

[0008] Reference can be made to the diagram of an embodiment showing the system configuration for executing the natural language information processing method shown in FIG. 1 by the above-mentioned cloud server.

[0009] The cloud server 100 shown in the figure is realized by a computer system, a database, and a network, and includes various functional modules realized by the cooperation of software and hardware. As shown in the figure, it has a natural language processing module 101, a machine learning module 103, an external system interface module 105, and a user interface module 107. The natural language processing module 101 is a module for executing natural language information processing methods. The natural language processing module 101 realizes a chatbot with natural language information processing capabilities. The machine learning module 103 executes machine learning algorithms. In addition to learning the natural language model, it can learn the activities of users on the network by deep learning methods to obtain information preferred by users, and provide the chatbot with dialogue content that conforms to the user's preferences. The external system interface module 105 provided by the cloud server 100 executes the connection with external systems (such as the first external system 111 and the second external system 112) (for example, via the network 10), the circuit for obtaining data through an application program interface (API), and related software. The user interface module 107 provided by the cloud server 100 executes a server that enables the user device 150 to connect to the cloud server 100 through the network connection function in the user interface module 107 to provide network services, and executes an application program for obtaining services corresponding to the services provided by the cloud server 100 in the user device 150.

[0010] According to the configuration shown in the figure, the cloud server 100 is provided with a built-in or externally connected database, and the cloud server 100 provides data services. The video database 110 shown in the figure provides video content uploaded and shared by each end user stored by the user device 150 via the network 10. What the user uploads and shares may include text content and image content. In the user database 120, user data including user personal data, text content, image content, and video content to be uploaded is stored, and user activity data in the network services provided by the cloud server 100, such as network activities such as the content viewed, followed, liked button, shared, and reserved subscription, is acquired, and thus a user profile can be formed. When continuously generating interactive content over time, the user database 120 can store and update user data including the description of the user's historical interactive records in chronological order so as to be an interactive record provided to the natural language model and learned by the machine learning algorithm. The vector database 130 is a database for storing structured information obtained by performing operations on various text content, image content, and video content using a vector algorithm. The structured information is used to compare various data adapted to the personalization of the user.

[0011] According to the system configuration schematic diagram shown in the figure, the cloud server 100 obtains data of external systems such as the first external system 111 and the second external system 112 schematically shown in the figure through a connection in the network 10 or a specific protocol. The first external system 111 and the second external system 112 are, for example, servers installed by a government or a company for providing open data. The cloud server 100 can obtain real-time information that meets the needs, such as real-time weather, real-time road conditions, real-time news, and network information related to the real-time position, through the application program interfaces provided by the external systems via the external system interface module 105.

[0012] An application program that can obtain the services provided by the cloud server 100 in the user device 150, for example, runs the social media service provided by the cloud server 100. The user device 150 runs the corresponding social media application program and obtains the social media service through the user interface module 107. In particular, the cloud server 100 provides a natural language chat robot through the natural language processing module 101, and the user can interact with the chat robot through the dialogue interface 115. On the other hand, the cloud server 100 can learn the activity data of the user using various services of the cloud server 100 through the machine learning module 103, and the machine learning module 103 can learn the user's interest characteristics and construct user data. The activity data includes the activity data obtained by the user using the social media application program and the dialogue interface 115.

[0013] As mentioned here, various text contents, image contents, and video contents acquired by the cloud server 100 are unstructured data, and may be converted into vectorized data by an encoder method so as to facilitate obtaining the meanings of the various text contents, image contents, and video contents and be advantageous for data search. Furthermore, the vectorized vectorized data is used for comparison with the user's keyword search, and the distance function is used to calculate the distance between the search keyword and the vectorized data in the database. The closer the distance is, the more similar the data is. Thereby, the user can search for data by the vector database 130.

[0014] According to the embodiment, the vector database 130 of the cloud server 100 can support multi-mode search services such as text content and image content. The vector database 130 provides structured data. For example, after various text contents, image contents, and video contents are texturized, vectorized data is obtained by the operation of a vector algorithm, and the vectorized data can be applied to the search service and can be used in a natural language processing program. The natural language processing program uses a natural language model to correspond the vectorized data to a vector space. Taking the word input by the user as an example, a word vector is obtained by the operation of a vector algorithm.

[0015] In an embodiment, the natural language information processing method provided in this specification is actually a chatbot executed on the cloud server 100. The chatbot can interact with a user in a natural language including text and voice. In addition to answering the information input by the user, it may also obtain user data in advance by the cloud server 100 before the interaction, and obtain the user's personality and habits from the user data. It may also obtain real-time status from external systems (the first external system 111, the second external system 112). For example, it may obtain the weather and news of the location according to the user's location. Thereby, the content answered by the chatbot can not only target the user's preferences, but also reflect the actual situation.

[0016] Referring to FIG. 2, which is a diagram of an embodiment showing a data structure for executing the natural language information processing method, the data is divided into social media platform data 21, user data (for example, user description file) 23, and user activity data 25.

[0017] The social media platform data 21 is data not publicly available in the system, and the cloud server 100 obtains viewer data 211 accessed by the user to various contents provided by the system for executing the natural language information processing method and creator data 212 of the user who provides various contents of the system. The social media platform data 21 further includes business data 213 for the system to provide enterprise data to enterprises for advertising and various location data 214 related to obtaining geographical locations for the system to provide appropriate services.

[0018] User data 23 is data that can be made public in the system and includes data edited by the user himself / herself. The data edited by the user himself / herself includes viewer data 231 obtained from various user activity data. The viewer data 231 may include the user's preference data as a viewer obtained through machine learning, such as recent preference data, past preference data, and preference data related to location.

[0019] The creator data 232 in the user data 23 is the relevant data of the user as a creator and includes the type preferred by the creator and the location-related data of the creator obtained by the system through machine learning. The creator data 232 may include, for example, the data of the user as a creator, and the type preferred by the creator obtained through learning and the location including the geographical location or a specific location of a place.

[0020] The business data 233 in the user data 23 includes the business type of the enterprise and product features obtained by the system through machine learning when the user is an enterprise.

[0021] The user activity data 25 is data that is not made public in the system and includes statistical data of the user's activities in various services provided by the cloud server 100 and data obtained through machine learning. The user activity data 25 mainly includes viewer data 251, creator data 252, and business data 253.

[0022] The viewer data 251 is the viewing rate, viewing time, and activity data such as following, preference, comment, and reserved subscription of the services provided by the user on the cloud server 100. The creator data 252 is the statistical data of the user as a critator, such as the followers of the channel or account, the number of views of the created content, and the account viewing rate. The business data 253 is the followers, the number of content views, and the overall impression data obtained when the user is an enterprise.

[0023] The social media platform data 21, user data 23, and user activity data 25 collected and learned by the above cloud server 100 serve as the basis for the dialogue service realized using the natural language processing and generative artificial intelligence mentioned in this specification. In the above cloud server 100, by performing operations on the above various types of data by the processing circuit, it is realized that a chatbot that conforms to the user's personalization and real-time needs can be provided.

[0024] As an example, in the cloud server 100, the natural language model to be executed performs a vector algorithm on the content input by the user through the dialogue interface, the user's preferences, and real-time environment information, marks the obtained text, calculates the vector of each word, and obtains relevant content after investigating the database based on the vector distance between words, so as to generate dialogue content that conforms to the user's preferences and real-time environment information. In the process of online dialogue, a program for performing machine translation, document summarization, document generation, etc. on the texturized data using the conversion model may also be executed. Referring to the description of the flow of the embodiment shown in FIG. 4, by obtaining the semantic meaning during the user's dialogue, the chatbot may be made to generate dialogue content.

[0025] In the flow shown in FIG. 3, first, the user connects to the cloud server 100 using an application program executed on the user device and browses the content therein using the provided service. The graphical user interface for operating and browsing the text content, image content, and video content provided by the cloud server 100 (e.g., by the user interface module 107 shown in FIG. 1) to the user is a map interface 50 with an electronic map as the background, which is launched after the application program is executed by the user device as shown in FIG. 5. FIG. 5 schematically shows that the link icons are displayed on the map interface 50 based on the geographical information related to each video content. The link points related to a plurality of geographical locations as shown in the figure may include one or more video link points. In the drawing, video link points 501, 502, and 503 are taken as examples.

[0026] According to an embodiment, as shown in FIG. 5, while browsing the content of the map interface 50, the user may start an online interaction program (step S301) by clicking on the interaction link icon of the interaction 512 with a touch or a specific hand gesture, or by clicking on the link point for providing interaction in the map interface 50, through link icons such as play 511, interaction 512, help 513, search 514, and user main page 515, etc., shown as a plurality of icons below.

[0027] On the one hand, FIG. 6 provides a schematic diagram of another link method for starting an online interaction program. For example, after the user selects any video link point (video link points 501, 502, or 503) of the map interface 50, as shown in FIG. 6, the screen launches a video playback page 60 that plays a video shared or produced by a certain user. In FIG. 6, the geographical location 601 of the video is further displayed, and a plurality of link icons such as preferences 603, interaction 604, collection 605, share 606, etc. are displayed on the sidebar. The user may also launch the online interaction program by clicking on the interaction link icon of interaction 604 (step S301).

[0028] Next, after launching the interaction interface, the user inputs text content, image content, or specific video content through the interaction interface (for example, inputs a link to share video content), and the cloud server 100 may receive the content input by the user through the user interface module 107 (step S303). According to an embodiment, the above online interaction program is actually a chatbot using a natural language model, and may interact with the user through the interaction interface and perform a natural language information processing method on the content input by the user each time. By way of example, reference may be made to the interaction interface 70 shown in FIG. 7, the interaction interface 80 shown in FIG. 8, the interaction interface 90 shown in FIG. 9, etc. Each interaction interface shown in the examples provides an input field for the user to input content and an interaction display area for displaying the interaction content output by the chatbot and the content input by the user.

[0029] At this time, the cloud server 100 acquires the content input by the user through the user interface module 107. The content received through the dialogue interface may be text content, voice content, or video content. When the received content is voice content or video content, the voice content or video content may be converted into text content by a text conversion program, and then semantic analysis may be performed to obtain semantic features (step S305). During the execution of the above program, the cloud server 100 acquires user data from the user database 120 and acquires real-time environment information from an external system (through the external system interface module 105 shown in FIG. 1) (step S307).

[0030] After that, it may be possible to determine (or screen after investigating the database) the content that conforms to the semantic features of the content input by the user, the user's preferences obtained from the user data, and the real-time environment information (step S309). After performing natural language model processing in the online dialogue program, dialogue content is generated (step S311). Then, the dialogue content is introduced into the online dialogue program and output to the dialogue interface (step S313).

[0031] Furthermore, when the natural language model of the cloud server 100 operates, multi-dimensional information is recorded using a database or system memory, and the multi-dimensional information may include historical dialogue records of the same online dialogue program. Thereby, before the chatbot generates a dialogue, in addition to considering the semantic features, the user's preferences, and the real-time environment information during the user's dialogue as in step S309, the historical dialogue record in the user's current online dialogue program is further considered (step S315). As a result, the dialogue content generated by the natural language model (step S311) has dialogue content that conforms to the current situation.

[0032] For example, in the historical conversation records of the same online conversation program of a user, there are often current situations that can reflect the user's current emotions and needs. Therefore, as shown in the embodiment of FIG. 1, the natural language processing module 101 in the cloud server 100 can use the natural language model to simultaneously consider the user's semantics, preferences, real-time environment information, and historical conversation records, and can continue the same conversation situation when generating conversation content. For example, when generating a conversation in natural language while continuing the same conversation theme, in terms of terms, having the same tone as the previous conversation content (reflecting the user's emotions such as joy, anger, sorrow, and happiness), the chatbot may learn the user's emotional expression method from the historical conversation records.

[0033] As a related figure, refer to FIG. 7. As shown in FIG. 7, in the dialogue interface 70, there are dialogue contents 701, 702, 703 between the user and the chatbot. The chatbot further investigates the database based on the user semantic features obtained from the dialogue content 702 to provide recommended video content 704, and provides an input field 705 below the dialogue interface 70 where the user can further input dialogue content.

[0034] As another mode, there is the dialogue interface 80 shown in FIG. 8. As shown in this example, when starting the online conversation program, the system directly provides natural language dialogue contents 801, 802, 804 based on the user's preferences and real-time information, directly provides recommended video content 803, and the user may subsequently use the input field 805 in the dialogue interface 80 to respond to the above dialogue content.

[0035] As another embodiment, the dialogue content generated by the natural language model executed in the chatbot according to the semantic features of the content input by the user, the user's preferences, and the real-time environment information may further include providing a plurality of recommendation options, a plurality of recommended video contents, and / or a plurality of recommended friend links. Refer to the example of FIG. 9.

[0036] In an online dialogue program, the dialogue interface 90 shown in FIG. 9 includes dialogue content 901 generated based on the semantic features of the user obtained by the chatbot. Since the semantics in this example enable the chatbot to determine that the user has selected a specific item, a plurality of recommendation options 902 are provided. In particular, the chatbot provides recommendation options to the user based on the real-time environmental information obtained by the system from an external system. For example, the chatbot may provide the recommendation options 902 based on real-time weather, road conditions, time, and the user's location. When the time is exactly meal time, referring to the user's eating habits, meal options are provided by restaurants that are open and close to the user's location.

[0037] In addition, when the user expresses the desire to watch video content, the above recommendation options 902 may be a plurality of recommended video contents. When the user expresses the desire to find friends with similar hobbies, the above recommendation options 902 may be a plurality of recommended friend links.

[0038] Furthermore, when the user subsequently uses the input field 906 to input dialogue content 903 in response to some of the recommendation options 902, the chatbot further answers with dialogue content 904 based on the semantics of the dialogue content 903, and further provides a plurality of recommended contents 905 based on the semantics of the above dialogue content. Continuing with the above example, when the user answers that they want one of the meals, the chatbot provides options for restaurants corresponding to the meal the user wants to eat based on the real-time weather, road conditions, and the user's location obtained by the system from an external system. When the weather is bad or the road is congested, corresponding options for restaurants that are easy for the user to go to are recommended.

[0039] As yet another example, reference may be made to the flowchart of another embodiment of the natural language information processing method shown in FIG. 4.

[0040] In the flow shown in FIG. 4, the user starts an online interaction program through an application program (step S401), interacts with the chatbot, the system receives the interaction content input by the user (step S403), and further obtains the semantic features of the user. As an example, in the cloud server 100, conversion operations and vector algorithm operations may be performed using a natural language processing module to obtain semantic features (step S405).

[0041] Here, as an example of the natural language information processing method provided by the present invention, after learning natural language using artificial intelligence technology and understanding natural language, text classification and grammar analysis are performed. When processing the interaction content input by the user, a deep learning method of a conversion model (transformer model, proposed by Google's Brain team in 2017) may be used to process the user input having natural language content in chronological order. If the input content is not text content, it is necessary to textify it and then obtain the text. In this way, in the online interaction program, this conversion model may be used to perform machine translation, document summarization, document generation, etc. TM After obtaining the semantic features of the user interaction content, according to the user's preferences and the user's current position obtained by the system or analyzing the position that the user is interested in from the interaction content, real-time environmental information is obtained in real time from an external system according to that position (step S407). Note that the real-time environmental information may include information obtained in real time from one or more external systems, such as real-time weather, real-time road conditions, real-time news, and real-time location-related network information (POI (point of interface) on the map, evaluation of POI, etc.), or any combination thereof.

[0042] After obtaining the semantic features of the user interaction content, according to the user's preferences and the user's current position obtained by the system or analyzing the position that the user is interested in from the interaction content, real-time environmental information is obtained in real time from an external system according to that position (step S407). Note that the real-time environmental information may include information obtained in real time from one or more external systems, such as real-time weather, real-time road conditions, real-time news, and real-time location-related network information (POI (point of interface) on the map, evaluation of POI, etc.), or any combination thereof.

[0043] After that, the system uses the vector database to calculate the closest answer based on the user's semantic features, preferences, and real-time environmental information, or by adding historical conversation records (step S409). It should be noted here that the data in the vector database is structured information obtained using a vector algorithm. The system may obtain words with relatively similar semantics from the content obtained by the vector distance. For example, the vector distance between the word "computer" that appears in the conversation content and the word "calculation" in the database is relatively close, while the vector distance between the word "computer" and the word "run" is relatively far.

[0044] In this embodiment, for the content input by the user, the content the user is interested in, and the real-time environmental information, according to the needs, add the historical conversation records, execute the vector algorithm, then mark the obtained text, calculate the vectors of each word, and obtain the relevant content by the vector distance between words, so as to generate conversation content that conforms to the user's preferences and real-time environmental information. Further, in the embodiment, when the vector algorithm is executed on the historical conversation records described in the cloud server 100, conversation content that conforms to the user's current mood may be generated. For example, the same topic in the historical conversation records can be continued, or terms corresponding to the mood obtained by analysis can be used.

[0045] Furthermore, the system further investigates the video database 110 based on the above information to obtain suitable video content (step S411). The chatbot generates conversation content by natural language processing and artificial intelligence technology for generating natural language (step S413) and outputs the conversation content to the conversation interface (step S415). Also, in the embodiment, during the chat process, when the system continues the above steps, the chatbot may interact with the user in natural language (text or voice) and provide real-time content (video content, text content) that the user is interested in.

Description of Reference Numerals

[0046] 10: Network 100: Cloud Server 101: Natural Language Processing Module 103: Machine Learning Module 105: External System Interface Module 107: User Interface Module 110: Video Database 111: First External System 112: Second External System 115: Dialogue Interface 120: User Database 130: Vector Database 150: User Device 21: Social Media Platform Data 211: Viewer Data 212: Creator Data 213: Business Data 214: Location Data 23: User Data 25: User Activity Data 231, 251: Viewer Data 232, 252: Creator Data 233, 253: Business Data 50: Map Interface 501, 502, 503: Video Link Point 511: Playback 512, 604: Dialogue 513: Help 514: Search 515: User Main Page 60: Video Playback Page 601: Geographical Location 603: Preferences 605: Collection 606: Share 70, 80, 90: Dialogue Interface 701, 702, 703, 801, 802, 804, 901, 903, 904: Dialogue Content 704, 803: Recommended video content 705, 805, 906: Input fields 902: Recommendation options 905: Recommended content S301~S315, S401~S415: Natural language information processing flow

Claims

1. A natural language information processing method executed in a cloud server, comprising: Launching an online interactive program and receiving content input by a user through an interactive interface; Obtaining semantic features of the content input by the user; Obtaining user data and real-time environmental information; determining content that matches the semantic features of the content input by the user, the user's preferences obtained from the user data, and the real-time environment information, and then performing natural language model processing in the online dialogue program to generate dialogue content; introducing the dialogue content into the online dialogue program and outputting it to the dialogue interface.

2. The content received by the dialogue interface is text content, audio content, or video content; The natural language information processing method according to claim 1 , wherein when the received content is audio content or video content, the audio content or video content is converted into text by a text conversion program, and then semantic analysis is performed to obtain semantic features of the text.

3. The natural language information processing method according to claim 2 , wherein the natural language model executed on the cloud server generates the dialogue content by executing a program for machine translation, document summarization, and document generation using a conversion model.

4. The natural language information processing method of claim 3, wherein in the cloud server, a vector algorithm is executed on the content input by the user, the user's preferences, and the real-time environmental information to mark the acquired text, calculate the vector of each word, and acquire related content based on the vector distance between words, thereby generating dialogue content that matches the user's preferences and the real-time environmental information.

5. The method of claim 4 , further comprising: executing the vector algorithm on historical dialogue records stored in the cloud server to generate the dialogue content.

6. The natural language information processing method of claim 1 , wherein the real-time environmental information includes one or any combination of real-time weather, real-time road conditions, real-time news, and real-time location-related network information obtained in real time from one or more external systems.

7. The method of claim 1 , wherein the dialogue content includes text content generated after performing the natural language model processing and video content obtained by examining a video database.

8. the online dialogue program operates a chat robot using the natural language model, dialogues with the user through the dialogue interface, and executes the natural language information processing method for the content input by the user each time; The natural language information processing method of any one of claims 1 to 7, wherein the dialogue content generated by the natural language model executed in the chat robot based on the semantic features of the content input by the user, the user's preferences, and the real-time environmental information includes providing a plurality of recommendation options, a plurality of recommended video contents, and / or a plurality of recommended friend links.

9. A natural language information processing system implemented by a computer system, comprising: A cloud server that executes a natural language information processing method by a processing circuit, The natural language information processing method includes: Launching an online interactive program and receiving content input by a user through an interactive interface; Obtaining semantic features of the content input by the user; Obtaining user data and real-time environmental information; determining content that matches the semantic features of the content input by the user, the user's preferences obtained from the user data, and the real-time environment information, and then performing natural language model processing in the online dialogue program to generate dialogue content; introducing the dialogue content into the online dialogue program and outputting the dialogue content to the dialogue interface.

10. The natural language information processing system of claim 9, wherein the cloud server provides an external system interface for connecting to one or more external systems to obtain the real-time environmental information including one or any combination of real-time weather, real-time road conditions, real-time news, and real-time location-related network information.

Images