Commodity review content processing method, and electronic device
By analyzing the relationship between product categories and decision factor labels, using AI models to generate detailed evaluation content, the problem of low quality of user evaluation content is solved and the reference value and decision value of evaluation is improved.
Patent Information
- Application Number
- PCT/CN2024/094681
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2023-05-31
- Filing Date
- 2024-05-22
- Publication Date
- 2024-12-05
AI Technical Summary
In the existing product evaluation system, the quality of user evaluation content is low. When filling out the evaluation, the user has limited written expression ability or is unwilling to spend too much time, which makes the evaluation content simple and difficult to provide effective reference for other users.
By analyzing the relationship between product categories and decision factor labels, we generate recommended evaluation content, and use AI big models to process the basic evaluation content to increase the detailedness and relevance of evaluation.
It improves the quality and reference value of user evaluation content, makes evaluation more decision-making value, and helps other users better understand product characteristics and user experience.
Smart Images

Figure CN2024094681_05122024_PF_FP_ABST
Abstract
Description
Product review content processing method and electronic device
[0001] This application claims priority to the Chinese patent application filed with the China Patent Office on May 31, 2023, with application number 202310638276.1 and application name “Product Evaluation Content Processing Method and Electronic Device”, all contents of which are incorporated by reference into this application. Technical Field
[0002] The present application relates to the field of information processing technology, and in particular to a method for processing product evaluation content and an electronic device. Background Art
[0003] In a product information service system, after a user purchases a specific product, they can leave a review of the product through the portal provided in the system, including their own experience using the product, etc. This user review content can provide a reference for other users. However, in actual applications, when users leave a review for a product they have already purchased, they may only score some required rating items (for example, description match score, merchant service score, logistics service score, etc.) due to limited written expression skills or not wanting to spend too much time thinking about how to express themselves. The main body of the review may be very brief. For example, only content such as "very good" or "not bad" is written, and this kind of review content may be difficult to provide effective reference for other users.
[0004] In some systems, in order to improve the quality of user evaluation content, some reference tags can be provided to users based on the category of the product when they fill in the user evaluation content. For example, for food products, specific reference tags may include freshness, taste, packaging quality, etc., and users can enter structured evaluation content by selecting such tags. However, users still need to think about how to express the freshness, taste, etc. of specific products, so the probability of being actually used by users is relatively low. In addition, even if users use such reference tags to express themselves, the specific evaluation content displayed will appear too mechanical and stereotyped due to the unified format (for example, each tag corresponds to a paragraph, etc.), and therefore it is easy to be ignored by readers.
[0005] Therefore, how to improve the quality of user evaluation content of products in the system has become a technical problem that needs to be solved by those skilled in the art.
[0006] Summary of the Invention
[0007] This application provides a product review content processing method and electronic device, which can improve the quality of user review content of products in the system.
[0008] This application provides the following solutions:
[0009] A method for processing product review content, comprising:
[0010] By analyzing user reviews of multiple products in a product category, the corresponding relationship between product categories and decision factor labels is determined. The decision factor label includes at least one text label that is the focus of users' attention when making decisions about products in the same product category.
[0011] After the target user initiates a request to fill in the user evaluation content for the target product object, the basic evaluation content input by the target user is obtained;
[0012] The basic evaluation content is processed according to the decision factor label corresponding to the commodity category to which the target commodity belongs to generate suggested evaluation text content, so as to publish user evaluation content for the target commodity according to the suggested evaluation text content.
[0013] The analysis of user reviews corresponding to multiple products to determine the corresponding relationship between product categories and decision factor labels includes:
[0014] Perform word segmentation on multiple user reviews collected under the same product category and perform semantic recognition on the resulting text tags;
[0015] After clustering multiple text labels according to the semantic recognition results, at least one text label that is the focus of user attention under the same product category is determined and determined as the decision factor label when the user makes a decision on the product category.
[0016] Among them, also include:
[0017] Determining target tone or style information selected by the target user;
[0018] The processing of the basic evaluation content according to the decision factor label corresponding to the commodity category to which the target commodity belongs includes:
[0019] The basic evaluation content is processed according to the decision factor label corresponding to the product category to which the target product belongs and the target tone or style information to generate recommended evaluation content with the target tone or style.
[0020] The basic evaluation content includes basic evaluation text content and basic picture / video content;
[0021] The processing of the basic evaluation content according to the decision factor label corresponding to the commodity category to which the target commodity belongs includes:
[0022] Generating textual expressions of the basic image / video content by performing model understanding processing on the basic image / video content;
[0023] According to the decision factor label corresponding to the commodity category to which the target commodity belongs, the basic evaluation text content and the text expression content generated based on the basic image / video content are processed to generate the recommended evaluation text content.
[0024] The basic evaluation content includes basic evaluation text content and basic picture / video content;
[0025] The method further comprises:
[0026] In response to the request for optimizing the basic image / video content, the basic image / video content is optimized according to the product category to which the target product belongs, so that the optimized image / video content is published.
[0027] A method for processing product review content, comprising:
[0028] After the target user initiates a request to fill in the user evaluation content for the target product object, the basic evaluation content input by the target user is obtained;
[0029] Submitting the basic evaluation content to a server, so that the server processes the basic evaluation content according to the decision factor tag corresponding to the product category to which the target product belongs, and generates recommended evaluation text content; wherein the decision factor tag is determined by analyzing user evaluation content of multiple products associated with the product category, and the decision factor tag includes: at least one text tag that is the focus of the user's attention when making a decision about products in the same product category;
[0030] The suggested evaluation text content is displayed so that user evaluation content is published for the target product based on the suggested evaluation text content.
[0031] A method for providing product information, comprising:
[0032] By collecting and analyzing multiple user reviews of a product, the decision factor label corresponding to the product is determined. The decision factor label includes at least one text label that is the focus of the user's attention when making a decision on the same product.
[0033] In the process of providing target users with information about target products, user evaluation summary content is generated for the target products based on the decision factor labels corresponding to the target products and the target user information, so as to provide the product evaluation summary content to the target users.
[0034] When collecting user evaluation content, the scope of collection includes the current product information service system, other product information service systems and / or social media systems;
[0035] The method further comprises:
[0036] For product reviews collected from different systems, we analyze the images and / or text descriptions of the corresponding products to determine whether they correspond to the same product.
[0037] Among them, also include:
[0038] Determine the sentiment tag corresponding to the decision factor tag, so as to generate the product evaluation summary content in combination with the sentiment tag.
[0039] Among them, also include:
[0040] By analyzing the product reviews published by multiple publishers with the same / similar characteristics as the target user for the target product, determining the decision factor labels when the users with the same / similar characteristics make decisions about the product;
[0041] Generating a user evaluation summary for the target product based on the decision factor label corresponding to the target product and the target user information includes:
[0042] Based on the decision factor label corresponding to the target product and the decision factor label when the user with the same / similar characteristics makes a decision on the product, a user evaluation summary content is generated for the target product, and the user evaluation summary content includes purchase recommendation information provided for the target user.
[0043] A method for providing product information, comprising:
[0044] In the process of providing information related to a target product to a target user, obtaining a user review summary of the target product, wherein the user review summary is generated based on a decision factor tag corresponding to the target product and information about the target user, and the decision factor tag is determined by collecting and analyzing multiple user reviews corresponding to the target product, and the decision factor tag includes at least one text tag that is a focus of the user's attention during the user's decision-making process for the same product;
[0045] Output the user evaluation summary content.
[0046] A method for processing product review content, comprising:
[0047] After the target user initiates a request to fill in user evaluation content for the target product object, in response to the request for assisting in evaluation content processing, basic evaluation content input by the target user is obtained, wherein the basic evaluation content includes basic picture / video content;
[0048] The basic picture / video content is optimized according to the category to which the target product belongs, so that the optimized picture / video content can be published.
[0049] The optimizing process of the basic image / video content includes:
[0050] If the basic picture / video content includes multiple pictures, the order of the multiple pictures is adjusted by performing model understanding on the multiple pictures and according to the product category to which the target product belongs.
[0051] The optimizing process of the basic image / video content includes:
[0052] The subject of the basic image / video content is identified, and the subject type information of the identified subject is determined. The identified subject is optimized according to the product category to which the target product belongs and the optimization processing rule information of each subject type corresponding to the product category.
[0053] The optimizing process of the basic image / video content includes:
[0054] If the basic video content includes voice content, the voice content is converted into text content by performing voice recognition on the voice content, and a subtitle file of the video content is generated according to the text content, so as to add subtitles to the basic video content.
[0055] The text content is translated into multiple languages, and subtitle files in corresponding languages are generated respectively, so as to provide subtitle information in corresponding languages for users with different language requirements in a multilingual scenario.
[0056] The optimizing process of the basic image / video content includes:
[0057] If the basic picture / video content includes a facial image, an AI facial image is generated through an artificial intelligence (AI) algorithm to replace and display the original facial image in the basic picture / video content.
[0058] The optimizing process of the basic image / video content includes:
[0059] The privacy information and background contained in the basic image / video content are removed or blurred, and / or meaningless segments in the video content are cropped.
[0060] A product review content processing device, comprising:
[0061] A preprocessing unit is configured to analyze user reviews of multiple products by product category to determine a correspondence between product categories and decision factor labels; wherein the decision factor label includes at least one text label that is a focus of user attention when making a decision about products in the same product category;
[0062] A basic evaluation content acquisition unit is used to acquire the basic evaluation content input by the target user after the target user initiates a request to fill in the user evaluation content for the target product object;
[0063] The suggested evaluation content generating unit is used to process the basic evaluation content according to the decision factor label corresponding to the commodity category to which the target commodity belongs, and generate suggested evaluation text content, so as to publish user evaluation content for the target commodity according to the suggested evaluation text content.
[0064] A product review content processing device, comprising:
[0065] A basic evaluation content acquisition unit is used to acquire the basic evaluation content input by the target user after the target user initiates a request to fill in the user evaluation content for the target product object;
[0066] A basic evaluation content submission unit is configured to submit the basic evaluation content to a server, so that the server processes the basic evaluation content according to a decision factor tag corresponding to the product category to which the target product belongs, and generates a suggested evaluation text content; wherein the decision factor tag is determined by analyzing user evaluation content of multiple products associated with the product category, and the decision factor tag includes: at least one text tag that is the focus of the user's attention when the user makes a decision on products in the same product category;
[0067] The suggested evaluation text content display unit is used to display the suggested evaluation text content so as to publish user evaluation content for the target product based on the suggested evaluation text content.
[0068] A device for providing product information, comprising:
[0069] A preprocessing unit is configured to collect and analyze multiple user reviews of a product, and determine a decision factor label corresponding to the product. The decision factor label includes at least one text label that is a focus of the user's attention during the user's decision-making process for the same product.
[0070] The product evaluation summary content generation unit is used to generate user evaluation summary content for the target product based on the decision factor label corresponding to the target product and the information of the target user in the process of providing information about the target product to the target user, so as to provide the product evaluation summary content to the target user.
[0071] A device for providing product information, comprising:
[0072] A user evaluation summary content acquisition unit is configured to acquire user evaluation summary content regarding a target product in the process of providing information related to the target product to the target user, wherein the user evaluation summary content is generated based on a decision factor label corresponding to the target product and information about the target user. The decision factor label is determined by collecting and analyzing multiple user evaluation contents corresponding to the target product. The decision factor label includes at least one text label that is a focus of the user's attention during the user's decision-making process for the same product.
[0073] The user evaluation summary content output unit is used to output the user evaluation summary content.
[0074] A product review content processing device, comprising:
[0075] A basic evaluation content acquisition unit is configured to, after a target user initiates a request to fill in user evaluation content for a target product object, respond to a request to assist in evaluation content processing and acquire basic evaluation content input by the target user, the basic evaluation content including basic picture / video content;
[0076] The optimization processing unit is used to optimize the basic picture / video content according to the category to which the target product belongs, so as to publish the optimized picture / video content.
[0077] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of any of the aforementioned methods.
[0078] An electronic device, comprising:
[0079] one or more processors; and
[0080] A memory associated with the one or more processors, the memory being used to store program instructions, wherein the program instructions, when read and executed by the one or more processors, execute the steps of any of the aforementioned methods.
[0081] According to the specific embodiments provided in this application, this application discloses the following technical effects:
[0082] Through the embodiment of the present application, the user evaluation content of the product can be analyzed in advance based on the product category to determine the decision factor label corresponding to the specific product category. This decision factor label can be a text label that is the focus of the user's attention when the user makes a decision on the product in the specific product category. In this way, in the case where the user needs to submit user evaluation content for a target product, after the user submits the basic evaluation content, the recommended evaluation text content can be generated based on this basic user evaluation content and the decision factor label corresponding to the specific product category. In this way, since the basic evaluation content output by the user can be rewritten, the decision factor label corresponding to the specific product category can be fully utilized during the rewriting process. This decision factor label is determined by information mining from the user evaluation content corresponding to the specific product category. Therefore, the recommended evaluation text content generated by the algorithm can not only be more complete and detailed, but also express more text labels related to the focus of the user's decision, so that the user evaluation content has more reference value in helping other users make decisions, that is, to improve the quality of the user evaluation content.
[0083] Of course, any product implementing the present application does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS
[0084] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0085] FIG1 is a schematic diagram of a system architecture provided in an embodiment of the present application;
[0086] FIG2 is a flow chart of a first method provided in an embodiment of the present application;
[0087] FIG3 is a schematic diagram of a first interface provided in an embodiment of the present application;
[0088] FIG4 is a schematic diagram of a second interface provided in an embodiment of the present application;
[0089] FIG5 is a schematic diagram of a third interface provided in an embodiment of the present application;
[0090] FIG6 is a flow chart of a second method provided in an embodiment of the present application;
[0091] FIG7 is a flowchart of a third method provided in an embodiment of the present application;
[0092] FIG8 is a schematic diagram of a fourth interface provided in an embodiment of the present application;
[0093] FIG9 is a flowchart of a fourth method provided in an embodiment of the present application;
[0094] FIG10 is a flowchart of a fifth method provided in an embodiment of the present application;
[0095] FIG11 is a schematic diagram of a first device provided in an embodiment of the present application;
[0096] FIG12 is a schematic diagram of a second device provided in an embodiment of the present application;
[0097] FIG13 is a schematic diagram of a third device provided in an embodiment of the present application;
[0098] FIG14 is a schematic diagram of a fourth device provided in an embodiment of the present application;
[0099] FIG15 is a schematic diagram of a fifth device provided in an embodiment of the present application;
[0100] FIG16 is a schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0101] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field are within the scope of protection of this application.
[0102] In an embodiment of the present application, in order to improve the quality of user evaluation content of products in the system, AIGC (Artificial Intelligence Generated Content) technology can be applied to scenarios that help users process and rewrite user evaluation content of products. That is, in an embodiment of the present application, if a user needs to publish evaluation content for a certain product, after the user briefly evaluates the product, a more complete and detailed text content can be generated by AIGC while retaining the user's original intention. In this way, the user can publish user evaluation content based on the text content, so that the specific published user evaluation content is more valuable for reference, which also improves the quality of the user evaluation content.
[0103] When using AI models to help users process or rewrite user reviews, in addition to generating text that uses natural semantics, it's also crucial to ensure that the generated text closely reflects the actual product. This means that users may use different wording when writing reviews for different product categories. For example, apparel is often reviewed using terms like "good-looking," "soft material," and "very comfortable," while food products are often rated with "good taste," "fresh," and so on. Therefore, to achieve these goals, AI large-scale parameter models (abbreviated as AI large models) and big data mining technologies can be used. AI large models are pre-trained on large datasets containing vast amounts of data using neural networks. These models typically possess strong model understanding capabilities for text and image information, and are also capable of producing content, including text, images, video, and audio. Due to their powerful model understanding capabilities, these AI large models have found applications in a wide range of fields. Big data mining primarily involves extracting the required information from large datasets. In an embodiment of the present application, it may involve taking the evaluation content about products in the system or even the entire network as a data set, and mining the required information from it to assist in AI generation of user evaluation content, etc.
[0104] Specifically, user reviews can be collected first. Then, the collected user reviews are analyzed and processed by product category to obtain the decision factor labels used by users when making decisions about specific product categories. A decision factor label can specifically be at least one text label that is a user's primary focus when making decisions about products in the same product category, as well as text labels that are frequently used by users when evaluating products in that product category. For example, for products in the clothing category, specific decision factor labels may include "good-looking," "comfortable," and so on. Specifically, these decision factor labels can be determined by analyzing multiple product reviews in the same product category. For example, in one approach, semantic label identification can be performed on the collected multiple user reviews. Then, by clustering the semantic labels, multiple semantic labels within the same product category that meet the user's attention criteria can be determined. These multiple semantic labels can then be saved as the decision factor labels used by users when making decisions about that product category. After obtaining the above decision factor labels, when a user needs to review a target product, a suggested review text expressed in natural language can be generated based on the basic evaluation content entered by the user and the above decision factor labels. In this way, the user can evaluate the specific product based on this AI-generated suggested review text.
[0105] From the perspective of system architecture, the embodiments of the present application can provide the above functions in the commodity information service system. Specifically, referring to Figure 1, the specific system may include a client and a server, wherein the client may exist in the form of an application (App) or a page such as Web, H5, etc., and mainly runs on the user's terminal device, including mobile devices such as mobile phones. The server mainly runs in the server, and may specifically include a cloud server or an independent server, etc. Among them, the client is mainly used for displaying the front-end page and interacting with the user, and the server can provide specific data services. In the embodiments of the present application, it may specifically include analysis of user evaluation content to obtain the correspondence between commodity categories and decision factor labels. In addition, it may also include calling a large AI model to generate text content, etc., and return it to the client for display, etc.
[0106] The specific implementation scheme provided in the embodiments of this application is introduced in detail below.
[0107] Example 1
[0108] First, from the perspective of the aforementioned server, this embodiment 1 provides a method for processing product review content. Referring to FIG2 , the method may include:
[0109] S201: Determine the correspondence between product categories and decision factor labels by analyzing user evaluation content corresponding to multiple products in units of product categories; wherein the decision factor label includes: at least one text label that is the focus of user attention when the user makes a decision on products in the same product category.
[0110] Specifically, in an embodiment of the present application, user reviews of multiple products can be collected. Such user reviews can include text, images, videos, and so on. The collected user reviews can then be analyzed by product category to determine the primary focus of users when making decisions about each product category. These focus points can be expressed using specific text labels, and the text labels corresponding to the user focus points can be called decision factor labels.
[0111] For example, for the clothing category, we can collect user evaluation information for multiple different clothing products. Then, by analyzing and processing these evaluation contents, we can determine the main concerns of users when purchasing clothing products. The text labels corresponding to these concerns, for example, include "good-looking", "comfortable", "good style", etc. These text labels can be used as decision factor labels corresponding to the clothing category. Of course, in the actual implementation, the granularity of the specific product category can be determined according to actual needs. For example, the clothing category can be subdivided into men's clothing, women's clothing, children's clothing, etc., and the decision factor labels can even be determined at the leaf category granularity to ensure that the generated text content has higher quality.
[0112] Specifically, there are multiple approaches to analyzing and processing user reviews. For example, one approach involves performing word segmentation on multiple user reviews collected from the same product category and performing semantic recognition on the resulting text tags. Then, after clustering the multiple text tags based on the semantic recognition results, the distribution of the clustered text tags can be used to identify at least one text tag within the same product category that is a focus of user attention. This tag can then be used as a decision factor when the user makes a decision about the product category.
[0113] Specifically, user evaluation content may include text content or pictures or video content. For the text content, word segmentation can be performed directly, that is, a paragraph or a long sentence can be cut into multiple words or phrases, and some modal particles can be removed, which can be used as text labels. For pictures or videos, the AI algorithm can be used to first understand the model of the image content, and the information understood from the image can be converted into text content for expression. Furthermore, multiple text labels can be obtained by performing word segmentation on the text content. After obtaining the text labels, semantic analysis can be performed on each text label separately, and after clustering the multiple text labels based on the semantic analysis results, at least one text label that is the focus of the user's attention under the same product category can be determined, and determined as the decision factor label when the user makes a decision on the product category.
[0114] For example, after word segmentation processing of multiple user reviews corresponding to a certain product category, the resulting text labels include "good-looking," "beautiful," "too beautiful," and so on. After semantic recognition, these text labels are found to have the same or similar semantics. Therefore, they can be clustered into one category. A unified text label can also be generated for text labels of the same category to express them. For example, the above-mentioned "good-looking," "beautiful," "too beautiful," and so on can be collectively referred to as "good-looking," and so on. After clustering, the number of occurrences of the text labels included in each category in the user reviews can be used to determine whether the text label words in a specific category are the focus of users' attention when making decisions about the product category. For example, if the text labels in the aforementioned categories, such as "good-looking," "comfortable," and "good style," appear the most frequently in the multiple user reviews corresponding to a certain product category, then these text labels are the focus of users' attention when making decisions about the product category and can therefore be used as decision factor labels for the product category, and so on.
[0115] S202: After the target user initiates a request to fill in user evaluation content for the target product object, basic evaluation content input by the target user is obtained.
[0116] After determining the decision factor labels corresponding to the specific product categories in the aforementioned steps, the correspondence between the specific product categories and the decision factor labels can be saved for use when generating recommended evaluation content for users through AI.
[0117] Specifically, the user may first initiate a request to fill in user evaluation content for a target product (for example, an entry for filling in user evaluation content for a specific order or product may be provided in a page such as an order list page). Afterwards, a user evaluation content editing interface may be provided for the user, through which the user may input some basic evaluation content, such as basic evaluation text content, and may also include some pictures, videos (for example, photos and videos taken by the user of the actual product after receiving the actual product; for products such as clothing, pictures of the user's upper body may also be included, etc.) and other image content.
[0118] For example, as shown in Figure 3(A), in this user review content editing interface, several rating items can be provided for users to rate. For example, specific rating items may include comprehensive evaluation, description consistency (i.e., whether the actual product matches the information described in the product details page, or the degree of consistency), merchant service (usually referring to the merchant's delivery speed, etc.), logistics service (usually referring to logistics delivery efficiency, etc.), and other aspects. Users can rate based on their actual experience. For example, multiple score levels can be provided for each rating item, and users can choose which score level to assign to a specific rating item. The higher the score level of the rating item, the higher the score and the higher the user's satisfaction with the rating item, etc. In addition, the above editing interface can also provide users with an input box for entering the content of the review. In addition to rating the above rating items, users can also enter the specific content of the review. For example, in the example shown in Figure 3(B), the user purchased a pair of shoes. The specific content of the review body entered may include: "These sneakers are very comfortable, authentic, and the logistics are fast, but the seller is not very friendly," etc. Furthermore, options such as "Upload Pictures" and "Upload Videos" can be provided, allowing users to upload pictures or videos for review content. Furthermore, within the aforementioned editing interface, users can be provided with an option to provide a "Comprehensive Review." Through this option, users can provide a "good review," "average review," "bad review," and other emotional ratings for specific products. Users can make their selections through this option, and therefore, this emotional information also constitutes the user's basic review content.
[0119] After receiving the above-mentioned basic evaluation content, the AI big model can assist in generating a paragraph of text content based on the basic evaluation content input by the user. In other words, the user can simply score some rating items, and then the AI big model will assist in generating a paragraph of text content to be used as the evaluation text or as a reference. Alternatively, after simply filling in some evaluation text content, the AI big model can assist in rewriting it to make the actual published evaluation text content fuller and more able to bring more reference value to other users, while not causing excessive occupation of user time costs, etc.
[0120] Of course, in a specific implementation, in order to avoid interfering with the process of the user actually inputting the user evaluation content, the above-mentioned process of generating the target text content can be performed after receiving the user's confirmation operation. That is to say, an operation option for asking the user can be provided in the above-mentioned editing interface. For example, as shown at 31 in Figure 3(B), a dialogue content of "Would you like me to rewrite it a little?" can be provided in the editing interface, and options such as "later" and "OK" can be provided. If the user selects "OK", it means that the user confirms that the AI large model needs to help rewrite it, and then the step of generating the target text content can be triggered.
[0121] In the above-mentioned manner, the user can first enter the evaluation content editing interface through the "Evaluation" entry provided for the specific order / product object in the order list page, and after providing some basic evaluation content, initiate a request for AI-assisted generation of evaluation content through the operation options in the evaluation content editing interface. Alternatively, in another manner, in order to further improve efficiency, it is also possible to directly provide an "Evaluation" entry in the order list page and other pages, and directly provide "good review", "bad review" and other operation options for expressing emotional types, or directly provide operation options for scoring rating items. In this way, the user can directly make a simple basic evaluation through the operation option, and then trigger AI-assisted generation of evaluation content without having to enter the evaluation content editing interface to trigger it again, so that the user can more efficiently obtain AI-assisted generated user evaluation content, etc.
[0122] S203: Processing the basic evaluation content according to the decision factor label corresponding to the product category to which the target product belongs to generate suggested evaluation text content, so as to publish user evaluation content for the target product according to the suggested evaluation text content.
[0123] After obtaining the basic evaluation content provided by the user, the basic evaluation content can be processed according to the decision factor label corresponding to the product category to which the target product belongs to generate recommended evaluation text content expressed in natural language.
[0124] That is, in the embodiments of the present application, a target algorithm model (e.g., the aforementioned AI big model, etc., which is mainly introduced in this specification as an example) can be used to generate a suggested evaluation text for the user, wherein the information input into the AI big model can include the basic evaluation content input by the user, and can also include the decision factor label information corresponding to the commodity category to which the current specific target commodity belongs. In this way, the AI big model can generate a piece of suggested evaluation text content coherently expressed in natural language based on the above two aspects of information.
[0125] For example, in the example shown in Figure 3 above, the user purchased a pair of sneakers. The basic evaluation text entered by the user is: "These sneakers are very comfortable, authentic, and the logistics are fast, but the seller doesn't pay much attention to me." In addition, the decision factor labels corresponding to the "sneakers" category include: "comfortable", "light", "good foot feel", etc. After inputting the above two aspects of information into the AI big model, the AI big model can generate a more complete and richer expression. For example, as shown in Figure 4, the recommended evaluation content generated by AI can be:
[0126] "I purchased this pair of sneakers and am very satisfied. First, they are very comfortable, light, and feel great on the feet. Second, I can confirm that they are authentic. The quality and details of the shoes are excellent. Finally, the delivery speed was very fast. I received the goods shortly after placing the order.
[0127] However, the only downside is the seller's poor customer service. They seem to ignore me and are slow to respond to messages. However, this did not affect my shopping experience or my evaluation of the product itself.
[0128] In short, if you want a pair of comfortable, high-quality sneakers and don't mind poor customer service from the seller, then this shoe is a good choice."
[0129] As can be seen, the AI big model can, based on a thorough understanding of the basic user review content, refine the specific text or express it in a richer and more coherent manner. For example, from a relatively brief user-entered basic review, it can analyze the user's overall satisfaction with the product, the reasons for "satisfaction" or "dissatisfaction," and the factors involved. The AI big model's text generation capabilities can then be used to generate text content that is fluently expressed in natural language. The richness of the language in the text can be enhanced based on the decision factor labels corresponding to specific product categories. For example, in the aforementioned example, the user's basic review content can be analyzed to indicate that the user is very satisfied with the product they purchased, citing three reasons: "comfortable," "authentic," and "fast delivery." Therefore, when generating the recommended review text, the AI can express it in these three aspects, inserting content related to decision factor labels for the "sneakers" category, such as "lightweight" and "good feel." Furthermore, the AI-generated text can also address user comments in the basic review, such as "but the seller doesn't really talk to me," through the use of transitional speech. Finally, you can also provide content of a summary or suggestive conclusion nature to make the expression more complete and more valuable for reference to other users, etc.
[0130] It should be noted that in the above-mentioned process of rewriting the review text based on the basic review content input by the user, the specific basic review content input by the user may include not only the basic review text content but also basic image / video content. In this case, the basic image / video content can also be processed by model understanding to generate textual representations of the basic image / video content. Then, the basic review text content and the textual representations generated based on the basic image / video content can be processed according to the decision factor labels corresponding to the product category to which the target product belongs to generate the recommended review text content expressed in natural language. In other words, relevant textual content can also be identified or mined from user-uploaded images, videos, and other content. For example, for clothing products, a user uploads a picture of the upper body effect of the product they tried on. Although the basic review text content does not mention whether it "fits" or not, if this is reflected in the picture, the AI-generated text content can also include descriptions related to "fit," and so on.
[0131] Furthermore, in specific implementations, personalized customization of tone or writing style can be provided. That is, for the same target product and the same basic review content, different suggested review texts corresponding to different tones and writing styles can be generated. Specifically, users can be provided with optional tone and writing style options, allowing them to select a specific tone, writing style, etc. Alternatively, users can be provided with a dialogue portal with the AI, allowing them to express their needs to the AI through a conversation. For example, after displaying the AI-generated suggested review text, the user can enter it through a keyboard or voice input, saying, in a more coquettish tone, "These sneakers are very comfortable, authentic, and the delivery is fast, but the seller is not very helpful." The AI model can then process the basic review content based on the decision factor labels corresponding to the product category to which the target product belongs, as well as the target tone or writing style information, to generate suggested review content expressed in natural language and with the target tone or writing style. That is, the information input into the AI model, in addition to the aforementioned basic evaluation content and decision factor labels, can also include information such as tone and writing style. For example, as shown in Figure 5, for the above example, the text content of "more coquettish tone" generated by AI can be:
[0132] "Wow! This pair of sneakers is really super super! What I bought is authentic, and the logistics speed is also very fast. I was a little surprised when I received the goods! But the only regret is that the seller seems to be indifferent."
[0133] In short, this method allows users to personalize the tone, writing style, etc. of the evaluation content according to their own preferences.
[0134] After the above-mentioned suggested evaluation text content is generated through the AI big model, it can be displayed through the aforementioned evaluation content editing interface. In addition, the user can be provided with options such as "Refresh" and "Use It". If the user is not satisfied with the currently generated suggested evaluation text content, he can choose "Refresh". At this time, the AI model can regenerate the text content. If you choose "Use It", the text content generated by the AI can be displayed in the input box for entering the evaluation text content, and the evaluation text content previously entered by the user can be replaced, so that the evaluation text content can be determined based on the AI-generated text content and published. Of course, users can also use the input box to perform some editing and other processing on the text content generated by the AI before submitting it for publication.
[0135] In addition, in a specific implementation, the embodiments of the present application can not only help users rewrite the text evaluation content they input, but also help users optimize uploaded pictures, videos, and other content. That is, in a specific implementation, after a user uploads picture / video content for evaluating a target product, if a request is received to optimize the basic picture / video content, the basic picture / video content can be optimized based on the product category to which the target product belongs, so that the optimized picture / video content can be published.
[0136] In other words, when optimizing image / video content, it can also be done based on specific product category information. Different product category information can correspond to different optimization rules. Specifically, the optimization of images / videos can include the following aspects:
[0137] 1. Rearrange multiple images so that images that better reflect the characteristics of specific product categories are displayed first. For example, for clothing products, images of upper body effects of people can be prioritized, etc.
[0138] 2. Identify the subject of the image / video content, identify the subject type, and process the identified subject based on the specific product category information. For example, identified product subjects can be highlighted, while identified human subjects can be processed differently for different product categories. For example, in the case of clothing products, the characters in images or videos are usually used to show how the clothing looks on the body, so the human subjects can be highlighted or enlarged. However, in the case of digital products, the human subjects appearing in images or videos are generally not helpful in displaying the product characteristics, so they can be blurred, etc. The subject processing method corresponding to the specific product category can be input into the AI model using rule information, so after identifying the subject and type in the image or video, it can be processed according to the processing rule information corresponding to the specific product category.
[0139] 3. If the video content uploaded by the user includes voice content, the voice content can be converted into text content through voice recognition, and a subtitle file for the video content can be generated based on the text content to add subtitles to the basic video content. In multilingual scenarios, the text content obtained by specific voice recognition can also be translated into multiple languages, and subtitle files in corresponding languages can be generated for each language to provide subtitle information in the corresponding language for users with different language needs in multilingual scenarios.
[0140] 4. In some cases, users may be willing to upload pictures of themselves trying on clothes, etc., but may not want other users to see their faces. In the prior art, users usually need to use some props to cover part of their faces during the photo shooting process, or blur part of their faces after taking the photos, etc. However, this situation will affect the integrity of the picture display effect. In response to this situation, in an embodiment of the present application, if the picture / video content uploaded by the user includes a facial image, an AI facial image can also be generated by an AI algorithm to replace the original facial image in the picture / video content uploaded by the user. In other words, the real facial image can be replaced with the AI facial image to protect the user's privacy.
[0141] 5. In addition, the pictures or videos uploaded by users may also include some private information such as phone numbers (for example, if the shooting background happens to include phone number information, it may be captured in the picture or video), or the shooting background may be messy, which may affect the presentation of the product, etc. Therefore, when performing the optimization process, the private information and background contained in the pictures / videos uploaded by users can also be removed or blurred. Furthermore, the videos uploaded by users may also include some meaningless content. Therefore, when performing the optimization process, the meaningless segments in the video content can also be identified and cropped, that is, such meaningless segments can be removed to improve the video quality and avoid the occupation of system resources by meaningless video content.
[0142] The above introduces a variety of optimization processing methods for image / video content. In actual applications, other optimization processing methods can also be included, which are not listed here one by one. It should be noted that in the specific implementation, the specific processing of the pictures / videos uploaded by users can be determined by the AI algorithm. Alternatively, in another way, a variety of optional optimization processing solutions can be provided to users for selection based on actual conditions. Accordingly, the AI algorithm can perform optimization processing according to the optimization processing solution selected by the user. For example, some users may not need to replace facial images, so they do not need to select this option when selecting the optimization processing solution, and so on.
[0143] In summary, through the embodiments of the present application, the user evaluation content of the product can be analyzed in advance based on the product category to determine the decision factor label corresponding to the specific product category. This decision factor label can be a text label that is the focus of the user's attention when the user makes a decision on the product of the specific product category. In this way, in the case where the user needs to submit user evaluation content for a target product, after the user submits the basic evaluation content, the recommended evaluation text content can be generated based on this basic user evaluation content and the decision factor label corresponding to the specific product category. In this way, since the basic evaluation content output by the user can be rewritten, the decision factor label corresponding to the specific product category can be fully utilized during the rewriting process. This decision factor label is determined by information mining from the user evaluation content corresponding to the specific product category. Therefore, the recommended evaluation text content generated by the algorithm can not only be more complete and detailed, but also express more text labels related to the focus of the user's decision, so that the user evaluation content has more reference value in helping other users make decisions, that is, to improve the quality of the user evaluation content.
[0144] Example 2
[0145] The second embodiment corresponds to the first embodiment and provides a method for processing product review content from the perspective of the client. Referring to FIG6 , the method may include:
[0146] S601: After a target user initiates a request to fill in user evaluation content for a target product object, basic evaluation content input by the target user is obtained;
[0147] S602: Submitting the basic evaluation content to the server, so that the server processes the basic evaluation content based on the decision factor tag corresponding to the product category to which the target product belongs, and generates recommended evaluation text content; wherein the decision factor tag is determined by analyzing user evaluation content of multiple products associated with the product category, and the decision factor tag includes: at least one text tag that is the focus of the user's attention when making a decision about products in the same product category;
[0148] S603: Displaying the suggested evaluation text content so as to publish user evaluation content for the target product based on the suggested evaluation text content.
[0149] Example 3
[0150] In the above embodiments 1 and 2, AI models and big data mining technologies are used to help users generate more valuable user reviews when they need to submit user reviews for a target product. In this embodiment 3, the aforementioned AI models and big data mining technologies can also be applied to help users quickly understand the user reviews of a product.
[0151] Specifically, in actual applications, the user review content associated with a product may include many items, especially for some popular products or products with relatively high sales, the number of user review content may be even greater. At this time, the user may need to spend a long time to view them. Therefore, the user may usually exit or close the page after viewing a few items. This may lead to a one-sided impression of the product due to the lack of viewing the complete review content, or some review content that is truly useful to the current user may not be seen by the user because it is ranked relatively low, resulting in failure to play a real role in helping the user make shopping decisions, etc.
[0152] To address this situation, in an embodiment of the present application, a model understanding can be performed on multiple user evaluation contents associated with the target product based on the aforementioned AI big model. Based on the model understanding results, the multiple target evaluation contents can be summarized and summarized. The AI big model produces target text content for summarizing and expressing the multiple user evaluation contents. In addition, in the process of summarizing and summarizing the multiple user evaluation contents, the text content can be generated based on the decision factor label information corresponding to the aforementioned specific product category, so that the generated text content can more accurately reflect the text labels related to the focus of the user's decision. For example, it is possible to identify and summarize the user's emotional expression information such as positive, negative, or neutral evaluation of the product, or it is possible to identify and summarize the characteristics, advantages, disadvantages, etc. of the product mentioned in the user's review; or it is possible to perform commonality or similarity analysis on multiple user evaluation contents to classify and summarize them. Furthermore, it is possible to summarize and summarize the matching relationship with the user's preferences based on the current target user's preferences or some basic attributes (including the group to which they belong), and so on.
[0153] Furthermore, when generating the above text content, the user's characteristics can also be taken into account. By analyzing user reviews posted by publishers with the same or similar characteristics as the current user, the generated text content can include content related to purchase recommendations. For example, if a user has a pear-shaped body, and other users who have purchased the current product with similar shapes have commented that it suits their body type, this information can be highlighted in the generated text content.
[0154] Furthermore, when collecting user reviews for the same product, in addition to collecting user reviews from the current product information service system, we can also collect user reviews from other product information service systems. We can also include user reviews related to the product posted on social media systems, and so on. By integrating the user review information collected from these multiple systems, we can produce a more complete overview of the user reviews for the product, helping users to more comprehensively understand the user reviews of the product and make more accurate shopping decisions.
[0155] In summary, in Example 3 of the present application, one or more paragraphs of summarizing text content can be generated based on the user evaluation content of the same product. The specific text content can be a coherent expression that conforms to natural language grammar to facilitate user understanding. Furthermore, by providing this target text content to the user, the user can quickly browse the user evaluation content through this target text content. Moreover, due to its summarizing nature, after seeing this target text content, it is equivalent to having read all the user evaluation content and drawn a conclusion. In this way, it can help users improve their browsing efficiency of the user evaluation content and have a more complete and comprehensive understanding of the user evaluation content, thereby better helping users make purchasing decisions.
[0156] Specifically, referring to FIG7 , the third embodiment of the present application provides a method for providing product information from the perspective of a server. The method may include:
[0157] S701: Taking a product as a unit, collect and analyze multiple user evaluation contents corresponding to the product to determine the decision factor label corresponding to the product; wherein the decision factor label includes: at least one text label that is the focus of the user's attention when the user makes a decision on the same product.
[0158] Similar to Example 1, when determining the decision factor label for a specific product, multiple user reviews corresponding to the same product can be segmented and semantic recognition can be performed on the resulting multiple text labels. Then, the multiple text labels can be clustered according to the semantic recognition results to determine at least one text label that is the focus of the user's review and determine it as the decision factor label for the user's decision making regarding the product. For example, for a certain clothing product, if most users use labels such as "good fit" and "comfortable" when evaluating the product, this indicates that most users purchased the product for these reasons. Therefore, these labels can serve as the decision factor label for the product.
[0159] Among them, when collecting user evaluation content, the specific collection scope may include not only the current product information service system, but also other product information service systems, and even some social media systems, etc. In this way, through cross-platform information collection, richer user evaluation content about the same product can be obtained. It should be noted here that in order to mark the same product, the specific product information system can usually add identifications such as product IDs to the products, and the same ID corresponds to the same product, thereby realizing the identification of the same product. However, since the marking methods of products may be different between different systems, it is impossible to determine whether the products in different systems are the same product by using product IDs. In this case, it is possible to determine whether they correspond to the same product by analyzing the image and / or text description information of the corresponding product, and then analyze and process the user evaluation content about the same product collected in different systems.
[0160] It should be noted that, in the specific implementation, in the process of determining the decision factor label for a specific product, the corresponding emotional label can also be determined, so that the user evaluation content of the product can be more accurately summarized based on the emotional label. Among them, the so-called emotional labels can include positive, negative or neutral, etc. In the specific implementation, on the basis of performing word segmentation processing on the user evaluation content, obtaining multiple text labels and performing semantic analysis, the specific text labels can be scored with emotional scores through a specific algorithm (which can be specifically expressed through emotional vectors, etc.), and then according to a pre-set threshold, it is determined whether it is a positive, negative or neutral evaluation, etc. In this way, it is convenient to count the number and distribution of user evaluation content with various emotional labels corresponding to the same product. For example, for a certain product, it can be determined whether the product has received more positive or negative evaluations, etc.
[0161] S702: In the process of providing information about a target product to a target user, a user evaluation summary is generated for the target product based on the decision factor label corresponding to the target product and the target user's information, so as to provide the product evaluation summary to the target user.
[0162] After obtaining the decision factor labels for a product, they can be used to generate a user review summary for that product. Of course, in a specific implementation, when generating a user review summary for the same product, not only the decision factor labels mentioned above can be considered, but also the user's personal information. That is, in the embodiments of this application, the user review summary for the same product can be personalized when providing information about a target product to a specific user. That is, when different users browse the same product, the specific user review summary they see can be different.
[0163] There are various situations in which information about the target product can be provided to the target user, and accordingly, the user evaluation summary content can be provided on various pages. For example, when the system recommends information related to the target product to the target user, a user evaluation summary content about the specific product can be provided on the product recommendation information page. Alternatively, when the user initiates a search and the search results page needs to include information about the target product, a user evaluation summary content about the specific product can be provided on the search results page. Alternatively, when the user initiates a browse request for a product details page, a user evaluation summary content about the product can be provided on the product details page, and so on.
[0164] Specifically, in order to better achieve the personalized generation of the above-mentioned user evaluation summary content, it is also possible to analyze the product evaluation content published for the target product by multiple publishers with the same / similar characteristics as the target user to determine the decision factor labels when the users with the same / similar characteristics make decisions on the product; then, based on the decision factor labels corresponding to the target product and the decision factor labels when the users with the same / similar characteristics make decisions on the product, a user evaluation summary content expressed in natural language can be generated for the target product. At this time, the specific user evaluation summary content can include purchase recommendation information provided for the target user. Among them, the characteristic information about the user can include the user's body characteristics, preference characteristics, etc., which can be obtained through relevant data provided by the user, or from the user's historical behavior records, etc.
[0165] For example, referring to FIG8 , regarding a certain clothing product, the generated user review summary may include:
[0166] “This product’s size S is perfect for your body type (12). 50 people with similar body types have bought it and all found it to be a good fit. It is made of pure cotton and will not hurt the skin (37). Your favorite celebrity has also worn similar style clothing (9). However, the logistics of this product may be a bit slow. It may take 5 to 8 days to be delivered to your address (67).”
[0167] It can be seen that the above user evaluation summary content is mainly personalized around the information of the current target user. During the generation process, the user's body characteristics, whether there are favorite celebrities and other preference information, common delivery addresses and other information are taken into account, so that shopping suggestions can be given in terms of size, etc., and reference information on logistics can also be given, etc. Of course, in specific implementation, the characteristic information about the user can also include multiple types, such as whether there are some brand preferences, etc. Among them, when generating the above user evaluation summary content, for some conclusion-style content, the specific corresponding number of user evaluations can also be marked. For example, the number marked after each sentence in the above example represents the number of user evaluation content that has expressed the same sentence, and so on.
[0168] In specific implementation, in order to meet the display requirements in a multilingual scenario, the above user evaluation summary content can also be translated into the language required by the target user, so as to provide the user with the user evaluation summary content based on the language, and so on.
[0169] In short, by determining the decision factor labels of users in the process of making decisions on specific products at the product granularity, and combining them with the user's personal characteristic information, we can summarize the user evaluation content of the same product and generate personalized user evaluation overview content for specific users, thereby helping users make shopping decisions more efficiently.
[0170] Example 4
[0171] This fourth embodiment corresponds to the third embodiment and provides a method for providing product information from the perspective of a client. Referring to FIG9 , the method may include:
[0172] S901: In the process of providing information related to a target product to a target user, a user evaluation summary content about the target product is obtained. The user evaluation summary content is generated based on a decision factor label corresponding to the target product and information about the target user. The decision factor label is determined by collecting and analyzing multiple user evaluation contents corresponding to the target product. The decision factor label includes: at least one text label that is the focus of the user's attention during the user's decision-making process for the same product.
[0173] The specific user evaluation summary content can be generated by the server and provided to the client.
[0174] S902: Output the user evaluation summary content.
[0175] Specifically, the client can output the user review summary in a variety of ways. For example, if the user review summary is text, it can be displayed on a specific page. Alternatively, to meet the needs of people with visual impairments or reading disabilities, the server can generate the user review summary in audio format, and the client can play the audio version of the user review summary, etc.
[0176] Example 5
[0177] In the aforementioned first embodiment, the optimization processing of user-uploaded content such as pictures and videos is mentioned during the process of user-submitted user evaluation content. Therefore, this fifth embodiment also provides a method for processing product evaluation content. Referring to FIG10 , the method may include:
[0178] S1001: After a target user initiates a request to fill in user evaluation content for a target product object, in response to a request for assisting in evaluation content processing, basic evaluation content input by the target user is obtained, the basic evaluation content including basic image / video content;
[0179] S1002: Optimizing the basic image / video content according to the category to which the target product belongs, so as to publish the optimized image / video content.
[0180] Specifically, the optimization processing of the basic picture / video content may include multiple aspects. For example, if the basic picture / video content includes multiple pictures, the order of the multiple pictures is adjusted by performing model understanding on the multiple pictures and according to the product category to which the target product belongs.
[0181] Alternatively, the subject of the basic image / video content can be identified, and the subject type information of the identified subject can be determined. The identified subject can be optimized based on the product category to which the target product belongs and the optimization processing rule information of each subject type corresponding to the product category.
[0182] Alternatively, if the basic video content includes voice content, the voice content can be converted into text content through voice recognition, and a subtitle file for the video content can be generated based on the text content to add subtitles to the basic video content. In specific implementations, the text content can also be translated into multiple languages, and subtitle files in corresponding languages can be generated for each language to provide subtitle information in corresponding languages to users with different language requirements in multilingual scenarios.
[0183] Alternatively, if the basic picture / video content includes a facial image, an AI facial image is generated through an artificial intelligence (AI) algorithm to replace and display the original facial image in the basic picture / video content.
[0184] Alternatively, the privacy information and background contained in the basic image / video content are removed or blurred, and / or meaningless segments in the video content are cropped, etc.
[0185] For the parts not described in detail in Examples 2 to 5, please refer to the description in Example 1 and other parts of this specification, and no further details will be given here.
[0186] It should be noted that the embodiments of the present application may involve the use of user data. In actual applications, user-specific personal data can be used in the scheme described herein within the scope permitted by applicable laws and regulations, subject to the requirements of applicable laws and regulations of the country where the user is located (for example, with the user's explicit consent, effective notification to the user, etc.).
[0187] Corresponding to the first embodiment, the embodiment of the present application further provides a product review content processing device, as shown in FIG11 , which may include:
[0188] The pre-processing unit 1101 is configured to analyze user reviews of multiple products in a product category-by-product basis to determine a correspondence between product categories and decision factor labels. The decision factor label includes at least one text label that is a focus of user attention when making a decision about products in the same product category.
[0189] The basic evaluation content acquisition unit 1102 is used to acquire the basic evaluation content input by the target user after the target user initiates a request to fill in the user evaluation content for the target product object;
[0190] The suggested evaluation content generating unit 1103 is used to process the basic evaluation content according to the decision factor label corresponding to the commodity category to which the target commodity belongs, and generate suggested evaluation text content, so as to publish user evaluation content for the target commodity according to the suggested evaluation text content.
[0191] The specific pre-processing unit can be used for:
[0192] Perform word segmentation on multiple user reviews collected under the same product category and perform semantic recognition on the resulting text tags;
[0193] After clustering multiple text labels according to the semantic recognition results, at least one text label that is the focus of user attention under the same product category is determined and determined as the decision factor label when the user makes a decision on the product category.
[0194] In addition, the device may further include:
[0195] a target tone or style information determination unit, configured to determine the target tone or style information selected by the target user;
[0196] The suggestion evaluation content generating unit can be specifically used to:
[0197] The basic evaluation content is processed according to the decision factor label corresponding to the product category to which the target product belongs and the target tone or style information to generate recommended evaluation content with the target tone or style.
[0198] The basic evaluation content includes basic evaluation text content and basic picture / video content;
[0199] The suggestion evaluation content generating unit can be specifically used to:
[0200] Generating textual expressions of the basic image / video content by performing model understanding processing on the basic image / video content;
[0201] According to the decision factor label corresponding to the commodity category to which the target commodity belongs, the basic evaluation text content and the text expression content generated based on the basic image / video content are processed to generate the recommended evaluation text content.
[0202] In addition, if the basic evaluation content includes basic evaluation text content and basic picture / video content, the device may further include:
[0203] The picture / video optimization processing unit is used to respond to the request for optimizing the basic picture / video content and optimize the basic picture / video content according to the product category to which the target product belongs, so as to publish the optimized picture / video content.
[0204] Corresponding to the second embodiment, the embodiment of the present application further provides a product review content processing device, as shown in FIG12 , which may include:
[0205] The basic evaluation content acquisition unit 1201 is used to acquire the basic evaluation content input by the target user after the target user initiates a request to fill in the user evaluation content for the target product object;
[0206] The basic evaluation content submission unit 1202 is configured to submit the basic evaluation content to the server, so that the server processes the basic evaluation content according to the decision factor tag corresponding to the product category to which the target product belongs, and generates recommended evaluation text content. The decision factor tag is determined by analyzing user evaluation content of multiple products associated with the product category, and the decision factor tag includes at least one text tag that is the focus of the user's attention when making a decision about products in the same product category.
[0207] The suggested evaluation text content display unit 1203 is used to display the suggested evaluation text content so as to publish user evaluation content for the target product based on the suggested evaluation text content.
[0208] Corresponding to the third embodiment, the embodiment of the present application further provides a device for providing product information. Referring to FIG13 , the device may include:
[0209] The pre-processing unit 1301 is configured to collect and analyze multiple user reviews of a product, and determine a decision factor label corresponding to the product. The decision factor label includes at least one text label that is a focus of the user's attention when making a decision about the product.
[0210] The product evaluation summary content generation unit 1302 is used to generate user evaluation summary content for the target product based on the decision factor label corresponding to the target product and the information of the target user in the process of providing information about the target product to the target user, so as to provide the product evaluation summary content to the target user.
[0211] The pre-processing unit can be used to:
[0212] Perform word segmentation on multiple user reviews corresponding to the same product, and perform semantic recognition on the resulting multiple text tags;
[0213] After clustering the multiple text labels according to the semantic recognition result, at least one text label belonging to the user evaluation focus is determined and determined as the decision factor label when the user makes a decision on the product.
[0214] When collecting user evaluation content, the scope of collection includes the current product information service system, other product information service systems and / or social media systems;
[0215] At this time, the device may further include:
[0216] The same product identification unit is used to determine whether the product evaluation content collected from different systems corresponds to the same product by analyzing the image and / or text description information of the corresponding product.
[0217] In addition, the device may further include:
[0218] The emotion tag determination unit is used to determine the emotion tag corresponding to the decision factor tag, so as to generate the product evaluation summary content in combination with the emotion tag.
[0219] Furthermore, the device may further include:
[0220] a label determination unit, configured to analyze product reviews published by multiple publishers with the same or similar characteristics as the target user for the target product, and determine a decision factor label when the users with the same or similar characteristics make a decision on the product;
[0221] The product review summary content generation unit can be specifically used to:
[0222] Based on the decision factor label corresponding to the target product and the decision factor label when the user with the same / similar characteristics makes a decision on the product, a user evaluation summary content is generated for the target product, and the user evaluation summary content includes purchase recommendation information provided for the target user.
[0223] Corresponding to the fourth embodiment, the embodiment of the present application further provides a device for providing product information. Referring to FIG14 , the device may include:
[0224] The user evaluation summary content acquisition unit 1401 is configured to acquire user evaluation summary content regarding a target product in the process of providing information related to the target product to the target user. The user evaluation summary content is generated based on a decision factor label corresponding to the target product and information about the target user. The decision factor label is determined by collecting and analyzing multiple user evaluation contents corresponding to the target product. The decision factor label includes at least one text label that is a focus of the user's attention during the user's decision-making process regarding the same product.
[0225] The user evaluation summary content output unit 1402 is configured to output the user evaluation summary content.
[0226] Corresponding to the fifth embodiment, the embodiment of the present application further provides a product review content processing device, as shown in FIG15 , which may include:
[0227] The basic evaluation content acquisition unit 1501 is configured to, after a target user initiates a request to fill in a user evaluation content for a target product object, respond to a request for assisting in evaluation content processing and acquire the basic evaluation content input by the target user, the basic evaluation content including basic image / video content;
[0228] The optimization processing unit 1502 is used to optimize the basic image / video content according to the category to which the target product belongs, so as to publish the optimized image / video content.
[0229] Specifically, the optimization processing unit can be used to:
[0230] If the basic picture / video content includes multiple pictures, the order of the multiple pictures is adjusted by performing model understanding on the multiple pictures and according to the product category to which the target product belongs.
[0231] Alternatively, the subject is identified on the basic image / video content, and the subject type information of the identified subject is determined, and the identified subject is optimized according to the product category to which the target product belongs and the optimization processing rule information of each subject type corresponding to the product category.
[0232] Alternatively, if the basic video content includes voice content, the voice content is converted into text content by performing voice recognition, and a subtitle file of the video content is generated according to the text content, so as to add subtitles to the basic video content.
[0233] The text content may be translated into multiple languages, and subtitle files in the corresponding languages may be generated respectively, so as to provide subtitle information in the corresponding languages for users with different language requirements in a multilingual scenario.
[0234] Alternatively, if the basic picture / video content includes a facial image, an AI facial image is generated through an artificial intelligence (AI) algorithm to replace and display the original facial image in the basic picture / video content.
[0235] Alternatively, the privacy information and background contained in the basic image / video content are removed or blurred, and / or meaningless segments in the video content are cropped.
[0236] In addition, an embodiment of the present application further provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps of any one of the methods in the aforementioned method embodiments are implemented.
[0237] And an electronic device comprising:
[0238] one or more processors; and
[0239] A memory associated with the one or more processors, the memory being used to store program instructions, wherein the program instructions, when read and executed by the one or more processors, execute the steps of the method described in any one of the aforementioned method embodiments.
[0240] 16 exemplarily illustrates the architecture of an electronic device, which may include a processor 1610, a video display adapter 1611, a disk drive 1612, an input / output interface 1613, a network interface 1614, and a memory 1620. The processor 1610, the video display adapter 1611, the disk drive 1612, the input / output interface 1613, the network interface 1614, and the memory 1620 may be communicatively connected via a communication bus 1630.
[0241] Among them, the processor 1610 can be implemented by a general-purpose CPU (Central Processing Unit, processor), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, etc., to execute relevant programs to implement the technical solutions provided in this application.
[0242] The memory 1620 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1620 can store an operating system 1621 for controlling the operation of the electronic device 1600, and a basic input and output system (BIOS) for controlling the low-level operations of the electronic device 1600. In addition, a web browser 1623, a data storage management system 1624, and a product evaluation content processing system 1625, etc. can also be stored. The above-mentioned product evaluation content processing system 1625 can be an application program that specifically implements the operations of the aforementioned steps in the embodiment of the present application. In short, when the technical solution provided by the present application is implemented by software or firmware, the relevant program code is stored in the memory 1620 and is called and executed by the processor 1610.
[0243] The input / output interface 1613 is used to connect input / output modules to implement information input and output. The input / output modules can be configured as components within the device (not shown in the figure) or can be externally connected to the device to provide corresponding functions. Input devices may include a keyboard, mouse, touch screen, microphone, various sensors, etc., and output devices may include a display, speaker, vibrator, indicator light, etc.
[0244] The network interface 1614 is used to connect to a communication module (not shown) to enable communication between the device and other devices. The communication module can communicate via a wired method (such as USB, network cable, etc.) or a wireless method (such as mobile network, WiFi, Bluetooth, etc.).
[0245] The bus 1630 comprises a pathway for transmitting information between the various components of the device (eg, the processor 1610 , the video display adapter 1611 , the disk drive 1612 , the input / output interface 1613 , the network interface 1614 , and the memory 1620 ).
[0246] It should be noted that although the above device only shows a processor 1610, a video display adapter 1611, a disk drive 1612, an input / output interface 1613, a network interface 1614, a memory 1620, a bus 1630, etc., in a specific implementation, the device may also include other components necessary for normal operation. In addition, it will be understood by those skilled in the art that the above device may also include only the components necessary to implement the solution of the present application, and does not necessarily include all the components shown in the figure.
[0247] Through the description of the above embodiments, it can be seen that those skilled in the art can clearly understand that the present application can be implemented by means of software plus a necessary general hardware platform. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which can be stored in a storage medium such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in various embodiments of the present application or certain parts of the embodiments.
[0248] Each embodiment in this specification is described in a progressive manner. The same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments. In particular, for system or system embodiments, since they are basically similar to method embodiments, the description is relatively simple. For relevant parts, refer to the partial description of the method embodiment. The system and system embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without expending creative work.
[0249] The above is a detailed introduction to the product review content processing method and electronic device provided by this application. This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core ideas of this application. At the same time, for those skilled in the art, based on the ideas of this application, there may be changes in the specific implementation methods and application scope. In summary, the contents of this specification should not be understood as limiting this application.
Claims
1. A method for processing product evaluation content, characterized in that: include: By analyzing the user evaluation contents corresponding to multiple commodities in commodity categories, the corresponding relationship between commodity categories and decision factor labels is determined; wherein the decision factor labels include: at least one text label that is the focus of user attention during the process of user decision-making on commodities in the same commodity category; After the target user initiates a request to fill in the user evaluation content for the target commodity object, the basic evaluation content input by the target user is obtained; The basic evaluation content is processed according to the decision factor label corresponding to the commodity category to which the target commodity belongs to generate suggested evaluation text content, so as to publish user evaluation content for the target commodity according to the suggested evaluation text content.
2. The method according to claim 1, characterized in that The method of analyzing the user evaluation contents corresponding to the multiple commodities to determine the corresponding relationship between the commodity categories and the decision factor labels includes: Perform word segmentation on multiple user reviews collected under the same product category, and perform semantic recognition on the obtained text tags; After clustering multiple text labels according to the semantic recognition result, at least one text label that is the focus of user attention under the same product category is determined, and is determined as a decision factor label when the user makes a decision on the product category.
3. The method according to claim 1, characterized in that Also includes: Determining the target tone or style information selected by the target user; The processing of the basic evaluation content according to the decision factor label corresponding to the commodity category to which the target commodity belongs includes: The basic evaluation content is processed according to the decision factor label corresponding to the commodity category to which the target commodity belongs, and the target tone or style information, so as to generate recommended evaluation content with the target tone or style.
4. The method according to claim 1, characterized in that: The basic evaluation content includes basic evaluation text content and basic picture / video content; The processing of the basic evaluation content according to the decision factor label corresponding to the commodity category to which the target commodity belongs includes: Generating textual expression content of the basic picture / video content by performing model understanding processing on the basic picture / video content; According to the decision factor label corresponding to the commodity category to which the target commodity belongs, the basic evaluation text content and the text expression content generated based on the basic image / video content are processed to generate the recommended evaluation text content.
5. The method according to claim 1, characterized in that The basic evaluation content includes basic evaluation text content and basic picture / video content; The method further comprises: In response to a request to optimize the basic image / video content, the basic image / video content is optimized according to the commodity category to which the target commodity belongs, so as to publish the optimized image / video content.
6. A method for processing product evaluation content, characterized in that: include: After the target user initiates a request to fill in the user evaluation content for the target commodity object, the basic evaluation content input by the target user is obtained; Submitting the basic evaluation content to the server so that the server processes the basic evaluation content according to the decision factor label corresponding to the commodity category to which the target commodity belongs, and generates a suggested evaluation text content; wherein the decision factor label is determined by analyzing the user evaluation content of multiple commodities associated with the commodity category, and the decision factor label includes: at least one text label that is the focus of the user's attention during the user's decision-making process for commodities in the same commodity category; The suggested evaluation text content is displayed so as to publish user evaluation content for the target product according to the suggested evaluation text content.
7. A method for providing product information, characterized in that: include: Taking a product as a unit, by collecting and analyzing multiple user evaluation contents corresponding to the product, the decision factor label corresponding to the product is determined; wherein the decision factor label includes: at least one text label that is the focus of the user's attention during the user's decision-making process for the same product; In the process of providing target users with information about target products, user evaluation summary content is generated for the target products according to the decision factor labels corresponding to the target products and the information of the target users, so as to provide the product evaluation summary content to the target users.
8. The method according to claim 7, characterized in that When collecting user evaluation content, the scope of collection includes the current product information service system, other product information service systems and / or social media systems; The method further comprises: For the product evaluation contents collected from different systems, it is determined whether they correspond to the same product by analyzing the image and / or text description information of the corresponding product.
9. The method according to claim 7, characterized in that: Also includes: By analyzing the product evaluation contents published for the target product by multiple publishers having the same / similar characteristics as the target user, determining the decision factor labels when the users having the same / similar characteristics make decisions on the product; Generating user evaluation summary content for the target product according to the decision factor label corresponding to the target product and the information of the target user includes: Based on the decision factor label corresponding to the target product and the decision factor label when the user with the same / similar characteristics makes a decision on the product, a user evaluation summary content is generated for the target product, and the user evaluation summary content includes purchase recommendation information provided for the target user.
10. A method for providing product information, characterized in that: include: In the process of providing target users with information related to target products, obtaining user evaluation summary content about the target products, wherein the user evaluation summary content is generated according to the decision factor label corresponding to the target products and the information of the target users, and the decision factor label is determined by collecting and analyzing multiple user evaluation contents corresponding to the target products, and the decision factor label includes: at least one text label that is the focus of user attention in the process of the user making a decision on the same product; Output the user evaluation summary content.
11. A method for processing product evaluation content, characterized in that: include: After the target user initiates a request to fill in the user evaluation content for the target commodity object, in response to the request for assisting in the evaluation content processing, the basic evaluation content input by the target user is obtained, the basic evaluation content including basic picture / video content; According to the category to which the target product belongs, the basic picture / video content is optimized so as to publish the optimized picture / video content.
12. The method according to claim 11, characterized in that The optimizing process of the basic picture / video content includes: If the basic picture / video content includes multiple pictures, the order of the multiple pictures is adjusted by performing model understanding on the multiple pictures and according to the commodity category to which the target commodity belongs; and / or, Performing subject recognition on the basic image / video content, and determining subject type information of the recognized subject, and optimizing the recognized subject according to the commodity category to which the target commodity belongs and optimization processing rule information of each subject type corresponding to the commodity category; and / or, If the basic video content includes voice content, converting the voice content into text content by performing voice recognition on the voice content, and generating a subtitle file of the video content according to the text content, so as to add subtitles to the basic video content; and / or, Translating the text content into multiple languages and generating subtitle files in corresponding languages respectively, so as to provide subtitle information in corresponding languages for users with different language requirements in a multi-language scenario; and / or, If the basic picture / video content includes a face image, an AI face image is generated by an artificial intelligence (AI) algorithm to replace the original face image in the basic picture / video content; and / or, The privacy information and background contained in the basic image / video content are removed or blurred, and / or the meaningless segments in the video content are cropped.
13. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method described in any one of claims 1 to 12 are implemented.
14. An electronic device, characterized in that: include: one or more processors; as well as A memory associated with the one or more processors, the memory being used to store program instructions, wherein the program instructions, when read and executed by the one or more processors, execute the steps of the method described in any one of claims 1 to 12.
Citation Information
Patent Citations
Procurement user evaluation method and system for purchaser non-beneficiary industries
CN109101553A
Data generation method and device, medium and electronic equipment
CN111783445A
Evaluation text pushing method and device
CN112818082A
Comment generation method and device, electronic equipment and storage medium
CN114970494A
Data generation, display and processing method and device
CN115018564A
