Comment display method and device, equipment, storage medium and program product

Abstract

The invention provides a comment display method and device, equipment, a storage medium and a program product. Relates to the technical field of computers. The method comprises the steps that in response to a comment screening operation, to-be-displayed comment information and user behavior data of the comment information are obtained, and the comment information comprises a comment score and a comment text; based on the evaluation score and the user behavior data, emotion analysis is carried out on the comment text to obtain an emotion evaluation result of the comment text, and the emotion evaluation result comprises an emotion tendency and a corresponding emotion intensity value; based on a screening type indicated in the comment screening operation, screening out comment information of which the display association degree exceeds a preset threshold value, the display association degree being used for representing the correlation between the screening type and the emotional tendency of the comment text; and sorting and displaying the screened comment information based on the user behavior data of the comment information and the difference coefficient. The comment display effect is improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a comment display method, device, equipment, storage medium and program product. Background Art

[0002] In the digital age, user reviews have become an important source of information with extremely high information value. Product reviews on online shopping platforms, order comments on group buying websites, user posts on social media, etc., all contain users' real feelings and opinions, which can better help businesses understand market demand, shape brand image, and improve products or services. It can also help users understand real product information and make better decisions.

[0003] In real-world review scenarios, a star rating mechanism is usually used. However, the inconsistency between ratings and reviews is common, that is, users give full-star praise but the reviews are negative, or users comment with positive emotions but give the opposite low-score bad reviews, which leads to high-quality reviews being buried or low-quality reviews being mistaken for good. Therefore, low-quality or useless reviews are often recommended to users, and the review display effect is poor. Summary of the invention

[0004] The present application provides a comment display method, device, equipment, storage medium and program product, which can solve the technical problem that the comments currently recommended to users are of low quality or useless, and improve the comment display effect.

[0005] In a first aspect, the present application provides a comment display method, comprising:

[0006] In response to the comment filtering operation, obtaining comment information to be displayed and user behavior data of the comment information, wherein the comment information includes an evaluation score and a comment text, and the user behavior data is used to indicate behavior data related to the comment information after it is published;

[0007] Based on the evaluation score and the user behavior data, sentiment analysis is performed on the comment text to obtain a sentiment evaluation result of the comment text, wherein the sentiment evaluation result includes a sentiment tendency and a corresponding sentiment intensity value;

[0008] Based on the screening type indicated in the comment screening operation, screening out comment information with a display relevance exceeding a preset threshold, the display relevance being used to characterize the correlation between the screening type and the sentiment tendency of the comment text;

[0009] The screened comment information is sorted and displayed based on the user behavior data and the difference coefficient of the comment information, and the difference coefficient is used to characterize the difference between the evaluation score and the emotion intensity value.

[0010] In a possible design, the sentiment analysis of the comment text is performed based on the evaluation score and the user behavior data to obtain the sentiment evaluation result of the comment text, including:

[0011] Performing entity recognition on the review text to obtain at least one review object and corresponding text content, wherein the text content is used to describe the evaluation content of the review object on any attribute;

[0012] Using a pre-established sentiment analysis model to perform sentiment analysis on the text content of the review object to obtain a sentiment analysis result of the review object, wherein the sentiment analysis result includes an entity sentiment tendency and a corresponding sentiment score;

[0013] The comment object, its corresponding attributes, and sentiment analysis results are combined into a triple;

[0014] Based on the comment information and the user behavior data, determining a combination weight of the triplet, the combination weight being used to characterize the degree of influence of the triplet on the evaluation score, the comment text and the user behavior data;

[0015] Based on the triples and the combined weights of the triples, a sentiment evaluation result of the comment text is determined.

[0016] In a possible design, determining the combination weight of the triplet based on the comment information and the user behavior data includes:

[0017] Determine the score influence of the corresponding review object based on the sentiment analysis result in the triple and the evaluation score, wherein the score influence value is used to represent the influence degree of the review object on the evaluation score;

[0018] Extracting semantic features from the comment text to obtain the subject described by the comment text, and determining the semantic relevance between the attribute in the triple and the subject;

[0019] Extracting relevant features of the comment object from the user behavior data;

[0020] Based on the association feature, the score influence and the semantic relevance, determine a first relevance and a second relevance of the triple, wherein the first relevance is used to characterize the degree of association between the score influence and the user behavior data, and the second relevance is used to characterize the degree of association between the semantic relevance and the user behavior data;

[0021] Determining the strength of the relationship between the triple and the user behavior data based on the first degree of association and the second degree of association of the triple, wherein the strength of the relationship is used to characterize the degree of influence of the triple on the user behavior data;

[0022] Based on the score influence, semantic relevance and relationship strength of the triples, a combination weight of the triples is determined.

[0023] In a possible design, the user behavior data includes interactive behavior data and / or conversion behavior data on the comment information;

[0024] The interactive behavior data is used to represent the behavior data generated after the user interacts with the comment information;

[0025] The conversion behavior data includes search behavior data and decision behavior data, wherein the search behavior data includes search terms related to the review object in the review information and browsing operation data of the display content displayed in response to the search terms, and the browsing operation data includes the number of browsing times, the corresponding browsing content and the browsing time;

[0026] The decision-making behavior data is used to indicate the behavior data generated by the decision-making behavior regarding the comment object after the comment information is released; the decision-making behavior data includes conversion comment information, and the conversion comment information is used to indicate the evaluation information formed after the decision-making behavior data occurs.

[0027] In a possible design, the screened comment information is sorted and displayed based on the user behavior data and the difference coefficient of the comment information, including:

[0028] Determining the comment influence of the comment information based on the user behavior data of the comment information;

[0029] Determining the comment type of the filtered comment information based on the comment influence and the difference coefficient;

[0030] The review information is sorted from high to low according to the priority of the review type to form a target display sequence.

[0031] In a possible design, determining the comment influence of the comment information based on the user behavior data of the comment information includes:

[0032] comparing the semantic consistency between the sentiment evaluation result of the review information and the sentiment evaluation result of the converted review information, and determining the user satisfaction of the review information based on the comparison result;

[0033] Performing feature processing on the interactive behavior data, the browsing operation data, and the decision-making behavior data to obtain a user interest level of the comment information, where the user interest level is used to indicate the degree of interest of the user in the comment information;

[0034] Based on the user satisfaction and the user interest, the comment influence of the comment information is determined.

[0035] In a possible design, determining the comment type of the filtered comment information based on the comment influence and the difference coefficient includes:

[0036] When the difference coefficient is less than a preset difference threshold, and the comment influence is greater than a preset influence threshold, the corresponding comment information is identified as a first-category target comment with the highest priority;

[0037] When the comment influence is greater than the influence threshold, but the difference coefficient is greater than the difference threshold, the corresponding comment information is identified as a second-category target comment;

[0038] When the difference coefficient is less than the difference threshold, but the comment influence is less than the influence threshold, the corresponding comment information is identified as a third-category target comment;

[0039] Among them, the priority of the second category target comments is higher than the priority of the third category target comments.

[0040] In a second aspect, the present application provides a review display device, comprising:

[0041] A data acquisition module, for acquiring, in response to a comment screening operation, comment information to be displayed and user behavior data of the comment information, wherein the comment information includes an evaluation score and a comment text, and the user behavior data is used to indicate behavior data related to the comment information after it is published;

[0042] A sentiment analysis module, configured to perform sentiment analysis on the comment text based on the evaluation score and the user behavior data, and obtain a sentiment evaluation result of the comment text, wherein the sentiment evaluation result includes a sentiment tendency and a corresponding sentiment intensity value;

[0043] A data screening module, for screening out the review information having a display relevance exceeding a preset threshold value based on the screening type indicated in the review screening operation, wherein the display relevance is used to characterize the correlation between the screening type and the sentiment tendency of the review text;

[0044] A display module is used to sort and display the screened comment information based on the user behavior data and the difference coefficient of the comment information, and the difference coefficient is used to represent the difference between the evaluation score and the emotion intensity value.

[0045] In a third aspect, an embodiment of the present application provides an electronic device, comprising: at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the comment display method described in the first aspect and various possible designs of the first aspect.

[0046] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, in which computer execution instructions are stored. When a processor executes the computer execution instructions, the comment display method described in the first aspect and various possible designs of the first aspect is implemented.

[0047] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program. When the computer program is executed by a processor, it implements the comment display method described in the first aspect and various possible designs of the first aspect.

[0048] The comment display method, device, equipment, storage medium and program product provided by the present application perform a multi-dimensional analysis of the sentiment tendency of the comment text through user behavior data and evaluation scores, and take into account the degree of influence of the comment text on the user behavior data and evaluation scores, which is conducive to a more accurate understanding of the sentiment tendency of the comment text and improves the accuracy of sentiment analysis. According to the screening type indicated by the comment screening operation, the comment information whose sentiment tendency of the comment text is consistent with the screening type is screened out, and is not limited to the situation where the comment text that meets the screening type cannot be displayed due to the inconsistency between the evaluation score and the comment text, thereby improving the generalizability of the comment information. According to the user behavior data and the difference coefficient, the comment information that meets the screening type is further sorted, and at the same time, the subjective perspective of the user's true intention and rating preference, as well as the objective perspective of the impact of the comment information on subsequent user behavior, are taken into account, so that the displayed comment information has a high reference value and reliability, which can solve the technical problem of low-quality or useless comments currently recommended to users, and can further improve the effect of comment display. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0050] Figure 1 A schematic diagram of the computer system architecture provided in the embodiment of the present application;

[0051] Figure 2 A flowchart of a comment display method provided in an embodiment of the present application;

[0052] Figure 3 A schematic diagram of the sentiment analysis process provided in the embodiment of the present application;

[0053] Figure 4 A schematic diagram of the structure of a sentiment analysis model provided for an exemplary embodiment of the present application;

[0054] Figure 5 A schematic diagram of a process for determining a combination weight provided in an embodiment of the present application;

[0055] Figure 6 A schematic diagram of a flow chart for sorting review information provided in an embodiment of the present application;

[0056] Figure 7 A schematic diagram of a process for determining the impact of a comment provided in an embodiment of the present application;

[0057] Figure 8 A structural schematic diagram of a review display device provided in an embodiment of the present application;

[0058] Fig. 9 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application.

[0059] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0060] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0061] In the technical solution of this application, the collection, storage, use, processing, transmission, provision and disclosure of information such as financial data or user data involved shall comply with the provisions of relevant laws and regulations and shall not violate public order and good morals.

[0062] It should be noted that in the embodiments of the present application, certain software, components, models and other existing solutions in the industry may be mentioned, and they should be regarded as exemplary. Their purpose is only to illustrate the feasibility of implementing the technical solution of the present application, but it does not mean that the applicant has or will necessarily use the solution.

[0063] First, the terms involved in this application are explained:

[0064] Sentiment analysis: refers to an important task in the field of natural language processing, which aims to understand and analyze the emotional tendencies expressed in text.

[0065] Emotional tendency: refers to the positive, negative or neutral emotional attitude towards an object, event, opinion or situation.

[0066] In the prior art, a star rating mechanism is usually set up to display the user's star rating results in the comment list or detail page. For example, in the list of good reviews, high-scoring reviews are usually displayed in a prominent position. However, the star rating mechanism has certain limitations. If it is judged only by ratings and scoring without considering the actual content of user reviews, it will lead to the inability to express the user's true feelings. At present, it is common for the rating to be inconsistent with the comment text, that is, the user gives a full-star praise but the comment is the opposite negative emotion (i.e., a high-scoring bad review), or the user's comment is positive but gives the exact opposite low-scoring bad review (i.e., a low-scoring good review), which causes high-quality reviews to be buried or low-quality reviews to be mistaken for being good.

[0067] Users usually rely on review content or ratings to understand product information in order to make better decisions. However, currently, the single-dimensional review display method often recommends low-quality or useless reviews to users, requiring users to spend more time to obtain the available information, increasing the complexity of the decision-making process and raising doubts about the authenticity of the reviews. Therefore, the existing review display effect is poor.

[0068] In response to the above technical problems, the present application proposes the following technical concept: by extracting comment information whose sentiment tendency is consistent with the filtering type indicated by the comment filtering operation, the extracted comment information can be sorted and displayed in combination with user behavior data, sentiment intensity value and evaluation score. It is not limited to the situation where the comment text that meets the filtering type cannot be displayed due to inconsistency between the evaluation score and the comment text, and improves the effect of comment display, thereby improving the efficiency of obtaining available information.

[0069] The execution subject of the method of the embodiment of the present application may be, for example, a computer system architecture including electronic devices (such as servers and terminals) with data processing capabilities.

[0070] Optional, Figure 1 The computer system architecture 100 provided in the embodiment of the present application includes a terminal 101 and a server 102. The terminal 101 and the server 102 can communicate through a network. The terminal 101 can be an electronic device installed with an application that supports displaying interactive content. The interactive content can be browsing, commenting, liking, quoting, etc. of multimedia information, and the multimedia information includes text, images, videos, and voice.

[0071] Exemplarily, the above application may be a terminal-side application of a multimedia platform, and the server 102 carries the multimedia platform, and the server 102 provides background services for the application installed and running in the terminal 101. For example, the multimedia platform may be a shopping e-commerce platform, a life service platform, a social application platform, an application download platform, etc.

[0072] In this application, if Figure 1 As shown, the user performs relevant operations through the terminal 101, such as selection operations, comment operations, like operations, purchase operations, browsing operations, etc., so that the terminal 101 collects comment information and user behavior data and communicates with the server 102. When the terminal 101 detects a comment filtering operation triggered by the user, the relevant instructions of the comment filtering operation are sent to the server 102. In response to the comment filtering operation, the server 102 processes the comment information to be displayed and the corresponding user behavior data, such as sentiment tendency analysis, comment filtering, and comment sorting, to obtain a target display sequence. The server 102 returns the target display sequence to the terminal 101 for comment display.

[0073] The program code processing method provided in this application is intended to solve the above technical problems in the prior art.

[0074] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0075] Figure 2 A flowchart of a comment display method provided in an embodiment of the present application, the comment display method includes steps S201 to S204.

[0076] S201. In response to a comment filtering operation, obtain comment information to be displayed and user behavior data of the comment information, wherein the comment information includes an evaluation score and a comment text, and the user behavior data is used to indicate behavior data related to the comment information after it is published.

[0077] In the present application, the comment filtering operation may be an operation for triggering the display of comment information according to the selected filtering type, for example, a triggering operation on a control of any filtering type. The filtering types may be good reviews, bad reviews, latest reviews, comment objects, attributes, and emotional descriptions.

[0078] It should be noted that when the filtering type is any review object, all reviews about the review object will be displayed, and the review object can be product A, dish B, etc.; when the filtering type is any attribute, all reviews about the attribute will be displayed, and the attribute can be quality, material, size, taste, etc.; when the filtering type is emotional description, comments related to the emotional description (that is, the emotional description of any attribute) will be displayed, and the emotional description can be comfortable material, good taste, good quality, etc.

[0079] The comment text may be in the form of text, and the evaluation score may be a rating value using a star rating mechanism. The comment information carries the publishing time, and the comment information to be displayed may be multiple comments sorted by publishing time, such as just published, in the last hour, in the last day, in the last week, in history, etc.

[0080] The user behavior data may be user behaviors associated with the comment information after the comment information is published, that is, relevant user behaviors occurring after the comment information is published.

[0081] In some embodiments, the user behavior data includes interactive behavior data and / or conversion behavior data for the comment information;

[0082] The interactive behavior data is used to represent the behavior data generated after the user interacts with the comment information;

[0083] The conversion behavior data includes search behavior data and decision behavior data, wherein the search behavior data includes search terms related to the review object in the review information and browsing operation data of the display content displayed in response to the search terms, and the browsing operation data includes the number of browsing times, the corresponding browsing content and the browsing time;

[0084] The decision-making behavior data is used to indicate the behavior data generated by the decision-making behavior regarding the comment object after the comment information is released; the decision-making behavior data includes conversion comment information, and the conversion comment information is used to indicate the evaluation information formed after the decision-making behavior data occurs.

[0085] Specifically, the user behavior data may be interactive behavior data, such as follow-up comment data, reply data, like data, reference data, browsing data, etc. The follow-up comment data refers to the data generated by adding comments to the comment information. The reply data refers to the data generated by replying to the comment information or follow-up comment data. The like data refers to the data generated by liking the comment information, and the like data may include the number of likes and the corresponding like time. The reference data may be the data generated by citing the comment information.

[0086] The user behavior data may also be conversion behavior data, which is used to characterize the conversion behavior that occurs after the comment content is published. The conversion behavior includes search behavior, purchase behavior, consultation behavior, download behavior, registration behavior, etc. related to the comment text (such as the comment object mentioned in the comment).

[0087] The search behavior data includes search terms related to the review object mentioned in the review, the displayed content displayed in response to the search terms, and browsing operation data used to characterize the browsing operation that occurs after clicking on any displayed content. The browsing operation data includes the number of browsing operations, the browsing content of each browsing, and the browsing time, which is used to characterize the length of time spent on the browsing content.

[0088] Decision-making behavior data includes data generated by decision-making behaviors (such as purchase behavior, consultation behavior, download behavior, and registration behavior). Specifically, decision-making behavior data can be the decision objects purchased, consulted, downloaded, and registered, and the decision status of the decision objects, such as the purchase quantity. The decision-making behavior data can also be conversion comment information generated after the decision-making behavior occurs, such as user comments, returns, complaints, etc.

[0089] For example, in the scenario of e-commerce shopping, if a user reads the review information "A brand thermos has good insulation effect", he / she will search for "A brand thermos". In this search behavior, the content displayed after the search may include related links of "A brand thermos", thermos of other brands similar to "B brand thermos" (such as "B brand thermos"), and other products of "A brand" (such as thermos cup sleeves, insulated lunch boxes, insulated bags, etc.). In this regard, the browsing operation data of the user under the displayed content is collected, such as the number of views, browsing content and browsing time.

[0090] Furthermore, in response to the above display content, the user purchased "Brand A Thermos Cup" and "Brand A Thermos Lunch Box", and recorded the corresponding decision-making behavior data (such as the products purchased and the quantity of products). Furthermore, the user posted corresponding user comments on the purchased products and / or initiated a return operation, and collected the corresponding conversion comment data (such as user comment information and return reasons).

[0091] S202: Based on the evaluation score and the user behavior data, sentiment analysis is performed on the comment text to obtain a sentiment evaluation result of the comment text, wherein the sentiment evaluation result includes a sentiment tendency and a corresponding sentiment intensity value.

[0092] In this application, sentiment tendency includes positive, negative and neutral, and sentiment intensity value can be the intensity or depth of emotion generated based on the sentiment tendency. In this way, sentiment analysis is performed on the comment text to obtain the sentiment tendency and sentiment intensity value of the comment text, thereby comparing the sentiment evaluation result of the comment text with the evaluation score, and filtering out the comment information whose evaluation score is inconsistent with the comment text.

[0093] Optionally, this application combines the evaluation score and user behavior data to perform sentiment analysis on the comment text to reflect the user's emotional tendency in multiple dimensions. Among them, considering the influence and importance of the content of the comment text on the user behavior data and the evaluation score, it is conducive to more accurately understanding the emotional tendency of the comment text and improving the accuracy of sentiment analysis.

[0094] In some embodiments, before sentiment analysis, the comment text to be displayed can be preprocessed by constructing a corpus. The corpus contains pre-collected sensitive words, illegal words, and updated network hot words and newly appeared illegal content obtained from other multimedia platforms (such as social platforms, e-commerce platforms). Then, with the help of the updated network hot words in the corpus, the comment text is segmented, and the network hot words can be identified to avoid being misjudged as illegal sensitive words and shielding, thereby ensuring the timeliness of the data. And find out whether there are sensitive words, illegal words or newly appeared illegal content in the text after segmentation, and if so, filter out the relevant comment text. Optionally, the comment text to be displayed is preprocessed by constructing a corpus. Therefore, the present embodiment improves the accuracy of sentiment analysis of the comment text by filtering out the comment information containing sensitive words and illegal words before sentiment tendency analysis.

[0095] S203: based on the screening type indicated in the comment screening operation, screening out the comment information whose display relevance exceeds a preset threshold, wherein the display relevance is used to characterize the correlation between the screening type and the sentiment tendency of the comment text.

[0096] In this embodiment, based on the screening type indicated in the comment screening operation, the comment information whose sentiment tendency of the comment text is consistent with the screening type or has a certain correlation (i.e., the display correlation exceeds a preset threshold) is screened out. For example, if the screening type is favorable comments, the screened comment information is the comment information whose sentiment tendency of the comment text is positive, such as the comment information with a high evaluation score and a positive comment text, and the comment information with a low evaluation score but a positive comment text.

[0097] In the prior art, after responding to the favorable comments indicated by the review screening operation, the review information with high evaluation scores is usually displayed. However, the review information with low-scoring favorable comments (i.e., the evaluation score is low but the review text is positive) is often displayed in the display list of negative comments and cannot be displayed in the display list of favorable comments. In addition, users may be affected by many factors when rating, such as personal preferences, scoring habits, expectations, etc., and these factors may not be fully reflected in the review text, resulting in incomplete correspondence between emotional tendencies and emotional intensity values ​​and evaluation scores. Therefore, it is difficult to fully and effectively display comments with reference value using a single-dimensional (such as evaluation scores based solely on star ratings) review display method.

[0098] In this regard, the present application filters out comment information that meets the screening type through the sentiment tendency of the comment text, that is, the true intention, and can extract corresponding comments that can provide information reference value, without being limited to the situation where the comment text that meets the screening type cannot be displayed due to inconsistency between the evaluation score and the comment text.

[0099] S204: Sort and display the screened comment information based on the user behavior data and the difference coefficient of the comment information, where the difference coefficient is used to characterize the difference between the evaluation score and the emotion intensity value.

[0100] In this application, after filtering out the review information that meets the screening type, the display ranking of the review information is analyzed in combination with the user behavior data, taking into account the degree of influence of the review information on user behavior, that is, the referenceability and usefulness of the review information. In this way, the review information with higher information value and influence is displayed first, the review display effect is improved, and thus the efficiency of users in the decision-making process is improved.

[0101] In addition, the display ranking of the comment information is analyzed by combining the difference between the evaluation score and the emotional intensity value of the comment text, taking into account the user's real intention and the subjective perspective of the rating preference. Therefore, based on user behavior data and the difference coefficient, the comment information that meets the screening type is further sorted, which is in line with the multi-dimensional (i.e. subjective and objective) analysis, avoiding the situation where the comments that are displayed first are highly subjective due to sorting based solely on the emotional intensity value of the comment text, and improving the display quality and display effect of the comments.

[0102] The comment display method provided in the embodiment of the present application performs a multi-dimensional analysis of the sentiment tendency of the comment text through user behavior data and evaluation scores, and takes into account the degree of influence of the comment text on the user behavior data and evaluation scores, which is conducive to a more accurate understanding of the sentiment tendency of the comment text and improves the accuracy of sentiment analysis. According to the screening type indicated by the comment screening operation, the comment information whose sentiment tendency of the comment text is consistent with the screening type is screened out, and is not limited to the situation where the comment text that meets the screening type cannot be displayed due to the inconsistency between the evaluation score and the comment text, thereby improving the generalizability of the comment information. According to the user behavior data and the difference coefficient, the comment information that meets the screening type is further sorted, and at the same time, the subjective perspective of the user's true intention and rating preference, as well as the objective perspective of the impact of the comment information on subsequent user behavior, are taken into account, so that the displayed comment information has a high reference value and reliability, which can solve the technical problem of low-quality or useless comments currently recommended to users, and can further improve the effect of comment display.

[0103] In one possible embodiment, currently the overall sentiment tendency of the comment text is usually judged, but there may be multiple sentiment points in the comment text (i.e., multiple comment objects and comment text content about the comment objects), and the sentiment tendency and sentiment intensity value of each sentiment point may be different, which may lead to fuzzy analysis results or even misleading conclusions. In this regard, in the sentiment analysis process, this embodiment performs sentiment analysis on at least one comment object involved in the comment text and the text content about the comment object to obtain the sentiment analysis results of each evaluation object, and then calculates the sentiment tendency and sentiment intensity value of the comment text according to the importance of each evaluation object. Specifically, Figure 3 The flowchart of sentiment analysis provided in the embodiment of the present application is as follows. The present embodiment improves the sentiment analysis method, which includes steps S301 to S305.

[0104] S301, performing entity recognition on the review text to obtain at least one review object and corresponding text content, wherein the text content is used to describe the evaluation content of the review object on any attribute;

[0105] S302, using a pre-established sentiment analysis model to perform sentiment analysis on the text content of the review object to obtain a sentiment analysis result of the review object, wherein the sentiment analysis result includes an entity sentiment tendency and a corresponding sentiment score;

[0106] S303, forming a triplet of the comment object, its corresponding attributes, and sentiment analysis results;

[0107] S304: determining a combination weight of the triplet based on the comment information and the user behavior data, wherein the combination weight is used to represent the influence of the triplet on the evaluation score, the comment text, and the user behavior data;

[0108] S305 : Determine a sentiment evaluation result of the comment text based on the triples and the combined weights of the triples.

[0109] It should be noted that the review text contains evaluations of multiple review objects, and these evaluations include both positive and negative reviews. For example, for restaurant reviews, there will be an evaluation of at least one food product in the review text. For example, the review text is "The fish and chicken taste good, but the vegetables are not fresh", which contains multiple review objects (fish, chicken, vegetables), but the sentiment tendencies of the review objects are different. For example, the sentiment tendencies of fish and chicken are positive, while the sentiment tendencies of vegetables are negative. If the sentiment tendency of the entire review text is judged, it may be difficult to accurately capture the overall sentiment tendency.

[0110] In this regard, this embodiment performs entity recognition on the review text to extract one or more review objects and text content related to the review object. The review object is the object being evaluated, such as the name of a product. Attributes can be certain attributes of the product, such as function, appearance, performance, quality, and the user's experience and feelings during use. Exemplarily, taking the scenario of a catering platform as an example, the review text is "The dishes taste good and the environment is comfortable", which contains two review objects. First, the review object is "dishes" and the attribute is taste. Second, the review object is "environment" and the attribute is experience (such as comfort).

[0111] In this embodiment, sentiment analysis is performed on the text content of each comment object to obtain the entity sentiment tendency and corresponding sentiment score of each comment object. Optionally, sentiment analysis is performed using a sentiment analysis model, which can be a neural network model such as a long short-term memory network (LSTM), a recurrent neural network, etc.

[0112] For example, the sentiment analysis model is a long short-term memory network. Figure 4 A schematic diagram of the structure of a sentiment analysis model provided by an exemplary embodiment of the present application, wherein the long short-term memory network includes a gating mechanism, namely a forget gate, an input gate, and an output gate, so as to effectively capture and process long-term dependencies. Among them, the forget gate is used to determine which information is forgotten or discarded, such as Figure 4As shown in the marked area 401 in FIG. 4 , the forget gate contains a sigmoid activation function (sigmoid1 in area 401). The forget gate outputs a value between 0 and 1 based on the input data and the previous hidden state, and the forget gate also controls the degree of retention of historical information. The input gate controls the input of new information, such as Figure 4 As shown in the marked area 402 in FIG. 4 . The input gate includes a sigmoid layer and a tanh layer, wherein the activation function of the sigmoid layer (sigmoid2 in area 402) is used to generate a value between 0 and 1, controlling which information will be updated. The tanh layer (tanh1 in area 402) generates new candidate values ​​for updating. The output gate controls the current hidden state and the final output, as shown in FIG. Figure 4 As shown in the area 403 marked in . The output gate includes a sigmoid activation function (sigmoid3 in area 403), which is used to determine which parts of the output state will be activated. The output gate also includes a tanh activation function (tanh2 in area 403), which is used to generate a candidate value between -1 and 1, and send the filtered and updated information to the next layer of hidden state, and generate the final output based on this hidden state. Among them, the Sigmoid function is a common S-type function, also known as the S-type growth curve. In deep learning, due to its monotonic and inverse monotonic properties, the Sigmoid function is often used as an activation function of a neural network to map variables between [0, 1].

[0113] Furthermore, each comment object in the comment text, the corresponding attributes and the sentiment analysis results obtained by analysis are combined into a triple of the comment object, and the comment text contains one or more triples. Furthermore, the combined weight of each triple is obtained to calculate the sentiment tendency and sentiment intensity value of the comment text based on the triple and its combined weight. Among them, each comment object may have different sentiment tendencies and sentiment intensities. This embodiment considers the importance of the evaluation of the comment object in the comment text (such as semantic sentiment intensity). Since the evaluation of each comment object may be different, the evaluation score (such as star rating) corresponding to each comment object is also different. This embodiment considers the influence of the rating object on the evaluation score (such as the correlation between the evaluation of the evaluation object and the evaluation score). In addition, considering the influence of each comment object on subsequent user behavior, the information value of the comment object can be reflected.

[0114] Therefore, this embodiment determines the sentiment analysis results of each comment object in the comment text separately, and analyzes the emotional intention expressed by each comment object separately. Then, considering the influence of the evaluation object on the overall evaluation score, comment text and subsequent user behavior, based on the weight of each comment object, the emotional tendency of the comment text is analyzed, and the weight of each comment object is flexibly adjusted, thereby improving the accuracy of sentiment analysis.

[0115] Based on the above embodiments, in some embodiments, for the combination weights in the sentiment analysis process, considering the influence of user behavior data, evaluation scores and comment texts on triples, this embodiment provides a solution for determining the combination weights. Figure 5 A schematic diagram of a process for determining a combination weight provided in an embodiment of the present application, the method includes steps S501 to S506.

[0116] S501, based on the sentiment analysis result in the triple and the evaluation score, determining the score influence of the corresponding review object, wherein the score influence value is used to represent the influence degree of the review object on the evaluation score;

[0117] S502, extracting semantic features from the comment text to obtain the topic described by the comment text, and determining the semantic relevance between the attributes in the triple and the topic;

[0118] S503, extracting the correlation features of the comment object in the user behavior data;

[0119] S504: Determine a first degree of association and a second degree of association of the triple based on the association feature, the score influence and the semantic relevance, wherein the first degree of association is used to characterize the degree of association between the score influence and the user behavior data, and the second degree of association is used to characterize the degree of association between the semantic relevance and the user behavior data;

[0120] S505: Determine the relationship strength between the triple and the user behavior data based on the first correlation degree and the second correlation degree of the triple, where the relationship strength is used to characterize the degree of influence of the triple on the user behavior data;

[0121] S506: Determine the combination weight of the triples based on the score influence, semantic relevance, and relationship strength of the triples.

[0122] In the prior art, if the evaluation of a certain review object in a review text occupies a larger space or is more prominent (such as using stronger sentiment words), then in the overall analysis, the evaluation of this review object may affect the evaluation of other products, resulting in uneven weight distribution of the review objects and inaccurate sentiment analysis of the review text.

[0123] In this regard, this embodiment extracts semantic features from the comment text to determine semantic features such as the topic of the comment text. The method of extracting the semantic features can be through a neural network model, function mapping, database matching, etc., which are not limited here. Then, the topic of the comment text and the attribute of the comment object in the triple are combined to determine the semantic influence of the attribute on the topic, thereby reflecting the data contribution of the triple to the overall comment text and the topic relevance.

[0124] Also, since there is an inconsistency between the comment text and the evaluation score, the actual sentiment tendency of the comment object does not completely correspond to the evaluation score. In this case, this embodiment takes into account the influence of the comment object on the evaluation score (such as the correlation between the evaluation of the evaluation object and the evaluation score), which can reflect the importance and reference value of the comment object.

[0125] For example, if the review content is "Mobile phone A has good performance, but poor delivery logistics", the evaluation score of this purchase is low, then the review object "Mobile phone delivery logistics" has a greater impact on the evaluation score, while the corresponding object "Mobile phone A" has a smaller impact on the evaluation score. Therefore, in the sentiment analysis of the review content, the combined weight of the "Mobile phone delivery logistics" triple is higher than the combined weight of the "Mobile phone A" triple, which can reflect the key factors affecting the evaluation score.

[0126] Furthermore, the associated features of each comment object in the user behavior data can be extracted by means of a neural network model or the like, such as search terms, purchase objects, download objects, user comments, reasons for returns, etc. Further, based on the associated features of the comment object and the score influence, the first degree of association between the two is determined, taking into account the impact of the comment object on the evaluation score on subsequent user behavior. For example, for a comment object with a higher score influence, its associated features in the user behavior data are more obvious (such as purchase behavior regarding the comment object or related products), then the first degree of association is higher. In this way, this embodiment can evaluate the degree of influence of factors affecting the evaluation score on user behavior, can reflect the importance of the comment object, and indirectly reflect the influence of the evaluation score on user behavior and the data value.

[0127] At the same time, based on the correlation characteristics and semantic relevance of the comment object, the second correlation between the two is determined. Taking into account the impact of the fit between the comment object and the comment topic on subsequent user behavior, it is possible to evaluate the degree of influence of factors affecting the comment text topic on user behavior, and reflect the importance and reference value of the comment object.

[0128] Furthermore, based on the first correlation and the second correlation of the triple, the strength of the relationship between the triple and the user behavior data is determined. Among them, the strength of the relationship can reflect the real influence of the comment object on the user behavior, that is, the influence of the comment object on the evaluation score and the topic of the comment text, and the correlation with the subsequent user behavior and the degree of influence. For example, in the face of high-scoring reviews, if the evaluation object A with a high score influence does not detect related search behavior, purchase behavior and other conversion behaviors in subsequent user behaviors, it is considered that the first correlation between the evaluation object A and the user behavior is low, indicating that the evaluation object A has a low influence on the user behavior (that is, the strength of the relationship is low), and there may be abnormal operation behaviors such as brushing good reviews.

[0129] Therefore, this embodiment determines the combined weight of the triples through the score influence, semantic relevance and relationship strength of the triples, taking into account the evaluation score, the topic of the comment text and the influence of the user behavior on the comment object. Combining the subjective factors of the comment object (such as the evaluation score determined by the user due to personal habits and preferences, the topic described in the comment text) and objective factors (such as the actual influence of the comment object on the user's behavior), a comprehensive assessment of the value and importance of the comment object is achieved, further improving the accuracy of sentiment analysis.

[0130] In some embodiments, after obtaining the sentiment evaluation results of the comment information, the comment information can be ranked based on factors such as data value and content authenticity to improve the impact of the comment on user behavior. Figure 6 This is a flow chart of comment information sorting provided in an embodiment of the present application. This embodiment provides a comment information sorting method, which includes steps S601 to S603.

[0131] S601, determining the comment influence of the comment information based on the user behavior data of the comment information;

[0132] S602, determining the comment type of the screened comment information based on the comment influence and the difference coefficient;

[0133] S603: Sort the comment information from high to low according to the priority of the comment type to form a target display sequence.

[0134] In this embodiment, by considering the impact of the review information on subsequent user behavior, such as the purchase behavior of the review object or related products involved in the review information, the review impact of the review information on the user behavior is determined, which can reflect the data contribution and content value of the review information. Then, based on the review impact, evaluation score and sentiment evaluation results, the review type of the review information can be determined, so that the review type can be sorted according to the priority to form a target display sequence.

[0135] Therefore, this embodiment can sort the comments according to the user's satisfaction with the comment information and the user's subjective evaluation of the comment information, thereby improving the display effect and facilitating the user to efficiently obtain the comments and the available information therein.

[0136] Based on the above embodiments, in some embodiments, the comment influence of the comment information is determined based on the user behavior data of the comment information, that is, the comment influence may include the user satisfaction and user interest of the user to the comment information. Specifically, Figure 7 This is a flow chart of determining the influence of a comment provided in an embodiment of the present application. This embodiment provides a method for determining the influence of a comment, and the method includes steps S701 to S703.

[0137] S701, comparing the semantic consistency between the sentiment evaluation result of the comment information and the sentiment evaluation result of the converted comment information, and determining the user satisfaction of the comment information based on the comparison result;

[0138] S702, performing feature processing on the interactive behavior data, the browsing operation data, and the decision-making behavior data to obtain a user interest level of the comment information, where the user interest level is used to indicate the degree of interest of the user in the comment information;

[0139] S703: Determine the comment influence of the comment information based on the user satisfaction and the user interest.

[0140] In this embodiment, if the publishing time of the comment information is before the publishing time of the converted comment information, the comment information (such as the user's evaluation of the comment object therein) may have a certain impact on the user, affecting the user's initial evaluation of the comment object indicated by the comment information. Based on this, the converted comment information generated when the user has a conversion behavior (such as the conversion behavior of the comment object in the comment information) is related to the previous comment information, and can reflect the accuracy and authenticity of the evaluation of the comment object in the previous comment information, as well as the user's satisfaction with the comment information. Therefore, this embodiment can determine the data value of the comment information and the degree of compliance with user needs by judging whether the semantics and emotions between the comment information and the converted comment information are consistent.

[0141] For example, user A posts a review about a pair of shoes. After user B browses the review, user B's opinion of the shoes may be influenced by user A's review. Then, after user B purchases the shoes, he posts a corresponding conversion review. If the conversion review is a positive review of the shoes, it means that the semantics and emotions of the two are consistent, and the comment is more directional and has higher information value. If the conversion review is a negative review of the shoes, it means that the semantics and emotions of the two are quite different, and there may be malicious comments such as fake orders and cash back for positive reviews, and the actual information value of the comment is low.

[0142] Furthermore, feature processing is performed on the user behavior data, and the feature processing may be performed by means of function transformation, that is, feature processing is performed on the interactive behavior data, browsing operation data, and decision-making behavior data to obtain like features (such as the number of likes and the number of dislikes), comment features (such as comment time and comment sentiment score), search features (such as the number of searches and the search time), browsing features (such as the number of views and the browsing duration), decision features (such as the purchase quantity and the purchase time), etc., to achieve operational quantification of user behavior, thereby determining the user's interest in the comment information.

[0143] Therefore, this embodiment determines the comment influence of the comment information through user satisfaction and user interest, which can reflect the data contribution and data value of the comment information from an objective perspective, and can screen out malicious comments such as fake orders and cash back for good reviews, thereby improving the effect of comment display.

[0144] Based on the above embodiments, in some embodiments, the comment type of the comment information is determined by the comment influence and the difference coefficient of the comment information, and then the comment information is sorted according to the comment type. Specifically, this embodiment provides a method for determining the comment type, which includes:

[0145] When the difference coefficient is less than a preset difference threshold, and the comment influence is greater than a preset influence threshold, the corresponding comment information is identified as a first-category target comment with the highest priority;

[0146] When the comment influence is greater than the influence threshold, but the difference coefficient is greater than the difference threshold, the corresponding comment information is identified as a second-category target comment;

[0147] When the difference coefficient is less than the difference threshold, but the comment influence is less than the influence threshold, the corresponding comment information is identified as a third-category target comment;

[0148] Among them, the priority of the second category target comments is higher than the priority of the third category target comments.

[0149] In this embodiment, on the one hand, when the difference coefficient between the comment text and the evaluation score is small, and the comment influence is large, it indicates that the comment information has high information value and high reference value, and the evaluation score is consistent with the real emotional tendency of the comment text, and the comment information is marked as the first type of target comment and is displayed first. Therefore, the target display sequence can display high-quality comment information and improve the comment display effect.

[0150] On the other hand, when facing comment information with a large coefficient of difference, the evaluation score of the comment information may be greatly affected by the user's personal habits and preferences, or it may be an abnormal scoring operation made by the user to highlight the comment text (such as to attract malicious comments from users). Therefore, when the comment has a large impact but a large coefficient of difference, it is marked as a second-class target comment and displayed after the first-class target comment. This embodiment can filter out malicious comments, and can accurately display available comment information regardless of whether it is affected by the user's personal habits and preferences.

[0151] On the other hand, when the coefficient of variation is small but the influence of the comments is small, there may be malicious comments such as fake positive reviews, or the comment information may be highly subjective. Therefore, they are identified as the third type of target comments and displayed after the second type of target comments.

[0152] Optionally, when the difference coefficient is large and the comment influence is small, the comment information can be considered to be a malicious comment with low information value, and can be filtered out and not displayed in the target display sequence.

[0153] Therefore, this embodiment sorts the comment information by comment influence and difference coefficient, which can display comments that may be malicious and highly subjective at a later position, give priority to displaying high-quality comments, and facilitate users to efficiently obtain comments and the available information therein, thereby improving the display effect.

[0154] Optionally, this embodiment may also include an associated display sequence displayed after the target display sequence. Among them, all browsing contents or browsing contents whose browsing time exceeds a preset value in the search behavior data may be extracted, and comment information related to the browsing contents may be obtained and added to the associated display sequence. Secondly, associated objects having the same attributes as the decision objects in the decision behavior data may be further extracted, and comment information related to the associated objects may be obtained and added to the associated display sequence. Therefore, this embodiment can be targeted and meet user needs.

[0155] Figure 8 A schematic diagram of the structure of a review display device provided in an embodiment of the present application, the review display device 800 includes:

[0156] The data acquisition module 801 is used to obtain the comment information to be displayed and the user behavior data of the comment information in response to the comment screening operation, wherein the comment information includes the evaluation score and the comment text, and the user behavior data is used to indicate the behavior data related to the comment information after it is released;

[0157] Sentiment analysis module 802, used to perform sentiment analysis on the comment text based on the evaluation score and the user behavior data to obtain a sentiment evaluation result of the comment text, wherein the sentiment evaluation result includes a sentiment tendency and a corresponding sentiment intensity value;

[0158] A data screening module 803 is used to screen out the comment information whose display relevance exceeds a preset threshold based on the screening type indicated in the comment screening operation, wherein the display relevance is used to characterize the correlation between the screening type and the sentiment tendency of the comment text;

[0159] The display module 804 is used to sort and display the screened comment information based on the user behavior data and the difference coefficient of the comment information, and the difference coefficient is used to represent the difference between the evaluation score and the emotion intensity value.

[0160] In some embodiments, the sentiment analysis module 802 includes:

[0161] An entity recognition unit, used to perform entity recognition on the review text to obtain at least one review object and corresponding text content, wherein the text content is used to describe the evaluation content of the review object on any attribute;

[0162] A model analysis unit, used to perform sentiment analysis on the text content of the review object using a pre-established sentiment analysis model to obtain a sentiment analysis result of the review object, wherein the sentiment analysis result includes an entity sentiment tendency and a corresponding sentiment score;

[0163] A triplet construction unit, used to form triples from the comment object and its corresponding attributes and sentiment analysis results;

[0164] a combination weight determination unit, used to determine the combination weight of the triplet based on the comment information and the user behavior data, wherein the combination weight is used to represent the influence of the triplet on the evaluation score, the comment text and the user behavior data;

[0165] The sentiment evaluation result calculation unit is used to determine the sentiment evaluation result of the comment text based on the triples and the combination weights of the triples.

[0166] Based on the above embodiments, in some embodiments, the combination weight determination unit includes:

[0167] A rating influence determination unit, used to determine the rating influence of the corresponding review object based on the sentiment analysis result in the triple and the evaluation score, wherein the rating influence value is used to represent the influence degree of the review object on the evaluation score;

[0168] A semantic feature extraction unit, used to extract semantic features from the comment text, obtain the subject described by the comment text, and determine the semantic relevance between the attribute in the triple and the subject;

[0169] A correlation feature extraction unit, used to extract correlation features for the comment object from the user behavior data;

[0170] a correlation calculation unit, configured to determine a first correlation and a second correlation of the triple based on the correlation feature, the score influence and the semantic correlation, wherein the first correlation is used to characterize the correlation between the score influence and the user behavior data, and the second correlation is used to characterize the correlation between the semantic correlation and the user behavior data;

[0171] a relationship strength calculation unit, used to determine the relationship strength between the triple and the user behavior data based on the first degree of association and the second degree of association of the triple, wherein the relationship strength is used to characterize the degree of influence of the triple on the user behavior data;

[0172] A combination weight calculation unit is used to determine the combination weight of the triples based on the score influence, semantic relevance and relationship strength of the triples.

[0173] In some embodiments, the user behavior data includes interactive behavior data and / or conversion behavior data for the comment information;

[0174] The interactive behavior data is used to represent the behavior data generated after the user interacts with the comment information;

[0175] The conversion behavior data includes search behavior data and decision behavior data, wherein the search behavior data includes search terms related to the review object in the review information and browsing operation data of the display content displayed in response to the search terms, and the browsing operation data includes the number of browsing times, the corresponding browsing content and the browsing time;

[0176] The decision-making behavior data is used to indicate the behavior data generated by the decision-making behavior regarding the comment object after the comment information is released; the decision-making behavior data includes conversion comment information, and the conversion comment information is used to indicate the evaluation information formed after the decision-making behavior data occurs.

[0177] In some embodiments, the display module 804 includes:

[0178] A comment influence determination unit, configured to determine the comment influence of the comment information based on user behavior data of the comment information;

[0179] A comment type identification unit, used to determine the comment type of the screened comment information based on the comment influence and the difference coefficient;

[0180] The target display sequence forming unit is used to sort the comment information from high to low according to the priority of the comment type to form a target display sequence.

[0181] In some embodiments, the review influence determination unit includes:

[0182] A user satisfaction calculation unit, used for comparing the semantic consistency between the sentiment evaluation result of the comment information and the sentiment evaluation result of the converted comment information, and determining the user satisfaction of the comment information based on the comparison result;

[0183] A user interest calculation unit, used to perform feature processing on the interactive behavior data, the browsing operation data, and the decision-making behavior data to obtain the user interest of the comment information, where the user interest is used to indicate the degree of interest of the user in the comment information;

[0184] The comment influence calculation unit is used to determine the comment influence of the comment information based on the user satisfaction and the user interest.

[0185] In some embodiments, the comment type identification unit includes:

[0186] A first identification unit, configured to identify the corresponding comment information as a first-category target comment with the highest priority when the difference coefficient is less than a preset difference threshold and the comment influence is greater than a preset influence threshold;

[0187] A second identification unit is used to identify the corresponding comment information as a second type of target comment when the comment influence is greater than the influence threshold but the difference coefficient is greater than the difference threshold;

[0188] A third identification unit, configured to identify the corresponding comment information as a third type of target comment when the difference coefficient is less than the difference threshold but the comment influence is less than the influence threshold;

[0189] Among them, the priority of the second category target comments is higher than the priority of the third category target comments.

[0190] The device of the embodiments of the present application can execute the method provided by the embodiments of the present application, and the implementation principles are similar. The actions performed by each module in the device of each embodiment of the present application correspond to the steps in the method of each embodiment of the present application. For the detailed functional description of each module of the device, please refer to the description in the corresponding method shown in the previous text, which will not be repeated here.

[0191] It should be noted that it should be understood that the division of the various modules of the above device is only a division of logical functions. In actual implementation, they can be fully or partially integrated into one physical entity, or they can be physically separated. And these modules can all be implemented in the form of software called by processing elements; they can also be all implemented in the form of hardware; some modules can also be implemented in the form of software called by processing elements, and some modules can be implemented in the form of hardware. For example, the sentiment analysis module 802 can be a separately established processing element, or it can be integrated in a chip of the above device. In addition, it can also be stored in the memory of the above device in the form of program code, and called and executed by a processing element of the above device. The function of the above sentiment analysis module 802. The implementation of other modules is similar. In addition, these modules can be fully or partially integrated together, or they can be implemented independently. The processing element here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each module above can be completed by an integrated logic circuit of hardware in the processor element or instructions in the form of software.

[0192] Fig. 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Fig. 9 As shown, the electronic device 90 may include: a transceiver 91 , a processor 92 , and a memory 93 .

[0193] The processor 92 executes the computer execution instructions stored in the memory, so that the processor 92 executes the scheme in the above embodiment. The processor 92 can be a general-purpose processor, including a central processing unit CPU, a network processor (NP), etc.; it can also be a digital signal processor DSP, an application-specific integrated circuit ASIC, a field programmable gate array FPGA or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components.

[0194] The memory 93 is connected to the processor 92 via a system bus and completes communication between them. The memory 93 is used to store computer program instructions.

[0195] The transceiver 91 may be used to obtain tasks to be executed and configuration information of the tasks to be executed.

[0196] The system bus can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The system bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus. The transceiver is used to realize the communication between the database access device and other computers (such as clients, read-write libraries, and read-only libraries). The memory may include random access memory (RAM) and may also include non-volatile memory.

[0197] The electronic device provided in the embodiment of the present application may be the terminal device of the above embodiment.

[0198] The embodiment of the present application also provides a chip for running instructions, which is used to execute the technical solution of the comment display method in the above embodiment.

[0199] An embodiment of the present application further provides a computer-readable storage medium, in which computer instructions are stored. When the computer instructions are executed on a computer, the computer executes the technical solution of the comment display method of the above embodiment.

[0200] An embodiment of the present application also provides a computer program product, which includes a computer program stored in a computer-readable storage medium. At least one processor can read the computer program from the computer-readable storage medium. When at least one processor executes the computer program, it can implement the technical solution of the comment display method in the above embodiment.

[0201] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of modules is only a logical function division, and there may be other division methods in actual implementation, such as multiple modules can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or modules, which can be electrical, mechanical or other forms.

[0202] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to implement the solution of this embodiment.

[0203] In addition, each functional module in each embodiment of the present application can be integrated into one processing unit, or each module can exist physically separately, or two or more modules can be integrated into one unit. The above-mentioned module-composed unit can be implemented in the form of hardware or in the form of hardware plus software functional units.

[0204] The above-mentioned integrated module implemented in the form of a software function module can be stored in a computer-readable storage medium. The above-mentioned software function module is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to perform some steps of the methods of various embodiments of the present application.

[0205] It should be understood that the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), etc. A general-purpose processor may be a microprocessor or any conventional processor. The steps of the method disclosed in the invention may be directly implemented as being executed by a hardware processor, or may be implemented by a combination of hardware and software modules in the processor.

[0206] The memory may include a high-speed RAM memory, and may also include a non-volatile storage NVM, such as at least one disk memory, and may also be a USB flash drive, a mobile hard disk, a read-only memory, a magnetic disk or an optical disk, etc.

[0207] The bus may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, the bus in the drawings of the present application is not limited to only one bus or one type of bus.

[0208] The above storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The storage medium can be any available medium that can be accessed by a general or special purpose computer.

[0209] An exemplary storage medium is coupled to a processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an application specific integrated circuit (ASIC). Of course, the processor and the storage medium can also exist as discrete components in an electronic control unit or a main control device.

[0210] Those skilled in the art can understand that all or part of the steps of implementing the above-mentioned method embodiments can be completed by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, the steps of the above-mentioned method embodiments are executed; and the aforementioned storage medium includes: ROM, RAM, disk or optical disk and other media that can store program codes.

[0211] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A comment display method, characterized in that: include: In response to the comment filtering operation, obtaining comment information to be displayed and user behavior data of the comment information, wherein the comment information includes an evaluation score and a comment text, and the user behavior data is used to indicate behavior data related to the comment information after it is published; Based on the evaluation score and the user behavior data, sentiment analysis is performed on the comment text to obtain a sentiment evaluation result of the comment text, wherein the sentiment evaluation result includes a sentiment tendency and a corresponding sentiment intensity value; Based on the screening type indicated in the comment screening operation, screening out comment information with a display relevance exceeding a preset threshold, the display relevance being used to characterize the correlation between the screening type and the sentiment tendency of the comment text; The screened comment information is sorted and displayed based on the user behavior data and the difference coefficient of the comment information, and the difference coefficient is used to characterize the difference between the evaluation score and the emotion intensity value.

2. The comment display method according to claim 1, characterized in that: The step of performing sentiment analysis on the comment text based on the evaluation score and the user behavior data to obtain a sentiment evaluation result of the comment text includes: Performing entity recognition on the review text to obtain at least one review object and corresponding text content, wherein the text content is used to describe the evaluation content of the review object on any attribute; Using a pre-established sentiment analysis model to perform sentiment analysis on the text content of the review object to obtain a sentiment analysis result of the review object, wherein the sentiment analysis result includes an entity sentiment tendency and a corresponding sentiment score; The comment object, its corresponding attributes, and sentiment analysis results are combined into a triple; Based on the comment information and the user behavior data, determining a combination weight of the triplet, the combination weight being used to characterize the degree of influence of the triplet on the evaluation score, the comment text and the user behavior data; Based on the triples and the combined weights of the triples, a sentiment evaluation result of the comment text is determined.

3. The comment display method according to claim 2, characterized in that: The determining the combination weight of the triplet based on the comment information and the user behavior data includes: Determine the score influence of the corresponding review object based on the sentiment analysis result in the triple and the evaluation score, wherein the score influence value is used to represent the influence degree of the review object on the evaluation score; Extracting semantic features from the comment text to obtain the subject described by the comment text, and determining the semantic relevance between the attribute in the triple and the subject; Extracting relevant features of the comment object from the user behavior data; Based on the association feature, the score influence and the semantic relevance, determine a first relevance and a second relevance of the triple, wherein the first relevance is used to characterize the degree of association between the score influence and the user behavior data, and the second relevance is used to characterize the degree of association between the semantic relevance and the user behavior data; Determining the strength of the relationship between the triple and the user behavior data based on the first degree of association and the second degree of association of the triple, wherein the strength of the relationship is used to characterize the degree of influence of the triple on the user behavior data; Based on the score influence, semantic relevance and relationship strength of the triples, a combination weight of the triples is determined.

4. The comment display method according to claim 3, characterized in that: The user behavior data includes interactive behavior data and / or conversion behavior data on the comment information; The interactive behavior data is used to represent the behavior data generated after the user interacts with the comment information; The conversion behavior data includes search behavior data and decision behavior data, wherein the search behavior data includes search terms related to the review object in the review information and browsing operation data of the display content displayed in response to the search terms, and the browsing operation data includes the number of browsing times, the corresponding browsing content and the browsing time; The decision-making behavior data is used to indicate behavior data generated by a decision-making behavior on the review object after the review information is published; The decision-making behavior data includes conversion comment information, and the conversion comment information is used to indicate evaluation information formed after the decision-making behavior data occurs.

5. The comment display method according to claim 4, characterized in that: The step of sorting and displaying the filtered comment information based on the user behavior data and the difference coefficient of the comment information includes: Determining the comment influence of the comment information based on the user behavior data of the comment information; Determining the comment type of the filtered comment information based on the comment influence and the difference coefficient; The review information is sorted from high to low according to the priority of the review type to form a target display sequence.

6. The comment display method according to claim 5, characterized in that: The determining the comment influence of the comment information based on the user behavior data of the comment information includes: comparing the semantic consistency between the sentiment evaluation result of the review information and the sentiment evaluation result of the converted review information, and determining the user satisfaction of the review information based on the comparison result; Performing feature processing on the interactive behavior data, the browsing operation data, and the decision-making behavior data to obtain a user interest level of the comment information, where the user interest level is used to indicate the degree of interest of the user in the comment information; Based on the user satisfaction and the user interest, the comment influence of the comment information is determined.

7. The comment display method according to claim 6, characterized in that: Determining the comment type of the filtered comment information based on the comment influence and the difference coefficient includes: When the difference coefficient is less than a preset difference threshold, and the comment influence is greater than a preset influence threshold, the corresponding comment information is identified as a first-category target comment with the highest priority; When the comment influence is greater than the influence threshold, but the difference coefficient is greater than the difference threshold, the corresponding comment information is identified as a second-category target comment; When the difference coefficient is less than the difference threshold, but the comment influence is less than the influence threshold, the corresponding comment information is identified as a third-category target comment; Among them, the priority of the second category target comments is higher than the priority of the third category target comments.

8. A review display device, characterized in that: include: A data acquisition module, for acquiring, in response to a comment screening operation, comment information to be displayed and user behavior data of the comment information, wherein the comment information includes an evaluation score and a comment text, and the user behavior data is used to indicate behavior data related to the comment information after it is published; A sentiment analysis module, configured to perform sentiment analysis on the comment text based on the evaluation score and the user behavior data, and obtain a sentiment evaluation result of the comment text, wherein the sentiment evaluation result includes a sentiment tendency and a corresponding sentiment intensity value; A data screening module, for screening out the review information having a display relevance exceeding a preset threshold value based on the screening type indicated in the review screening operation, wherein the display relevance is used to characterize the correlation between the screening type and the sentiment tendency of the review text; A display module is used to sort and display the screened comment information based on the user behavior data and the difference coefficient of the comment information, and the difference coefficient is used to represent the difference between the evaluation score and the emotion intensity value.

9. An electronic device, characterized in that: include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 7 when executed by a processor.

11. A computer program product, characterized in that The invention comprises a computer program, which implements the method according to any one of claims 1 to 7 when being executed by a processor.

Images