Interactive interface optimization suggestion generation method and device, equipment and storage medium

Abstract

The embodiment of the invention discloses an interactive interface optimization suggestion generation method and device, equipment and a storage medium, and the method comprises the steps: obtaining a plurality of comment texts related to the design of an interactive interface, and carrying out the sentiment analysis processing of the comment texts, and obtaining a negative sentiment text; determining an appeal object based on a keyword in the negative emotion text, and performing intention recognition processing on the negative emotion text to obtain appeal information; and performing semantic integration processing on the appeal object and the appeal information, and generating an interactive interface optimization suggestion based on a preset interface optimization strategy and a semantic integration processing result. The negative sentiment text in the complex statement can be recognized, and the accuracy of interactive interface optimization suggestions is improved.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a method, apparatus, device, and storage medium for generating interactive interface optimization suggestions. Background Technology

[0002] With the rapid popularization of mobile internet, the Internet of Things, and smart devices, users face a massive number of interactive interfaces every day. From app operation and web browsing to smart hardware control, the interface has become an essential channel for users to interact with the digital world. Therefore, in an environment of upgrading user needs, rapid technological iteration, and intensified market competition, optimizing the interface has become a necessity for survival and development. User reviews, as core data that directly reflects interface defects, service pain points, and user needs, are key evidence for optimizing interfaces and improving user experience. For example, e-commerce platforms need to optimize the display logic of product detail pages based on product reviews, and financial apps need to improve the operation process in the interface based on reviews to reduce user churn.

[0003] In related technologies, negative words about user interfaces are typically identified in user review data. An interface optimization suggestion is then generated based on keyword matching by setting up a mapping table between "negative words" and "improvement directions." However, this method has several drawbacks. It fails to recognize complex expressions such as irony and metaphors related to keywords, leading to suggestions that deviate from actual needs. Furthermore, the same keyword can have different meanings in different domains, resulting in inaccurate optimization suggestions. Summary of the Invention

[0004] This application provides a method, apparatus, device, and storage medium for generating interactive interface optimization suggestions. It solves the problem that during the generation process, the inability to recognize complex expressions such as irony and metaphor leads to suggestions that deviate from actual needs, resulting in inaccurate interactive interface optimization suggestions. By performing sentiment analysis on multiple comment texts from user comment data, it can accurately identify ironic or metaphorical statements in complex sentences, thus improving the accuracy of interactive interface optimization suggestions.

[0005] In a first aspect, embodiments of this application provide a method for generating interactive interface optimization suggestions, including: Obtain multiple comment texts associated with the interactive interface design, and perform sentiment analysis on the multiple comment texts to obtain the negative sentiment texts; The target audience is identified based on the keywords in the negative sentiment text, and the intent recognition process is performed on the negative sentiment text to obtain the appeal information. The appeal object and the appeal information are subjected to semantic integration processing, and an interactive interface optimization suggestion is generated based on a pre-set interface optimization strategy and a result of the semantic integration processing.

[0006] Optionally, the plurality of comment texts are subjected to sentiment analysis processing to obtain negative sentiment texts therefrom, including: The plurality of comment texts are subjected to semantic analysis, and negative sentiment words and degree words in each of the comment texts are extracted according to a result of the semantic analysis; The negative sentiment words and the degree words are positioned, and a positional relationship between the degree words and the negative sentiment words is determined; A negative sentiment value is evaluated according to a weighting coefficient associated with the degree words and the positional relationship, and a comment text with a negative sentiment value greater than a pre-set sentiment threshold is determined as a negative sentiment text.

[0007] Optionally, the negative sentiment words and the degree words are extracted according to a result of the semantic analysis, including: Each of the comment texts is subjected to word segmentation processing according to a result of the semantic analysis, a plurality of target keywords are obtained, and a plurality of candidate sentiment words, degree words and negative words in the plurality of target keywords are marked; A scene semantics of each of the candidate sentiment words is determined, the marked plurality of candidate sentiment words are screened according to the negative words and the scene semantics of each of the candidate words, and negative sentiment words are obtained.

[0008] Optionally, the negative sentiment words and the degree words are positioned, and the positional relationship between the degree words and the negative sentiment words is determined, including: A first position index corresponding to the negative sentiment words and a second position index corresponding to the degree words are respectively generated according to positions of the negative sentiment words and the degree words in a corresponding comment text; A word distance is calculated according to an index value of the first position index and an index value of the second position index, and in a case where the word distance is less than a pre-set word distance, a positional relationship between the degree words and the negative sentiment words is determined as an adjacent relationship.

[0009] Optionally, the negative sentiment value is evaluated according to the weighting coefficient associated with the degree words and the positional relationship, including: In a case where the positional relationship is the adjacent relationship, a positive weighting calculation is performed according to the weighting coefficient associated with the degree words and a quantity of the negative sentiment words, and a negative sentiment value is obtained; In a case where the positional relationship is a non-adjacent relationship, a reverse weighting calculation is performed according to the weighting coefficient associated with the degree words and the quantity of the negative sentiment words, and a negative sentiment value is obtained.

[0010] Optionally, the appeal object is determined based on the keyword in the negative sentiment text, and appeal information is obtained by performing intent recognition processing on the negative sentiment text, including: An entity keyword in the negative sentiment text is determined, and an appeal object matching the entity keyword is determined according to a preset dictionary library; Similarity between the negative sentiment text and each preset intent information is calculated to obtain a similarity value, and a preset intent information with the highest similarity value is determined as the appeal information of the negative sentiment text.

[0011] Optionally, the appeal object and the appeal information are subjected to semantic integration processing, and an interactive interface optimization suggestion is generated based on a preset interface optimization strategy and a result of the semantic integration processing, including: The appeal object and the appeal information are respectively subjected to expression normalization processing to obtain a target appeal object and a target appeal information; The target appeal object and the target appeal information are integrated according to a preset association logic to obtain optimized appeal information, and an interactive interface optimization suggestion matching the optimized appeal information is determined according to the preset interface optimization strategy.

[0012] In a second aspect, an embodiment of the present application provides an interactive interface optimization suggestion generation device, including: An acquisition module is configured to acquire a plurality of comment texts associated with interactive interface design; A text determination module is configured to perform sentiment analysis processing on the plurality of comment texts to obtain negative sentiment texts therefrom; An appeal determination module is configured to determine an appeal object based on a keyword in the negative sentiment text, and obtain appeal information by performing intent recognition processing on the negative sentiment text; An suggestion generation module is configured to perform semantic integration processing on the appeal object and the appeal information, and generate an interactive interface optimization suggestion based on a preset interface optimization strategy and a result of the semantic integration processing.

[0013] In a third aspect, an embodiment of the present application provides an electronic device, including: one or more processors; a storage device configured to store one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement the interactive interface optimization suggestion generation method of the first aspect.

[0014] In a fourth aspect, an embodiment of the present application provides a storage medium containing computer executable instructions, when the computer executable instructions are executed by a computer processor, the computer executable instructions are used to execute the interactive interface optimization suggestion generation method of the first aspect.

[0015] The embodiment of the present application obtains a plurality of comment texts associated with an interactive interface design, performs sentiment analysis on the plurality of comment texts to obtain negative sentiment texts, determines a solicitation object based on keywords in the negative sentiment texts, performs intent recognition on the negative sentiment texts to obtain solicitation information, performs semantic integration on the solicitation object and the solicitation information, and generates an interactive interface optimization suggestion based on a pre-set interface optimization strategy and a result of the semantic integration. In the above scheme, the sarcasm sentences or metaphor sentences in complex sentences can be accurately recognized by performing sentiment analysis on a plurality of comment texts in user comment data, and the accuracy of the interactive interface optimization suggestion is improved. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 is a flowchart of an interactive interface optimization suggestion generation method provided by the embodiment of the present application; Figure 2 is a flowchart of a negative sentiment text determination method provided by the embodiment of the present application; Figure 3 is a flowchart of a negative sentiment word and degree word determination method provided by the embodiment of the present application; Figure 4 is a flowchart of a degree word and negative sentiment word position relationship determination method provided by the embodiment of the present application; Figure 5 is a negative sentiment word and degree word labeling schematic diagram provided by the embodiment of the present application; Figure 6 is a structural schematic diagram of an interactive interface optimization suggestion generation device provided by the embodiment of the present application; Figure 7 is a structural schematic diagram of an interactive interface optimization suggestion generation device provided by the embodiment of the present application. DETAILED DESCRIPTION

[0017] To make the objectives, technical solutions and advantages of the present application clearer, the following further describes the specific embodiments of the present application with reference to the drawings. It should be understood that the specific embodiments described herein are only used to explain the present application, but not to limit the present application. In addition, it should be noted that, for the convenience of description, only parts related to the present application are shown in the drawings, but not all contents. Before discussing the example embodiments in more detail, it should be mentioned that some example embodiments are described as processes or methods depicted by flowcharts. Although the flowcharts describe the operations (or steps) as sequential processes, many of the operations can be implemented in parallel, concurrently or simultaneously. In addition, the order of the operations can be rearranged. The processes can be terminated when the operations are completed, but can also have additional steps not included in the drawings. The processes can correspond to methods, functions, procedures, subroutines, subprograms, etc.

[0018] The technical solutions in the embodiments of the present application will be clearly described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art are within the scope of protection of the present application.

[0019] The terms "first", "second", and the like in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than that illustrated or described herein, and the objects distinguished by "first", "second", etc. are generally a category and do not limit the number of objects, for example, the first object can be one or more. In addition, "and / or" in the specification and claims indicates at least one of the connected objects, and the character " / ", generally indicates that the objects before and after are in an "or" relationship.

[0020] The interactive interface optimization suggestion generation method, device, equipment and medium provided by the embodiments of the present application will be described in detail below in combination with the drawings and specific embodiments and their application scenarios.

[0021] The interactive interface optimization suggestion generation method provided by the embodiments of the present application is used in scenarios such as optimizing the game operation interface through comment text in a game platform, optimizing the detail page through product comments in an e-commerce platform, and improving the operation process in the interactive interface according to comments in a financial APP. Based on the above application scenarios, it can be understood that the execution subject of each step can be a computer device, which refers to any electronic device with data calculation, processing and storage capabilities, such as mobile phones, PC (Personal Computer), tablet computers and other terminal devices, or servers and other devices, which are not limited by the embodiments of the present application.

[0022] Figure 1 is a flowchart of an interactive interface optimization suggestion generation method provided by the embodiments of the present application, as shown in Figure 1 , comprising: Step S101, obtaining a plurality of comment texts associated with the interactive interface design, and performing sentiment analysis on the plurality of comment texts to obtain negative sentiment texts.

[0023] The interactive interface design can refer to an interface design with information interaction function, such as an operation interface of an APP, a browsing interface of a webpage, or a control interface of smart hardware, etc. The comment text can refer to text information of evaluation or suggestion of the user on the interactive interface design, which is direct feedback of the user after using the interactive interface, or can be user opinions related to the interactive interface design collected from other channels. The negative sentiment text can refer to text of dissatisfaction or complaint of the user on the interactive interface design in the comment text, which can provide an important basis for subsequent generation of interactive interface optimization suggestions.

[0024] In one embodiment, all comment texts related to the interactive interface design are collected, and sentiment analysis processing is performed on the plurality of comment texts. For example, sentiment tendency in the comment text can be identified through natural language processing technology, the comment text with problems is quickly located from a large number of user comment texts, and a basis is provided for subsequent analysis and processing.

[0025] In step S102, the appeal object is determined based on the keywords in the negative sentiment text, and the appeal information is obtained by performing intent recognition processing on the negative sentiment text.

[0026] The appeal object can refer to a specific object in the negative sentiment text, such as a function, a page layout, or an operation process, etc. The keywords can be words in the negative sentiment text that can reflect the user's emotion, which can reflect the user's dissatisfaction or complaint on which aspects of the interactive interface design. The appeal information can refer to specific content or suggestions that the user hopes to improve or optimize in the negative sentiment text.

[0027] In one embodiment, by analyzing the keywords, the target subject modified by each keyword is determined, and the target subject is determined as the appeal object of the user. By performing intent recognition processing on the negative sentiment text, such as using natural language processing technology, semantic understanding model, etc., the real intention and demand of the user are deeply understood, and key information is provided for subsequent generation of interactive interface optimization suggestions that meet the user's expectations.

[0028] In step S103, the appeal object and the appeal information are subjected to semantic integration processing, and the interactive interface optimization suggestions are generated based on the pre-set interface optimization strategy and the result of the semantic integration processing.

[0029] The semantic integration processing is to integrate and normalize the appeal object and the appeal information to form a clear and accurate optimization requirement description. The interface optimization strategy is based on past experience, industry standards, or user-developed guidelines to guide how to optimize the interface according to user feedback. The interactive interface optimization suggestions can refer to specific improvement suggestions or optimization schemes generated for user feedback problems and requirements, so as to improve the user experience of the interactive interface.

[0030] In one embodiment, the appeal object and the appeal information are semantically integrated to ensure the accuracy and consistency of the information, and the results of the semantic integration processing are combined to match corresponding optimization measures for different optimization needs. For example, if the appeal object focuses on the layout position of a certain function button, and the appeal information indicates that the user hopes that the button is more eye-catching or easy to click, the interface optimization strategy may suggest adjusting the size, color or position of the button to improve the user experience. To improve the pertinence and feasibility of the optimization suggestions, the user's historical usage behavior, preference settings and other information can be used to more comprehensively understand the user's needs, ensuring that the generated interactive interface optimization suggestions not only solve the current problem, but also meet the user's usage habits.

[0031] In another possible embodiment, a feedback mechanism can also be established to collect user evaluations of the optimized interface to continuously iterate and optimize the interface design, forming a virtuous cycle. In this way, the user experience of the interactive interface can be continuously improved, and user satisfaction and loyalty can be enhanced.

[0032] The embodiments of the present application obtain a plurality of comment texts associated with the interactive interface design, perform sentiment analysis processing on the plurality of comment texts to obtain negative sentiment texts therein; determine an appeal object based on keywords in the negative sentiment texts, and perform intent recognition processing on the negative sentiment texts to obtain appeal information; perform semantic integration processing on the appeal object and the appeal information, and generate interactive interface optimization suggestions based on pre-set interface optimization strategies and the results of the semantic integration processing. In the above scheme, the sarcasm sentences or metaphor sentences in complex sentences can be accurately identified by performing sentiment analysis on the plurality of comment texts in the user comment data, thereby improving the accuracy of the interactive interface optimization suggestions.

[0033] In one embodiment, the appeal object is determined based on the keywords in the negative sentiment text, and the appeal information is obtained by performing intent recognition processing on the negative sentiment text, including: determining entity keywords in the negative sentiment text, determining an appeal object matching the entity keywords according to a pre-set dictionary library; calculating the similarity between the negative sentiment text and each pre-set intent information to obtain a similarity value, and determining the pre-set intent information with the highest similarity value as the appeal information of the negative sentiment text.

[0034] The entity keywords can refer to words with clear reference meaning in the negative sentiment text, such as function names, page elements or operation processes, etc. These keywords can directly point to the specific object of the user feedback. The pre-set dictionary library is a database containing various possible appeal objects and their corresponding keywords. By matching the entity keywords with the information in the pre-set dictionary library, the appeal object of the user can be accurately identified.

[0035] In an embodiment, an entity keyword in the negative sentiment text is determined, and a solicitation object is determined by matching the entity keyword with information in a preset dictionary library. Similarities of the negative sentiment text and preset intention information are calculated to obtain similarity values, and preset intention information with the highest similarity value is determined as solicitation information of the negative sentiment text. For example, the negative sentiment text and the preset intention are encoded by using a pre-trained model to generate a vector containing context information, and a cosine similarity of the encoded vector is calculated. The preset intention information with the highest cosine similarity value is determined as the solicitation information of the negative sentiment text. The pre-trained model can be a BERT or RoBERTa model.

[0036] In the above scheme, the real intention and demand of the user can be accurately captured through in-depth analysis of the negative sentiment text, so that the pertinence and effectiveness of the interface optimization suggestion are improved.

[0037] In an embodiment, the solicitation object and the solicitation information are subjected to semantic integration processing, and an interactive interface optimization suggestion is generated based on a preset interface optimization strategy and a result of the semantic integration processing. The method includes: performing expression normalization processing on semantics of the solicitation object and the solicitation information respectively to obtain a target solicitation object and a target solicitation information; integrating the target solicitation object and the target solicitation information according to a preset association logic to obtain optimized solicitation information; and determining an interactive interface optimization suggestion matched with the optimized solicitation information according to the preset interface optimization strategy.

[0038] The normalization processing can refer to converting the solicitation object and the solicitation information into unified and standardized expression forms. The target solicitation object and the target solicitation information are the solicitation object and the solicitation information after the normalization processing. The preset association logic refers to rules or principles for describing relationships between the solicitation object and the solicitation information, which are formulated according to historical experience or industry standards. The optimized solicitation information is information obtained by integrating the target solicitation object and the target solicitation information, and can more comprehensively, directly and standardly reflect the demand and problem of the user.

[0039] In an embodiment, the appeal object and the appeal information are semantically normalized, such as unified into standard terms or expressions, to ensure the accuracy and consistency of the information. The target appeal object and the target appeal information are integrated into a standard appeal representation sentence according to a preset association logic, that is, the appeal information is optimized, such as the association structure of "target appeal object" + "target appeal information". The interactive interface optimization suggestion matched with the optimized appeal information is determined according to the pre-set interface optimization strategy, such as the interface layout principle and the interaction design criterion. For example, if the optimized appeal information is "a certain function button is not clear", the interface optimization strategy may suggest increasing the brightness or contrast of the button, or using animation effects to make the button clearer.

[0040] In the above scheme, the appeal object and the appeal information are normalized, which facilitates the quick and accurate matching of the corresponding interface optimization strategy, and improves the accuracy of the interactive interface optimization suggestion.

[0041] Figure 2 is a flowchart of a negative sentiment text determination method provided by an embodiment of the present application, as shown in Figure 2 , comprising: Step S1011, performing semantic analysis on the plurality of comment texts, and extracting negative sentiment words and degree words in each comment text according to the semantic analysis result.

[0042] The semantic analysis can refer to deep semantic understanding of the comment text, so as to accurately understand the meaning of the comment text and the meaning of each word in the comment text, and facilitate accurate extraction of the required words from the comment text. The negative sentiment words usually represent the dissatisfaction or complaint of the user, such as "terrible" and "disappointed", and the degree words are used to modify these negative sentiments to indicate the intensity, such as "very" and "extremely".

[0043] In an embodiment, the deep learning model, such as the recurrent neural network or the Transformer, can be used to model the end-to-end mapping between the text and the logical form, and the semantic association can be captured through the attention mechanism to determine the meaning of the comment text and the meaning of each word in the comment text. The negative sentiment words and the degree words are extracted based on the word meaning. In the extraction of the negative sentiment words and the degree words, the morphological analysis technology in natural language processing can be used, combined with the preset sentiment dictionary and the degree dictionary, to scan and match the comment text word by word, so as to quickly and accurately identify and extract the negative sentiment expression words and the degree words in the text.

[0044] Step S1012, locating the negative sentiment words and the degree words, and determining the positional relationship between the degree words and the negative sentiment words.

[0045] In an embodiment, the negative sentiment words and the degree words in the review text are respectively marked with different labels, the relative positional relationship between the negative sentiment words and the degree words in the review text is determined according to the marking order of different types of labels, and the positional relationship between the negative sentiment words and the degree words is analyzed, such as whether the degree words are immediately before or after the negative sentiment words, or whether other words are inserted between them.

[0046] Step S1013, evaluating the negative sentiment value according to the weighting coefficient associated with the degree word and the positional relationship, and determining the review text with a negative sentiment value greater than a preset sentiment threshold as a negative sentiment text.

[0047] The degree word can enhance or weaken the sentiment intensity of the negative sentiment word, so when calculating the sentiment score, the influence of the degree word on the sentiment intensity needs to be considered, and different degrees of degree words correspond to different weighting coefficients. The positional relationship can include adjacent positional relationship and non-adjacent positional relationship, the adjacent positional relationship can represent that the degree word is a word used to modify the negative sentiment word, and the non-adjacent positional relationship can represent that the degree word is not a word used to modify the negative sentiment word. The preset sentiment threshold can be set according to actual needs, which represents the boundary between positive sentiment and negative sentiment. By comparing the negative sentiment value of the review text with the preset threshold, it can be judged whether the text belongs to negative sentiment text. The sentiment intensity can represent the negative degree or the influence size on the user of the negative sentiment word. In order to perform more accurate sentiment analysis on the review text, a sentiment intensity value can be set for each negative sentiment word, which can be obtained in advance according to the definition of the word in the sentiment dictionary or based on historical data statistical analysis.

[0048] In an embodiment, the target weighting coefficient is determined according to the preset mapping relationship between the degree word and the weighting coefficient, if the positional relationship is the adjacent relationship, the sentiment value corresponding to the adjacent negative sentiment word is calculated according to the target weighting coefficient and the sentiment value corresponding to the adjacent negative sentiment word, and the sentiment values corresponding to all sentiment words in the review text are summed to obtain the total negative sentiment value. The total negative sentiment values corresponding to each review text are compared with the preset sentiment threshold respectively, and the review text with a negative sentiment value greater than the preset sentiment threshold is determined as a negative sentiment text.

[0049] The embodiment of the application extracts negative sentiment words and degree words in each comment text according to the semantic analysis result, positions the negative sentiment words and the degree words, determines the positional relationship between the degree words and the negative sentiment words, evaluates the negative sentiment value according to the weighting coefficient associated with the degree words and the positional relationship, and determines the comment text with a negative sentiment value greater than a preset sentiment threshold as a negative sentiment text. In the above scheme, the negative emotions in the user comments can be accurately captured through fine-grained sentiment analysis, and the negative sentiment is quantitatively evaluated in combination with the influence of the degree words, which not only improves the accuracy of sentiment analysis, but also provides more accurate data support for the generation of subsequent interface optimization suggestions. By comprehensively considering the positional relationship between the negative sentiment words and the degree words, the accuracy of sentiment analysis is further enhanced, ensuring that the real emotional tendency of the user can be accurately identified in complex sentences, laying a foundation for the generation of interface optimization suggestions, and helping to generate optimization schemes that better meet the user's expectations, thereby improving the user experience.

[0050] In one embodiment, the negative sentiment value evaluation according to the weighting coefficient associated with the degree word and the positional relationship includes: in the case of the adjacent relationship of the positional relationship, the positive weighting calculation is performed according to the weighting coefficient associated with the degree word and the number of negative sentiment words to obtain the negative sentiment value; in the case of the non-adjacent relationship of the positional relationship, the reverse weighting calculation is performed according to the weighting coefficient associated with the degree word and the number of negative sentiment words to obtain the negative sentiment value.

[0051] Wherein, the positive weighting can refer to giving higher weight to the negative sentiment word under the adjacent relationship. The reverse weighting refers to giving lower weight to the negative sentiment word under the non-adjacent relationship. Since the interval between the degree word and the negative sentiment word may mean that the correlation between them is weak, or other words have an impact on the negative sentiment, resulting in the emotional intensity of the negative sentiment word being weakened, therefore the number of negative sentiment words is calculated by reverse weighting.

[0052] In one embodiment, the number of negative sentiment words in the negative sentiment text is first determined, in the case of the adjacent relationship of the positional relationship, the number of negative sentiment words is calculated by positive weighting according to the weighting coefficient associated with the degree word to obtain the negative sentiment value of the negative sentiment text. Exemplarily, if the number of negative sentiment words is 3 and the weighting coefficient associated with the sentiment word is 0.6, then the negative sentiment value = 3 * (1 + 0.6) = 4.8. In the case of the non-adjacent relationship of the positional relationship, the number of negative sentiment words is calculated by reverse weighting according to the weighting coefficient associated with the degree word to obtain the negative sentiment value of the negative sentiment text. Exemplarily, if the number of negative sentiment words is 3 and the weighting coefficient associated with the sentiment word is 0.6, then the negative sentiment value = 3 * 0.6 = 1.8.

[0053] In a case where the position relation is an adjacent relation, the degree word associated weighting coefficient and the number of negative sentiment words are used for forward weighting calculation to obtain a negative sentiment value. In a case where the position relation is a non-adjacent relation, the degree word associated weighting coefficient and the number of negative sentiment words are used for reverse weighting calculation to obtain a negative sentiment value. In the above scheme, the sentiment intensity of different negative sentiment words can be distinguished more finely, and the evaluation of the negative sentiment value is more accurate.

[0054] Figure 3 is a flowchart of a negative sentiment word and degree word determination method provided by an embodiment of the present application, as shown in Figure 3 , comprising: Step S10111: performing word segmentation processing on each comment text according to a semantic analysis result to obtain a plurality of target keywords, and marking a plurality of candidate sentiment words, degree words and negative words in the plurality of target keywords.

[0055] The candidate sentiment words can refer to words that can express emotions in the comment text. These words can be positive or negative. The negative words are used to express the opposite or negative meaning, such as "not", "no", etc. They can reverse the sentiment tendency of the sentiment words.

[0056] In an embodiment, the meanings of each word in the comment text are determined according to the semantic analysis result, and the word segmentation processing is performed on each comment text to obtain a plurality of independent words, i.e., a plurality of target keywords, which are the basic units of the comment text. In the word segmentation processing, the comment text can be cut into independent words by using the word segmentation technology in natural language processing. The target keywords are matched with a preset sentiment dictionary to determine a plurality of candidate keywords that can express emotions. Similarly, the remaining words after determining the candidate sentiment words are matched with a preset degree word and a preset negative word to determine the degree word and the negative word in the comment text, and different types of identifiers are used to mark the candidate sentiment words, the degree words and the negative words.

[0057] Step S10112: determining the scene semantics of each candidate sentiment word, and screening the marked plurality of candidate sentiment words according to the negative words and the scene semantics of each candidate word to obtain negative sentiment words.

[0058] In one embodiment, since the same word can represent different emotional tendencies in different scenarios, the negative emotional words can be determined by screening the multiple candidate emotional words according to the scenario semantics of each candidate emotional word and the negation word. For example, the word "slow" can represent a negative emotion when describing the speed of system startup, meaning that the startup speed is too slow and the user experience is poor, but can represent a positive emotion when describing a soft music rhythm, meaning that the music is slow and peaceful. Therefore, when determining the negative emotional words, the context scenario in which the emotional word is located needs to be considered, and the candidate emotional words are screened in combination with the scenario semantics. If the scenario semantics in which the candidate emotional word is located is associated with a negative emotion, and the emotional word is not reversed by the negation word, the candidate emotional word is determined to be a negative emotional word.

[0059] According to the semantic analysis result, the embodiments of the present application perform word segmentation processing on each comment text to obtain multiple target keywords, mark multiple candidate emotional words, degree words and negation words in the multiple target keywords, determine the scenario semantics of each candidate emotional word, screen the marked multiple candidate emotional words according to the negation word and the scenario semantics of each candidate word, and obtain negative emotional words. In the above scheme, the context information in the comment text can be considered comprehensively, the negative emotional words can be accurately identified, the misjudgment caused by single word matching can be avoided, and the accuracy of the emotional analysis is improved. By introducing the processing of scenario semantics and negation words, the accuracy of negative emotional word recognition is enhanced, Figure 4 is a flowchart of a method for determining the positional relationship between a degree word and a negative emotional word provided by the embodiments of the present application, as shown in Figure 4 , comprising: Step S10121, generating a first position index corresponding to the negative emotional word and a second position index corresponding to the degree word according to the positions of the negative emotional word and the degree word in the corresponding comment text.

[0060] Step S10122, calculating the word distance according to the index value of the first position index and the index value of the second position index, and determining that the positional relationship between the degree word and the negative emotional word is adjacent relationship in the case that the word distance is less than a preset word distance.

[0061] The first position index can refer to the specific position information of the negative emotional word in the comment text, and the second position index represents the specific position information of the degree word in the comment text. The word distance refers to the relative distance of the negative emotional word and the degree word in the text, and the difference between the position indexes of the two can be obtained. The preset word distance is a threshold value set according to experience or statistical results, which is used to determine whether the negative emotional word and the degree word are adjacent. If the calculated word distance is the same as or similar to the preset word distance, it is considered that the degree word and the negative emotional word are adjacent, that is, the degree word is used to modify the negative emotional word.

[0062] For example, the position index of the marked comment text is: the word sequence of the sentence "clothes quality is particularly poor. Logistics is super slow" is [clothes, quality, particularly, poor,., logistics, super, slow], the index value sequence is [1, 2, 3, 4, 5, 6, 7], the first position index value corresponding to each negative sentiment word is pos_neg=3, pos_neg=7 respectively, and the second position index value corresponding to each degree word is pos_deg=2, pos_deg=6 respectively. Figure 5 is a marking diagram of a negative sentiment word and a degree word provided by an embodiment of the present application, as shown in Figure 5 The first position of the negative sentiment word and the second position of the degree word are marked by different markers. The preset word distance is set to 3, the word distance d = |pos_deg - pos_neg| is calculated, and in the case that the word distance is less than 3, the position relationship between the degree word and the negative sentiment word is determined to be adjacent, such as "particularly poor" d=1, "super slow" d=1.

[0063] According to the positions of the negative sentiment word and the degree word in the corresponding comment text, the first position index corresponding to the negative sentiment word and the second position index corresponding to the degree word are generated; the word distance is calculated according to the index value of the first position index and the index value of the second position index, and in the case that the word distance is the same as the preset word distance, the position relationship between the degree word and the negative sentiment word is determined to be adjacent. In the above scheme, the position relationship between the negative sentiment word and the degree word can be determined to accurately judge whether the degree word is used to modify the negative sentiment word, so as to more accurately evaluate the negative sentiment intensity. By calculating the word distance to determine the relationship between the negative sentiment word and the degree word, the problem of inaccurate sentiment analysis caused by misjudgment of the position relationship is avoided.

[0064] Figure 6 is a structural diagram of an interactive interface optimization suggestion generation device provided by an embodiment of the present application, as shown in Figure 6 It includes: The acquisition module 21 is configured to acquire a plurality of comment texts associated with the interactive interface design. The text determination module 22 is configured to perform sentiment analysis processing on the plurality of comment texts to obtain negative sentiment texts therefrom. The appeal determination module 23 is configured to determine an appeal object based on a keyword in the negative sentiment text, and perform intent recognition processing on the negative sentiment text to obtain appeal information. The suggestion generation module 24 is configured to perform semantic integration processing on the appeal object and the appeal information, generate an interactive interface optimization suggestion based on a pre-set interface optimization strategy and a result of the semantic integration processing.

[0065] The embodiment of the application obtains a plurality of comment texts associated with an interactive interface design, performs sentiment analysis on the plurality of comment texts to obtain negative sentiment texts, determines a solicitation object based on keywords in the negative sentiment texts, performs intent recognition on the negative sentiment texts to obtain solicitation information, performs semantic integration on the solicitation object and the solicitation information, and generates an interactive interface optimization suggestion based on a pre-set interface optimization strategy and a result of the semantic integration. In the above scheme, the sarcasm or metaphor in a complex sentence can be accurately recognized by performing sentiment analysis on a plurality of comment texts in user comment data, and the accuracy of the interactive interface optimization suggestion is improved.

[0066] In one possible embodiment, the text determination module 22 is specifically configured to: perform semantic analysis on the plurality of comment texts, and extract negative sentiment words and degree words in each of the comment texts according to a result of the semantic analysis; position the negative sentiment words and the degree words, and determine a positional relationship between the degree words and the negative sentiment words; evaluate a negative sentiment value according to a weighting coefficient associated with the degree words and the positional relationship, and determine a comment text with a negative sentiment value greater than a preset sentiment threshold as a negative sentiment text.

[0067] In one possible embodiment, the text determination module 22 is specifically configured to: perform word segmentation on each of the comment texts according to a result of the semantic analysis, obtain a plurality of target keywords, and mark a plurality of candidate sentiment words, degree words and negative words in the plurality of target keywords; determine scene semantics of each of the candidate sentiment words, filter the marked plurality of candidate sentiment words according to the negative words and the scene semantics of each of the candidate words, and obtain negative sentiment words.

[0068] In one possible embodiment, the text determination module 22 is specifically configured to: generate a first position index corresponding to the negative sentiment words and a second position index corresponding to the degree words according to positions of the negative sentiment words and the degree words in the corresponding comment texts; calculate a word distance according to an index value of the first position index and an index value of the second position index, and determine that the positional relationship between the degree words and the negative sentiment words is an adjacent relationship in a case where the word distance is less than a preset word distance.

[0069] In one possible embodiment, the text determination module 22 is specifically configured to: in a case where the positional relationship is the adjacent relationship, perform forward weighting calculation according to a weighting coefficient associated with the degree words and a quantity of the negative sentiment words to obtain a negative sentiment value; In a case where the position relation is a non-adjacent relation, a negative sentiment value is obtained by inversely weighting according to a weighting coefficient associated with the degree word and the number of negative sentiment words.

[0070] In one possible implementation, the appeal determination module 23 is specifically configured to: determine an entity keyword in the negative sentiment text, determine an appeal object matching the entity keyword according to a preset dictionary library; calculate a similarity between the negative sentiment text and each preset intention information to obtain a similarity value, and determine the preset intention information with the highest similarity value as the appeal information of the negative sentiment text.

[0071] In one possible implementation, the suggestion generation module 24 is specifically configured to: perform semantic normalization processing on the appeal object and the appeal information respectively to obtain a target appeal object and a target appeal information; integrate the target appeal object and the target appeal information according to a preset association logic to obtain optimized appeal information, and determine an interactive interface optimization suggestion matching the optimized appeal information according to a pre-set interface optimization strategy.

[0072] Embodiments of the present application further provide an electronic device, and the interactive interface optimization suggestion generation device can integrate the interactive interface optimization suggestion generation apparatus provided in the embodiments of the present application. Figure 6 FIG. 1 is a structural schematic diagram of an interactive interface optimization suggestion generation device provided in an embodiment of the present application, which will be described below with reference to Figure 6 The interactive interface optimization suggestion generation device includes an input apparatus 33, an output apparatus 34, a memory 32, and one or more processors 31; the memory 32 is configured to store one or more programs; when the one or more programs are executed by the one or more processors 31, the one or more processors 31 implement the interactive interface optimization suggestion generation method provided in the above-described embodiments. The input apparatus 33, the output apparatus 34, the memory 32, and the processor 31 can be connected through a bus or other means, Figure 6 for example, through the bus in the embodiment.

[0073] The memory 32 can be used to store software programs, computer executable programs and modules, such as program instructions / modules of the method for generating interaction interface optimization suggestions provided by any of the embodiments. The memory 32 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system and at least one application required by a function; and the data storage area can store data created according to the use of the device. In addition, the memory 32 can include a high-speed random access memory, and can also include a non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other non-volatile solid-state memory device. In some examples, the memory 32 can further include a memory remotely arranged with respect to the processor 31, which can be connected to the device through a network. Examples of the network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0074] The input device 33 can be used to receive input digital or character information, and to generate key signal inputs related to user settings and function controls of the device. The output device 34 can include a display device such as a display screen.

[0075] The processor 31 executes various function applications and data processing of the device by running software programs, instructions and modules stored in the memory 32, that is, implements the method for generating interaction interface optimization suggestions described above.

[0076] The device and the computer provided above can be used to execute the method for generating interaction interface optimization suggestions provided by any of the embodiments, and have corresponding functions and beneficial effects.

[0077] The embodiments of the present application also provide a storage medium storing computer executable instructions, which, when executed by a computer processor, are used to execute the method for generating interaction interface optimization suggestions provided by the above embodiments, and the method includes: Obtaining a plurality of comment texts associated with interaction interface design, and performing sentiment analysis on the plurality of comment texts to obtain negative sentiment texts; Determining a solicitation object based on keywords in the negative sentiment texts, and performing intent recognition on the negative sentiment texts to obtain solicitation information; Performing semantic integration processing on the solicitation object and the solicitation information, and generating interaction interface optimization suggestions based on a pre-set interface optimization strategy and a result of the semantic integration processing.

[0078] Storage medium - any type of memory device or storage device. The term "storage medium" is intended to include an installation medium, e.g., a CD-ROM, floppy disks, or tape apparatus; computer system memory or random access memory such as DRAM, DDR RAM, SRAM, EDO RAM, Rambus RAM, etc.; or a non-volatile memory such as a magnetic medium (e.g., a hard drive or optical storage); registers or other similar types of memory elements, etc. The memory medium can also include other types of storage medium and combinations thereof. In addition, the memory medium can reside in a first computer system's internal storage or external storage that is connected to the first computer system. The first computer system can also provide the program instructions to a second computer system for execution. The term "storage medium" can also include two or more memory mediums that can reside in different locations, e.g., in different computer systems that are connected over a network such as the Internet. The memory medium can store program instructions that implement the program instructions (e.g., as a computer program) that can be executed by one or more processors.

[0079] Of course, the storage medium provided by the embodiments of the present application includes computer executable instructions, and the computer executable instructions are not limited to the interactive interface optimization suggestion generation method as described above, but can also perform the related operations in the interactive interface optimization suggestion generation method provided by any of the embodiments of the present application.

[0080] The interactive interface optimization suggestion generation apparatus, device and storage medium provided in the above embodiments can execute the interactive interface optimization suggestion generation method provided by any of the embodiments of the present application, and the technical details not described in detail in the above embodiments can be referred to the interactive interface optimization suggestion generation method provided by any of the embodiments of the present application.

[0081] The above are only the preferred embodiments of the present application and the technical principles used. The present application is not limited to the specific embodiments described herein, and various obvious changes, readjustments and replacements made by those skilled in the art will not deviate from the protection scope of the present application. Therefore, although the present application has been described in more detail through the above embodiments, the present application is not limited to the above embodiments, and can include more other equivalent embodiments without deviating from the concept of the present application, and the scope of the present application is determined by the scope of the claims.

Claims

1. A method for generating interactive interface optimization suggestions, characterized in that, The method comprises the following steps: obtaining a plurality of comment texts associated with an interactive interface design, performing sentiment analysis on the plurality of comment texts to obtain negative sentiment texts therefrom; determining a solicitation object based on keywords in the negative sentiment texts, and performing intent recognition on the negative sentiment texts to obtain solicitation information; performing semantic integration processing on the solicitation object and the solicitation information, and generating an interactive interface optimization suggestion based on a pre-set interface optimization strategy and a result of the semantic integration processing.

2. The method of claim 1, wherein, The sentiment analysis on the plurality of comment texts to obtain negative sentiment texts therefrom comprises: performing semantic analysis on the plurality of comment texts, extracting negative sentiment words and degree words from each of the comment texts according to a result of the semantic analysis; positioning the negative sentiment words and the degree words, and determining a positional relationship between the degree words and the negative sentiment words; evaluating a negative sentiment value according to a weighting coefficient associated with the degree words and the positional relationship, and determining a comment text with a negative sentiment value greater than a pre-set sentiment threshold as a negative sentiment text.

3. The method of claim 2, wherein, The extraction of negative sentiment words and degree words from each of the comment texts according to a result of the semantic analysis comprises: performing word segmentation processing on each of the comment texts according to a result of the semantic analysis, obtaining a plurality of target keywords, and marking a plurality of candidate sentiment words, degree words and negative words in the plurality of target keywords; determining a scene semantics of each of the candidate sentiment words, screening the marked candidate sentiment words according to the negative words and the scene semantics of each of the candidate words, and obtaining negative sentiment words.

4. The method of claim 2, wherein, The positioning of the negative sentiment words and the degree words, and the determination of the positional relationship between the degree words and the negative sentiment words comprise: generating a first position index corresponding to the negative sentiment words and a second position index corresponding to the degree words according to positions of the negative sentiment words and the degree words in a corresponding comment text; calculating a word distance according to an index value of the first position index and an index value of the second position index, and determining that the positional relationship between the degree words and the negative sentiment words is an adjacent relationship in a case where the word distance is less than a pre-set word distance.

5. The method of claim 2, wherein, The evaluation of the negative sentiment value according to the weighting coefficient associated with the degree words and the positional relationship comprises: in a case where the positional relationship is the adjacent relationship, performing forward weighting calculation according to the weighting coefficient associated with the degree words and a number of the negative sentiment words to obtain a negative sentiment value; in a case where the positional relationship is a non-adjacent relationship, performing reverse weighting calculation according to the weighting coefficient associated with the degree words and the number of the negative sentiment words to obtain a negative sentiment value.

6. The method of claim 1-5, wherein, The determination of a solicitation object based on keywords in the negative sentiment texts, and the performance of intent recognition on the negative sentiment texts to obtain solicitation information comprise: determining an entity keyword in the negative sentiment texts, and determining a solicitation object matching the entity keyword according to a pre-set dictionary; calculating a similarity between the negative sentiment texts and each of pre-set intent information to obtain a similarity value, and determining a pre-set intent information with the highest similarity value as solicitation information of the negative sentiment texts.

7. The method of claim 1-5, wherein, The semantic integration processing is performed on the appeal object and the appeal information, an interactive interface optimization suggestion is generated based on a pre-set interface optimization strategy and a result of the semantic integration processing, and the interactive interface optimization suggestion includes: The semantic normalization processing is performed on the appeal object and the appeal information respectively to obtain a target appeal object and a target appeal information; The target appeal object and the target appeal information are integrated according to a pre-set association logic to obtain optimized appeal information, and an interactive interface optimization suggestion matched with the optimized appeal information is determined according to a pre-set interface optimization strategy.

8. An interactive interface optimization suggestion generation apparatus, characterized by, The method comprises the steps of: An acquisition module is configured to acquire a plurality of comment texts associated with interactive interface design; A text determination module is configured to perform sentiment analysis processing on the plurality of comment texts to obtain negative sentiment texts therefrom; An appeal determination module is configured to determine an appeal object based on keywords in the negative sentiment texts and to perform intent recognition processing on the negative sentiment texts to obtain appeal information; An suggestion generation module is configured to perform semantic integration processing on the appeal object and the appeal information, and to generate an interactive interface optimization suggestion based on a pre-set interface optimization strategy and a result of the semantic integration processing.

9. An electronic device, comprising: The device comprises one or more processors; a storage device configured to store one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors are caused to implement the interactive interface optimization suggestion generation method according to any one of claims 1-7.

10. A storage medium storing computer-executable instructions, wherein: The computer executable instructions, when executed by a computer processor, are used to perform the interactive interface optimization suggestion generation method according to any one of claims 1-7.

Images