A dialogue method, device, equipment and medium

Abstract

The present application relates to the field of data processing, and in particular to a dialogue method, device, equipment and medium, the method comprising: when a first user initiates a chat request for a second user, user features of both parties can be instantly obtained, and for each anchor point, it is determined whether the feature corresponding to the anchor point in the first user feature and the feature corresponding to the anchor point in the second user feature match, if so, it is determined that the anchor point matches successfully, otherwise, it is determined that the anchor point matches unsuccessfully; each anchor point is traversed in turn to obtain a matching anchor point that matches successfully; and after determining the matching anchor point, a plurality of candidate topic texts are first screened out, and then the candidate texts are personalized and sorted according to the features of both users using a sorting model to determine the optimal topic text, so that the efficiency of determining the final topic text is higher, and the optimal topic text is used as the opening words, thereby effectively promoting the smooth development of the dialogue between the two parties, improving the efficiency of starting the dialogue, and making the communication between users more natural and smooth.

Description

Technical Field

[0001] The present application relates to the field of data processing, and in particular to a dialogue method, device, equipment and medium. Background Art

[0002] In today's digital age, social networks have become an important medium for people to communicate and interact with each other. However, first interactions with strangers are often awkward, and users often feel unsure how to start a conversation or keep it going when chatting privately. This lack of confidence and interaction skills is particularly prevalent among young and novice users.

[0003] Icebreaking plays an important role in interpersonal communication, and successful icebreaking can significantly promote smooth communication. However, existing social networking platforms often lack effective tools to help users quickly and naturally start private chats. In this case, users have to figure out icebreaking topics on their own or use external templates, which not only increases the psychological pressure on users, but also makes the conversation seem less sincere and natural.

[0004] There are several major problems with current social software: a. Lack of icebreaker tools: Due to the lack of tool help, users often feel confused and at a loss when starting private chats. b. Insufficient personalization: Even if some platforms provide icebreaker topics, these topics are usually general templates, lacking personalization and pertinence, and are difficult to effectively arouse the other party's interest. c. Poor user experience: Due to the lack of suitable icebreaker tools or personalized suggestions, users have a poor experience in private chats, which can easily lead to conversation interruptions or embarrassment, thereby reducing user satisfaction and platform activity. d. Low interaction efficiency: Conversations that are not well started are often difficult to interact in depth, which in turn affects the emotional connection between users and the expansion of social networks.

[0005] Therefore, an intelligent private chat opening remarks system is needed to help users easily break the ice in an intelligent and personalized way, increase the fun and effectiveness of private chats, and thus improve the user satisfaction and engagement of the overall social networking platform. Summary of the invention

[0006] The purpose of this application is to provide a dialogue method, device, equipment and medium, which can automatically determine the optimal topic copy as the opening words, thereby effectively promoting the smooth development of the dialogue between the two parties, improving the efficiency of starting the dialogue, and making the communication between users more natural and smooth.

[0007] In a first aspect, a dialogue method is provided, comprising:

[0008] When the first user initiates a chat request for the second user, obtaining a first user feature corresponding to the first user and a second user feature corresponding to the second user;

[0009] Anchor point matching is performed on the first user feature and the second user feature to obtain a matching anchor point with successful matching; wherein the anchor point is a category of a feature, each anchor point corresponds to multiple topics, and each topic corresponds to at least one topic copy; wherein, for a certain anchor point, it is determined whether the feature corresponding to the anchor point in the first user feature matches the feature corresponding to the anchor point in the second user feature, if so, it is determined that the anchor point is matched successfully, otherwise, it is determined that the anchor point is matched unsuccessfully; each anchor point is traversed in turn to obtain a matching anchor point with successful matching;

[0010] According to the matching anchor point, multiple candidate topic copies corresponding to the matching anchor point are determined from a copy library; based on the first user feature and the second user feature, the multiple candidate topic copies are sorted using a sorting model to obtain a copy order; according to the copy order, an optimal topic copy is determined from the multiple candidate topic copies; the optimal topic copy is sent to a client device corresponding to the first user; wherein the optimal topic copy can be used as an opening remark for a conversation between the first user and the second user.

[0011] In a preferred example, the present application may be further configured as follows: determining, according to the matching anchor point, a plurality of candidate topic documents corresponding to the matching anchor point from a document library, including:

[0012] According to the matching anchor point, determining from a document library all first topic documents corresponding to the matching anchor point;

[0013] According to the user preference information corresponding to all the first topic texts, a plurality of candidate topic texts are determined from the first topic texts, wherein the user preference information includes: the user's selection information for the first topic text and the other user's reply information for the first topic text.

[0014] In a preferred example, the present application may be further configured as follows: according to the user preference information corresponding to each of all the first topic texts, a plurality of candidate topic texts are determined from all the first topic texts, including:

[0015] Determine the number of selections according to the selection information of the user for the first topic copy, and determine, according to the number of selections, the copy whose number of selections exceeds a preset number threshold as the first candidate topic copy;

[0016] According to the reply information of the other user to the first topic copy, determine the second candidate topic copy from all the first topic copies, wherein the second candidate topic copy is a copy in the reply information whose feedback time is less than a preset time and / or the user can reply;

[0017] Determine a plurality of candidate topic copywritings according to the first candidate topic copywriting and / or the second candidate topic copywriting;

[0018] or,

[0019] Determine the evaluation index corresponding to each first topic text according to the user preference information, wherein the evaluation index includes the selection rate, the reply rate and the conversation time information, wherein the conversation time information is determined according to the duration, the minimum duration and the maximum duration corresponding to each first topic text;

[0020] Determine the value corresponding to each first topic copy according to the selection rate, response rate, conversation time information and weight information;

[0021] The first topic copywriting with the highest preset number of numerical rankings is used as the candidate topic copywriting.

[0022] In a preferred example, the present application may be further configured as follows: the ranking model includes an XGBoost ranking model;

[0023] The training process of the XGBoost ranking model includes:

[0024] Acquire historical data corresponding to each of the plurality of users, the historical data of the users including: user registration information and user historical behavior information of the users with respect to the plurality of candidate topic copywritings;

[0025] The XGBoost model is trained according to the historical data corresponding to multiple users to obtain the XGBoost sorting model.

[0026] In a preferred example, the present application can be further configured as follows:

[0027] Get multiple anchor points;

[0028] Determine a topic copywriting set corresponding to each of the plurality of anchor points, wherein the topic copywriting set includes topic copywriting corresponding to each of the plurality of topics;

[0029] A copy library is generated according to the topic copy sets corresponding to the multiple anchor points.

[0030] In a preferred example, the present application can be further configured to: obtain multiple anchor points, including:

[0031] Acquire user information of multiple users, the user information including: user registration information and user behavior information;

[0032] Extract features based on user information corresponding to multiple users to obtain features corresponding to each of the multiple users;

[0033] Feature classification is performed based on the features corresponding to each of the multiple users to determine the multiple anchor points.

[0034] In a preferred example, the present application may be further configured as follows: further comprising at least one of the following;

[0035] Regularly obtain the latest user information of the target user, and update the user information corresponding to the target user based on the latest user information;

[0036] Regularly obtain feedback information on the topic copy for the target topic, and update the topic copy for the target topic based on the feedback information.

[0037] In a second aspect, a conversation device is provided, comprising:

[0038] An acquisition module, configured to acquire a first user feature corresponding to the first user and a second user feature corresponding to the second user when the first user initiates a chat request to the second user;

[0039] A matching module is used to perform anchor point matching on the first user feature and the second user feature to obtain a matching anchor point with successful matching; wherein the anchor point is a category of a type of feature, each anchor point corresponds to multiple topics, and each topic corresponds to at least one topic copy; wherein, for a certain anchor point, it is determined whether the feature corresponding to the anchor point in the first user feature matches the feature corresponding to the anchor point in the second user feature, if so, it is determined that the anchor point matches successfully, otherwise, it is determined that the anchor point matches unsuccessfully; each anchor point is traversed in turn to obtain a matching anchor point with successful matching;

[0040] A screening module is used to determine, from a document library, a plurality of candidate topic documents corresponding to the matching anchor point according to the matching anchor point; sort the plurality of candidate topic documents using a sorting model based on the first user feature and the second user feature to obtain a document sequence; and determine an optimal topic document from the plurality of candidate topic documents according to the document sequence;

[0041] A sending module is used to send the optimal topic copy to a client device corresponding to the first user; wherein the optimal topic copy can be used as an opening statement for a conversation between the first user and a second user.

[0042] In a third aspect, an electronic device is provided, including:

[0043] one or more processors;

[0044] Memory;

[0045] One or more applications, wherein one or more applications are stored in a memory and configured to be executed by one or more processors, and the one or more programs are configured to: perform operations corresponding to the dialogue method shown in any possible implementation method of the first aspect.

[0046] In a fourth aspect, a computer-readable storage medium is provided, wherein the storage medium stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, at least one program, a code set or an instruction set is loaded by a processor and executes the steps of the dialogue method shown in any possible implementation method of the first aspect.

[0047] In a fifth aspect, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements operations corresponding to the dialogue method shown in any possible implementation manner in the first aspect.

[0048] In summary, the method provided by this application includes the following beneficial technical effects:

[0049] When a first user initiates a chat request to a second user, the user features of both parties can be instantly obtained, and the common interests or related points of both parties can be quickly found through the anchor point matching technology. Specifically, for each anchor point, it is determined whether the features corresponding to the anchor point in the first user's features and the features corresponding to the anchor point in the second user's features match. If so, it is determined that the anchor point match is successful, otherwise, it is determined that the anchor point match fails; each anchor point is traversed in turn to obtain a matching anchor point that is successfully matched; and after determining the matching anchor point, a plurality of candidate topic texts are first screened out, and then the candidate texts are personalized and sorted according to the features of both users using a ranking model to determine the optimal topic text, so that the efficiency of determining the final topic text is higher, and the optimal topic text is used as the opening words, thereby effectively promoting the smooth development of the dialogue between the two parties, improving the efficiency of starting the dialogue, and making the communication between users more natural and smooth.

[0050] In addition, the present application also provides a dialogue device, equipment and medium, all of which have the above-mentioned beneficial technical effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the embodiments of the present application or the technical solutions of the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0052] Figure 1 This is a schematic diagram of an application scenario of a conversation method provided in an embodiment of the present application;

[0053] Figure 2 This is a flow chart of a conversation method provided by an embodiment of the present application;

[0054] Figure 3 It is a flowchart of a data processing provided by an embodiment of the present application;

[0055] Figure 4 is a structural diagram of a conversation device provided in an embodiment of the present application;

[0056] Figure 5 It is a structural schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0057] This specific embodiment is merely an explanation of the present application and is not a limitation of the present application. After reading this specification, a person skilled in the art may make non-creative modifications to the present embodiment as needed, but such modifications are protected by the patent law as long as they are within the scope of the present application.

[0058] It should be noted that in the optional embodiments of the present application, the object information and other related data involved, when the embodiments in the present application are applied to specific products or technologies, need to obtain the permission or consent of the object, and the collection, use and processing of the relevant data need to comply with the relevant laws, regulations and standards of the relevant countries and regions. In other words, if the embodiments of the present application involve data related to the object, it needs to be obtained with the authorization and consent of the object, the authorization and consent of the relevant departments, and in compliance with the relevant laws, regulations and standards of the country and region. If personal information is involved in the embodiments, the acquisition of all personal information needs to obtain the consent of the individual. If sensitive information is involved, the separate consent of the information subject needs to be obtained. The embodiments also need to be implemented with the authorization and consent of the object.

[0059] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0060] In addition, the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article, unless otherwise specified, generally means that the associated objects before and after are in an "or" relationship.

[0061] An embodiment of the present application provides a conversation method, which relates to the fields of artificial intelligence, intelligent friendship, etc.

[0062] In order to better understand the solution provided by the embodiment of the present application, the solution is described below in conjunction with a specific application scenario. The embodiment of the present application provides an intelligent private chat auxiliary opening speech generation system, which realizes the function of personalized and real-time generation of private chat opening speeches through multiple user data analysis and intelligent algorithms.

[0063] In one embodiment, please refer to Figure 1 , Figure 1 A scenario diagram of an application of a conversation method provided in an embodiment of the present application. The conversation method can be applied to a conversation system. The conversation method is executed by an electronic device. The electronic device can specifically be a server deployed with a platform for implementing a dating platform. The server can be an independent physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server that provides cloud computing services.

[0064] The electronic device and multiple terminal devices can communicate via a wireless network, and the terminal device is a device used by a registered user. The terminal device can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc., but is not limited thereto. Specifically, when a registered user initiates a private chat request, the electronic device will calculate the anchor point matching degree of both parties in real time, call the corresponding topic classification script library according to the matching result, and select and recommend the opening text with the highest matching degree to the registered user.

[0065] It is understandable that the above is only an example and is not limited to this embodiment.

[0066] The present application embodiment provides a dialogue method, such as Figure 2 As shown, the method includes:

[0067] S101. When a first user initiates a chat request for a second user, obtaining a first user feature corresponding to the first user and a second user feature corresponding to the second user;

[0068] Among them, the first user is the user who actively initiates a chat request in the chat scene, and the second user is the user who receives the chat request of the first user. The chat request can be a request generated by the first user based on clicking a chat button or sending a message. User characteristics are a set of information related to the user that can describe the user's characteristics, including but not limited to: user interests, lifestyles, geographic locations, personality categories MBTI or time information and other characteristics, wherein user interests, i.e., preference information, can be determined through behavior on the platform. The features in the user characteristics can be classified under specific anchor points. For example, the user characteristics include: location feature 1, location feature 2, preference feature 1, preference feature 2, distance feature 1 and personality feature 1. Location feature 1 and location feature 2 can be classified as the first anchor point, preference feature 1 and preference feature 2 can be classified as the second anchor point, distance feature 1 can be classified as the third anchor point, and personality feature 1 can be classified as the fourth anchor point.

[0069] In the embodiment of the present application, after a chat request is initiated, a first user feature of the first user and a second user feature of the second user can be retrieved from a database of the electronic device.

[0070] In some embodiments, before S101, the process may further include: collecting user data, generating user features based on the user data, and storing the user features.

[0071] Specifically, user data is obtained, which includes user registration data and user behavior data. Among them, the information provided by the user at the time of registration is captured, including user name, age, gender, occupation, hobbies, etc.; the user's browsing history, likes, comments, and published content on the platform are collected. Preprocessing of user data includes: data cleaning to remove noise and redundant data, and handle missing values ​​and outliers; data conversion to convert the user's unstructured data (such as text introductions, chat records, etc.) into processable structured data. User features are generated based on user data, including: generating user tags such as hobbies, occupations, and lifestyles through user behavior data and registration information; determining the user's geographic location based on the location information at the time of registration and the continuously updated location information, and calculating the distance between users; recording the specific time when the interaction occurs, such as morning, afternoon, evening, etc., and recording the user's active time period and the frequency of user visits to the platform and other characteristic information.

[0072] It is understandable that the user characteristics of the user may also be updated regularly to ensure the accuracy of the user portrait constructed by the user characteristics.

[0073] S102, performing anchor point matching on the first user feature and the second user feature to obtain a matching anchor point where the match is successful;

[0074] Among them, the anchor point is a category of features, each anchor point corresponds to multiple topics, and each topic corresponds to at least one topic copy;

[0075] In this step, when the first user initiates a private chat request, the electronic device captures the relevant information of the first user and the second user in real time and calculates the anchor point matching degree (such as interest similarity, geographical proximity, attraction signature matching degree, etc.) between the first user and the second user.

[0076] Among them, for a certain anchor point, determine whether the feature corresponding to the anchor point in the first user feature and the feature corresponding to the anchor point in the second user feature match, if so, determine that the anchor point matches successfully, otherwise, determine that the anchor point matches unsuccessfully; traverse each anchor point in turn to obtain a matching anchor point that matches successfully. Further, in order to improve the accuracy of matching, features with the same or similar semantics as the feature can be used for matching.

[0077] S103. According to the matching anchor point, determine multiple candidate topic copies corresponding to the matching anchor point from the copy library; based on the first user feature and the second user feature, use the sorting model to sort the multiple candidate topic copies to obtain a copy order; according to the copy order, determine the optimal topic copy from the multiple candidate topic copies.

[0078] According to the matching degree of the anchor point, the topic copy with high matching degree is retrieved from the corresponding speech library (copy library).

[0079] In the embodiments of the present application, the copy library refers to a database or collection containing multiple topic copies, which are usually classified and stored according to different anchor points so that they can be quickly retrieved and used when needed. The optimal topic copy refers to the topic copy selected from the copy library that is most relevant to the matching anchor point and has a certain appeal. In some possible implementations, each copy can calculate the click-through rate and / or reply rate of each copy dimension according to the user's click history; the optimal topic copy can be determined based on the click-through rate and / or reply rate.

[0080] In an embodiment of the present application, data is recalled from the copy library by matching the anchor point, and all the topic copy corresponding to the anchor point is obtained according to the matching degree of the anchor point, and the topic copy with a high matching degree is retrieved from the corresponding speech library. Then, a sorting model is used to sort multiple candidate opening remarks to facilitate the screening of the optimal topic copy, wherein the sorting model can be an XGBoost sorting model. It can be understood that the output of the XGBoost sorting model is the probability that the copy will be clicked, and then, the copy can be sorted according to the probability value to obtain the order of the copy. The input to the XGBoost sorting model is the user characteristics of both parties: such as age, constellation, registered city, etc., as well as the copy characteristics, which include the historical click-through rate of the copy. Reply rate, number of uses, etc.

[0081] After determining the matching anchor point, multiple candidate topic copywriting is first screened out, and then the ranking model is used to personalize the candidate copywriting according to the characteristics of both users to determine the optimal topic copywriting, which is more efficient in determining the final topic copywriting.

[0082] S104: Send the optimal topic copy to the client device corresponding to the first user; wherein the optimal topic copy can be used as the opening words of the conversation between the first user and the second user.

[0083] Give the best topic copy to the first user to increase the satisfaction of both parties in the conversation.

[0084] In some possible embodiments, feedback information may also be collected in the process of recommending topic texts to users, and then the topic texts may be further adjusted based on the feedback information, or a satisfaction label may be added to the topic texts.

[0085] In some possible embodiments, after S104, the method further includes: in response to the first user sending a request for the opening remarks of the optimal topic copy, sending the optimal topic copy to the client device corresponding to the second user.

[0086] It can be seen that in the embodiment of the present application, when the first user initiates a chat request to the second user, the user characteristics of both parties can be obtained instantly, and the common interest points or related points of both parties can be quickly found through the anchor point matching technology. Specifically, for each anchor point, it is determined whether the feature corresponding to the anchor point in the first user's feature and the feature corresponding to the anchor point in the second user's feature match. If so, it is determined that the anchor point match is successful, otherwise, it is determined that the anchor point match fails; each anchor point is traversed in turn to obtain a matching anchor point that is successfully matched; and after determining the matching anchor point, a plurality of candidate topic texts are first screened out, and then the candidate texts are personalized and sorted according to the characteristics of both users using the sorting model to determine the optimal topic text, so that the efficiency of determining the final topic text is higher, and the optimal topic text is used as the opening words, thereby effectively promoting the smooth development of the dialogue between the two parties, improving the efficiency of starting the dialogue, and making the communication between users more natural and smooth.

[0087] A possible implementation of the embodiment of the present application is to determine, according to the matching anchor point, a plurality of candidate topic documents corresponding to the matching anchor point from the document library, including:

[0088] According to the matching anchor point, all first topic copies corresponding to the matching anchor point are determined from the copy library;

[0089] According to the user preference information corresponding to all the first topic texts, a plurality of candidate topic texts are determined from all the first topic texts, wherein the user preference information includes: the user's selection information for the first topic text and the other user's reply information for the first topic text.

[0090] Among them, during the conversation process, users are more inclined to give topics that can improve the continuity of the conversation or achieve a benign conversation. The first topic copy is the copy corresponding to the matching anchor point stored in the copy library. User preference information refers to the personal preferences, interests or tendencies shown by users in interactive behaviors such as selection, browsing, and replying, which can reflect the user's preference for specific topics or content. Selection information refers to the record of the user's selection or click on a topic copy. Reply information refers to whether the other user responds or gives feedback to the first topic copy proposed by the user. The candidate topic copy is screened out from all the first topic copies, and is determined to be more likely to arouse user interest or promote conversation based on the majority of user preference information on the platform.

[0091] In some embodiments, determining candidate topic copywriting based on user preference information can be achieved in a variety of ways:

[0092] Optionally, based on the user's selection information for the first topic copy, the number of selections is determined, and based on the number of selections, the copy whose number of selections exceeds a preset threshold is determined as the first candidate topic copy; based on the other user's reply information for the first topic copy, the second candidate topic copy is determined from all the first topic copies, wherein the second candidate topic copy is a copy in the reply information with a feedback time less than a preset time length and / or a copy that the user can reply to; based on the first candidate topic copy and / or the second candidate topic copy, multiple candidate topic copies are determined.

[0093] Among them, the preset number threshold and preset duration can be set by the user according to actual conditions or experience.

[0094] Specifically, statistics are collected on the user's selection information, the number of times each first topic text is selected is calculated, and the text with the largest number of selections is selected as the first candidate; secondly, the reply information of the other user is analyzed, and the text that shows positive feedback in the reply or that the user can reply to is identified, and the text related to these interest points is determined as the second candidate; finally, the first candidate and the second candidate are combined to determine the final list of candidate topic texts. The candidate results may include both the first candidate and the second candidate, or may only include duplicate texts of the two, which is not limited in this embodiment.

[0095] Optionally, based on user preference information, determine evaluation indicators corresponding to each first topic copy, wherein the evaluation indicators include selection rate, reply rate and conversation time information, wherein the conversation time information is determined based on the duration, minimum duration and maximum duration corresponding to each first topic copy; determine the numerical value corresponding to each first topic copy based on the selection rate, reply rate, conversation time information and weight information; and select a preset number of first topic copies with the highest numerical rankings as candidate topic copies.

[0096] The weight information includes a first weight α corresponding to the selection rate, a second weight β corresponding to the reply rate, and a third weight γ corresponding to the conversation time information. The sum of the weights is 1. The embodiment does not limit the weight size, and the user can customize it.

[0097] Specifically, evaluation indicators are set, such as selection rate, reply rate, conversation duration, etc., and each first topic copy is quantitatively evaluated according to the evaluation indicators; a weighted calculation is performed based on the quantitative results of the indicators to obtain the corresponding value of each first topic copy, and the copies are then ranked; a certain number of candidate topic copy are selected from the top ranked copies.

[0098] Among them, the conversation time information = (the duration Ti corresponding to the first topic copy-the minimum duration Tmin) / (the maximum duration Tmax-the minimum duration Tmin), among which, the minimum duration Tmin is the minimum duration corresponding to all the first topic copies, the maximum duration is the maximum duration corresponding to all the first topic copies, and the duration corresponding to the first topic copy is the average duration corresponding to the first topic copy. The time parameters are normalized and the time scale is converted to the same scale as the selection rate and response rate.

[0099] Then, the score S calculation formula can be: It is understandable that other methods may be used to determine candidate topic texts based on user preference information, which are not limited here.

[0100] It can be seen that in the embodiment of the present application, based on the user preference information corresponding to all the first topic copies, the topic copies corresponding to the matching anchor points are recalled, and the candidate topic copy selection is more in line with the actual situation and is more likely to arouse user interest.

[0101] In a possible implementation manner of the embodiment of the present application, the sorting model includes an XGBoost sorting model; the training process of the XGBoost (eXtreme Gradient Boosting) sorting model includes:

[0102] Obtaining historical data corresponding to each of the multiple users, the historical data of the user including: user registration information and user historical behavior information of the user for multiple candidate topic copywritings;

[0103] The XGBoost model is trained according to the historical data corresponding to multiple users to obtain the XGBoost sorting model.

[0104] Among them, based on the available data provided by users on the platform, XGBoost is used as the sorting model to build a personalized recommendation algorithm.

[0105] The detailed steps are as follows:

[0106] Data preparation. Select a training data set, including user registration information, behavior data and interaction records, copywriting features, and whether the copywriting has been clicked by users. Divide the training set and test set to ensure the generalization ability and accuracy of the model.

[0107] Feature engineering. Generate feature vectors based on user tags, geographic distance, time information, etc. Use TF-IDF, Word2Vec and other methods to vectorize text features.

[0108] Model training. Use historical data to train the XGBoost model. Set model parameters, including learning rate, maximum depth, number of threads, etc., to tune and verify the model.

[0109] Model evaluation: Use the test data set to evaluate the model's accuracy, recall, F1-score, AUC and other indicators. Improve the model's stability and generalization ability through methods such as cross-validation and stratified sampling.

[0110] In some embodiments, iterative model optimization can also be performed: the model is regularly retrained and optimized with new data, including adjusting features and parameters.

[0111] It can be seen that in the embodiment of the present application, through learning a large amount of data during the training process, the XGBoost model is able to learn personalized recommendations, so as to use the XGBoost sorting model to perform high-precision personalized sorting of candidate topic copywriting.

[0112] A possible implementation of the embodiment of the present application further includes: S201-S203, wherein:

[0113] S201, obtaining multiple anchor points;

[0114] The anchor point is a category of a feature. For example, the registered address is anchor point 1, and the registered addresses under anchor point 1 include: Beijing, Shijiazhuang, and Chongqing.

[0115] In one achievable manner, multiple anchor points may be set by a technician, or determined based on historical data of platform users.

[0116] Thus, S201 obtains multiple anchor points, including:

[0117] Obtain user information of multiple users, including user registration information and user behavior information;

[0118] Extract features based on user information corresponding to multiple users to obtain features corresponding to each of the multiple users;

[0119] Feature classification is performed based on the features corresponding to the multiple users, and multiple anchor points are determined.

[0120] Specifically, data collection and preprocessing: collect data from user registration information, behavior data, interaction records and other aspects. User information includes: user registration information, behavior data. User registration information includes: registration location, registration duration, behavior data includes: user posting information, user browsing history on the platform, user likes, user comments, and user current location information. Clean and analyze the data to generate feature data such as user tags and geographic location information. Classify and organize the extracted feature data into different topic directions, such as gravity signature anchor points, time anchor points, love bell distance anchor points, etc. For example, gravity signature anchor points: focus on the matching of user personalized tags (interests, hobbies, occupations, etc.); time anchor points: classify according to the time period of mutual user communication (such as morning, afternoon, and evening); love bell distance anchor points: classify according to the geographical distance between users; hotspot anchor points: classify based on current news hotspots and popular topics.

[0121] The embodiment of the present application can conduct in-depth analysis of the registration information and behavior information of a large number of users, extract key features, and perform feature classification based on the key features, thereby determining a plurality of representative anchor points.

[0122] S202, determining a topic copywriting set corresponding to each of the plurality of anchor points, the topic copywriting set including topic copywriting corresponding to each of the plurality of topics;

[0123] When the anchor point is Chongqing, Chongqing corresponds to multiple topics, such as food, scenery, enterprises, specialties, etc. Each topic corresponds to at least one topic copy. The embodiment of the present application does not limit the type of topic copy, which can be in the form of text introduction, photo description, etc.

[0124] In one possible scenario, the topic text is entered by a technician, or is entered from a network database by allowing post-writing or rewriting, which is not limited in the embodiments of the present application.

[0125] S203: Generate a copy library according to the topic copy sets corresponding to the multiple anchor points.

[0126] Furthermore, the script library for hot anchor points needs to be updated regularly to keep the topic relevant.

[0127] It can be seen that in the embodiment of the present application, by collecting multiple anchor points and their corresponding topic copy sets, the system can flexibly respond to various dialogue needs and can cover topic copy for as many scenarios as possible, greatly enhancing the flexibility of the dialogue method.

[0128] A possible implementation of the embodiment of the present application also includes at least one of the following:

[0129] Regularly obtain the latest user information of the target user, and update the user information corresponding to the target user based on the latest user information;

[0130] Regularly obtain feedback information on topic copy for target topics, and update the topic copy for target topics based on the feedback information.

[0131] Specifically, the embodiments of the present application can continuously update data, such as regularly updating user tags and feature data to ensure the freshness and accuracy of the data. The embodiments of the present application can collect user feedback on recommended opening remarks, adjust the content of the speech library and the recommendation logic, and enhance user satisfaction.

[0132] It can be seen that in the embodiment of the present application, by timely obtaining the latest dynamics of the target user, the current needs and preferences of the user can be more accurately grasped; the timeliness of the copy is ensured by the feedback update of the topic copy. The continuous optimization mechanism enables the dialogue method to continuously adapt to changes and be more flexible.

[0133] Based on any of the above embodiments, see Figure 3 , Figure 3 A schematic diagram of a data processing flow provided for an embodiment of the present application includes:

[0134] Anchor point classification and topic direction: Determine the anchor point and topic direction, manually write the copy, and perform copy quality inspection. After passing the quality inspection, store it in the electronic device; introduce a variety of user data types (such as tags, registration time, current time, geographical distance, etc.) as anchor points, classify and organize them according to different topic directions, and manually write opening copy in various styles for each topic direction. Through the copy library management system, the diversity and applicability of the opening content can be achieved, and multi-dimensional personalized opening recommendations can be realized.

[0135] Data collection and preprocessing: Collect user data and preprocess the user data. Then, extract features from the preprocessed data and store them in electronic devices.

[0136] Sorting model: obtain training data, extract features from the training data, use the extracted features to train the model, and evaluate the model. After the evaluation is passed, the sorting model is obtained;

[0137] Real-time copy generation: Match the anchor points of the two parties in the conversation, and screen and recommend opening copy based on the matching anchor points; it can combine the multi-dimensional data of users on both sides, adopt weight and priority sorting algorithms, and ensure that the recommended opening copy has the best match and usage effect.

[0138] System optimization and maintenance: continuously update data, iteratively optimize the model, and adjust the copy based on user feedback.

[0139] It can be seen that the intelligent private chat assisted opening remarks generation system provided in the embodiment of the present application can generate personalized ice-breaking topics based on multi-dimensional user data; data-driven intelligent algorithms can realize real-time and personalized opening remarks generation; dynamic opening remarks recommendation strategies driven by user historical preferences and behavioral data; a speech text library management system for multiple topic directions to achieve diversity and applicability of opening remarks content. It has the following beneficial effects: significantly improve the success rate of user private chats, reduce the embarrassment of initial conversations, and improve the naturalness and sincerity of user interactions. Personalized opening remarks recommendations provide a better user experience, improve user satisfaction and platform activity. Accurate multi-dimensional data analysis and intelligent algorithms effectively improve the matching degree and effect of opening remarks.

[0140] The following is an introduction to a device provided in an embodiment of the present application. The device described below and the method described above can be referred to each other. The device of this embodiment is set in an electronic device. Figure 4 , Figure 4 It is a structural block diagram of a device of one embodiment of the present application, comprising:

[0141] An acquisition module 210, configured to acquire a first user feature corresponding to the first user and a second user feature corresponding to the second user when the first user initiates a chat request to the second user;

[0142] The matching module 220 is used to perform anchor point matching on the first user feature and the second user feature to obtain a successfully matched matching anchor point; wherein the anchor point is a category of a type of feature, each anchor point corresponds to multiple topics, and each topic corresponds to at least one topic copy; wherein, for a certain anchor point, it is determined whether the feature corresponding to the anchor point in the first user feature and the feature corresponding to the anchor point in the second user feature match, if so, it is determined that the anchor point matches successfully, otherwise, it is determined that the anchor point matches unsuccessfully; each anchor point is traversed in turn to obtain a successfully matched matching anchor point; the screening module 230 is used to determine, from the copy library, a plurality of candidate topic copies corresponding to the matching anchor point according to the matching anchor point; based on the first user feature and the second user feature, a sorting model is used to sort the plurality of candidate topic copies to obtain a copy order; according to the copy order, the optimal topic copy is determined from the plurality of candidate topic copies;

[0143] The sending module 240 is used to send the optimal topic copy to the client device corresponding to the first user; wherein the optimal topic copy can be used as the opening words of the conversation between the first user and the second user.

[0144] In one practicable manner, the screening module 230 is used to:

[0145] According to the matching anchor point, all first topic copies corresponding to the matching anchor point are determined from the copy library;

[0146] According to the user preference information corresponding to all the first topic texts, a plurality of candidate topic texts are determined from all the first topic texts, wherein the user preference information includes: the user's selection information for the first topic text and the other user's reply information for the first topic text.

[0147] In one practicable manner, the screening module 230 is used to:

[0148] Determine the number of selections according to the selection information of the user for the first topic copy, and determine, according to the number of selections, the copy whose number of selections exceeds a preset number threshold as the first candidate topic copy;

[0149] According to the reply information of the other user to the first topic copy, determine the second candidate topic copy from all the first topic copies, wherein the second candidate topic copy is a copy in the reply information whose feedback time is less than a preset time and / or the user can reply;

[0150] Determine a plurality of candidate topic copywritings according to the first candidate topic copywriting and / or the second candidate topic copywriting;

[0151] or,

[0152] Determine the evaluation index corresponding to each first topic text according to the user preference information, wherein the evaluation index includes the selection rate, the reply rate and the conversation time information, wherein the conversation time information is determined according to the duration, the minimum duration and the maximum duration corresponding to each first topic text;

[0153] Determine the value corresponding to each first topic copy according to the selection rate, response rate, conversation time information and weight information;

[0154] The first topic copywriting with the highest preset number of numerical rankings is used as the candidate topic copywriting.

[0155] In one achievable manner, the ranking model includes an XGBoost ranking model;

[0156] Also included: Training modules for:

[0157] Obtaining historical data corresponding to each of the multiple users, the historical data of the user including: user registration information and user historical behavior information of the user for multiple candidate topic copywritings;

[0158] The XGBoost model is trained according to the historical data corresponding to multiple users to obtain the XGBoost sorting model.

[0159] In one practicable manner, the method further includes: a document library generation module, which is used to:

[0160] Acquire multiple anchor points; determine the topic copy sets corresponding to each of the multiple anchor points, the topic copy sets include topic copy sets corresponding to each of the multiple topics; generate a copy library according to the topic copy sets corresponding to each of the multiple anchor points.

[0161] In one achievable manner, the copy library generation module is used to:

[0162] Obtain user information of multiple users, including user registration information and user behavior information;

[0163] Extract features based on user information corresponding to multiple users to obtain features corresponding to each of the multiple users;

[0164] Feature classification is performed based on the features corresponding to the multiple users, and multiple anchor points are determined.

[0165] In one achievable manner, the present invention further includes:

[0166] The first updating module is used to: regularly obtain the latest user information of the target user, and update the user information corresponding to the target user based on the latest user information;

[0167] The second updating module is used to: regularly obtain feedback information on the topic copy of the target topic, and update the topic copy of the target topic based on the feedback information.

[0168] An electronic device is provided in an embodiment of the present application, such as Figure 5 As shown, Figure 5 The electronic device 300 shown includes: a processor 301 and a memory 303. The processor 301 and the memory 303 are connected, such as through a bus 302. Optionally, the electronic device 300 may also include a transceiver 304. It should be noted that in actual applications, the transceiver 304 is not limited to one, and the structure of the electronic device 300 does not constitute a limitation on the embodiments of the present application.

[0169] The processor 301 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of this application. The processor 301 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0170] The bus 302 may include a path to transmit information between the above components. The bus 302 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus 302 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 5 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0171] The memory 303 may be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compressed optical disk, laser disk, optical disk, digital versatile disk, Blu-ray disk, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto.

[0172] The memory 303 is used to store the application code for executing the solution of the present application, and the execution is controlled by the processor 301. The processor 301 is used to execute the application code stored in the memory 303 to implement the contents shown in the above method embodiment.

[0173] Figure 5 The electronic device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0174] An embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer-readable storage medium is run on a computer, the computer can execute the corresponding content in the aforementioned method embodiment.

[0175] An embodiment of the present application provides a computer program product, including a computer program, which implements the corresponding contents of the aforementioned method embodiment when the computer program is executed by a processor.

[0176] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0177] The above are only some implementation methods of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A conversation method, characterized in that: include: When the first user initiates a chat request for the second user, obtaining a first user feature corresponding to the first user and a second user feature corresponding to the second user; Anchor point matching is performed on the first user feature and the second user feature to obtain a matching anchor point with successful matching; wherein the anchor point is a category of a feature, each anchor point corresponds to multiple topics, and each topic corresponds to at least one topic copy; wherein, for a certain anchor point, it is determined whether the feature corresponding to the anchor point in the first user feature matches the feature corresponding to the anchor point in the second user feature, if so, it is determined that the anchor point is matched successfully, otherwise, it is determined that the anchor point is matched unsuccessfully; each anchor point is traversed in turn to obtain a matching anchor point with successful matching; According to the matching anchor point, multiple candidate topic texts corresponding to the matching anchor point are determined from a text library; based on the first user feature and the second user feature, the multiple candidate topic texts are sorted using a sorting model to obtain a text order; according to the text order, an optimal topic text is determined from the multiple candidate topic texts; the optimal topic text is sent to a client device corresponding to the first user; wherein the optimal topic text can be used as an opening statement for a conversation between the first user and the second user; The step of determining, according to the matching anchor point, a plurality of candidate topic documents corresponding to the matching anchor point from a document library includes: According to the matching anchor point, determining from a document library all first topic documents corresponding to the matching anchor point; Based on user preference information corresponding to all first topic texts, multiple candidate topic texts are determined from all the first topic texts, the user preference information includes: user selection information for the first topic text and other user's reply information for the first topic text, the selection information includes at least dialogue selection rate, reply rate and dialogue time information, and the dialogue time information is determined based on the duration, minimum duration and maximum duration corresponding to each first topic text.

2. The dialogue method according to claim 1, characterized in that: According to the user preference information corresponding to each of the first topic texts, a plurality of candidate topic texts are determined from the first topic texts, including: Determine the number of selections according to the selection information of the user for the first topic copy, and determine, according to the number of selections, the copy whose number of selections exceeds a preset number threshold as the first candidate topic copy; According to the reply information of the other user to the first topic copy, determine the second candidate topic copy from all the first topic copies, wherein the second candidate topic copy is a copy in the reply information whose feedback time is less than a preset time and / or the user can reply; Determine a plurality of candidate topic copywritings according to the first candidate topic copywriting and / or the second candidate topic copywriting; or, Determine the evaluation index corresponding to each first topic copy according to the user preference information, wherein the evaluation index includes selection rate, response rate and conversation time information; Determine the value corresponding to each first topic copy according to the selection rate, response rate, conversation time information and weight information; The first topic copywriting with the highest preset number of numerical rankings is used as the candidate topic copywriting.

3. The dialogue method according to claim 1, characterized in that: The sorting model includes an XGBoost sorting model; The training process of the XGBoost ranking model includes: Acquire historical data corresponding to each of the plurality of users, the historical data of the users including: user registration information and user historical behavior information of the users with respect to the plurality of candidate topic copywritings; The XGBoost model is trained according to the historical data corresponding to multiple users to obtain the XGBoost sorting model.

4. The dialogue method according to any one of claims 1 to 3, characterized in that: Also includes: Get multiple anchor points; Determine a topic copywriting set corresponding to each of the plurality of anchor points, wherein the topic copywriting set includes topic copywriting corresponding to each of the plurality of topics; A copy library is generated according to the topic copy sets corresponding to the multiple anchor points.

5. The dialogue method according to claim 4, characterized in that: Get multiple anchor points, including: Acquire user information of multiple users, the user information including: user registration information and user behavior information; Extract features based on user information corresponding to multiple users to obtain features corresponding to each of the multiple users; Feature classification is performed based on the features corresponding to each of the multiple users to determine the multiple anchor points.

6. The dialogue method according to claim 1, characterized in that: Also includes at least one of the following; Regularly obtain the latest user information of the target user, and update the user information corresponding to the target user based on the latest user information; Regularly obtain feedback information on the topic copy for the target topic, and update the topic copy for the target topic based on the feedback information.

7. The dialogue method according to claim 1, characterized in that: Also includes: In response to the first user sending a request for the opening remarks of the optimal topic copy, the optimal topic copy is sent to the client device corresponding to the second user.

8. A conversation device, characterized in that: include: An acquisition module, configured to acquire a first user feature corresponding to the first user and a second user feature corresponding to the second user when the first user initiates a chat request to the second user; A matching module is used to perform anchor point matching on the first user feature and the second user feature to obtain a matching anchor point with successful matching; wherein the anchor point is a category of a type of feature, each anchor point corresponds to multiple topics, and each topic corresponds to at least one topic copy; wherein, for a certain anchor point, it is determined whether the feature corresponding to the anchor point in the first user feature matches the feature corresponding to the anchor point in the second user feature, if so, it is determined that the anchor point matches successfully, otherwise, it is determined that the anchor point matches unsuccessfully; each anchor point is traversed in turn to obtain a matching anchor point with successful matching; A screening module is used to determine, from a document library, a plurality of candidate topic documents corresponding to the matching anchor point according to the matching anchor point; sort the plurality of candidate topic documents using a sorting model based on the first user feature and the second user feature to obtain a document sequence; and determine an optimal topic document from the plurality of candidate topic documents according to the document sequence; A sending module, configured to send the optimal topic copy to a client device corresponding to the first user; wherein the optimal topic copy can be used as an opening statement for a conversation between the first user and a second user; The screening module is further used to determine, from a document library, all first topic documents corresponding to the matching anchor point according to the matching anchor point; Based on user preference information corresponding to all first topic texts, multiple candidate topic texts are determined from all the first topic texts, the user preference information includes: user selection information for the first topic text and other user's reply information for the first topic text, the selection information includes at least dialogue selection rate, reply rate and dialogue time information, and the dialogue time information is determined based on the duration, minimum duration and maximum duration corresponding to each first topic text.

9. An electronic device, characterized in that: include: one or more processors; Memory; One or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, and the one or more applications are configured to: execute the steps of the dialogue method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The storage medium stores at least one instruction, at least one program, code set or instruction set, and the at least one instruction, at least one program, code set or instruction set is loaded by the processor and executes the steps of the dialogue method according to any one of claims 1 to 7.

Images