CN113505293B

CN113505293B - Information pushing method and device, electronic equipment and storage medium

Info

Publication number: CN113505293B
Application number: CN202110662590.4A
Authority: CN
Inventors: 邓锐涛; 常向月; 刘云峰
Original assignee: Shenzhen Zhuiyi Technology Co Ltd
Current assignee: Shenzhen Zhuiyi Technology Co Ltd
Priority date: 2021-06-15
Filing date: 2021-06-15
Publication date: 2024-03-19
Anticipated expiration: 2041-06-15
Also published as: CN113505293A

Abstract

The embodiment of the application provides an information pushing method, an information pushing device, electronic equipment and a storage medium, and relates to the technical field of information. The method comprises the following steps: obtaining a dialogue log corresponding to an entry identifier, wherein the entry identifier is used for identifying a user interface for realizing dialogue with a user; clustering a plurality of user questions in the dialogue log to obtain a plurality of question clusters; and determining at least one push question based on the plurality of question clusters, the at least one push question for being pushed and displayed on the user interface. The push problem can be determined according to the user problem truly input by the user interface, so that the obtained push problem is more in line with the question intention of the user, and the accuracy of the push problem is improved.

Description

Information pushing method and device, electronic equipment and storage medium

Technical Field

The present disclosure relates to the field of information technologies, and in particular, to an information pushing method, an apparatus, an electronic device, and a storage medium.

Background

Along with the development of internet technology, the intelligent customer service technology establishes a quick and effective communication mode based on natural language for enterprises and mass users, and is widely applied to various industries at present. The intelligent customer service technology can push related questions to a user before the user asks the questions, so that the user is guided to input the questions, and the questions of the user can be responded quickly. However, the problem of pushing by the existing intelligent customer service technology is fixed, the problem of meeting the dialogue intention of the user is difficult to push, the pushing problem is not accurate enough, and the user experience is reduced.

Disclosure of Invention

In view of the above, the present application proposes an information pushing method, an apparatus, an electronic device, and a storage medium, so as to improve the above-mentioned drawbacks.

In a first aspect, an embodiment of the present application provides an information pushing method, including: obtaining a dialogue log corresponding to an entry identifier, wherein the entry identifier is used for identifying a user interface for realizing dialogue with a user; clustering a plurality of user questions in the dialogue log to obtain a plurality of question clusters; and determining at least one push question based on the plurality of question clusters, the at least one push question for being pushed and displayed on the user interface.

Further, the clustering processing is performed on the plurality of user questions in the dialogue log to obtain a plurality of question clusters, including: extracting each user question input by a user in the dialogue log; extracting semantic features of each user problem to obtain feature vectors of each user problem; and clustering the feature vectors of the user questions to obtain the question clusters.

Further, the determining at least one push question based on the plurality of question clusters includes: determining user questions corresponding to the center points of the question clusters as the candidate questions; and determining the at least one push question from the candidate questions corresponding to each question cluster.

Further, the determining at least one push question based on the plurality of question clusters includes: acquiring the distance between the feature vector of each user problem and the center point of the problem cluster; determining a user problem characterized by a feature vector smaller than a specified distance as a candidate problem; and determining the at least one push question from the candidate questions corresponding to each question cluster.

Further, the determining the at least one push question among the candidate questions corresponding to each question cluster includes: acquiring the number of user questions in each question cluster; sorting the candidate questions in descending order according to the number; and determining a specified number of the candidate questions ordered first as the at least one push question.

Further, after the determining the first-ranked specified number of the candidate questions as the at least one push question, the method further comprises: pushing the at least one push question and an order of arrangement of the at least one push question to the user interface to display the at least one push question in the order on the user interface.

Further, after the determining at least one push question based on the plurality of question clusters, the method further comprises: acquiring at least one preset specified problem and the priority of the specified problem; according to the priority, the at least one designated problem and the at least one push problem are arranged and combined to obtain a problem combination; and pushing the combination of questions to the user interface.

Further, after the determining at least one push question based on the plurality of question clusters, the method further comprises: acquiring at least one preset standard problem; acquiring semantic similarity of each standard problem and each push problem; and if the semantic similarity is larger than a specified threshold, updating the push problem corresponding to the semantic similarity into the standard problem.

Further, the obtaining a dialogue log corresponding to the entry identifier includes: updating the dialogue log corresponding to the entrance identifier into a dialogue log of a designated time every interval preset time; the clustering processing is performed on the plurality of user questions in the dialogue log to obtain a plurality of question clusters, including: clustering a plurality of user questions in the dialogue log at the appointed time to obtain updated question clusters; the determining at least one push question based on the plurality of question clusters includes: updating the at least one push question based on the updated plurality of question clusters.

In a second aspect, an embodiment of the present application provides an information pushing method, including: if a dialogue request of a user is detected, acquiring an entry identification of a user interface for realizing dialogue with the user; acquiring at least one push question corresponding to the entry identifier, wherein the at least one push question is a question determined based on a plurality of obtained question clusters after clustering a plurality of user questions, and the plurality of user questions are questions input by users in a dialogue log corresponding to the entry identifier; and displaying the push question on the user interface.

Further, after the obtaining the entry identifier of the user interface for implementing the dialogue with the user if the dialogue request of the user is detected, the method further includes: and acquiring a dialogue log of the user corresponding to the entry identifier, and sending the dialogue log to a server.

In a third aspect, an embodiment of the present application provides an information pushing apparatus, including: the log acquisition module is used for acquiring a dialogue log corresponding to an entry identifier, wherein the entry identifier is used for identifying a user interface for realizing dialogue with a user; the clustering processing module is used for carrying out clustering processing on a plurality of user problems in the dialogue log to obtain a plurality of problem clusters; and a question determination module for determining at least one push question based on the plurality of question clusters, the at least one push question being for being pushed and displayed on the user interface.

Further, the cluster processing module includes: a question extraction sub-module, configured to extract each of the user questions input by a user in the dialog log; the feature extraction sub-module is used for extracting semantic features of each user problem to obtain a feature vector of each user problem; and the feature clustering sub-module is used for carrying out clustering processing on the feature vectors of the user problems to obtain the problem clusters.

Further, the problem determination module includes: a central problem determination submodule, configured to determine a user problem corresponding to a central point of each problem cluster as the candidate problem; and a push question determination submodule, configured to determine the at least one push question from the candidate questions corresponding to each question cluster.

Further, the problem determination module includes: a distance acquisition sub-module, configured to acquire a distance between a feature vector of each user problem and a center point of the problem cluster; a candidate problem determination submodule for determining a user problem characterized by a feature vector smaller than a specified distance as a candidate problem; and a push question determination submodule, configured to determine the at least one push question from the candidate questions corresponding to each question cluster.

Further, the push problem determination submodule includes: a number acquisition unit, configured to acquire the number of user questions in each question cluster; the sorting unit is used for sorting the candidate questions in a descending order according to the number; and a push question determination subunit configured to determine a specified number of the candidate questions ordered first as the at least one push question.

Further, the information pushing device further includes: and the pushing module is used for pushing the at least one pushing problem and the arrangement sequence of the at least one pushing problem to the user interface so as to display the at least one pushing problem on the user interface in the arrangement sequence.

Further, the information pushing device further includes: the system comprises a specified problem acquisition module, a priority judgment module and a priority judgment module, wherein the specified problem acquisition module is used for acquiring at least one preset specified problem and the priority of the specified problem; the problem combination acquisition module is used for arranging and combining the at least one designated problem and the at least one push problem according to the priority to obtain a problem combination; and the question combination pushing module is used for pushing the question combination to the user interface.

Further, the information pushing device further includes: the standard problem acquisition module is used for acquiring at least one preset standard problem; the similarity acquisition module is used for acquiring the semantic similarity between each standard problem and each push problem; and the standard problem updating module is used for updating the push problem corresponding to the semantic similarity into the standard problem if the semantic similarity is larger than a specified threshold.

Further, the log obtaining module further includes: the log updating module is used for updating the dialogue log corresponding to the entrance identifier into a dialogue log of a designated time every interval for a preset time; the cluster processing module further includes: the problem cluster updating module is used for carrying out clustering processing on a plurality of user problems in the dialogue log at the appointed time to obtain a plurality of updated problem clusters; and the push question updating module is used for updating the at least one push question based on the updated plurality of question clusters.

In a fourth aspect, an embodiment of the present application provides an information pushing apparatus, including: the entrance identification acquisition module is used for acquiring an entrance identification of a user interface for realizing dialogue with a user if a dialogue request of the user is detected; the problem acquisition module is used for acquiring at least one push problem corresponding to the entrance identifier, wherein the at least one push problem is a problem determined based on a plurality of obtained problem clusters after clustering a plurality of user problems, and the plurality of user problems are problems input by users in a dialogue log corresponding to the entrance identifier; and the problem display module is used for displaying the push problem on the user interface.

Further, the entry identification acquisition module further includes: and the log sending module is used for obtaining the dialogue log of the user corresponding to the entry identifier and sending the dialogue log to a server.

In a fifth aspect, embodiments of the present application provide an electronic device, including: one or more processors; a memory; one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs configured to perform the method of the first aspect or the second aspect described above.

In a sixth aspect, embodiments of the present application provide a computer readable storage medium having stored therein program code that is callable by a processor to perform a method as described in the first or second aspects above.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the description of the embodiments will be briefly introduced below, it being obvious that the drawings in the following description are only some embodiments of the present application, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

FIG. 1 illustrates a schematic diagram of an application environment suitable for use in embodiments of the present application;

FIG. 2 illustrates a client application interface provided by an embodiment of the present application;

FIG. 3 illustrates a user interface provided by an embodiment of the present application for interacting with a customer;

fig. 4 is a schematic flow chart of an information pushing method according to an embodiment of the present application;

fig. 5 shows a processing schematic diagram of an information pushing method according to an embodiment of the present application;

fig. 6 is a schematic flow chart of an information pushing method according to another embodiment of the present application;

fig. 7 is a schematic flow chart of an information pushing method according to another embodiment of the present application;

fig. 8 shows a schematic flow chart of step S407 in fig. 7;

fig. 9 is a schematic flow chart of an information pushing method according to still another embodiment of the present application;

Fig. 10 is a schematic flow chart of an information pushing method according to another embodiment of the present application;

fig. 11 is a schematic flow chart of an information pushing method according to still another embodiment of the present application;

fig. 12 is a schematic flow chart of an information pushing method according to still another embodiment of the present application;

fig. 13 is a schematic flow chart of an information pushing method according to still another embodiment of the present application;

FIG. 14 is a block diagram of an information pushing device according to an embodiment of the present application;

FIG. 15 is a block diagram illustrating an information pushing device according to another embodiment of the present application;

fig. 16 shows a block diagram of an electronic device for performing an information pushing method according to an embodiment of the present application;

fig. 17 shows a storage unit for storing or carrying program code implementing the information push method according to the embodiment of the present application.

Detailed Description

The following description of the embodiments of the present application will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are only some, but not all, of the embodiments of the present application. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are within the scope of the present disclosure.

With the progress of technology, artificial intelligence technology has become more popular, and services such as reservation, consultation, etc. have been changed from artificial services to machine services in daily life. Compared with manual service, the service efficiency is greatly improved through machine service, and great convenience is brought to daily life of people.

Typically intelligent customer service may provide services to users by way of conversations with the users. Specifically, the user may input a question on a user interface that dialogues with the intelligent customer service, which pushes an answer to the user by identifying the question entered by the user. In order to enable a user to conveniently acquire a desired answer, it is very critical to guide the user to input questions, specifically, some questions can be pushed on a user interface for user reference so as to guide the user to input the questions, thereby shortening the time for the user to input the questions and clarify the questions. However, the problem of pushing to users in intelligent customer service is usually fixed and requires manual addition by business personnel. Because users served by intelligent customer service can have a very large number of problems in practice, business personnel are required to add preset problems as much as possible, a large amount of labor cost is consumed, accuracy and efficiency cannot be guaranteed, the problems meeting the conversation intention of the users are difficult to push, the users usually need to input the problems for many times to communicate with the customer service personnel to obtain answers, user experience is poor, and loss of the users in the intelligent customer service conversation process is caused.

The inventor researches the user behavior when the user interacts with the intelligent customer service, discovers that a plurality of interactive interfaces usually exist in the client in practical application, conversation inlets of the intelligent customer service are respectively arranged on the plurality of interactive interfaces, and the user can enter the user interface for carrying out conversation with the intelligent customer service by triggering the conversation inlets. By considering the user behavior more, the inventor finds that when a problem is generated in the process of interacting with the current interactive interface, the user can trigger the dialogue entrance of the current interactive interface, so as to perform dialogue with the intelligent customer service. That is, the problem entered by the user is generally related to the business data of the interactive interface corresponding to the current conversation entrance. For example, when the interactive page is a credit card application page, user questions entering into a smart customer service session are typically related to credit card information; when the interactive interface is a shopping page, user questions entering into an intelligent customer service session are typically related to payment information; when the interactive interface is a personal information interface, user questions entering into an intelligent customer service session are typically related to personal information; when the interactive interface is a financial interface, the user's problem of entering an intelligent customer service session is often related to investment financial. That is, the user questions input by the user who performs the intelligent customer service dialogue through the same interactive interface may be relatively close, if the user questions input by the user can be analyzed according to the interactive interface, the user intention of the interactive interface, that is, the questions which the user inputs questions to be solved, are accurately identified, the questions related to the user intention are pushed to the user during the dialogue, the accuracy of the problem pushing can be improved, and the use experience of the user is improved.

In order to improve the above problems, the inventors propose an information pushing method, an information pushing device, an electronic device, and a storage medium in the embodiments of the present application. The pushing problem can be determined by analyzing the user problem on the user interface, the pushing problem accords with the user intention, and the pushing accuracy can be improved.

Referring to fig. 1, fig. 1 shows a schematic view of an application environment suitable for use in an embodiment of the present application. The information pushing method provided by the embodiment of the application can be applied to the information pushing system 10 shown in fig. 1. The information push system 10 includes a plurality of or at least one terminal device 100 and a server 200, where the terminal device 100 and the server 200 are located in a wireless network or a wired network, and the terminal device 100 and the server 200 perform data interaction, and the interaction data includes, but is not limited to, audio, video, text, images, and the like.

The terminal device 100 may be a mobile terminal device, and may include a smart phone, a tablet computer, an electronic book reader, a laptop portable computer, a car-mounted computer, a wearable mobile terminal, and so on, for example. The server 200 may be a single server, a server cluster, a server center formed by a plurality of servers, a local server, or a cloud server. The server 200 may be used to provide background services to users, which may include, but are not limited to, information push, etc., and is not limited herein.

In some embodiments, a client application may be installed on the terminal device 100, and a user may communicate with the server 200 based on the client application (e.g., APP, etc.). Specifically, the terminal device 100 may acquire input information of a user, based on the communication between the client application on the terminal device 100 and the server 200, the server 200 may process the received input information of the user, the server 200 may also return corresponding output information to the terminal device 100 according to the information, and the terminal device 100 may perform an operation corresponding to the output information. The input information of the user may be voice information, touch operation information based on a screen, gesture information, motion information, and the like, and the output information may be images, videos, characters, audio, and the like, which are not limited herein.

Specifically, the client application program on the terminal device includes a plurality of interactive interfaces, different content is displayed on different interactive interfaces, icons or keys of a conversation entrance may be set on the interactive interfaces, and in response to a triggering operation of a user on the conversation entrance, the current interactive interface may be switched to a user interface for conversation with the user. That is, the user interface for talking to the user is the next level interface to the current interactive interface. For example, fig. 2 illustrates a client application interface provided by an embodiment of the present application. The interactive interface in fig. 2 is a credit card application interface, and a "customer service" button is displayed on the upper end of the credit card application interface, so as to enter a user interface for talking with a customer service in response to the operation of the button by the user, i.e. the dialogue request of the user. As shown in fig. 3, fig. 3 illustrates a user interface provided by an embodiment of the present application for talking to a customer. The intelligent customer service interface of fig. 3 is provided with a preset display area for "guessing you want to ask", a push question is displayed on the preset display area, and a text or voice input area is arranged at the bottom of the intelligent customer service interface. Alternatively, the user may quickly enter the push question by clicking on the push question or speaking the push question with speech.

The server 200 can collect user data of a user at a client application, and in particular, the user data may include an entry identifier corresponding to a user session request, and a session log of the user. Alternatively, the server 200 may push information for the user of the terminal device 100 according to the collected dialogue log; the server 200 may also transmit the collected conversation log to the terminal device 100, so that the terminal device 100 determines push information from the conversation log.

The above application environments are merely examples for facilitating understanding, and it is to be understood that embodiments of the present application are not limited to the above application environments.

The information pushing method, the device, the electronic equipment and the medium provided by the embodiment of the application are described in detail through specific embodiments.

Referring to fig. 3, fig. 3 is a flow chart of an information pushing method according to an embodiment of the present application, which is applied to the server. The information push method includes S201 to S203.

S201, a dialogue log corresponding to the entry identification is acquired.

The portal identification is used to identify a user interface for implementing a dialog with a user.

The user interface is an interface for implementing a dialog function. Specifically, conversation inlets are arranged on some interactive interfaces, wherein each user interface for realizing conversation corresponds to one conversation inlet, the conversation inlets are arranged on the upper-level interactive interface of the user interface, each conversation inlet corresponds to one inlet identification, and the inlet identification can identify the user interface for realizing conversation with a user.

As an embodiment, each dialog portal has a portal identification, that is to say the user interface and the portal identification are in one-to-one correspondence. As another implementation, multiple conversation portals may correspond to the same portal identification. Wherein the plurality of conversation portals may be interactive interfaces having similar content. For example, the similar content may be a similar service type, and the plurality of conversation entries have the same identifier on an interactive interface of a notebook computer, a mobile phone, or the like for purchasing an electronic product; multiple conversation portals have the same identification on a child book, a historic book, a textbook, or the like interactive interface for purchasing the book.

Wherein, the dialogue log corresponding to the entrance identifier obtained by the server comprises: the portal identification corresponds to a plurality of historical conversation logs generated by a plurality of historical users on a user interface, and each historical conversation log records the real conversation process of one historical user and intelligent/manual customer service. A history user is a user who has performed a conversation on the user interface. The conversation log is a conversation record of the user in text format. Alternatively, the server may directly read the text format dialogue log from the database, or may record the voice format dialogue log, and obtain the text format dialogue log through voice recognition.

Specifically, the dialog log corresponding to each user includes at least one user message and at least one system message. Optionally, an object identification may be included in the conversation log, the object identification identifying whether the message was entered by a user or by the system. As an embodiment, the system messages may be entered by the customer service robot, that is to say the dialogue log comprises a dialogue record of the user with the intelligent customer service. As another implementation, the system message may be entered by a human customer service, i.e., the conversation log includes a conversation record of the user with the human customer service.

It is understood that the dialogue log is not limited to the customer service log, but may be a dialogue text log in other scenarios. For example, the dialogue log may be questions set by one party under the question-and-answer platform, answers provided by multiple parties, and the like; chat logs of non-question-answer type, etc. are also possible.

S202: and clustering the plurality of user questions in the dialogue log to obtain a plurality of question clusters.

After the dialogue log of the portal identification is obtained, a plurality of user questions in the dialogue log may be extracted. The user question can reflect the user's dialog intention, i.e., the meaning that the user inputs the question to express, i.e., the question that the user inputs to be solved.

In particular, the server may extract user messages entered by the user from the dialog log based on the object identification. As one embodiment, each user message entered by the user may be determined as each user question, thereby enabling a more comprehensive analysis of the messages entered by the user. As another embodiment, question detection may be performed on the user message, for example, detecting whether a query word is included or not, obtaining question input by the user, and determining each question input by the user as each user question, so that the intention of the user question can be analyzed more accurately.

In some embodiments, user questions may also be filtered before clustering the plurality of user questions. Specifically, user questions may be filtered, and questions determined to be boring questions may be filtered based on intent recognition. Through preliminary screening of user problems, operation cost can be reduced, and memory consumption is reduced.

The clustering process may be implemented by a variety of algorithms. For example, K-means algorithms, also known as K-average or K-means clustering algorithms, graph clustering algorithms, such as label propagation algorithms, etc., hierarchical clustering algorithms, density-based clustering, mesh-based clustering, etc., may be used. It can be appreciated that the clustering process can adopt an unsupervised learning clustering algorithm, and the unsupervised learning does not need to manually mark a large number of samples in advance, so that the limitation of the granularity of category division due to the overhigh training cost can be avoided.

In some implementations, the clustering process can be text clustering process based on semantic similarity. The plurality of user questions may be clustered based on semantics to obtain a plurality of question clusters. Wherein each question cluster includes at least one user question having similar semantics. In particular, please refer to the following examples.

In other embodiments, the clustering process may be text clustering process based on the type of service. For example, in a banking scenario, the type of business may be financial, bank card, credit card, etc. And obtaining the problem clusters corresponding to the service types through clustering.

In still other embodiments, the server may classify a plurality of user questions according to preset labels to obtain labels of each user question, and then perform clustering processing based on semantic matching on user questions with the same label to obtain at least one question cluster corresponding to each label. The tag may be a service type related to the user interface, or may be a user portrait, such as a user age, a member level, a gender, etc. For example, the preset label may be an age group of the user, and after obtaining the user questions corresponding to the user in each age group, the plurality of user questions in each age group are clustered respectively to obtain a plurality of question clusters corresponding to each age group.

S203: at least one push question is determined based on the plurality of question clusters.

Wherein the at least one push question is for being pushed and displayed on a user interface. The user interface is an interface for implementing a dialog with the user corresponding to the portal identification. An information display area is arranged on the user interface and is used for displaying push problems. Optionally, when the push question is a real user question in the dialogue log, after determining the push question, error correction processing can be performed on the push question based on a dictionary and rules, so as to correct wrongly written characters or wrong grammar in the push question, and push the error corrected question.

In some embodiments, the server may determine one or more target problem clusters among the plurality of problem clusters, and then determine a push problem based on user problems of the target problem clusters.

As one approach, the target problem cluster may be determined based on a tag of the current business scenario or a tag of the current interactive user. For example, the user portrait of the current interactive user can be obtained, the problem cluster with the same user portrait label is determined as the target problem cluster, and the attribute of the current interactive user is considered when the push problem is determined, so that the push problem is more targeted, and the push accuracy is higher.

As another way, the target problem cluster may be determined according to the number of user problems in each problem cluster, and the problem cluster with the larger number is determined as the target problem cluster, so that the push problem is a user problem of high-frequency questioning of the user. By determining the target problem clusters first, the power consumption required for processing all the problem clusters can be reduced, and particularly under the condition of more problem clusters, the calculated amount can be greatly reduced, and the calculation efficiency is improved.

In some embodiments, the server may determine a candidate issue corresponding to each of the issue clusters, and determine at least one push issue among the candidate issues corresponding to each of the issue clusters. Specifically, please refer to the following embodiments, which are not described herein.

In some embodiments, after determining the push question, an answer to the push question may also be determined and pushed.

As one way, a plurality of standard questions, which are questions having a standard format and contents, which can be accurately recognized, and answers to each of the standard questions may be preset. The standard question with the highest semantic similarity with the push question can be determined, and the answer of the standard question is determined as the answer of the push question. Alternatively, the standard questions may also be updated by the operator based on the push questions. For example, when there is no standard question similar to the push question, the standard question corresponding to the push question is configured and the answer is configured.

Alternatively, answers to the push questions may be determined in the conversation log based on the contextual relationship of the user messages and the system messages in the conversation log. As an implementation manner, a system message corresponding to the push question input by the user can be searched in the dialogue log, and the system message is determined as an answer to the push question. As another embodiment, a question cluster to which the push question belongs may be obtained, a system message set corresponding to each user question in the question cluster is taken as an answer cluster, and a system message corresponding to a center point of the answer cluster is determined as an answer to the push question.

Referring to fig. 5, fig. 5 is a schematic processing diagram of an information pushing method according to an embodiment of the present application. As shown in fig. 5, a session entry "customer service" button is provided on the credit card application interface, and the session entry has an entry identifier 1. The server acquires a conversation log corresponding to the entrance identifier 1, clusters a plurality of user questions in the conversation log to obtain a plurality of question clusters, determines at least one push question 1 based on the plurality of question clusters, and pushes the push question 1 to a user interface for conversation corresponding to the entrance identifier 1 for display. And the terminal equipment responds to the triggering operation of the user on the dialogue entrance of the entrance mark 1 on the credit card application interface, and switches the current credit card application interface into the user interface corresponding to the entrance mark 1, namely, the intelligent customer service interface with the pushing problem 1 is displayed. Dialogue inlets are respectively arranged on the shopping mall interface and the personal information interface as shown in fig. 5, the dialogue inlets respectively correspond to inlet identifiers, and similarly, push questions corresponding to the inlet identifiers can be respectively pushed and displayed for user interfaces corresponding to other inlet identifiers.

It will be appreciated that the push questions on each user interface are derived from user questions corresponding to the entry identifications of that user interface, with different entry identifications corresponding to different push questions. Because users who trigger a conversation through the same conversation portal typically have the same conversation intent, user input user questions tend to be closer as well. Therefore, a plurality of question clusters obtained by clustering a plurality of user questions of the same portal identifier are used for determining the pushing questions based on the question clusters, and the pushing questions are displayed on a user interface of the portal identifier, so that the user intention triggering the conversation portal can be hit more accurately, and the accuracy of the pushing questions is improved.

In practical application, the process of determining push problems can be automatically performed by a server, and a large number of conversation entrances can be put in a business scene, so that user problems in the conversation scene corresponding to each conversation entrance can be divided into finer granularity, push problems on a user interface corresponding to the entrance identification are met with a larger probability, the intention of a user triggering the conversation entrance is hit, and the time for inputting the problems and clarifying the problems by the user is greatly shortened.

According to the information pushing method provided by the embodiment of the application, a dialogue log corresponding to an entrance identifier is obtained, and the entrance identifier is used for identifying a user interface for realizing dialogue with a user; clustering a plurality of user questions in a dialogue log to obtain a plurality of question clusters; at least one push question is determined based on the plurality of question clusters, the at least one push question for being pushed and displayed on the user interface. The pushing problem can be determined based on the entry identification according to the user problem input by the historical user, so that the obtained pushing problem is more in line with the questioning intention of the user, and the accuracy of the pushing problem is improved.

Referring to fig. 6, fig. 6 is a flowchart of an information pushing method according to an embodiment of the present application, which is applied to the server. The method includes S301 to S305.

S301: a dialog log corresponding to the portal identification is obtained.

S302: each user question entered by the user in the dialog log is extracted.

S303: and extracting semantic features of each user problem to obtain feature vectors of each user problem.

The feature vector of each user question can be used for representing the semantics of the user question, and the feature vector is a mathematical representation converted from a text form through semantic feature extraction of the user question. Specifically, the server may perform semantic feature extraction on the user problem based on a preset semantic understanding model, input the user problem into the semantic understanding model, and output a feature vector of the user problem. As an embodiment, the semantic understanding model may be a transformer-based bi-directional coded representation (Bidirectional Encoder Representation from Transformers, BERT) network or the like model. The specific embodiment for semantic feature extraction is not limited herein.

In some embodiments, the server may perform word segmentation processing on each user question, and divide the user question into independent words, so as to obtain word segmentation phrases of each question; extracting semantic features of word groups of each user problem, calculating word vectors of the word groups of each user problem, and obtaining feature vectors of sentences of the whole user problem according to the word vectors of the word groups of each user problem. Optionally, before word segmentation, error correction processing can be performed on the user problem, and the wrongly written or wrongly grammatical words in the user problem can be corrected, so as to obtain more accurate feature vectors.

As an embodiment, the feature vector may be generalized by using feature context or feature synonym relationships. Generalization refers to replacing some similar words with the same representation, e.g. "several money" is generalized to "how much money". The semantics of the user problem can be accurately represented through generalization processing, and the accuracy of clustering processing in subsequent processing is improved.

S304: and clustering the feature vectors of the user problems to obtain a plurality of problem clusters.

The server can cluster the feature vectors according to the semantic similarity of the feature vectors of the user problem to obtain a plurality of problem clusters. The number of the problem clusters can be preset, can be automatically determined in the clustering process by a clustering algorithm, and can be preset or randomly selected. Specifically, the distance between the feature vector of each user problem and each cluster center is calculated, the cluster center to which the user problem belongs is determined according to the distance, and the cluster center is updated. The distance between the feature vectors can represent the semantic similarity between the user questions corresponding to the feature vectors. The larger the distance, the smaller the semantic similarity. Alternatively, the distance may be a cosine distance, euclidean distance, manhattan distance, or the like.

And clustering the feature vectors of the user problems to obtain a plurality of problem clusters. The distance between the feature vectors of the user questions in the same question cluster is as small as possible, that is, the semantics of the user questions in the same question cluster are relatively close. The distance between the feature vectors of the user questions among different question clusters is as large as possible, namely the semantic difference expressed by the user questions among different question clusters is large.

Because of the diversity of language expressions, for the same question, there may be multiple user questions, and extracting the user log may result in multiple user questions containing the same semantics. By clustering, user problems with similar semantics can be classified, which is beneficial to managing the user problems. For example, after the clustering process, it is possible to "how to open a credit card? "," how do credit cards transact? "how do me want to do a credit card? "these semantically identical or similar user questions are grouped together into a question cluster.

S305: at least one push question is determined based on the plurality of question clusters.

It should be noted that, in the foregoing method steps, those portions not described in detail refer to the foregoing embodiments, and are not described herein again.

According to the information pushing method, semantic feature extraction is carried out on each user problem through extracting each user problem input by a user in a dialogue log, so that feature vectors of each user problem are obtained, and clustering processing is carried out on the feature vectors of the user problems, so that a plurality of problem clusters are obtained. The feature vector of each user problem is obtained through semantic feature extraction, and then the feature vector of the user problem is clustered to obtain a plurality of problem clusters, so that the user problem in each problem cluster has similar semantics, and further the push problem corresponding to the user intention is determined based on the obtained plurality of problem clusters.

Referring to fig. 7, fig. 7 is a flowchart of an information pushing method according to an embodiment of the present application, which is applied to the server. The method includes S401 to S406.

S401: a dialog log corresponding to the portal identification is obtained.

S402: each user question entered by the user in the dialog log is extracted.

S403: and extracting semantic features of each user problem to obtain feature vectors of each user problem.

S404: and clustering the feature vectors of the user problems to obtain a plurality of problem clusters.

S405: and determining the user problem corresponding to the central point of each problem cluster as a candidate problem.

And the center point of the problem cluster is a clustering center obtained after clustering. Specifically, the center point is a feature vector of the user problem located at the center of the cluster, and can represent the center of semantic distribution in each problem cluster. That is, the user question corresponding to the center point of the question cluster may represent the semantics of the question cluster. By determining the user problem corresponding to the center point of each problem cluster as a candidate problem, one candidate problem corresponding to each problem cluster can be obtained, and the candidate problem can cover the semantics of more user problems in the problem cluster.

S406: and determining at least one push problem in the candidate problems corresponding to each problem cluster.

In some embodiments, the server may determine one or more target problem clusters among the plurality of problem clusters, and determine a push problem from candidate problems for the target problem clusters. The specific implementation of determining the target problem cluster is referred to the foregoing embodiments, and will not be described herein.

In some embodiments, after obtaining the candidate questions corresponding to each question cluster, the candidate questions of the multiple question clusters may be ranked, and the push questions may be determined according to the ranking order. Specifically, the candidate questions can be scored through the dimensions of the length of the candidate questions, the semantic consistency of the candidate questions, the number of preset keywords appearing in the candidate questions, the occurrence frequency of the candidate questions in the dialogue log and the like, the candidate questions are sorted in descending order according to the score, and the designated number of questions sorted in front are determined to be push questions. The length, semantic consistency and other dimensionalities of the candidate problems can represent the readability, and the higher the score is, the better the corresponding readability is; the keywords can be preset words related to the service scene, and the number of the preset keywords appearing in the candidate questions can represent the association degree of the user questions and the service scene; the frequency of occurrence of candidate questions may characterize the question hotness. By determining the candidate problem with a higher score as the push problem, a more appropriate push problem can be determined by integrating various factors.

In some embodiments, as shown in fig. 8, S406 may include S4061 to S4063.

S4061: the number of user questions in each question cluster is obtained.

The number of user questions in each question cluster is counted separately. Since the user questions are real questions entered by the user in the history dialogue, each question cluster may characterize a semantic meaning, i.e., a user intent, and the number of user questions may characterize the hotness of the question cluster. It should be noted that the repeated occurrence of the problem will be counted repeatedly. For example, the problem cluster includes 2 duplicate user problems, "how credit card is applied", and the number of user problems in the problem cluster is 2.

S4062: and sorting the candidate questions in a descending order according to the number.

And according to the number of the user questions in each question cluster, descending order of the question clusters, and taking the order of the question clusters as the order of the candidate questions of the question clusters. For example, the problem cluster 1 includes 10 user problems, the candidate problem is problem 1, the problem cluster 2 includes 5 user problems, the candidate problem is problem 2, and the order after the ordering is problem 1 and problem 2.

As one implementation, candidate questions in a question cluster may be ranked according to the distance of the candidate question from a center point, or a scoring value for the candidate question. For one embodiment, for the same number of problem clusters, the order may be determined according to the average grading value of the candidate problems in the problem clusters, and the problem cluster with higher grading value is ranked in front.

S4063: a specified number of candidate questions ordered first are determined as at least one push question.

Wherein the designated number is a preset number. As one embodiment, the specified number is the maximum of the number of questions that can be displayed on the user interface. After the candidate questions are ordered in a descending order, the designated number of the candidate questions ordered in front can be determined to be push questions, so that the push questions are high in heat, and users frequently ask questions, and therefore accuracy of the push questions hitting user intention is improved. As another implementation, the specified number may be operator configurable.

In some implementations, step S4073 may be performed by a server, which may push the push question to the terminal device after determining the at least one push question to display the push question on a user interface of the terminal device. As one way, after pushing the pushing problem to the terminal device, if the server obtains the request information sent by the terminal device, the server pushes other problems except the pushing problem to the terminal device. The request information is used for representing a request which is input by a user and is used for acquiring more push questions after the push questions are displayed on the user interface. For example, the request information may be sent in response to a "swap question" operation triggered by the user on the interactive interface of the push question.

In other embodiments, step S4073 may also be performed by the terminal device, and the server may send the candidate questions and the ranking order to the terminal device, and the terminal device determines the designated number of candidate questions ranked first as at least one push question.

In some implementations, after determining the first-ranked specified number of candidate questions as the at least one push question, the at least one push question and a ranking of the at least one push question may be pushed to the user interface to display the at least one push question in the ranking on the user interface. For example, the push questions can be displayed on the display interface from top to bottom or from left to right according to the arrangement sequence, so that the questions with higher heat and frequently asked questions of the user are displayed at positions which are easy to see first according to the reading habit of the user, the display of the push questions is more in accordance with the reading habit of the user, and the user experience is improved.

According to the information pushing method, the user problem corresponding to the center point of each problem cluster is determined to be the candidate problem, at least one pushing problem is determined in the candidate problem corresponding to each problem cluster, the user problem in the semantic distribution center in each problem cluster can be determined to be the candidate problem, and the candidate problem can cover more semantics of the user problem in each problem cluster, so that the pushing problem is more in line with the intention of the user.

Referring to fig. 7, fig. 7 is a flowchart of an information pushing method according to an embodiment of the present application, which is applied to the server. The method includes S501 to S507.

S501: a dialog log corresponding to the portal identification is obtained.

S502: each user question entered by the user in the dialog log is extracted.

S503: and extracting semantic features of each user problem to obtain feature vectors of each user problem.

S504: and clustering the feature vectors of the user problems to obtain a plurality of problem clusters.

And S505, acquiring the distance between the feature vector of each user problem and the center point of the problem cluster.

For each problem cluster, a feature vector for each user problem in the problem cluster may be calculated, a distance from a center point of the problem cluster. The distance between the feature vectors can represent the semantic similarity between the user questions corresponding to the feature vectors. The larger the distance, the smaller the semantic similarity. Alternatively, the distance may be a cosine distance, euclidean distance, manhattan distance, or the like.

S506: user questions characterized by feature vectors less than a specified distance are determined as candidate questions.

The specified distance may be used to characterize a semantic similarity threshold. As one embodiment, the specified distance may be a preset distance. As another implementation, the specified distance may be an average distance between the center point of each problem cluster and other points of the problem cluster.

After the distance between the feature vector of each user problem in the problem cluster and the center point of the problem cluster is obtained, determining the feature vector with the distance smaller than the specified distance, and determining the user problem represented by the feature vector as a candidate problem corresponding to the problem cluster. That is, the candidate questions are questions in each question cluster that are close to the center of the semantic distribution and can be used to represent the semantics of the question cluster.

It should be noted that the candidate questions include user questions corresponding to the center points of the question clusters. In some embodiments, if there is no feature vector smaller than the specified distance in the question cluster, the candidate question of the question cluster is the user question corresponding to the center point.

S507: and determining at least one push problem in the candidate problems corresponding to each problem cluster.

In some embodiments, the candidate questions corresponding to each question cluster may be respectively ranked, the candidate question ranked first is determined as the target candidate question of the question cluster, and at least one push question is determined in the target candidate question corresponding to each question cluster. The candidate questions may be ranked according to their scoring values, so that the target candidate questions are determined in each question cluster by considering both semantics and various factors. Reference is made to the foregoing examples for a specific implementation of ranking according to scoring values. Each question cluster corresponds to one push question, so that the semantic difference among the push questions is larger, and more user intentions are covered when the same number of push questions are displayed.

In some implementations, S507 may include obtaining a number of user questions in each question cluster; sorting the candidate questions in a descending order according to the number; a specified number of candidate questions ordered first are determined as at least one push question. Alternatively, the candidate questions may be each candidate question in each question cluster, or may be target candidate questions.

According to the information pushing method, the distance between the feature vector of each user problem and the center point of the problem cluster is obtained; determining a user problem characterized by a feature vector smaller than a specified distance as a candidate problem; and determining at least one push problem in the candidate problems corresponding to each problem cluster. Because the center of each question cluster can represent the semantic distribution center, candidate questions are determined according to the distance from the center point, more semantics of user questions can be covered in each question cluster, and therefore pushing questions are more in line with the intention of users.

Referring to fig. 10, fig. 10 is a flowchart of an information pushing method according to an embodiment of the present application, which is applied to the server. The method includes S601 to S606.

S601: a dialog log corresponding to the portal identification is obtained.

S602: and clustering the plurality of user questions in the dialogue log to obtain a plurality of question clusters.

S603: at least one push question is determined based on the plurality of question clusters.

S604: and acquiring at least one preset specified problem and the priority of the specified problem.

The specified problem is a preset problem. For example, the specified question may be configured by at least one question of the operator. As one embodiment, the assignment problem is the same for each entry identification. As another embodiment, a specified problem may be set for each entry identification separately.

Specifically, the priority is used to determine the order of the assignment questions and push questions. For example, the priority may be that the specified question is arranged before the push question, or that the specified question is arranged after the push question. As another example, the priority is used to determine a location at which to insert the specified issue into the at least one push issue.

S605: and arranging and combining at least one designated problem and at least one push problem according to the priority to obtain a problem combination.

And ranking and combining the at least one specified question and the at least one push question according to the priority to obtain a question combination comprising the at least one specified question and the at least one push question. For example, when a priority is assigned to a problem and a priority is assigned to a problem 1 and a priority is assigned to a problem 2, the problems are combined to be problem 1 and problem 2.

As one embodiment, the specified questions and the priorities of the specified questions are stored in the server, and S604 and S605 are performed by the server. As another embodiment, the specified questions and the priorities of the specified questions are stored in the terminal device, and the server may push the push questions to the terminal device, which performs S604 and S605.

S606: the combination of questions is pushed to the user interface.

The problem combinations are pushed to the user interface, the problem combinations can be displayed according to the arrangement sequence of the problem combinations, push problems obtained from dialogue logs of users and preset appointed problems can be displayed on the user interface, and therefore requirements for displaying specific content in some business scenes are met, and application scenes of the information push method in the embodiment of the application are enriched.

According to the information pushing method provided by the embodiment of the invention, after at least one pushing problem is determined based on a plurality of problem clusters, at least one preset designated problem and the priority of the designated problem are obtained, at least one designated problem and at least one pushing problem are arranged and combined according to the priority to obtain a problem combination, and the problem combination is pushed to a user interface, so that the problem combination of the pushing problem and the designated problem can be displayed on the user interface, the requirement of specific content display in some business scenes is met, and the application scene of information pushing is enriched.

Referring to fig. 11, fig. 11 is a flowchart of an information pushing method according to an embodiment of the present application, which is applied to the server. The method includes S701 to S706.

S701: a dialog log corresponding to the portal identification is obtained.

S702: and clustering the plurality of user questions in the dialogue log to obtain a plurality of question clusters.

S703: at least one push question is determined based on the plurality of question clusters.

S704: at least one preset standard problem is acquired.

The standard question is a preset question in the form of text with standard format and standard content, has definite semantics and can be accurately identified by a server or a terminal device. Optionally, for each standard question, its corresponding standard answer may also be configured.

In some implementations, the standard questions may be questions stored in a common question solution (Frequently Asked Questions, FAQ) knowledge base after unified processing of user questions with similar semantics. Since different users may have different expressions for the same problem, some common standard problems may be defined in the FAQ knowledge base for subsequent problem determination operations. For example, "how to increase credit card line" is a standard problem, and similar user problems corresponding thereto include "how little credit card line is handled", "how much credit card line is made a bit higher", and so on.

S705: and acquiring the semantic similarity of each standard problem and each push problem.

For each standard question, the semantic similarity of the standard question to each push question may be obtained based on text matching.

As an implementation mode, semantic extraction can be respectively carried out on the standard problem and the pushing problem based on a preset semantic understanding model, vector representations of the standard problem and the user problem are obtained, and then the semantic similarity is determined according to the distance between the vector representations, and the smaller the distance is, the larger the semantic similarity is. For example, the semantic similarity of each standard question to each push question may be determined by a deep semantic model (Deep Structured Semantic Models, DSSM).

As another implementation manner, cross features between the input standard problem and the push problem can be extracted through a preset depth model to obtain a matching signal tensor, and then the matching signal tensors are aggregated to obtain the semantic similarity. For example, the predetermined depth model may be an ARC-II (Convolutional Neural Network Architectures for Matching Natural Language), matchPyramid (Text Matching as Image Recognition) or the like model.

The embodiment of the application is not limited to a specific implementation of how to obtain the semantic similarity.

S706: if the semantic similarity is larger than the specified threshold, updating the push problem corresponding to the semantic similarity into a standard problem.

The specified threshold is a preset semantic similarity threshold. If the semantic similarity is larger than the specified threshold, determining that the semantic similarity corresponds to the push problem and the semantic approximation of the standard problem, updating the push problem into the standard problem, and pushing the updated push problem to the user interface. As an implementation manner, when the standard questions are configured with standard answers, the standard answers corresponding to the standard questions can be pushed to the user interface, and when the user touches or inputs the pushed questions on the interactive interface, the standard answers are displayed on the terminal device.

In some embodiments, if the semantic similarity is less than or equal to the specified threshold, the correction processing may be performed on the push problem based on the dictionary and the rule, so as to correct the wrongly written word or the wrong grammar in the push problem, and update the push problem to the corrected problem, thereby improving the readability of the push problem.

Because the standard questions are preset, the push questions before updating are questions input by the real user, and the standard questions are more smooth and easier to accurately identify relative to the semantics of the questions input by the real user. By replacing the push questions with standard questions with similar semantics, the readability of the push questions can be improved, the accuracy of the question identification can be improved, and the situation that the push questions generated by real user questions possibly have unsmooth semantics can be avoided.

According to the information pushing method provided by the embodiment of the application, after at least one pushing problem is determined based on a plurality of problem clusters, the semantic similarity of each standard problem and each pushing problem is obtained through obtaining at least one preset standard problem, if the semantic similarity is larger than a specified threshold, the pushing problem corresponding to the semantic similarity is updated to be the standard problem, and the standard problem can be used for replacing the pushing problem under the condition that the semantics are similar, so that the legibility of the pushing problem is improved.

Referring to fig. 12, fig. 12 is a flowchart of an information pushing method according to an embodiment of the present application, which is applied to the server. The method includes S801 to S803.

S801: every preset time interval, the dialogue log corresponding to the entrance identifier is updated to be the dialogue log of the appointed time.

The preset time is a preset time interval for updating the dialogue log, and the frequency of updating the dialogue log can be represented. The designated time is a preset time length of the acquired dialogue log. The preset time and the specified time may be set by default by the server or may be configured by an operator. In particular, for the embodiment of obtaining the dialogue log corresponding to the specified time of the entry identifier, please refer to the foregoing embodiment, which is not described herein.

As an embodiment, the specified time may be a length of time from the current time. For example, the preset time is 30 minutes, the specified time is 12 hours from the current time, and the dialogue log corresponding to the entry mark is updated every 30 minutes to the dialogue log of the entry mark in the past 12 hours. By updating the dialogue log at preset time intervals, the dialogue log with better timeliness can be obtained, and data support is provided for the follow-up more accurate determination of user intention. Alternatively, the server may maintain a dialogue log corresponding to the designated time of the portal identification every predetermined time, and remove the previous dialogue log, thereby saving storage space.

In some embodiments, after the dialogue log corresponding to the specified time of the entry identifier is obtained, it may be determined whether the user log at the specified time satisfies a preset update condition, and if so, the dialogue log is updated to the dialogue log at the specified time. As one way, the preset update condition may be a preset quantity threshold for determining whether the user question in the dialogue log at the specified time is greater than the quantity threshold. By presetting the update conditions, the subsequent processing is not performed under the condition that the dialogue log does not meet the conditions, so that the data processing efficiency is improved.

S802: and clustering the plurality of user questions in the dialogue log at the appointed time to obtain a plurality of updated question clusters.

The server may perform clustering processing on a plurality of questions in the dialogue log at a specified time, and the embodiment of obtaining updated plurality of question clusters may refer to the foregoing embodiment, which is not described herein again.

S803: at least one push question is updated based on the updated plurality of question clusters.

The embodiment of updating at least one push problem based on the updated plurality of problem clusters may refer to the foregoing embodiment, and will not be described herein. The server can push updated push problems to the terminal equipment, so that the timeliness change of the user intention can be captured more accurately under the condition of considering the timeliness information, and the accuracy of information push is higher.

According to the information pushing method, the dialogue logs corresponding to the entrance identifiers are updated to be the dialogue logs with the appointed time through preset time intervals, the plurality of user problems in the dialogue logs with the appointed time are clustered to obtain updated plurality of problem clusters, at least one pushing problem is updated based on the updated plurality of problem clusters, and the pushing problem can be updated every preset time interval, so that the pushing problem is determined in real time, and timeliness of the pushing problem is improved.

Referring to fig. 13, fig. 13 is a flowchart of an information pushing method according to an embodiment of the present application, which is applied to the above terminal device. The method includes S901 to S903.

S901: if a dialogue request of a user is detected, an entry identification of a user interface for realizing dialogue with the user is acquired.

The dialogue request is used for representing the operation of triggering dialogue entrance in the man-machine interaction process of the user. Specifically, conversation inlets are arranged on some interactive interfaces of the client application program, and in response to a conversation request of a user in a man-machine interaction process, the current interactive interface can be jumped to a user interface for realizing conversation with the user. Wherein each user interface implementing a dialog corresponds to a dialog portal, each dialog portal corresponds to a portal identification, and the portal identification can be used to identify the user interface. When a user's request for a conversation is detected, a conversation portal triggered by the user may be determined, and a portal identification of the conversation portal may be determined as a portal identification of a user interface implementing the conversation with the user.

In some embodiments, after step S901, the information pushing method may further include acquiring a dialogue log of the user corresponding to the portal identification, and transmitting the dialogue log to the server.

Specifically, after the portal identifier is acquired, a dialogue log of the user corresponding to the portal identifier on the current terminal device may be acquired, the terminal device establishes a connection with the server, and the dialogue log is sent to the server. Wherein the conversation log is a historical conversation record of the user in text format, the conversation log may include an entry identification of a user interface generating the conversation, at least one user message, and at least one system message. Optionally, the dialogue log may also include the time of each message or the user identification of the current terminal device, etc. As an embodiment, the terminal device may send the session log of the present session to the server after each session is ended.

S902: at least one push question corresponding to the entry identification is acquired.

The at least one push question is a question determined based on the obtained plurality of question clusters after clustering a plurality of user questions, wherein the plurality of user questions are questions input by users in a dialogue log corresponding to the entry identifier. Optionally, the terminal device may further obtain an answer corresponding to the push question.

In some embodiments, the terminal device may also obtain the order of push questions from the server.

The specific implementation of the server sending the push problem refers to the foregoing embodiments, and will not be described herein.

S903: the push question is displayed on the user interface.

The user interface is for implementing a dialog with a user. An information display area is arranged on the user interface and is used for displaying the acquired pushing problem. As an embodiment, the push questions may be respectively arranged on a plurality of keys in the information display area, and the push question text may be displayed on each key. If the click operation of the user on the push question on the interactive interface is obtained, the push question is input on the interactive interface quickly, that is, the click operation of the push question can be used for replacing the user typing or voice input push question. Optionally, after the push question is quickly input, the terminal device may obtain and display an answer to the push question.

In some implementations, at most a specified number of push questions can be displayed on the user interface. Wherein the specified number may be used to limit the number of push questions per acquisition. As one implementation mode, a push trigger mark for acquiring push questions can be arranged on the user interface, and the terminal equipment responds to the operation of a user on the push trigger mark and can send request information for acquiring more push questions to the server so as to acquire more push questions to update the push questions currently displayed. For example, the user operation to push the trigger may be clicking a button on the user interface to "change question".

In some embodiments, after the terminal device obtains the ranking order of the push questions, the terminal device may display the push questions in the ranking order on the user interface. For example, push questions may be displayed in a top-to-bottom order, or left-to-right order. The arrangement sequence can represent the heat of the pushing problem, and according to the reading habit and the arrangement sequence of the user, the problem frequently asked by the user can be displayed at a position which is easy to see, so that the method and the device more accord with the reading habit of the user, and improve the user experience.

According to the information pushing method, if the dialogue request of the user is detected, the entrance identification of the user interface for realizing the dialogue with the user is obtained, at least one pushing problem corresponding to the entrance identification is obtained, the at least one pushing problem is a problem determined by a plurality of obtained problem clusters after clustering processing is carried out on the plurality of user problems, the plurality of user problems are problems input by the user in the dialogue log corresponding to the entrance identification, the pushing problem is displayed on the user interface, the pushing problem corresponding to the entrance identification can be displayed on the user interface, the pushing problem is determined according to the historical user problems of the same entrance identification, the pushing problem is enabled to accord with user intention, and user experience is improved.

It should be understood that the foregoing examples are merely illustrative of the application of the method provided in the embodiments of the present application in one specific context, and are not intended to limit the embodiments of the present application. More different applications can be realized based on the method provided by the embodiment of the application.

Referring to fig. 14, fig. 14 is a block diagram illustrating a structure of an information pushing device according to an embodiment of the present application. The information pushing apparatus 1400 includes: a log acquisition module 1410, a cluster processing module 1420, and a problem determination module 1430, wherein: a log obtaining module 1410, configured to obtain a dialogue log corresponding to an entry identifier, where the entry identifier is used to identify a user interface for implementing a dialogue with a user; the clustering processing module 1420 is configured to perform clustering processing on a plurality of user questions in the dialogue log to obtain a plurality of question clusters; a question determination module 1430 for determining at least one push question based on the plurality of question clusters, the at least one push question for being pushed and displayed on the user interface.

Further, the cluster processing module 1420 includes: a question extraction sub-module, configured to extract each of the user questions input by a user in the dialog log; the feature extraction sub-module is used for extracting semantic features of each user problem to obtain a feature vector of each user problem; and the feature clustering sub-module is used for carrying out clustering processing on the feature vectors of the user problems to obtain the problem clusters.

Further, the problem determination module 1430 includes: a central problem determination submodule, configured to determine a user problem corresponding to a central point of each problem cluster as the candidate problem; and a push question determination submodule, configured to determine the at least one push question from the candidate questions corresponding to each question cluster.

Further, the problem determination module 1430 includes: a distance acquisition sub-module, configured to acquire a distance between a feature vector of each user problem and a center point of the problem cluster; a candidate problem determination submodule for determining a user problem characterized by a feature vector smaller than a specified distance as a candidate problem; and a push question determination submodule, configured to determine the at least one push question from the candidate questions corresponding to each question cluster.

Further, the information pushing device 1400 further includes: and the pushing module is used for pushing the at least one pushing problem and the arrangement sequence of the at least one pushing problem to the user interface so as to display the at least one pushing problem on the user interface in the arrangement sequence.

Further, the information pushing device 1400 further includes: the system comprises a specified problem acquisition module, a priority judgment module and a priority judgment module, wherein the specified problem acquisition module is used for acquiring at least one preset specified problem and the priority of the specified problem; the problem combination acquisition module is used for arranging and combining the at least one designated problem and the at least one push problem according to the priority to obtain a problem combination; and the question combination pushing module is used for pushing the question combination to the user interface.

Further, the information pushing device 1400 further includes: the standard problem acquisition module is used for acquiring at least one preset standard problem; the similarity acquisition module is used for acquiring the semantic similarity between each standard problem and each push problem; and the standard problem updating module is used for updating the push problem corresponding to the semantic similarity into the standard problem if the semantic similarity is larger than a specified threshold.

Further, the log obtaining module 1410 further includes: the log updating module is used for updating the dialogue log corresponding to the entrance identifier into a dialogue log of a designated time every interval for a preset time; the cluster processing module further includes: the problem cluster updating module is used for carrying out clustering processing on a plurality of user problems in the dialogue log at the appointed time to obtain a plurality of updated problem clusters; and the push question updating module is used for updating the at least one push question based on the updated plurality of question clusters.

Referring to fig. 15, fig. 15 is a block diagram illustrating a structure of an information pushing device according to an embodiment of the present application. The information pushing apparatus 1500 will be described below with respect to the structural block diagram shown in fig. 15, and includes: an entry identification acquisition module 1510, a problem acquisition module 1520, and a problem display module 1530, wherein: an entry id acquisition module 1510, configured to acquire an entry id of a user interface for implementing a dialogue with a user if a dialogue request of the user is detected; a question obtaining module 1520, configured to obtain at least one push question corresponding to the portal identifier, where the at least one push question is a question determined based on a plurality of obtained question clusters after clustering a plurality of user questions, and the plurality of user questions are questions input by a user in a dialogue log corresponding to the portal identifier; and a question display module 1530 configured to display the push question on the user interface.

Further, the entry id acquisition module 1510 further includes: and the log sending module is used for obtaining the dialogue log of the user corresponding to the entry identifier and sending the dialogue log to a server.

It will be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working process of the apparatus and modules described above may refer to the corresponding process in the foregoing method embodiment, which is not repeated herein.

In several embodiments provided herein, the coupling of the modules to each other may be electrical, mechanical, or other.

Referring to fig. 16, a block diagram of an electronic device according to an embodiment of the present application is shown. The electronic device 1600 may be a smart phone, tablet, electronic book, etc. capable of running applications. The electronic device 1600 in this application may include one or more of the following components: processor 1610, memory 1620, and one or more application programs, wherein the one or more application programs may be stored in memory 1620 and configured to be executed by the one or more processors 1610, the one or more program(s) configured to perform a method as described in the foregoing method embodiments.

Processor 1610 may include one or more processing cores. Processor 1610 uses various interfaces and lines to connect various portions of the overall electronic device 1600, performing various functions of the electronic device 1600 and processing data by executing or executing instructions, programs, code sets, or instruction sets stored in memory 1620, and invoking data stored in memory 1620. Alternatively, the processor 1610 may be implemented in at least one hardware form of digital signal processing (Digital Signal Processing, DSP), field programmable gate array (Field-Programmable Gate Array, FPGA), programmable logic array (Programmable Logic Array, PLA). Processor 1610 may integrate one or a combination of several of a central processing unit (Central Processing Unit, CPU), an image processor (Graphics Processing Unit, GPU), and a modem, etc. The CPU mainly processes an operating system, a user interface, an application program and the like; the GPU is used for being responsible for rendering and drawing of display content; the modem is used to handle wireless communications. It will be appreciated that the modem may not be integrated into the processor 1610, but may be implemented solely by a single communication chip.

The Memory 1620 may include a random access Memory (Random Access Memory, RAM) or a Read-Only Memory (Read-Only Memory). The memory 1620 may be used to store instructions, programs, code sets, or instruction sets. The memory 1620 may include a stored program area and a stored data area, wherein the stored program area may store instructions for implementing an operating system, instructions for implementing at least one function (such as a touch function, a sound playing function, an image playing function, etc.), instructions for implementing the various method embodiments described below, etc. The storage data area may also store data created by the electronic device 1600 in use (e.g., phonebook, audiovisual data, chat log data), and the like.

Referring to fig. 17, a block diagram of a computer readable storage medium according to an embodiment of the present application is shown. The computer readable storage medium 1700 has stored therein program code that can be invoked by a processor to perform the methods described in the method embodiments described above.

The computer readable storage medium 1700 may be an electronic memory such as a flash memory, an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (erasable programmable read only memory, EPROM), a hard disk, or a ROM. Optionally, computer readable storage medium 1000 includes a non-volatile computer readable medium (non-transitory computer-readable storage medium). The computer readable storage medium 1700 has storage space for program code 1710 that performs any of the method steps described above. The program code can be read from or written to one or more computer program products. The program code 1710 may be compressed, for example, in a suitable form.

Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present application, and are not limiting thereof; although the present application has been described in detail with reference to the foregoing embodiments, one of ordinary skill in the art will appreciate that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not drive the essence of the corresponding technical solutions to depart from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An information pushing method is characterized by comprising the following steps:

obtaining a dialogue log corresponding to an entry identifier, wherein the entry identifier is used for identifying a user interface for realizing dialogue with a user;

clustering a plurality of user questions in the dialogue log to obtain a plurality of question clusters; and

Determining at least one push question based on the plurality of question clusters, the at least one push question for being pushed and displayed on the user interface;

the determining at least one push question based on the plurality of question clusters includes:

acquiring a user portrait of a current interactive user, determining a problem cluster with the same user portrait label as a target problem cluster, and determining the push problem according to the user problem of the target problem cluster;

After the determining at least one push question based on the plurality of question clusters, the method further comprises:

acquiring at least one preset specified problem and the priority of the specified problem;

according to the priority, the at least one designated problem and the at least one push problem are arranged and combined to obtain a problem combination; and

Pushing the combination of questions to the user interface.

2. The method of claim 1, wherein clustering the plurality of user questions in the dialogue log to obtain a plurality of question clusters comprises:

extracting each user question input by a user in the dialogue log;

extracting semantic features of each user problem to obtain feature vectors of each user problem; and

And clustering the feature vectors of the user questions to obtain the question clusters.

3. The method of claim 2, wherein the determining at least one push question based on the plurality of question clusters comprises:

determining user questions corresponding to the center points of the question clusters as candidate questions; and

And determining the at least one push question from the candidate questions corresponding to each question cluster.

4. The method of claim 2, wherein the determining at least one push question based on the plurality of question clusters comprises:

acquiring the distance between the feature vector of each user problem and the center point of the problem cluster;

determining a user problem characterized by a feature vector smaller than a specified distance as a candidate problem; and

5. The method according to claim 3 or 4, wherein said determining the at least one push question among the candidate questions corresponding to each question cluster comprises:

acquiring the number of user questions in each question cluster;

sorting the candidate questions in descending order according to the number; and

Determining a specified number of the candidate questions ordered first as the at least one push question.

6. The method of claim 5, wherein after the determining the first-ordered specified number of the candidate questions as the at least one push question, the method further comprises:

pushing the at least one push question and an order of arrangement of the at least one push question to the user interface to display the at least one push question in the order on the user interface.

7. The method of claim 1, wherein after the determining at least one push question based on the plurality of question clusters, the method further comprises:

acquiring at least one preset standard problem;

acquiring semantic similarity of each standard problem and each push problem; and

And if the semantic similarity is larger than a specified threshold, updating the push problem corresponding to the semantic similarity into the standard problem.

8. The method of claim 1, wherein the obtaining a dialog log corresponding to an entry identification comprises:

updating the dialogue log corresponding to the entrance identifier into a dialogue log of a designated time every interval preset time;

the clustering processing is performed on the plurality of user questions in the dialogue log to obtain a plurality of question clusters, including:

clustering a plurality of user questions in the dialogue log at the appointed time to obtain updated question clusters;

updating the at least one push question based on the updated plurality of question clusters.

9. An information pushing method is characterized by comprising the following steps:

If a dialogue request of a user is detected, acquiring an entry identification of a user interface for realizing dialogue with the user;

acquiring at least one push question corresponding to the entrance identifier, wherein the at least one push question is a question determined according to a user question of a target question cluster after clustering a plurality of user questions to obtain a plurality of question clusters, the plurality of user questions are questions input by users in a dialogue log corresponding to the entrance identifier, and the target question cluster is a question cluster with the same user portrait label as a user portrait of a current interactive user in the plurality of question clusters; and

Displaying the push question on the user interface;

receiving a question combination, wherein the question combination is based on the acquisition of at least one preset specified question and the priority of the specified question; a combination obtained by arranging and combining the at least one designated problem and the at least one push problem according to the priority;

the question combination is displayed on the user interface.

10. The method of claim 9, wherein after the obtaining of the entry identification of the user interface for implementing a conversation with the user if the user's conversation request is detected, the method further comprises:

And acquiring a dialogue log of the user corresponding to the entry identifier, and sending the dialogue log to a server.

11. An information pushing apparatus, characterized by comprising:

the log acquisition module is used for acquiring a dialogue log corresponding to an entry identifier, wherein the entry identifier is used for identifying a user interface for realizing dialogue with a user;

the clustering processing module is used for carrying out clustering processing on a plurality of user problems in the dialogue log to obtain a plurality of problem clusters; and

A question determination module for determining at least one push question based on the plurality of question clusters, the at least one push question being for being pushed and displayed on the user interface;

the problem determination module is specifically configured to obtain a user portrait of a current interactive user, determine a problem cluster with the same user portrait tag as a target problem cluster, and determine the push problem according to a user problem of the target problem cluster;

the information pushing device further comprises a specified problem acquisition module, a problem combination acquisition module and a problem combination pushing module, wherein:

the specified problem acquisition module is used for acquiring at least one preset specified problem and the priority of the specified problem;

The problem combination acquisition module is used for arranging and combining the at least one designated problem and the at least one push problem according to the priority to obtain a problem combination;

the question combination pushing module is used for pushing the question combination to the user interface.

12. An information pushing apparatus, characterized by comprising:

the entrance identification acquisition module is used for acquiring an entrance identification of a user interface for realizing dialogue with a user if a dialogue request of the user is detected;

the problem acquisition module is used for acquiring at least one push problem corresponding to the entrance identifier, wherein the at least one push problem is a problem determined according to user problems of a target problem cluster after a plurality of user problems are clustered to obtain a plurality of problem clusters, the plurality of user problems are problems input by users in a dialogue log corresponding to the entrance identifier, and the target problem cluster is a problem cluster with the same user portrait label as a user portrait of a current interactive user in the plurality of problem clusters; and

The problem display module is used for displaying the push problem on the user interface;

the problem acquisition module is further used for receiving a problem combination, wherein the problem combination is based on acquisition of at least one preset specified problem and priority of the specified problem; a combination obtained by arranging and combining the at least one designated problem and the at least one push problem according to the priority;

The question display module is further configured to display the question combination on the user interface.

13. An electronic device, comprising:

one or more processors;

a memory;

one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs configured to perform the method of any of claims 1-10.

14. A computer readable storage medium, characterized in that the computer readable storage medium has stored therein a program code, which is callable by a processor for performing the method according to any one of claims 1-10.