WO2018201964A1

WO2018201964A1 - Processing method for session information, server, and computer readable storage medium

Info

Publication number: WO2018201964A1
Application number: PCT/CN2018/084639
Authority: WO
Inventors: 金星明; 李鹏; 李科
Original assignee: 腾讯科技（深圳）有限公司
Priority date: 2017-05-05
Filing date: 2018-04-26
Publication date: 2018-11-08
Also published as: CN106911812A

Abstract

Disclosed are a processing method for session information, a server, and a computer readable storage medium. The method comprises: parsing session information from a session request launched by a terminal; classifying and clustering the session information to produce at least one type of feature parameter; distributing the session information to an information feedback platform corresponding to the at least one type of feature parameter; producing credible first feedback information on the basis of feedback information transmitted by the information feedback platform, returning the credible first feedback information via a session request response to the terminal; tracking a current session state of the terminal on the basis of a state tracking policy, determining a response operation for the next step on the basis of session information collected from the current session state, and executing the response operation then returning a response result to the terminal.

Description

Method for processing session information, server and computer readable storage medium

Cross-reference to related applications

The present application is filed on the basis of the Chinese Patent Application No. PCT Application No.

Technical field

The present application relates to information processing technologies, and in particular, to a method for processing session information, a server, and a computer readable storage medium.

Background technique

The interactive feedback of information can be implemented in a variety of ways, from face-to-face interaction between people, online interaction or offline offline interaction. For example, if the user asks a question, the answer to the question can be obtained in all three ways. However, the specificity, accuracy, real-time, and interactive effects of the answers given in different ways have their own drawbacks. For example, face-to-face interaction between people, if the customer service staff is just not there, then the interaction can not continue, the interaction effect is worse; for example, online online interaction or offline offline interaction, does not have the diversity of customer service staff to deal with the problem And the questionable and targeted, usually choose the answer from the set fixed options.

How to make the effect of information interaction feedback artificially intelligent, at least solve the above drawbacks is a technical problem to be solved. However, in the related art, there is no effective solution to this problem.

Summary of the invention

In view of this, the embodiment of the present application provides a method for processing session information, a server, and a computer readable storage medium, which solve at least the problems existing in the prior art.

A method for processing session information in an embodiment of the present application, the method includes:

Parsing session information from a session request initiated by the terminal;

Performing classification and clustering processing on the session information to obtain at least one category feature parameter;

Distributing the session information to an information feedback platform corresponding to the at least one category feature parameter;

Obtaining the trusted first feedback information according to the feedback information sent by the information feedback platform, and returning the trusted first feedback information to the terminal by using the session request response;

The current session state of the terminal is tracked according to the state tracking policy, and the next response operation is determined according to the session information collected from the current session state, and the response result is returned to the terminal after the response operation is performed.

A server in an embodiment of the present application, the server includes:

a receiving unit, configured to parse the session information from the session request initiated by the terminal;

a class processing unit configured to perform classification and clustering processing on the session information to obtain at least one category feature parameter;

a distribution unit, configured to distribute the session information to an information feedback platform corresponding to the at least one category feature parameter;

The response unit is configured to obtain trusted first feedback information according to the feedback information sent by the information feedback platform, and return the trusted first feedback information to the terminal by using a session request response;

The status tracking unit is configured to track the current session state of the terminal according to the status tracking policy, determine a next response operation according to the session information collected from the current session state, and return the response result after performing the response operation. To the terminal.

A computer readable storage medium of the embodiment of the present application, wherein a computer program is stored thereon, wherein the computer program is executed by the processor to implement the step of processing the session information.

A server in an embodiment of the present application, the server includes:

a memory configured to store a computer program capable of running on a processor;

And a processor configured to perform the step of processing the session information described above when the computer program is run.

A method for processing session information according to an embodiment of the present application, the method is performed by a server, the server includes one or more processors and a memory, and one or more programs, wherein the one or more programs The program is stored in a memory, and the program may include one or more units each corresponding to a set of instructions, the one or more processors being configured to execute the instructions; the method comprising:

Parsing session information from a session request initiated by the terminal;

The method for processing session information in the embodiment of the present application includes: parsing session information from a session request initiated by a terminal; classifying and clustering the session information to obtain at least one category feature parameter; and distributing the session information Giving an information feedback platform corresponding to the at least one category feature parameter; obtaining trusted first feedback information according to the feedback information sent by the information feedback platform, and returning the trusted first feedback information to the The terminal tracks the current session state of the terminal according to the state tracking policy, determines a response operation of the next step according to the session information collected from the current session state, and returns the response result to the terminal after performing the response operation. .

In the embodiment of the present application, the session information is classified and clustered to obtain at least one category feature parameter; the session information is distributed to the information feedback platform corresponding to the at least one category feature parameter. The processing (classification and clustering) of the session information (including the problem) obtains the category feature parameters, and gives the corresponding information feedback platform according to the category feature parameters to obtain a targeted accurate answer (the answer is not set, nor fixed) ), is a real-time interactive mechanism, the interactive mode with questions and answers is targeted, and the interaction effect is good. Moreover, the feedback information sent by the information feedback platform is not immediately fed back to the terminal, but after the trusted first feedback information is obtained according to the feedback information, the trusted first feedback information is returned to the terminal through the session request response. Since the answer is credentially analyzed and then fed back to the terminal, the accuracy of the reply is ensured.

DRAWINGS

1 is a schematic diagram of hardware entities of each party performing information interaction in an embodiment of the present application;

2 is a schematic diagram of an implementation process of a method according to Embodiment 1 of the present application;

3 is a schematic diagram of a system architecture according to Embodiment 1 of the present application;

4 is a schematic diagram of a hardware structure of a server according to Embodiment 1 of the present application;

Figure 5-6 is a plurality of schematic diagrams of the state machine;

FIG. 7 is a flowchart of a session process applied to an application scenario of an embodiment of the present application.

detailed description

The implementation of the technical solution will be further described in detail below with reference to the accompanying drawings.

A mobile terminal implementing various embodiments of the present application will now be described with reference to the accompanying drawings. In the following description, the suffixes such as "module", "component" or "unit" used to denote elements are merely illustrative of the embodiments of the present application, and do not have a specific meaning per se. Therefore, "module" and "component" can be used in combination.

In the following detailed description, numerous specific details are set forth in the However, it is apparent to those skilled in the art that the present invention may be practiced without these specific details. In other instances, well-known methods, procedures, components, circuits, and networks are not described in detail to avoid unnecessarily obscuring aspects of the embodiments.

In addition, although the terms "first", "second", etc. are used herein to describe various elements (or various thresholds or various applications or various instructions or various operations), etc., these elements (or thresholds) Or application or instruction or operation) should not be limited by these terms. These terms are only used to distinguish one element (or threshold or application or instruction or operation) and another element (or threshold or application or instruction or operation). For example, the first operation may be referred to as a second operation, and the second operation may also be referred to as a first operation, without departing from the scope of the present application, the first operation and the second operation are operations, but the two are not the same The operation is only.

The steps in the embodiments of the present application are not necessarily processed in the order of the steps described. The steps may be selectively arranged to be reordered according to requirements, or the steps in the embodiment may be deleted, or the steps in the embodiment may be added. The description of the steps in the application examples is only an optional combination of the steps, and does not represent a combination of the steps in the embodiments of the present application. The order of the steps in the embodiments is not to be construed as limiting the present application.

The term "and/or" in the embodiments of the present application refers to any and all possible combinations including one or more of the associated listed items. It is also to be understood that the phrase "including/comprising", when used in the specification, is intended to mean the presence of the stated features, integers, steps, operations, components and/or components, but does not exclude one or more other features, integers The presence or addition of steps, operations, elements and/or components and/or groups thereof.

The intelligent terminal (such as a mobile terminal) of the embodiment of the present application can be implemented in various forms. For example, the mobile terminal described in the embodiments of the present application may include, for example, a mobile phone, a smart phone, a notebook computer, a digital broadcast receiver, a personal digital assistant (PDA, Personal Digital Assistant), a tablet computer (PAD), a portable multimedia player ( Mobile terminals such as PMP (Portable Media Player), navigation devices, and the like, and fixed terminals such as digital TVs, desktop computers, and the like. In the following, it is assumed that the terminal is a mobile terminal. However, those skilled in the art will appreciate that configurations in accordance with embodiments of the present application can be applied to fixed type terminals in addition to components that are specifically for mobile purposes.

With the development of Internet technology, communication technology, intelligent terminals and other fields, users' lives are becoming more and more intelligent. New artificial intelligence (AI) technology is becoming an important entry point for data, services, and products to access human life. With AI technology, users can interact with smart chat messages through a background server accessed by a chat bot or chat page. For example, in a shopping life scene of a user, usually calling the agent customer service, the agent customer service is a real person, that is to say, this traditional information interaction is a real conversation between people, if the agent is busy, It is impossible to respond to user inquiries in a timely manner. There is no such problem in smart chat, you can use virtual customer service to replace the real person. The virtual customer service is online at any time, can provide consultation at any time, and provide the default answer to the user for selection.

There are also certain problems with AI technology: the answer is fixed. Even if there are preset answers (fixed answers) based on historical analysis in the database, on the one hand, the questions asked by the users have breadth and depth, and it is difficult to cope with unpredicted situations. The default answers are not comprehensive; In terms of aspect, the default answer is provided to the user for selection. There is a difference between the chat mode and the answer mode in the real person interaction scene, which is not intelligent enough and needs technical perfection.

1 is a schematic diagram of hardware entities of each party performing information interaction in the embodiment of the present application. FIG. 1 includes: a terminal 1, a server 2, and a server 31-33. The server 31-33 is a specific implementation of multiple information feedback platforms, and only It only refers to the role, and does not limit the number of servers. The terminal 1 can be composed of a plurality of terminals 11-13, and exchanges information with the server 2 through wireless or wired means. As shown in FIG. 1 , the processing logic 10 of the embodiment of the present application is processing logic executed by the server, including: S1, parsing session information in the session request, and classifying and clustering the session information to obtain at least a category feature parameter, the session information is distributed to the at least one information feedback platform according to the at least one category feature parameter; S2, the feedback identified from the information credibility parameter according to the feedback information and the information credibility parameter Selecting the first feedback information with the highest credibility, and comparing the first feedback information with a threshold, and when the first feedback information is trusted, the first feedback information is passed The session request response is returned to the terminal; S3, tracking the current session state of the terminal according to the state tracking policy, determining a next response operation according to the session information collected from the current session state, and performing the response operation The response result is then returned to the terminal.

The example of FIG. 1 is only a system architecture example of the embodiment of the present application. The embodiment of the present application is not limited to the system structure described in FIG. 1 , and various embodiments of the method of the present application are proposed based on the system architecture described in FIG. 1 . .

In this document, multiple information feedback platforms may be multiple servers independent of the server 2; multiple information feedback platforms may also be multiple servers of the server 2, which may be referred to as multiple subsystems, specifically system.

As shown in FIG. 2, the method includes: receiving a session request initiated by a terminal, and parsing session information (101) from a session request initiated by the terminal. After entering the AI chat scenario, after the user asks a question, a session request is generated according to the proposed question, and the session request is sent to the server, and the server responds to the question raised by the user according to the policy and operation of the processing logic 10 in FIG. The response is fed back to the user via the session request response. In the entire AI chat scenario, the response to the session request and the session request can be understood as a session established between the foreground and the background, real-time monitoring of the session established by the terminal where the user is located and the server, and maintaining and tracking the change of the session state, according to the session state. Changes, adjust the server's strategy and respond. For example, in an example, a state tracking policy can be adopted, specifically, a state machine is used to maintain and track user session state, and a form fill method can be used to collect user requirements. For the form filling method, the purpose is: each time Try to extract as much information as possible from the user's question, instead of extracting only the information corresponding to the state each time, so the conversation is more natural.

A state machine is a mathematical model that represents a finite number of states and behaviors such as transitions and actions between these states. It reflects the input change from the beginning of the system to the present time, the transition indicates the state change, and describes it with the conditions that must be met to ensure that the transfer occurs; the action is a description of the activity to be performed at a given time. It includes various types of actions: 1) entry action: when entering state; 2) exit action: when exiting state; 3) input action: depending on current state and input condition; 4) transfer action: in specific When transferring. For example, a button command parsing program can be regarded as a state machine. Specifically, it is initially in the A state. When a button is triggered, it switches to the B state, and then triggers another button to switch to the C state, or returns to A state. This is an example of a simple button state machine.

In the embodiment of the present application, in the case that the state machine tracks the user session, the state of the state machine may include: a normal state, at least one information confirmation state, and an abnormal state. Under normal conditions, it is divided into two situations: when the user needs are clear and the user's demand is not clear. 1) When the user needs are clear, distribute it to each question and answer subsystem, and then choose the answer to return to the user; 2) When the user's demand is not clear, use the guiding method to further subdivide and clarify the user's needs. The question and answer mode is adopted to continue to collect the questions raised by the user, the server continues to analyze, and finally, the most accurate answer is directly fed back to the user without requiring the user to make a selection; 3) the embodiment of the present application also provides an exception. The processing strategy, when at least one information confirmation status fails to confirm, the session cannot be carried out, indicating that when the user question exceeds the domain range that the system can answer, the normal state is transferred to the abnormal state, and the session information is not required to be distributed to each question and answer subsystem. Instead, it responds according to the preset answer, and can perform the exception processing "smoothly" to avoid session interruption.

There are many AI chat scenarios applicable to the embodiments of the present application, such as an intelligent voice search engine, intelligent navigation and parking, intelligent robot service, and intelligent voice comparison. In a smart chat scene, when the user turns on the navigation mode, during the driving process, since the user is driving, the hands are occupied. At this time, through the intelligent chat of human-computer interaction, the best navigation route to the destination can be known in time. , or avoid congestion and re-select a navigation route and so on.

The method for processing the session information in the embodiment of the present application further includes: classifying and clustering the session information, obtaining at least one category feature parameter, and distributing the session information to the at least one category feature parameter. Information feedback platform (102). The information feedback platform is at least one, and respectively corresponds to different category feature parameters of the at least one category feature parameter. The user is classified, and after being clustered, it is distributed to different channels to obtain an initial answer. For example, splitting the information content included in the first information may result in multiple content items. There are multiple information feedback platforms, such as a knowledge base based Q&A feedback platform and a Q&A feedback platform based on Frequently Asked Questions (FAQ).

For the Q&A feedback platform of the knowledge base, the data of different knowledge types is stored in the form of a database, which belongs to a structured knowledge set within the limited domain, and the data is provided for query use, and the data is more comprehensive than the data of the FAQ question and answer feedback platform. For the FAQ question and answer feedback platform, the stored data is the user's "frequently asked questions" or "common questions."

The classification and clustering of users include: 1) classifying these content items first, first obtaining a relatively large classification, that is, realizing the first screening of the information content. 2) Because, the classification obtained by the first screening, to some extent, there may be information crossover or high information similarity. At this time, these situations can be analyzed and clustered to achieve The second screening of information content. With the embodiment of the present application, the multiple information items in the first information can be specifically distributed to one or more of the multiple information feedback platforms, which can improve the accuracy and processing efficiency of the information feedback.

The method for processing the session information in the embodiment of the present application further includes: obtaining the trusted first feedback information according to the feedback information sent by the information feedback platform, and returning the trusted first feedback information to the terminal by using the session request response. (103). Specifically, the feedback information and the information credibility parameter sent by the at least one information feedback platform are received, and the information credibility parameter is used to identify the credibility of the feedback information. The first feedback information with the highest credibility is selected from the feedback information identified by the information credibility parameter, and the first feedback information is compared with a threshold to obtain a comparison result. In the embodiment of the present application, the server can obtain the analysis result of the credibility analysis from at least one information feedback platform, and can also perform the calculation on the server itself, that is, 1) the server only receives the feedback information sent by at least one information feedback platform (eg, The server receives the feedback information sent by at least one information feedback platform, performs credibility analysis on the feedback information, and obtains at least one analysis result (such as information credibility parameter); 2) the server only Receiving feedback information sent by at least one information feedback platform (such as the answer to the question posed by the user) and the analysis result of the credibility analysis made by at least one information feedback platform, and performing a round of credibility analysis on the server , get at least one analysis result (such as information credibility parameters). Then, the server compares the feedback information with the highest credibility (such as the first feedback information) in the feedback information identified by the information credibility parameter with the threshold to determine whether the answer with the highest credibility is credible.

In the method for processing session information in the embodiment of the present application, when it is determined that the first feedback information is trusted according to the comparison result, the first feedback information is returned to the terminal by using a session request response. In the process of the embodiment of the present application, the answer is directly fed back to the user through the analysis of the credibility of the answer, so that the accurate and closest to the human interaction has a question and answer mode, and no user selection is required.

The method for processing the session information in the embodiment of the present application further includes: tracking a current session state of the terminal according to the state tracking policy, determining a next response operation according to the session information collected from the current session state, and executing the location The response result is returned to the terminal (104) after the response operation. In the embodiment of the present application, the server tracks the current session state of the terminal according to the state tracking policy, and the state machine can track the current session state, and the difference between “clear for user requirements” and “unclear user requirements” is tracked. , further distinguish different processing strategies. When the state tracking policy is implemented by the state machine, it needs to pay attention to: 1) state; 2) response operation (or response action) of the server. During the maintenance and tracking of the state machine, it is the maintenance and tracking of the information state and system actions of the session at that time. In addition to the state machine, the server can track the current session state of the terminal, and can also maintain state maintenance such as session update and session deletion in the current session state of the terminal. An example of state maintenance is: which step the current conversation progresses to and which processing strategy should be used.

An example of a state machine adopting the embodiment of the present application is that the state machine includes four states, namely: a normal state, a model confirmation state, an attribute confirmation state, and an abnormal state. In the process of state tracking, in the normal state, when the user needs are clear, the session information is distributed to the subsystem for processing; in the normal state, but when the user requirements are not clear, the user needs are confirmed by guiding and analyzing the details, and the user is clarified. After the demand, the session information is distributed to the subsystem for processing. In the process of analyzing the details, if the vehicle enters the vehicle confirmation status or the attribute confirmation status, the confirmation is performed. If the confirmation fails, the session information is out of the range that can be answered. , jumps from the normal state to the abnormal state.

According to the embodiment of the present application, the problem is distributed to different reply channels according to the user's problem classification, and the consultation width of the question answer is provided. After the answer is collected, the user is not immediately informed to make a selection, but the answer is first credible. After the analysis, the answer with high possibility is directly fed back to the user, and the depth of the answer is provided, so that the answer is in the comprehensive coverage breadth and depth, and the accuracy of the reply is also improved. In this way, the answer with the most accurate answer is directly fed back to the user, and the user does not need to make a selection, thereby avoiding the user's choice. In the embodiment of the present application, there is a chat mode with answers and answers, which is to interact with real people. In the scene, there is a real simulation of the answer and chat mode, which makes the AI human-machine chat more intelligent, which is the perfection of AI technology.

In the embodiment of the present application, the feedback information with the highest credibility in the feedback information identified by the information credibility parameter (such as the first feedback information) is compared with a threshold, and the obtained comparison result is: the first If the feedback information exceeds the threshold, it is determined that the first feedback information is trusted, and the first feedback information is returned to the terminal by using a session request response. For example, the answer with the highest credibility is selected to be compared with the entire system preset threshold T, and when the threshold T is exceeded, the answer to the question is returned to the user.

In the embodiment of the present application, the feedback information with the highest credibility in the feedback information identified by the information credibility parameter (such as the first feedback information) is compared with a threshold, and the obtained comparison result is: the first When the feedback information does not exceed the threshold, the prompt information is returned to the terminal, and the prompt information is used to indicate that the session information cannot be answered currently. For example, the answer with the highest credibility is compared with the preset threshold T of the entire system. When the threshold T is not exceeded, the problem cannot be solved, and the user prompt is given that the question cannot be answered currently. Then, the manual service is transferred, and the second feedback information obtained by the manual service is returned to the terminal through the session request response. For example, from AI customer service to manual customer service to answer, after answering the question, update the answer to the FAQ library, so that you can answer the next time you encounter a similar problem; and return the manual answer to the user.

In the embodiment of the present application, a situation in which the state machine is in a normal state is a response operation that is clear to the user's needs. The current session state of the terminal is tracked according to the state tracking policy. When the current session state is clear, the currently collected session information is distributed to at least one information feedback platform. Receiving feedback information sent by at least one information feedback platform, selecting third feedback information from the feedback information, and returning the third feedback information to the terminal. For example, when the user needs are clear, it is distributed to each question and answer subsystem, and then the preferred answer is returned to the user. It is a scheme that the server selects and directly feeds the answer of the preferred choice to the user, without the user having to choose. Specifically, the third feedback information may also be selected according to the feedback information and the information credibility parameter sent by the at least one information feedback platform, and select the highest credibility from the feedback information identified by the information credibility parameter. The third feedback information directly feeds back the third feedback information to the user. Alternatively, the third feedback information is compared with the threshold T. When the threshold T is exceeded, it is determined that the third feedback information is authentic, and the third feedback information is returned to the terminal.

In the embodiment of the present application, another case for the normal state of the state machine is a response operation that is unclear to the user's needs. The current session state of the terminal is tracked according to the state tracking policy. When the current session state is unclear, the needs of the collected session information are analyzed, and the analysis result is fed back to the terminal. For example, the user needs to like a certain brand, but it is not clear from the beginning of the conversation: it is interested in which specific series of the brand, through the history or other data analysis methods to know that the user may have a sense of "series one" Interest, the relevant information of "Series One" is fed back to the terminal. Then, the collecting terminal continues to perform the demand analysis on the session information in the session response according to the session response reported by the analysis result, until an explicit user requirement is obtained from the session information, the demand requirement analysis is ended, and the currently collected session information is obtained. Distributed to at least one information feedback platform. Receiving feedback information sent by at least one information feedback platform, selecting fourth feedback information from the feedback information, and returning the fourth feedback information to the terminal. For example, when the user's demand is not clear, the user's needs are further segmented and clarified in a guided manner, that is, through multiple interactions between the terminal and the server, the server performs demand analysis, and the terminal responds to the demand analysis result until the user needs are specified. . When the user needs are clear, distribute it to each question and answer subsystem, and then choose the answer to return to the user. It is a solution that the server chooses and directly feeds the answer of the preferred choice to the user, without the user's choice. Specifically, the fourth feedback information may also select the highest credibility from the feedback information identified by the information credibility parameter according to the feedback information and the information credibility parameter sent by the at least one information feedback platform. The fourth feedback information directly feeds the fourth feedback information to the user. Alternatively, the fourth feedback information is compared with the threshold T. When the threshold T is exceeded, the fourth feedback information is determined to be authentic, and the fourth feedback information is returned to the terminal.

In the embodiment of the present application, the response operation is performed for the abnormal state of the state machine. The current session state of the terminal is tracked according to the state tracking policy. When the current session state is abnormal, the fifth feedback information is selected from the preset feedback information, and the fifth feedback information is returned to the terminal. For example, when the dialogue cannot be carried out, if the user's question exceeds the scope of the field that the system can answer, the preset feedback information is used to respond to the question. For example, the response that can be answered is “good weather today”.唠嗑 Feedback, this kind of feedback does not necessarily solve the problem in a targeted way, but it can keep the conversation going and can handle the exception smoothly.

In the embodiment of the present application, it is a specific scenario of vehicle consultation. In addition to the normal state and the abnormal state, it is necessary to confirm the vehicle type and attributes that the user focuses on in the context. At this time, the state machine also includes two states of vehicle type confirmation and attribute confirmation. Tracking the current session state of the terminal according to the state machine, the state machine having a normal state, at least one confirmation state (such as a vehicle type confirmation state, an attribute confirmation state), and an abnormal state; when the current session state is the normal state, Distributing the currently collected session information to the at least one information feedback platform, receiving the feedback information sent by the at least one information feedback platform, selecting the sixth feedback information from the feedback information, and returning the sixth feedback information to the terminal . Determining whether it is necessary to jump to the at least one acknowledgement state according to the currently collected session information. When the jump to the at least one acknowledgement state is not confirmed, the jump to the current session state is the abnormal state.

In the embodiment of the present application, a dialog manager implemented by a state machine has four states: a normal state, an attribute confirmation state, a model confirmation state, and an abnormal state. 1) In the normal state, the form auto-filling technology may be adopted to collect the session information from the current session state, and the form is automatically filled, and each time as much information as possible is extracted from the user's statement, instead of The information corresponding to the state is extracted only by the state each time, so that the AI dialogue is more natural. The automatic filling function of the form is to obtain data through the background or text file, and automatically fill the data, which reduces the repeated labor of the user and reduces the labor cost. The embodiment of the present application may also reconstruct the problem raised by the user, including referring to the digestion and omitting the recovery (such as the subject omitting completion), and then forwarding the problem to multiple subsystems. If the subsystem returns a trusted answer, then Return the answer to the user, the system is still in a normal state, no jump. Among them, the reference digestion refers to the problem of determining which noun phrase the pronoun points to. According to the pointing, it can be divided into anaphora and pre-finger. The anaphora is the pronoun of the pronoun in front of the pronoun. The pre-predicate is the pronoun of the pronoun after the pronoun. According to the type of reference, it can be divided into three categories: personal pronouns, demonstrative pronouns, fixed descriptions, omissions, partial-total representations, common noun phrases, and so on. For example, replace the indicated pronouns in the user question according to the context. For the omission of recovery, it means that the omitted subject, object, etc. are completed. An example of subject omission completion is that there is no subject in the question, but from the context analysis, we can know that the question is my own, that is, the subject is "I". Or, in the context, the previous question is "How much is the A car?", the question of the connection is: "Is there a sunroof?", then the server should know the subject or "A car." 2) If the system only recognizes the model at this time and does not recognize the attribute, go to the “Property Confirmation” status to confirm the attribute that the user cares about; 3) If the model is not confirmed, jump to “Model Confirmation” Status to confirm the type of car that the user cares about; 4) If the model and attributes are unrecognizable or the problem is beyond the system area, then go to the "abnormal state". Among them, the vehicle type confirmation is used to confirm the vehicle type. Generally, according to the user's question, the relevant model is given to the user. After the user selects and the model is successfully confirmed, the problem can be distributed to the subsystem and returned to the "normal state". . If the first confirmation is not successful, the second confirmation can be made. At this time, it is still in the “model confirmation” state, and the processing flow is the same as above. If the secondary model confirmation fails, go to "Exception Status". In the abnormal state, when the state is mainly used for the user problem is not recognized, the user prompt is given according to the preset. If the "model" cannot be confirmed, the system default model is given and the user is prompted to confirm the selection. Around the model and attributes, when both are confirmed, you can give the correct answer.

The system for processing session information in the embodiment of the present application, as shown in FIG. 3, includes a terminal 41, a server 42, and a server 43-44, wherein the server 42 parses session information, classification, and clustering from a session request initiated by the terminal 41. After the session information is distributed to at least one server (such as the information feedback platform) 43-44, the feedback information processed by the at least one server 43-44 is obtained, and the trusted information is obtained according to the feedback information sent by the at least one server 43-44. a feedback information, the trusted first feedback information is returned to the terminal by the session request response, specifically, the first feedback information with the highest credibility is found from the feedback information, and compared with the threshold to confirm the first The feedback is trustworthy. If the first feedback information is authentic, the first feedback information is directly fed back to the terminal 41, and the feedback information is selected by the server 42 without requiring the terminal to select itself. The server 42 tracks the current session state of the terminal according to the state tracking policy, determines a response operation of the next step according to the session information collected from the current session state, and returns the response result to the terminal after performing the response operation. .

The server 42 includes: a receiving unit 421 configured to parse the session information from the session request initiated by the terminal; the category processing unit 422 is configured to perform classification and clustering processing on the session information to obtain at least one category feature parameter; 423, configured to distribute the session information to an information feedback platform corresponding to the at least one category feature parameter; the response unit 424 is configured to obtain trusted first feedback information according to the feedback information sent by the information feedback platform, Returning the trusted first feedback information to the terminal by using the session request response; the status tracking unit 425 is configured to track the current session state of the terminal according to the status tracking policy, according to the collected from the current session state. The session information determines a response operation of the next step, and the response result is returned to the terminal after performing the response operation.

By adopting the embodiment of the present application, the answer is directly fed back to the user through the analysis of the credibility of the answer, so that the accurate and closest to the real person interaction has a question and answer mode, and no user selection is needed.

The response unit further includes: a processing subunit configured to receive an information credibility parameter sent by the information feedback platform, where the information credibility parameter is used to identify the credibility of the feedback information; and according to the information The reliability parameter obtains the first feedback information with the highest credibility in the feedback information. The comparison subunit is configured to compare the first feedback information with a threshold, and when the first feedback information is trusted, obtain the trusted first feedback information.

In an embodiment of the present application, the response unit is further configured to: when the comparison result is that the first feedback information exceeds a threshold, determine that the first feedback information is authentic, and the The first feedback information is returned to the terminal through a session request response.

In an embodiment of the present application, the server further includes: an information prompting unit, configured to: when the comparison result is that the first feedback information does not exceed a threshold, returning prompt information to the terminal, the prompt information The second feedback information obtained by the manual service is returned to the terminal by the session request response.

In an embodiment of the present application, the state tracking unit is further configured to: when the current session state is clear, the current session information is distributed to at least one information feedback platform; and at least one information feedback is received. The feedback information sent by the platform selects third feedback information from the feedback information, and returns the third feedback information to the terminal.

In an embodiment of the present application, the status tracking unit is further configured to: when the current session state is unclear, the demand analysis of the collected session information is performed, and the analysis result is fed back to the terminal; The session response reported by the analysis result continues to perform demand analysis on the session information in the session response until an explicit user requirement is obtained from the session information, the demand requirement analysis is ended, and the currently collected session information is distributed to at least one piece of information. a feedback platform; receiving feedback information sent by at least one information feedback platform, selecting fourth feedback information from the feedback information, and returning the fourth feedback information to the terminal.

In the first embodiment of the present application, the status tracking unit is further configured to: when the current session state is abnormal, select the fifth feedback information from the preset feedback information, and use the fifth feedback information. Return to the terminal.

In an embodiment of the present application, the state tracking unit is further configured to: track, according to a state machine, a current session state of the terminal, where the state machine has a normal state, at least one acknowledgement state, and an abnormal state; When the session state is the normal state, the currently collected session information is distributed to the at least one information feedback platform, and the feedback information sent by the at least one information feedback platform is received, and the sixth feedback information is selected from the feedback information, and the And returning to the terminal according to the currently collected session information, determining whether it is necessary to jump to the at least one acknowledgement state, and jumping to the at least one acknowledgement state, failing to confirm, then jumping The current session state is the abnormal state.

A computer readable storage medium of the embodiment of the present application, wherein a computer program is stored thereon, wherein the computer program is executed by the processor to implement the steps of the session information processing method in the above embodiment.

A server of the embodiment of the present application, as shown in FIG. 4, the server includes: a memory 61 for storing a computer program capable of running on a processor; and a processor 62 for executing when the computer program is executed The steps of the session information processing method as in the above embodiment. The server may further include: an external communication interface 63, configured to perform information interaction with a peripheral device such as a terminal, for example, the server receives the session request sent by the terminal, and obtains session information from the session request. Responding to the session request, selecting the first feedback information with the highest credibility from the plurality of feedback information and returning to the terminal. The server may further include: an internal communication interface 64, which may specifically be a bus interface such as a PCI bus.

Taking a real application scenario as an example, the embodiment of the present application is as follows:

With the embodiment of the present application, a restricted domain dialog manager is designed. The dialog manager has two tasks: 1) maintaining the state of the system, called the dialog model; 2) being responsible for selecting the appropriate action based on the system state to reply to the user, called action selection. It can be seen that the dialog manager plays a vital role in the chat system, directly determining the quality of the user experience, and even directly determining whether the system can work normally. The dialog manager of the embodiment of the present application takes a query type chat scenario as an example, and can use a state machine to maintain a user session state, and use a form fill method to collect user requirements. When the user needs are clear, distribute it to each question and answer subsystem, and then choose the answer to return to the user. When the user's needs are not clear, the user's needs are further subdivided and clarified by means of guidance; when the dialogue cannot be carried out, if the user's question exceeds the scope of the domain that the system can answer, the exception processing can be performed "smoothly".

An example of a dialog manager is shown in Figure 5, which is based on a finite state machine. The situation of each conversation segment can be seen as one state corresponding to s0-3 in the figure. Each interaction of the dialogue process is regarded as a state transition, that is, each state node represents the information state and system action of the dialogue at that time, and each connection arc represents each operation of the user, corresponding to e1-3 respectively. . Throughout the course of the conversation, from start to finish can be seen as a path of state transitions between the connection start node S1 and the end node S3 in the state diagram.

With the above state machine, it is necessary to pre-design all possible states and the conditions for the states to jump to each other before the response operation can be performed. In each such round of dialogue, there is a state and edge corresponding to the state diagram, the overall system is completely controllable, and the entire conversation process is relatively clear. However, on the one hand, since the jump can only be performed based on the pre-defined state, the information collection can only be step by step. Therefore, the response operation is relatively rigid and the user experience is not natural enough. On the other hand, if the content of the domain is complex, the state diagram is difficult to guarantee that there is no flaw, it takes a lot of manpower to implement, and the finite state has structural shortcomings, that is, it is difficult to cope with unpredicted situations, if the user responds. If the design is out of state, the dialogue is difficult to proceed normally.

The dialog manager of the embodiment of the present application simplifies the state description of the above state machine, and maintains and tracks the current conversation state with as few states as possible. At the same time, considering that the user's response may exceed the known range, so an independent exception handling state is added, and the form filling method is combined, and each time as much information as possible is extracted from the user's statement, instead of each time. Only the information corresponding to the state is extracted by state, so the conversation is more natural. It also supports the integration of multiple Q&A subsystems, with strong scalability and support for context tracking and referencing. FIG. 6 shows an example of a state machine according to an embodiment of the present application. The state diagram of the dialog manager is as shown above, and has only four states: a normal state, an attribute confirmation state, a model confirmation state, and an abnormal state. Under normal conditions, the problem can be reconstructed, including referencing and omitting recovery, and then forwarding the problem to the subsystem. If the question and answer subsystem returns a trusted answer, the answer is returned to the user. At this time, it is still in a normal state and does not travel. If, at this time, the system only recognizes the model and does not recognize the attribute, it goes to the attribute confirmation status to confirm the attribute that the user cares about. Similarly, if the model is not confirmed, it will jump to the model confirmation status to confirm the user. The car model of interest. If the model and attributes are not recognized or the problem is beyond the system area, then the status is turned to abnormal. Among them, for the vehicle confirmation status, this status is used to confirm the vehicle type. Generally, according to the user's question, the relevant vehicle model is given to the user. After the user selects and the vehicle model is successfully confirmed, the problem can be distributed to the question and answer subsystem and returned to the normal state. If, for the first time, there is no successful confirmation, there will be a second confirmation. At this time, it is still in the vehicle confirmation status. If the secondary vehicle fails to confirm, it goes to the abnormal state. For the attribute confirmation status, the process of confirming the status with the model is not described here. For the abnormal state, this state is mainly used to give the user prompt according to the preset when the user's problem is not recognized. If the "model" cannot be confirmed, the system default model is given and the user is prompted to confirm the selection. It can be seen that it is built around the model and attributes. When both are confirmed, the server can give back to the user the correct answer.

As shown in FIG. 7 , the processing flow of the embodiment of the present application includes a dialog manager, a FAQ-based question and answer subsystem, and a knowledge base-based question and answer subsystem. The processing flow includes:

Step 301: The user asks a question;

Step 302: The dialog manager separately distributes the problem to the FAQ-based question answering subsystem and the knowledge base based question answering subsystem;

Step 303: The FAQ-based question answering subsystem and the knowledge base-based question answering subsystem respectively return the credibility of the answer and the answer;

Step 304: Determine whether the answer with the highest credibility is trusted by comparing with the preset threshold, if yes, perform step 305; if not, perform step 306;

Step 305, directly returning an answer;

Step 306, manually participating in answering the question;

Step 307, automatically update the question to the FAQ library, so that the next time you encounter a similar problem, you can answer and return the manual answer to the user.

In the embodiment of the present application, the session information processing system includes the following functions: 1) system state maintenance, such as which step the current conversation progresses; 2) refers to digestion, such as replacing the indicated pronoun in the user question according to the context; 3) Question distribution, distribute the problem to the sub-sub-question system; 4) Return the answer, judge the confidence of the answer, and respond accordingly; 5) Exclude the completion function such as the subject, such as the previous question is "How much is the A car?" Then, ask, "Is there a skylight?" At this time, the system should know whether the subject is still "A car."

Taking the pre-sales field as an example, the state maintenance is described as follows:

In this field, users generally consult on specific characteristics of a particular model, such as "How much is the Chevrolet Cruze?", "Is this car with a sunroof?" Therefore, for each round of problems of the user, as long as the specific models and attributes can be located, they can be combined into a complete problem to be distributed to the subsystem, and the answer is returned.

Refers to the implementation of omitting completion functions such as digestion and subject: when the subject or attribute is omitted, the dialog manager can implement automatic replacement or completion. The vehicle and the attributes that are focused by the user in the context are maintained by the dialog manager. When the vehicle model or attribute is omitted from the user problem, the system completes the model according to the vehicle model and attribute records of the maintenance context.

Implementation of problem distribution and answer return: When the model and attributes are confirmed, the dialog manager only needs to send the complete question to the Q&A subsystem interface. After the question and answer subsystem receives the question, the answer and The credibility of the answer is returned to the dialog manager.

The dialog manager judges according to the answer and the credibility, and selects the answer with the highest credibility to compare with the whole system preset threshold T. When the threshold T is exceeded, the result is returned to the user; otherwise, the user prompt is given, and the question cannot be answered currently.

The embodiment of the present application is applicable to all the restricted areas of the intelligent chat system (or intelligent customer service system framework), can provide services by using a common browser/server mode (BS) architecture, and can also add voice recognition at the input and output ends. And text-to-sound technology (TTS) to provide services in a communication network such as a telephone. As long as it involves the simulation of the interactive scene of everyone, AI human-computer interaction can be used to simulate the implementation. The embodiment of the present application can effectively reduce the difficulty and complexity of system development; has multiple question and answer subsystems and strong scalability; has strong portability and can quickly migrate from one restricted domain to another Restricted area.

In the several embodiments provided by the present application, it should be understood that the disclosed apparatus and method may be implemented in other manners. The device embodiments described above are merely illustrative. For example, the division of the unit is only a logical function division. In actual implementation, there may be another division manner, such as: multiple units or components may be combined, or Can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, or direct coupling, or communication connection of the components shown or discussed may be indirect coupling or communication connection through some interfaces, devices or units, and may be electrical, mechanical or other forms. of.

The units described above as separate components may or may not be physically separated, and the components displayed as the unit may or may not be physical units, that is, may be located in one place or distributed to multiple network units; Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of the embodiment.

In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may be separately used as one unit, or two or more units may be integrated into one unit; the above integration The unit can be implemented in the form of hardware or in the form of hardware plus software functional units.

A person skilled in the art can understand that all or part of the steps of implementing the above method embodiments may be completed by using hardware related to the program instructions. The foregoing program may be stored in a computer readable storage medium, and the program is executed when executed. The foregoing storage device includes the following steps: the foregoing storage medium includes: a mobile storage device, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk. A medium that can store program code.

Alternatively, the above-described integrated unit of the present application may be stored in a computer readable storage medium if it is implemented in the form of a software functional unit and sold or used as a stand-alone product. Based on such understanding, the technical solution of the embodiments of the present application may be embodied in the form of a software product in essence or in the form of a software product stored in a storage medium, including a plurality of instructions. A computer device (which may be a personal computer, server, or network device, etc.) is caused to perform all or part of the methods described in various embodiments of the present application. The foregoing storage medium includes various media that can store program codes, such as a mobile storage device, a ROM, a RAM, a magnetic disk, or an optical disk.

The foregoing is only a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto, and any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in the present application. It should be covered by the scope of protection of this application. Therefore, the scope of protection of the present application should be determined by the scope of the claims.

Industrial applicability

Claims

A method for processing session information, the method comprising:

Parsing session information from a session request initiated by the terminal;

Performing classification and clustering processing on the session information to obtain at least one category feature parameter;

Distributing the session information to an information feedback platform corresponding to the at least one category feature parameter;

Obtaining the trusted first feedback information according to the feedback information sent by the information feedback platform, and returning the trusted first feedback information to the terminal by using the session request response;

The current session state of the terminal is tracked according to the state tracking policy, and the next response operation is determined according to the session information collected from the current session state, and the response result is returned to the terminal after the response operation is performed.
The method according to claim 1, wherein the obtaining the trusted first feedback information according to the feedback information sent by the information feedback platform comprises:

Receiving, by the information feedback platform, an information credibility parameter, where the information credibility parameter is used to identify the credibility of the feedback information;

Obtaining, according to the information credibility parameter, the first feedback information with the highest credibility in the feedback information;

Comparing the first feedback information with a threshold, and when the first feedback information is trusted, obtaining trusted first feedback information.
The method of claim 2, wherein the returning the trusted first feedback information to the terminal by the session request response comprises:

When the first feedback information exceeds the threshold, it is determined that the first feedback information is trusted, and the first feedback information is returned to the terminal by using a session request response.
The method of claim 2, wherein the method further comprises:

When the first feedback information does not exceed the threshold, the prompt information is returned to the terminal, and the prompt information is used to indicate that the session information cannot be answered currently;

Transferring to the manual service, the second feedback information obtained by the manual service is returned to the terminal through the session request response.
The method according to any one of claims 1 to 4, wherein the tracking of the current session state of the terminal according to the state tracking policy, determining the next response according to the session information collected from the current session state Operations, including:

When the current session state is clear, the currently collected session information is distributed to the information feedback platform;

Receiving feedback information sent by the information feedback platform, selecting third feedback information from the feedback information, and returning the third feedback information to the terminal.
The method according to any one of claims 1 to 4, wherein the tracking of the current session state of the terminal according to the state tracking policy, determining the next response according to the session information collected from the current session state Operations, including:

When the current session state is unclear, the needs of the collected session information are analyzed, and the analysis result is fed back to the terminal;

The collecting terminal continues to perform the demand analysis on the session information in the session response according to the session response reported by the analysis result, until an explicit user requirement is obtained from the session information, the requirement analysis is ended, and the currently collected session information is distributed to the Information feedback platform;

Receiving feedback information sent by the information feedback platform, selecting fourth feedback information from the feedback information, and returning the fourth feedback information to the terminal.
The method according to any one of claims 1 to 4, wherein the tracking of the current session state of the terminal according to the state tracking policy, determining the next response according to the session information collected from the current session state Operations, including:

When the current session state is abnormal, the terminal feedback information is selected from the preset feedback information, and the fifth feedback information is returned to the terminal.
The method of any of claims 1 to 4, wherein the method further comprises:

Tracking the current session state of the terminal according to the state machine, the state machine having a normal state, at least one acknowledgement state, and an abnormal state;

When the current session state is the normal state, the currently collected session information is distributed to the information feedback platform, and the feedback information sent by the information feedback platform is received, and the sixth feedback information is selected from the feedback information. Returning the sixth feedback information to the terminal;

Determining whether it is necessary to jump to the at least one acknowledgement state according to the currently collected session information. When the jump to the at least one acknowledgement state is not confirmed, the jump to the current session state is the abnormal state.
A server, the server comprising:

a receiving unit, configured to parse the session information from the session request initiated by the terminal;

a class processing unit configured to perform classification and clustering processing on the session information to obtain at least one category feature parameter;

a distribution unit, configured to distribute the session information to an information feedback platform corresponding to the at least one category feature parameter;

The response unit is configured to obtain trusted first feedback information according to the feedback information sent by the information feedback platform, and return the trusted first feedback information to the terminal by using a session request response;

The status tracking unit is configured to track the current session state of the terminal according to the status tracking policy, determine a next response operation according to the session information collected from the current session state, and return the response result after performing the response operation. To the terminal.
The server according to claim 9, wherein the response unit further comprises:

The processing sub-unit is configured to receive the information credibility parameter sent by the information feedback platform, where the information credibility parameter is used to identify the credibility of the feedback information;

And obtaining, according to the information credibility parameter, the first feedback information with the highest credibility in the feedback information;

The comparison subunit is configured to compare the first feedback information with a threshold, and when the first feedback information is trusted, obtain the trusted first feedback information.
The server according to claim 10, wherein the response unit is further configured to:

When the first feedback information exceeds the threshold, it is determined that the first feedback information is trusted, and the first feedback information is returned to the terminal by using a session request response.
The server of claim 10, wherein the server further comprises:

The information prompting unit is configured to: when the first feedback information does not exceed the threshold, return the prompt information to the terminal, where the prompt information is used to indicate that the session information cannot be answered currently;

The manual feedback unit is configured to transfer to the manual service, and the second feedback information obtained by the manual service is returned to the terminal through the session request response.
The server according to any one of claims 9 to 12, wherein the status tracking unit is further configured to:

When the current session state is clear, the currently collected session information is distributed to the information feedback platform;

Receiving feedback information sent by the information feedback platform, selecting third feedback information from the feedback information, and returning the third feedback information to the terminal.
The server according to any one of claims 9 to 12, wherein the status tracking unit is further configured to:

When the current session state is unclear, the needs of the collected session information are analyzed, and the analysis result is fed back to the terminal;

The collecting terminal continues to perform the demand analysis on the session information in the session response according to the session response reported by the analysis result, until the explicit user requirement is obtained from the session information, the demand requirement analysis is ended, and the currently collected session information is distributed to the session information. The information feedback platform;

Receiving feedback information sent by the information feedback platform, selecting fourth feedback information from the feedback information, and returning the fourth feedback information to the terminal.
The server according to any one of claims 9 to 12, wherein the status tracking unit is further configured to:

Tracking the current session state of the terminal according to the state machine, the state machine having a normal state, at least one acknowledgement state, and an abnormal state;

When the current session state is the normal state, the currently collected session information is distributed to the information feedback platform, and the feedback information sent by the information feedback platform is received, and the sixth feedback information is selected from the feedback information. Returning the sixth feedback information to the terminal;

Determining whether it is necessary to jump to the at least one acknowledgement state according to the currently collected session information. When the jump to the at least one acknowledgement state is not confirmed, the jump to the current session state is the abnormal state.
A computer readable storage medium having stored thereon a computer program that, when executed by a processor, implements the steps of the method of any one of claims 1-8.
A server, the server comprising:

a memory configured to store a computer program capable of running on a processor;

A processor, configured to perform the steps of the method of any one of claims 1 to 8 when the computer program is run.
A method of processing session information, the method being performed by a server, the server comprising one or more processors and a memory, and one or more programs, wherein the one or more programs are stored in a memory The program may include one or more units each corresponding to a set of instructions, the one or more processors being configured to execute instructions; the method comprising:

Parsing session information from a session request initiated by the terminal;

Performing classification and clustering processing on the session information to obtain at least one category feature parameter;

Distributing the session information to an information feedback platform corresponding to the at least one category feature parameter;

Obtaining the trusted first feedback information according to the feedback information sent by the information feedback platform, and returning the trusted first feedback information to the terminal by using the session request response;

The current session state of the terminal is tracked according to the state tracking policy, and the next response operation is determined according to the session information collected from the current session state, and the response result is returned to the terminal after the response operation is performed.
The method according to claim 18, wherein the obtaining the trusted first feedback information according to the feedback information sent by the information feedback platform comprises:

Receiving, by the information feedback platform, an information credibility parameter, where the information credibility parameter is used to identify the credibility of the feedback information;

Obtaining, according to the information credibility parameter, the first feedback information with the highest credibility in the feedback information;

Comparing the first feedback information with a threshold, and when the first feedback information is trusted, obtaining trusted first feedback information.
The method of claim 19, wherein the returning the trusted first feedback information to the terminal by the session request response comprises:

When the first feedback information exceeds the threshold, it is determined that the first feedback information is trusted, and the first feedback information is returned to the terminal by using a session request response.
The method of claim 19, wherein the method further comprises:

When the first feedback information does not exceed the threshold, the prompt information is returned to the terminal, and the prompt information is used to indicate that the session information cannot be answered currently;

Transferring to the manual service, the second feedback information obtained by the manual service is returned to the terminal through the session request response.
The method according to any one of claims 18 to 21, wherein the tracking of the current session state of the terminal according to the state tracking policy, determining the next response according to the session information collected from the current session state Operations, including:

When the current session state is clear, the currently collected session information is distributed to the information feedback platform;

Receiving feedback information sent by the information feedback platform, selecting third feedback information from the feedback information, and returning the third feedback information to the terminal.
The method according to any one of claims 18 to 21, wherein the tracking of the current session state of the terminal according to the state tracking policy, determining the next response according to the session information collected from the current session state Operations, including:

When the current session state is unclear, the needs of the collected session information are analyzed, and the analysis result is fed back to the terminal;

The collecting terminal continues to perform the demand analysis on the session information in the session response according to the session response reported by the analysis result, until an explicit user requirement is obtained from the session information, the requirement analysis is ended, and the currently collected session information is distributed to the Information feedback platform;

Receiving feedback information sent by the information feedback platform, selecting fourth feedback information from the feedback information, and returning the fourth feedback information to the terminal.
The method according to any one of claims 18 to 21, wherein the tracking of the current session state of the terminal according to the state tracking policy, determining the next response according to the session information collected from the current session state Operations, including:

When the current session state is abnormal, the terminal feedback information is selected from the preset feedback information, and the fifth feedback information is returned to the terminal.
The method of any one of claims 18 to 21, wherein the method further comprises:

Tracking the current session state of the terminal according to the state machine, the state machine having a normal state, at least one acknowledgement state, and an abnormal state;

When the current session state is the normal state, the currently collected session information is distributed to the information feedback platform, and the feedback information sent by the information feedback platform is received, and the sixth feedback information is selected from the feedback information. Returning the sixth feedback information to the terminal;

Determining whether it is necessary to jump to the at least one acknowledgement state according to the currently collected session information. When the jump to the at least one acknowledgement state is not confirmed, the jump to the current session state is the abnormal state.