CN108197167A - Human-computer dialogue processing method, equipment and readable storage medium storing program for executing - Google Patents

Abstract

The invention discloses a man-machine dialogue processing method, comprising the following steps: a reception robot receives a question raised by a user, and based on its own knowledge base, identifies whether the question is a task-type question; if it is a task-type question, the reception robot takes the question The corresponding task is distributed to a processing robot that has the ability to execute the task, so that the processing robot can perform the task based on its own knowledge base; if it is not a task-type question, the reception robot will return the answer to the question by searching its own knowledge base. The invention also discloses a man-machine dialogue device and a computer-readable storage medium. The invention realizes the conversion from question receiving to task distribution processing, not only improves the service capability and efficiency of the dialogue robot, but also improves user service experience.

Description

Man-machine dialogue processing method, device and readable storage medium

technical field

The present invention relates to the technical field of man-machine dialogue, in particular to a man-machine dialogue processing method, device and readable storage medium.

Background technique

With the advancement of science and technology and the rapid growth of enterprise business volume, in order to provide users with more convenient and efficient one-to-one service, more and more enterprises have introduced man-machine dialogue technology on the basis of manual customer service to adapt to business and service requirements.

The knowledge base is an important part of the human-machine dialogue system, which usually consists of patterns, keywords, questions and corresponding answers. During the process of man-machine dialogue, the dialogue robot is usually only in the knowledge base If the dialogue machine cannot accurately and quickly identify the user's question, it will not only be inefficient in finding the answer, but may even return the wrong answer, thereby reducing the user's experience in the process of man-machine dialogue.

Contents of the invention

The main purpose of the present invention is to provide a man-machine dialogue processing method, device and readable storage medium, aiming at solving the technical problem of how to improve the service capability and service efficiency of the dialogue robot during the process of man-machine dialogue.

In order to achieve the above object, the present invention provides a method for man-machine dialogue processing. The dialogue robot includes a reception robot and a processing robot. The man-machine dialogue processing method includes the following steps:

The reception robot receives the question raised by the user, and based on its own knowledge base, identifies whether the question is a task-type question;

If it is a task-type problem, the reception robot will distribute the task corresponding to the problem to the processing robot that has the ability to execute the task, so that the processing robot can perform the task based on its own knowledge base;

If it is not a task-type question, the reception robot will return the answer to the question by searching its own knowledge base.

Optionally, before the reception robot identifies whether the question raised by the user is a task-type question, the man-machine dialogue processing method further includes:

The reception robot cleans the question raised by the user to remove invalid words, and/or words, and/or phrases in the question; and

Complete acronyms and/or colloquial words in cleaned questions.

Optionally, the man-machine dialogue processing method also includes:

The reception robot performs intent recognition on the cleaned and completed question to determine the type of the question, and the type at least includes: explanation type, reason type, and time type.

Optionally, the knowledge base includes at least: a chat knowledge base, a knowledge graph base, a business knowledge base, and a task knowledge base;

The gossip knowledge base includes a preset collection of daily question and answer terms; the knowledge graph library includes a number of knowledge for reasoning; the business knowledge base includes a preset set of questions and answers;

The task knowledge base includes:

Question instruction set, used to identify the mapping relationship between task-type questions and corresponding execution instructions;

Instruction state set, which is used to identify the mapping relationship between the instruction and the corresponding task state sequence;

The state operation set is used to identify the mapping relationship between the task state and the corresponding operation;

A collection of basic operations, including a variety of basic operations.

Optionally, the man-machine dialogue processing method also includes:

After receiving the task distributed by the reception robot, the processing robot executes the task distributed by the reception robot based on the task knowledge base in its own knowledge base, and returns the task execution result as an answer.

Optionally, if the question is not a task-type question, the step of returning the answer to the question by the reception robot by searching its own knowledge base includes:

If it is not a task-type question, the reception robot will perform question matching and retrieval sorting on the question based on the business knowledge base in its own knowledge base, and perform knowledge reasoning on the question based on the knowledge graph database in its own knowledge base. get the result set;

Judging whether there is an answer that meets the preset requirements in the result set;

If there is an answer that meets the preset requirements, the answer is returned, and other questions similar to the question are recommended.

Optionally, after the step of judging whether there is an answer that meets preset requirements in the result set, the man-machine dialogue processing method further includes:

If there is no answer that meets the preset requirements, the reception robot searches the chat knowledge base in its own knowledge base;

If there is a corresponding answer in the gossip knowledge base, the answer is returned; otherwise, a preset default answer is returned.

Further, in order to achieve the above object, the present invention also provides a human-machine dialogue device, which includes: a memory, a processor, and a human-machine dialogue stored on the memory and operable on the processor. A dialogue program, when the man-machine dialogue program is executed by the processor, implements the steps of the man-machine dialogue processing method described in any one of the above.

Optionally, the man-machine dialogue device further includes: a knowledge base optimization program stored in the memory, and the processor is configured to execute the knowledge base optimization program to perform any one or more of the following operations item:

Use the question automatic expansion template to expand the diversity of the questions in the knowledge base, and associate the expanded results with the corresponding questions;

Use the problem understanding classification template to classify the problems in the knowledge base according to the semantic understanding;

Analyze and organize the dialogue logs between the user and the human customer service, and/or the dialogue logs between the user and the dialogue robot, so as to learn new knowledge points from the dialogue logs and save them in the knowledge base or strengthen them based on the dialogue logs Existing knowledge points saved in the knowledge base.

Further, in order to achieve the above object, the present invention also provides a computer-readable storage medium, on which a man-machine dialogue program is stored, and when the man-machine dialogue program is executed by a processor, any of the above-mentioned The steps of the man-machine dialogue processing method described in item.

In the present invention, on the one hand, the dialogue robots are divided into reception robots and task processing robots according to their job roles; Responsible for receiving and identifying questions raised by users, and then realize the preliminary division of user questions. If it is a task-type question, distribute the task corresponding to the question to the corresponding task processing robot for execution, so as to realize further subdivision of user questions in terms of processing , and if it is not a task-type question, the answer to the question is returned by searching its own knowledge base. Through the division of the above two aspects and the use of task distribution to deal with user questions, this not only reduces the workload of the dialogue robot to match and find answers in the knowledge base, improves work efficiency, but also improves the professional service capabilities of the dialogue robot .

Description of drawings

Fig. 1 is a schematic structural diagram of the device hardware operating environment involved in the solution of the embodiment of the present invention;

Fig. 2 is a schematic diagram of the functional architecture of the dialogue robot in an embodiment of the man-machine dialogue processing method of the present invention;

Fig. 3 is a schematic flow chart of an embodiment of the man-machine dialogue processing method of the present invention;

Fig. 4 is a schematic diagram of the content structure of an embodiment of the knowledge base used in the man-machine dialogue processing method of the present invention.

The realization of the purpose of the present invention, functional characteristics and advantages will be further described in conjunction with the embodiments and with reference to the accompanying drawings.

Detailed ways

It should be understood that the specific embodiments described here are only used to explain the present invention, not to limit the present invention.

As shown in FIG. 1 , FIG. 1 is a schematic structural diagram of the hardware operating environment of the device involved in the solution of the embodiment of the present invention.

The human-machine dialogue device in the embodiment of the present invention can be a PC, a server, or a device with a display or recording and playback function such as a smart phone, a tablet computer, a portable computer, or a smart toy. The process of the man-machine dialogue can be a voice dialogue, It can also be a dialogue in the form of text, or a dialogue in the form of voice + text.

As shown in FIG. 1 , the man-machine dialogue device may include: a processor 1001 , such as a CPU, a communication bus 1002 , a user interface 1003 , a network interface 1004 , and a memory 1005 . Wherein, the communication bus 1002 is used to realize connection and communication between these components. The user interface 1003 may include a display screen (Display), an input unit such as a keyboard (Keyboard), and the optional user interface 1003 may also include a standard wired interface and a wireless interface. Optionally, the network interface 1004 may include a standard wired interface and a wireless interface (such as a WI-FI interface). The memory 1005 can be a high-speed RAM memory, or a stable memory (non-volatile memory), such as a disk memory. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001 .

Optionally, the man-machine dialogue device may also include a camera, an RF (Radio Frequency, radio frequency) circuit, a sensor, an audio circuit, a WiFi module, and the like.

Those skilled in the art can understand that the hardware structure of the human-machine dialogue device shown in Figure 1 does not constitute a limitation to the human-machine dialogue device, and may include more or less components than those shown in the illustration, or combine certain components, or different component arrangements.

As shown in FIG. 1 , the memory 1005 as a computer-readable storage medium may include an operating system, a network communication module, a user interface module, and computer programs, such as man-machine dialogue programs, knowledge base optimization programs, and the like. Among them, the operating system is a program that manages and controls man-machine dialogue equipment and software resources, and supports the operation of network communication modules, user interface modules, man-machine dialogue programs, knowledge base optimization programs, and other programs or software; network communication modules are used to manage and control the network interface 1002; the user interface module is used to manage and control the user interface 1003.

In the hardware structure of the man-machine dialogue device shown in Figure 1, the network interface 1004 is mainly used to connect to the system background and perform data communication with the system background; the user interface 1003 is mainly used to connect to the client (client) and perform data communication with the client Communication; the man-machine dialogue device invokes the man-machine dialogue program stored in the memory 1005 through the processor 1001 to perform the following operations:

The reception robot receives the question raised by the user, and based on its own knowledge base, identifies whether the question is a task-type question;

If it is a task-type problem, the reception robot will distribute the task corresponding to the problem to the processing robot that has the ability to execute the task, so that the processing robot can perform the task based on its own knowledge base;

If it is not a task-type question, the reception robot will return the answer to the question by searching its own knowledge base.

Further, the man-machine dialogue device invokes the man-machine dialogue program stored in the memory 1005 through the processor 1001 to perform the following operations:

The reception robot cleans the question raised by the user to remove invalid words, and/or words, and/or phrases in the question; and

Complete acronyms and/or colloquial words in cleaned questions.

Further, the man-machine dialogue device invokes the man-machine dialogue program stored in the memory 1005 through the processor 1001 to perform the following operations:

The reception robot performs intent recognition on the cleaned and completed question to determine the type of the question, and the type at least includes: explanation type, reason type, and time type.

Further, the man-machine dialogue device invokes the man-machine dialogue program stored in the memory 1005 through the processor 1001 to perform the following operations:

After receiving the task distributed by the reception robot, the processing robot executes the task distributed by the reception robot based on the task knowledge base in its own knowledge base, and returns the task execution result as an answer.

Further, the man-machine dialogue device invokes the man-machine dialogue program stored in the memory 1005 through the processor 1001 to perform the following operations:

If it is not a task-type question, the reception robot will perform question matching and retrieval sorting on the question based on the business knowledge base in its own knowledge base, and perform knowledge reasoning on the question based on the knowledge graph database in its own knowledge base. get the result set;

Judging whether there is an answer that meets the preset requirements in the result set;

If there is an answer that meets the preset requirements, the answer is returned, and other questions similar to the question are recommended.

Further, the man-machine dialogue device invokes the man-machine dialogue program stored in the memory 1005 through the processor 1001 to perform the following operations:

If there is no answer that meets the preset requirements, the reception robot searches the chat knowledge base in its own knowledge base;

If there is a corresponding answer in the gossip knowledge base, the answer is returned; otherwise, a preset default answer is returned.

Further, the man-machine dialogue device invokes the knowledge base optimization program stored in the memory 1005 through the processor 1001 to perform any one or more of the following operations:

Use the question automatic expansion template to expand the diversity of the questions in the knowledge base, and associate the expanded results with the corresponding questions;

Use the problem understanding classification template to classify the problems in the knowledge base according to the semantic understanding;

Analyze and organize the dialogue logs between the user and the human customer service, and/or the dialogue logs between the user and the dialogue robot, so as to learn new knowledge points from the dialogue logs and save them in the knowledge base or strengthen them based on the dialogue logs Existing knowledge points saved in the knowledge base.

Based on the above-mentioned hardware structure of the man-machine dialogue device, various embodiments of the man-machine dialogue processing method of the present invention are proposed.

Referring to FIG. 2 , FIG. 2 is a schematic diagram of the functional architecture of the dialogue robot in an embodiment of the man-machine dialogue processing method of the present invention.

As shown in Figure 2, the human-machine dialogue specifically refers to the dialogue interaction between the user and the machine. In this embodiment, the dialogue robot is specifically a virtual object virtualized in the man-machine dialogue program for dialogue with the user. Based on the different job roles, the dialogue robot can be divided into a reception robot (receptionist in the figure) and a processing robot (receptionist in the figure). Shown as a processor), the job roles of reception robots and processing robots are similar to human customer service and human technicians in reality, so I won’t go into details.

For example, the wealth management app used by the user has the function of an intelligent voice assistant. The user can have a conversation with the intelligent voice assistant. If the user asks a general question, the reception robot will answer it directly. If it is a professional question, such as opening an account, borrowing money , purchase of wealth management products, etc., the reception robot will transfer the question to a processing robot with corresponding business capabilities to answer.

In this embodiment, the reception robot and the processing robot have their own knowledge bases, and each robot answers user questions based on its own knowledge base. It should be noted that the knowledge bases used by the reception robot and the processing robot may have the same content and structure, or may have different content and structure, or may have the same structure but different content.

The reception robot and the processing robot can learn from each other's knowledge base. The reception robot is mainly responsible for identifying whether the user's question is a task-type question, and if so, distributes the task corresponding to the question to a processing robot that has the ability to perform the task, and then the processing robot takes over the role of the reception robot to continue performing the task, and After the subsequent interaction with the user until the task is completed, it is re-delivered to the reception robot for the next user question processing. For example, the account opening task requires the user to enter the name, ID number, and mobile phone number. Return to the user while re-delivering the current conversation to the reception bot.

In addition, those skilled in the art can understand that, in order to ensure the service quality and security of man-machine dialogue, in the process of man-machine dialogue, according to the operator's business process design or the user's personal needs, human customer service can be introduced in the background of the dialogue robot Participation, such as manual review for account opening business.

Referring to FIG. 3 , FIG. 3 is a schematic flow chart of the first embodiment of the method for man-machine dialogue processing according to the present invention. In this embodiment, the man-machine dialogue processing method includes the following steps:

Step S10, the reception robot receives the question raised by the user, and based on its own knowledge base, identifies whether the question is a task-type question;

What needs to be emphasized in the present invention is that the question raised by the user does not refer to a question in an absolute sense, such as "how is the weather today?" Can be: questions, requests, orders, explanations, advice, compliments, etc. For example, the user's question can be "you are so cute", "how are you", "how to get to the bus stop", "please sing a song", "close the fund account for me" and so on.

At the same time, this embodiment is not limited to the form in which the user asks the question, for example, a text form or a voice form. If the question raised by the user is in voice form, after receiving the voice question, the reception robot needs to perform voice recognition to obtain the corresponding text form of the question.

In this embodiment, the specific content and structure of the knowledge base used by the reception robot are not limited, but the reception robot can identify whether the question raised by the user is a task-type question based on its own knowledge base.

In this embodiment, the questions raised by users are divided into two categories: task-type questions and non-task-type questions. Task-type questions specifically mean that the dialogue robot needs to answer the user's questions by performing tasks. For example, when a user asks "Can I open an account?", the reception robot's answer cannot be just "yes", because this answer does not solve the user's problem, so the reception robot should also perform the task-type question The corresponding tasks can truly solve the problems raised by users.

There is no limit to the manner in which the reception robot identifies whether the question raised by the user is a task-type question. For example, identifying whether a task-type question can be converted into identifying whether a task is to be performed, and a task question template can be used to judge whether a task is to be performed. For example, pre-set question templates for handling all tasks supported by the robot, and one question template can correspond to multiple ways of asking questions. For example, the task question template of the account opening task can be asked in the following ways: "Please open an account for me", "I want to open an account", "Open an account", "Is it okay to open an account" and so on.

For another example, the identification of task-type questions is triggered by keywords and patterns, such as the pattern: "*reserved*meeting room*". When the user's question matches this pattern, it will trigger the task execution of the reserved meeting room. For example, the user asks "I want to reserve the meeting room on the 7th floor from 9 am to 10 am tomorrow", this sentence triggers the mode of booking the meeting room above, and then enters the task process of meeting room reservation, which requires some necessary Parameters such as time, floor, etc., and just this sentence contains time and floor, so they are identified and filled in the task interface of booking a conference room, and then the processing robot initiates the basic operation of booking a conference room.

Step S20, if it is a task-type problem, the reception robot distributes the task corresponding to the problem to a processing robot that has the ability to execute the task, so that the processing robot can execute the task based on its own knowledge base;

In this embodiment, when the reception robot identifies a question raised by a user, it can not only identify whether the question is a task-type question, but also determine the task corresponding to the task-type question. For example, if the user asks "Is it possible to open an account", the reception robot will recognize the question as a task-type question and determine that the task corresponding to the question is "account opening task".

In this embodiment, the manner of determining the task-type problem is not limited. For example, all the tasks supported by the processing robot can be pre-set, and the task keywords corresponding to each task can be set. If the task-type question contains the task keyword, the task corresponding to the task-type question can be determined, as shown in the following table 1.

Table 1

task type task keyword open an account open an account, open an account, account loan borrow, borrow money, loan, loan Buy buy, buy

If the reception robot recognizes that the question raised by the user is a task-type question, the reception robot distributes the task corresponding to the question to a processing robot that has the ability to execute the task. In this embodiment, different processing robots correspond to different tasks, that is, different processing robots have different task execution capabilities.

In this embodiment, the processing robot specifically executes the task based on its own knowledge base, and the specific execution manner is not limited. It should be further explained that the processing robot may continue to have a dialogue with the user during the process of performing the task, and the form of the dialogue may be a form of dialogue between the user and the processing robot directly.

Step S30, if it is not a task-type question, the reception robot returns the answer to the question by searching its own knowledge base.

If the question raised by the user is not a task-type question, the question does not need to be distributed to the processing robot for processing, but the reception robot directly returns the answer to the user by searching its own knowledge base. It should be noted that the answer returned by the reception robot may or may not be in its own knowledge base, but eventually the reception robot will return a reply to the user.

In this embodiment, on the one hand, the dialogue robots are divided into reception robots and task processing robots according to their job roles; The robot is responsible for receiving and identifying the questions raised by the user, and then realizes the preliminary division of the user's questions. If it is a task-type question, the task corresponding to the question will be distributed to the corresponding task processing robot for execution, so as to realize further detailed processing of the user's questions. If it is not a task-type question, it returns the answer to the question by searching its own knowledge base. Through the division of the above two aspects and the use of task distribution to deal with user questions, this not only reduces the workload of the dialogue robot to match and find answers in the knowledge base, improves work efficiency, but also improves the professional service capabilities of the dialogue robot .

Further optionally, in the second embodiment of the man-machine dialogue processing method of the present invention, the reception robot also performs the following preprocessing before identifying whether the question raised by the user is a task-type question:

(1) The reception robot cleans the questions raised by the users, so as to remove invalid words, and/or words, and/or phrases in the questions;

In this embodiment, question cleaning mainly refers to eliminating irrelevant and meaningless invalid words, words or phrases in the question. For example, "I want to ask", "Please tell me", these phrases are not only meaningless to the whole question, but even interfere with the robot's answer to the question.

In this embodiment, there is no limit to the implementation manner of problem cleaning. For example, the invalid words and phrases library is pre-set, including invalid words, phrases, short sentence sets and sentence patterns. Partially removed. For example, the colloquial auxiliary words often used by users: "ah, what", "is it possible", "does it work", etc., in the process of question cleaning, the semantic division of the question is carried out, and based on the invalid word lexicon, after the semantic division The content of the question is cleaned, thereby removing invalid words, and/or words, and/or phrases in the question.

(2) Complete the abbreviations and/or colloquial words in the cleaned questions;

In this embodiment, the question completion mainly refers to completing the abbreviations or colloquial words in the user's question, so as to make the content of the question clearer. For example, "When should I repay the money I borrowed" should be completed as "When should I repay the money I borrowed".

In this embodiment, there is no limit to the way of implementing question completion. For example, the pre-set question completion word library is composed of a collection of acronyms and complete word mappings and sentence patterns. When a question matches a certain sentence pattern and the abbreviations in this pattern, the acronym will be replaced. for complete words. For example, the completion methods corresponding to commonly used colloquial words and abbreviations can be set in advance, and the original question can also be replaced in terms of asking method, so as to realize the completion of the original question.

(3) The reception robot conducts intent recognition on the cleaned and completed question to determine the type of the question, and the type includes at least: explanation type, reason type, and time type.

In this embodiment, intent recognition mainly refers to classifying the user's questions on intentions, such as "interpretation-type" questions, "reason-type" questions, and "time-type" questions.

This embodiment is not limited to the implementation manner of the intention recognition. For example, first semantically divide the question, such as executive subject, executive action, time adverb, mood adverb, etc., and then determine the intention of the user's question based on the semantic division, such as "When will I repay the loan" can be classified as " Questions about time” and “how to open an account” can be classified as “explanation” questions, and “why you can’t get a loan” can be classified as “reasons”.

Of course, in this embodiment, the questions raised by users can be further classified. For example, "explanation category" can be further divided into categories such as "business condition explanation", "business content explanation" and "business processing result explanation".

In this embodiment, before the reception robot judges whether the question raised by the user is a task-type question, it first performs preprocessing on the received question, such as cleaning, completion, and intention recognition, so as to improve the dialogue robot's ability to understand the question, and then Improve the efficiency and accuracy of finding answers to questions.

Referring to FIG. 4 , FIG. 4 is a schematic diagram of the content structure of an embodiment of the knowledge base used in the man-machine dialogue processing method of the present invention.

As shown in FIG. 4 , in this embodiment, the knowledge base includes at least: a gossip knowledge base, a knowledge graph base, a business knowledge base, and a task knowledge base.

(1) Chat knowledge base

The chat knowledge base mainly stores a set of preset daily question and answer terms, such as "You are so cute" corresponds to "Thank you for the compliment"; "Can you sing a song" corresponds to "Okay, what song do you want to hear", etc. It is used to answer some daily language dialogues of users.

The daily question-and-answer terms in the chat knowledge base are mainly general data obtained through offline learning, so manual editing by the knowledge base administrator is not required.

(2) Knowledge graph library

The knowledge graph library stores several knowledge for reasoning, such as knowledge graphs related to account opening, knowledge graphs related to borrowing, etc., or knowledge graphs related to user dialogue topics, such as weather knowledge graphs, car knowledge graphs, star knowledge Maps, etc., are mainly used for knowledge reasoning in the reply process.

The knowledge map is mainly general data obtained through offline learning, so manual editing by the knowledge base administrator is not required.

(3) Business knowledge base

The business knowledge base stores a set of preset questions and answers. The content of the knowledge base is mainly related to the business, so it is mainly formed by the knowledge base administrator through manual editing. For example, if the question is about working hours, the corresponding answers are Monday to Friday, 9 am to 17 pm; if the question is the loan repayment period, the corresponding answers are 5 years, 10 years, 15 years, 20 years, 30 years.

(4) Task knowledge base

The task knowledge base is mainly used to provide support when dealing with tasks corresponding to task-type problems handled by robots. The task knowledge base mainly includes the following collection contents:

Question instruction set, used to identify the mapping relationship between task-type questions and corresponding execution instructions;

Instruction state set, which is used to identify the mapping relationship between the instruction and the corresponding task state sequence;

The state operation set is used to identify the mapping relationship between the task state and the corresponding operation;

A collection of basic operations, including a variety of basic operations, such as receiving, obtaining, uploading, sending, querying, deleting, prompting, compressing, displaying, etc.

In this embodiment, the task to be executed is decomposed into multiple execution steps and executed sequentially. Specifically: problem—>task instruction—>task status sequence, task status—>operation, and finally achieve the completion of the task.

For example, the instruction is "book a meeting room", the status includes "floor", "start time", "end time", "number of participants", etc., and the operations include "query the floor where the meeting room is located" and "query all meeting room details" , "Query the details of a meeting room on a certain floor", "Book a meeting room", "Cancel a meeting room", etc.

The structural form of the knowledge base adopted in this embodiment strengthens the dialogue robot's ability to understand the questions raised by the user, so that the user can interact with the robot in a more natural way of asking questions.

Optionally, based on the content structure of the knowledge base in the previous embodiment, in the third embodiment of the human-machine dialogue processing method of the present invention, after the processing robot receives the task distributed by the reception robot, based on the task in its own knowledge base, The knowledge base executes the tasks distributed by the reception robot and returns the task execution results as answers.

For example, the task-type question is "Please help me open an account", and the corresponding task is "Account Opening Task". The process of processing the robot to perform this task is as follows:

1. Through the question "Please help me open an account", determine the corresponding instruction as "Execute the task of opening an account";

2. Through the command "execute account opening task", determine the corresponding task status sequence as: user basic information, information verification, account opening account information;

2.1 The corresponding operation of the user's basic information is as follows: Prompt the user to enter the basic information, package the information and upload it to the background;

2.2 The corresponding operations for information verification are as follows: submit information, obtain audit rules, verify information based on audit rules, and return verification results;

2.3 The corresponding operations for account opening and account information are as follows: generate account information, extract relevant information in the account information, and issue the relevant information.

Further optionally, in an embodiment of the man-machine dialogue device of the present invention, in order to further optimize and enhance the answering ability of the dialogue robot, in this embodiment, the man-machine dialogue device further includes: a knowledge base optimization program stored in the memory 1005 , the processor 1001 is configured to execute a knowledge base optimization program to perform any one or more of the following operations:

Operation 1: Use the question automatic expansion template to expand the diversity of questions in the knowledge base, and associate the expanded results with the corresponding questions; this operation is mainly to automatically expand the questions edited by the knowledge base administrator. Then the set of questions expressing the same meaning is automatically expanded.

For example, on the issue of account opening, some users may ask "how to apply for a card", "what is the process of applying for a card", "what are the conditions and requirements for applying for a card" and so on. For example, after a user edits "What is Weiweidai", this operation will automatically expand the question to "Introduction to Weiweidai", "Explanation of Weiweidai" and so on.

Operation 2: Use the question understanding classification template to classify the questions in the knowledge base according to the semantic understanding; this operation can improve the processing speed and accuracy of the robot's reply.

For example, first semantically divide the question, such as executive subject, executive action, time adverb, mood adverb, etc., and then determine the intention of the user's question based on the semantic division, such as "When will I repay the loan" can be classified as " Questions about time” and “how to open an account” can be classified as “explanation” questions, and “why you can’t get a loan” can be classified as “reasons”. As long as the category granularity of user questions is fine enough, the accuracy of question retrieval can be greatly improved.

Operation 3: Analyze and organize the dialogue logs between the user and the human customer service, and/or the dialogue logs between the user and the dialogue robot, so as to learn new knowledge points from the dialogue logs and save them in the knowledge base or based on Dialogue logs reinforce existing knowledge points that have been saved in the knowledge base.

This operation realizes the automatic learning of the knowledge base, learns new knowledge points to expand the knowledge base, and consolidates existing knowledge points to improve the robot's ability to answer questions. There is no limit to the automatic learning method. For example, machine learning methods such as learning templates, text generation, and seq2seq are used to mine new knowledge points (such as question and answer sets) from dialogue life. It should be further explained that since the reception robot and the processing robot use their own knowledge bases, in order to improve the answering ability of the dialogue robot, the knowledge bases of both parties can learn from each other, so that the reception robot or processing robot can have Response ability of the other party, so as to realize the efficiency and accuracy of the reply process as a whole.

Further optionally, in the fourth embodiment of the human-computer dialogue processing method of the present invention, based on the first embodiment of the method of the present invention, the implementation of the above-mentioned step S30 includes the following steps:

Step S301, if it is not a task-type question, then the reception robot performs question matching and retrieval sorting on the question based on the business knowledge base in its own knowledge base, and performs a search and sorting on the question based on the knowledge graph database in its own knowledge base. Knowledge reasoning, get the result set;

Step S302, judging whether there is an answer that meets the preset requirements in the result set;

Step S303, if there is an answer that meets the preset requirements, return the answer and recommend other questions similar to the question.

In this embodiment, if the question raised by the user is not a task-type question, the reception robot searches for the corresponding answer to the question based on the business knowledge base and the knowledge graph database in its own knowledge base.

(1) Based on the business knowledge base, perform question matching and retrieval sorting;

In this embodiment, the reception robot matches the questions raised by the user with the questions in the business knowledge base one by one to obtain a corresponding matching result set; The question set is sorted according to the weight of the matching words, and the corresponding retrieval result set is obtained;

(2) Carry out knowledge reasoning based on the knowledge graph library.

In this embodiment, the reception robot will also perform knowledge reasoning on the questions raised by the user based on the knowledge graph library, and obtain a corresponding reasoning result set.

The above matching result set, retrieval result set and reasoning result set constitute the reply result set of the reception robot, and then the reception robot needs to further judge whether there is an answer that meets the preset requirements in the reply result set, and if so, return the answer to the user.

It should be noted that, in this embodiment, there is no limit to the setting of the preset requirement, which is specifically set according to actual needs. For example, the preset requirement can be set as follows: if there is one answer, it will be directly returned as the answer to the user's question, and if there are multiple answers, the optimal answer in the matching result set will be preferred, and the matching result set and the retrieval result set will be selected. Select the optimal result with the inference result set.

In addition, in order to improve service capabilities, when answering user questions, other questions similar to those raised by users are further recommended to users.

Further optionally, in the fifth embodiment of the man-machine dialog processing method of the present invention, based on the fourth embodiment of the method of the present invention, in this embodiment, the man-machine dialog processing method further includes:

Step S304, if there is no answer that meets the preset requirements, the reception robot searches the gossip knowledge base in its own knowledge base;

Step S305, if there is a corresponding answer in the gossip knowledge base, return the answer, otherwise return the preset default answer.

In this embodiment, if the reception robot cannot obtain the answer from the business knowledge base and the knowledge graph database, it will further search for the answer in the chat knowledge base, and the search method can be question matching, retrieval and sorting, etc. If the corresponding gossip result is found in the gossip knowledge base, it will be returned to the user as an answer, otherwise, the default answer preset by the man-machine dialogue device will be selected and returned to the user, such as "I'm sorry that no answer that meets the requirements has been found, we Will keep learning...", and so on.

The present invention also provides a computer-readable storage medium.

The computer-readable storage medium of the present invention stores a man-machine dialogue program, and when the man-machine dialogue program is executed by a processor, the steps in any embodiment of the above-mentioned man-machine dialogue processing method are realized.

Further, a knowledge base optimization program may also be stored on the computer-readable storage medium of the present invention, and when the knowledge base optimization program is executed by a processor, any one or more of the following operations may be performed:

Use the question automatic expansion template to expand the diversity of the questions in the knowledge base, and associate the expanded results with the corresponding questions;

Use the problem understanding classification template to classify the problems in the knowledge base according to the semantic understanding;

Analyze and organize the dialogue logs between the user and the human customer service, and/or the dialogue logs between the user and the dialogue robot, so as to learn new knowledge points from the dialogue logs and save them in the knowledge base or strengthen them based on the dialogue logs Existing knowledge points saved in the knowledge base.

It should be noted that, in this document, the term "comprising", "comprising" or any other variation thereof is intended to cover a non-exclusive inclusion such that a process, method, article or apparatus comprising a set of elements includes not only those elements, It also includes other elements not expressly listed, or elements inherent in the process, method, article, or device. Without further limitations, an element defined by the phrase "comprising a ..." does not preclude the presence of additional identical elements in the process, method, article, or apparatus comprising that element.

The serial numbers of the above embodiments of the present invention are for description only, and do not represent the advantages and disadvantages of the embodiments.

Through the description of the above embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus a necessary general-purpose hardware platform, and of course also by hardware, but in many cases the former is better implementation. Based on such an understanding, the essence of the technical solution of the present invention or the part that contributes to the prior art can be embodied in the form of software products, and the computer software products are stored in a storage medium (such as ROM/RAM, disk, CD) contains several instructions to make a terminal (which may be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) execute the methods described in various embodiments of the present invention.

Embodiments of the present invention have been described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementations, and the above-mentioned specific implementations are only illustrative, rather than restrictive, and those of ordinary skill in the art will Under the enlightenment of the present invention, without departing from the gist of the present invention and the scope of protection of the claims, many forms can also be made, and any equivalent structure or equivalent process transformation made by using the description and drawings of the present invention, or Directly or indirectly used in other relevant technical fields, these all belong to the protection of the present invention.

Claims (10)

1. a kind of human-computer dialogue processing method, which is characterized in that dialogue robot includes reception robot and handling machine people, institute Human-computer dialogue processing method is stated to include the following steps：

It receives robot and receives the problem of user proposes, and based on knowledge library, identify whether the problem is Task problem；

If Task problem, then robot is received by the corresponding task of the problem, is distributed to and possesses the Mission Capability Handling machine people, so that handling machine people performs the task based on knowledge library；

If not Task problem, then robot is received by searching for knowledge library, returns to the answer of the problem.

2. human-computer dialogue processing method as described in claim 1, which is characterized in that reception robot is proposed in identification user Before whether problem is Task problem, the human-computer dialogue processing method further includes：

Reception robot the problem of being proposed to user, cleans, for by word, and/or word invalid in problem, and/or phrase It removes；And

Initialism and/or colloquial style word in the problem of by after cleaning carry out completion.

3. human-computer dialogue processing method as claimed in claim 2, which is characterized in that the human-computer dialogue processing method is also wrapped It includes：

It receives the problem of robot is to after cleaning and completion and carries out intention assessment, for determining the type of the problem, the type It includes at least：Explain type, cause type, time type.

4. human-computer dialogue processing method as described in claim 1, which is characterized in that the knowledge base includes at least：Chat is known Know library, knowledge mapping library, professional knowledge library and task knowledge library；

The chat knowledge base includes preset daily question and answer term set；The knowledge mapping library includes several for reasoning Knowledge；The professional knowledge library includes the problem of preset and answer set；

The task knowledge library includes：

Problem instruction set, for identifying the mapping relations between Task problem and the instruction of corresponding execution；

Command status set, for identifying instruction and the mapping relations between corresponding task status sequence；

State operational set, for identifying the mapping relations between task status and the operation of corresponding execution；

Basic operation set includes a variety of basic operations.

5. human-computer dialogue processing method as claimed in claim 4, which is characterized in that the human-computer dialogue processing method is also wrapped It includes：

After the receiving the distribution of reception robot of the task, handling machine people is held based on the task knowledge library in knowledge library The task of row reception robot distribution, and returned task action result as answer.

6. human-computer dialogue processing method as claimed in claim 4, which is characterized in that if not the Task problem, then connect Robot is treated by searching for knowledge library, the step of answer for returning to the problem includes：

If not Task problem, then receive robot based on the professional knowledge library in knowledge library, to the problem respectively into Row question matching and retrieval ordering and based on the knowledge mapping library in knowledge library, carry out knowledge reasoning to the problem, obtain To result set；

Judge in the result set with the presence or absence of the answer for meeting preset requirement；

If in the presence of the answer for meeting the preset requirement, the answer is returned to, and recommend the other problems similar to the problem.

7. human-computer dialogue processing method as claimed in claim 6, which is characterized in that it is described judge in the result set whether After the step of answer for meeting preset requirement, the human-computer dialogue processing method further includes：

If there is no the answer for meeting the preset requirement, the chat knowledge base in robot lookup knowledge library is received；

If there is corresponding answer in the chat knowledge base, the answer is returned, otherwise returns to preset acquiescence answer.

8. a kind of man-machine dialogue equipment, which is characterized in that the man-machine dialogue equipment includes：It memory, processor and is stored in On the memory and the interactive program(me) that can run on the processor, the interactive program(me) is by the processor The step of human-computer dialogue processing method as described in any one of claim 1 to 7 is realized during execution.

9. man-machine dialogue equipment as claimed in claim 8, which is characterized in that the man-machine dialogue equipment further includes：It is stored in Knowledge base optimization program on the memory, it is following to perform that the processor is configured to perform the knowledge base optimization program It is any one or more in operation：

Template is expanded using problem automatically, the diversity that way to put questions is carried out to the problems in knowledge base expands, and by the result of expansion It is associated with correspondence problem；

Classification model is understood using problem, is classified to the problems in knowledge base according to semantic understanding；

The conversation log between conversation log, and/or user and dialogue robot between user and artificial customer service is divided Analysis and arrange, for from the new knowledge point of conversation log learning and be saved in knowledge base or based on conversation log strengthen knowledge The existing knowledge point preserved in library.

10. a kind of computer readable storage medium, which is characterized in that it is man-machine right to be stored on the computer readable storage medium Program is talked about, the human-computer dialogue as described in any one of claim 1 to 7 is realized when the interactive program(me) is executed by processor The step of processing method.