CN112528002A

CN112528002A - Dialog recognition method and device, electronic equipment and storage medium

Info

Publication number: CN112528002A
Application number: CN202011545960.8A
Authority: CN
Inventors: 王猛; 赵筱军; 黄庆伟
Original assignee: Beijing Baidu Netcom Science and Technology Co Ltd
Current assignee: Beijing Baidu Netcom Science and Technology Co Ltd
Priority date: 2020-12-23
Filing date: 2020-12-23
Publication date: 2021-03-19
Anticipated expiration: 2040-12-23
Also published as: CN112528002B

Abstract

The application discloses a dialog recognition method, a dialog recognition device, electronic equipment and a storage medium, and relates to the field of Natural Language Processing (NLP). The implementation scheme is as follows: acquiring a target dialogue in at least one round of dialogue, and determining a target problem set to be identified from the at least one round of dialogue; carrying out entity word recognition on the target conversation to obtain and store a plurality of entity words in the target conversation; matching each problem in the target problem set with a plurality of stored entity words respectively to determine the entity words corresponding to each problem; and determining answers of the questions according to the entity words corresponding to the questions. Therefore, when the user answers a plurality of questions in one round of conversation, the system can automatically identify the answers of the questions, reduce the unnecessary interaction times and improve the use experience of the user.

Description

Dialog recognition method and device, electronic equipment and storage medium

Technical Field

The present application relates to the technical field of AI (Artificial Intelligence) such as NLP (Natural Language Processing), and in particular, provides a dialog recognition method, apparatus, electronic device, and storage medium.

Background

With the continuous evolution of human society informatization and the continuous rise of human service cost, people increasingly hope to communicate with computers through natural language, and a man-machine conversation system becomes a product born under the historical background. By means of the man-machine conversation system, a human can converse with a machine by using a natural language, and commands or consults a computer through the conversation to complete specific operations, such as commanding intelligent hardware to complete short message reading and replying, weather inquiring, car renting, ticket booking, route arrangement and the like.

Disclosure of Invention

The application provides a method and a device for recognizing a conversation, an electronic device and a storage medium.

According to an aspect of the present application, there is provided a dialog recognition method including:

acquiring a target dialogue in at least one round of dialogue, and determining a target problem set to be identified from the at least one round of dialogue;

carrying out entity word recognition on the target conversation to obtain and store a plurality of entity words in the target conversation;

matching each question in the target question set with a plurality of stored entity words respectively to determine the entity words corresponding to each question;

and determining answers of the questions according to the entity words corresponding to the questions.

According to another aspect of the present application, there is provided a dialog recognition device including:

the conversation acquisition module is used for acquiring a target conversation in at least one round of conversation;

the set determining module is used for determining a target problem set to be identified from the at least one round of conversation;

the recognition module is used for carrying out entity word recognition on the target conversation to obtain and store a plurality of entity words in the target conversation;

the matching module is used for matching each question in the target question set with a plurality of stored entity words respectively so as to determine the entity words corresponding to each question;

and the answer determining module is used for determining the answer of each question according to the entity word corresponding to each question.

According to yet another aspect of the present application, there is provided an electronic device including:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein the content of the first and second substances,

the memory stores instructions executable by the at least one processor to enable the at least one processor to perform a dialog recognition method according to the above-described embodiments of the present application.

According to yet another aspect of the present application, there is provided a non-transitory computer readable storage medium of computer instructions for causing a computer to perform a dialog recognition method set forth in the above-described embodiments of the present application.

According to yet another aspect of the present application, there is provided a computer program product, wherein instructions of the computer program product, when executed by a processor, perform a dialog recognition method as set forth in the above-described embodiments of the present application.

It should be understood that the statements in this section do not necessarily identify key or critical features of the embodiments of the present application, nor do they limit the scope of the present application. Other features of the present application will become apparent from the following description.

Drawings

The drawings are included to provide a better understanding of the present solution and are not intended to limit the present application. Wherein:

FIG. 1 is a schematic view of a partially continuous entity collection unit flow configuration;

fig. 2 is a schematic flowchart of a dialog recognition method according to an embodiment of the present application;

fig. 3 is a flowchart illustrating a dialog recognition method according to a second embodiment of the present application;

fig. 4 is a schematic flowchart of a dialog recognition method according to a third embodiment of the present application;

fig. 5 is a schematic flowchart of a dialog recognition method according to a fourth embodiment of the present application;

fig. 6 is a schematic structural diagram of a dialog recognition device according to a fifth embodiment of the present application;

FIG. 7 shows a schematic block diagram of an example electronic device that may be used to implement embodiments of the present application.

Detailed Description

The following description of the exemplary embodiments of the present application, taken in conjunction with the accompanying drawings, includes various details of the embodiments of the application for the understanding of the same, which are to be considered exemplary only. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the present application. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.

With the continuous development of natural language processing technology, the multi-turn conversation technology is increasingly applied to scenes such as intelligent customer service and intelligent outbound, so that the cost of an enterprise is greatly reduced while the customer service efficiency is improved.

At present, a multi-turn conversation system is based on a slot filling method (abbreviated as a slot filling method) or an improved version of the slot filling method, realizes multi-turn conversations according to a series of flow node jump controls of a conversation tree, and completes conversation interaction according to a fixed-order business flow for collecting entity words. However, in the actual conversation process, the user replies to the expression collected by the entity words, and is easy to carry some other entity word information as the channel associated information of the conversation according to the conversation habit, and the channel associated information is often random and out of order. For example, in a car rental scene, the entity word collection process preset by the man-machine conversation system sequentially inquires about three entity attributes of a place, a car type and time, and when a first entity word is collected, a user may possibly make a prejudgment on the entity words to be collected by a subsequent collection node and carry the prejudged entity words according to the context of the current conversation, or may have subconsciously recorded some entity words to be collected subsequently through several times of complete conversation interaction. For example, the human-computer interaction system issues a question: "ask for a question about which car rental is to be rented? "user reply: "I rent the large-scale car on Beijing bang" or "I rent the large-scale car on Beijing bang tomorrow".

However, the conventional man-machine interaction system can only recognize one problem in one round of conversation, for example, only can recognize "beijing" in the above example, but cannot recognize two entity words of time ("tomorrow") and vehicle type ("large vehicle"), and in this way, even if the user carries other entity words (such as tomorrow and large vehicle) in one problem, the man-machine interaction system cannot recognize, so that multiple rounds of conversation interaction can be completed only according to a fixed-order business process to recognize all entity words.

That is to say, in the prior art, one round of dialog can only identify one question, and is not suitable for the situation that a user answers a plurality of questions simultaneously in one round of dialog, so that the number of interactions of the user is increased, and the user experience is reduced.

Therefore, the dialogue identification method is mainly used for solving the technical problems that in the prior art, one-wheel dialogue can only identify one question and cannot be suitable for the situation that a user answers a plurality of questions simultaneously in one-wheel dialogue, the interaction times of the user are increased, and the use experience of the user is reduced.

According to the conversation identification method, the target conversation in at least one conversation is obtained, and the target problem set to be identified in at least one conversation is determined; carrying out entity word recognition on the target conversation to obtain and store a plurality of entity words in the target conversation; matching each problem in the target problem set with a plurality of stored entity words respectively to determine the entity words corresponding to each problem; and determining answers of the questions according to the entity words corresponding to the questions. Therefore, when the user answers a plurality of questions in one round of conversation, the system can automatically identify the answers of the questions, reduce the unnecessary interaction times and improve the use experience of the user.

A dialog recognition method, apparatus, electronic device, and storage medium according to embodiments of the present application are described below with reference to the accompanying drawings. Before describing embodiments of the present invention in detail, for ease of understanding, common terminology will be introduced first:

intent, which is a business operation to be performed by the user, such as renting a car, querying weather, booking tickets, etc., is similar to the function in the code. The intention can be distinguished into a top-level intention and a sub-intention, the top-level intention is an intention which can be triggered at any time in a conversation process, and the sub-intention can be triggered only in a corresponding scene.

A scene, consisting of one intent and all dialog interactions (such as gathering physical words, clarifying, confirming, sub-intent, etc.) under that intent, switches scenes or ends scenes only when the user expresses a new intent or ends the dialog.

The entity words refer to parameters required for completing business actions, such as time, place, vehicle type and the like, are similar to parameters in functions, and one intention and a plurality of entity words can complete the handling of one business.

Conversation, a series of conversations of the same user with the man-machine conversation system within a certain time period is a conversation, such as a telephone in a telephone customer service.

In the dialog, a question and a response of a user and a man-machine dialog system are a dialog turn, namely, the dialog turn comprises a question and a corresponding answer or response of the question.

And the entity collection unit is node configuration required for completing an entity word collection process in the conversation process. The entity collection unit generally comprises a father node and two child nodes, wherein the father node and the child nodes are nodes in a dialogue tree, the father node is used for being used as a node selection entry condition to judge whether entity words to be collected are collected by dialogue context, under the condition that the entity words are not collected by the dialogue context, the first child node judges whether content expressed by a user contains the entity words, under the condition that the entity words are contained, assignment of variables in the context is completed to collect, under the condition that the entity words are not contained, the second child node is executed, the child node jumps to the father node again to initiate the collection of the entity words, the whole collection process comprises three nodes, and is a closed loop until the collection of the entity words is completed.

And the entity collection unit set consists of entity collection units. The entity collection unit set includes a plurality of consecutive entity collection units, that is, the number of consecutive entity collection units included in the entity collection unit set needs to be greater than 1. For example, referring to fig. 1, fig. 1 is a schematic view of a flow configuration of a locally continuous entity collection unit set. Assuming that the fourth entity collection unit from the left is an illegal entity collection unit, the first, second and third entity collection units from the left form an entity collection unit set, and only in the same entity collection unit set, the entity word information temporarily stored in the memory during the conversation process is effectively available for the entity words to be collected in the current entity collection unit set, while the memories between different entity collection unit sets are independent.

The dialog tree is layered according to a parent-child relationship, and a top-level node in fig. 1 refers to a first parent node of the dialog tree.

For example, if the entity collection unit set 1 is applied to a car rental scene, the entity collection unit set 1 may identify entity words in each dialog in the car rental scene, and if the entity collection unit set 2 is applied to a weather query scene, the entity collection unit set 2 may identify entity words in each dialog in the weather query scene. The memory between the entity collection unit set 1 and the entity collection unit set 2 is independent, and the entity words temporarily stored in the memory are not influenced mutually.

And the entity collecting unit detector is used for searching whether a legal entity collecting unit set exists in the layer where the current node is positioned, stopping detection when the legal entity collecting unit set does not exist, and storing the detected entity collecting unit set and the detected node track information after the detection is finished each time.

The validity of the entity collection unit set can be determined according to the number of the continuous entity collection units contained in the entity collection unit set, when the number of the continuous entity collection units contained in the entity collection unit set is greater than 1, the entity collection unit set is determined to be valid, and when the number of the continuous entity collection units contained in the entity collection unit set is less than or equal to 1, the entity collection unit set is determined to be invalid.

As shown in fig. 1, a complete detection process is to perform detection from the first entity collection unit from the left to the right, since the fourth entity collection unit from the left is an illegal entity collection unit structure, the first detection result is an entity collection unit set composed of the first entity collection unit from the left, the second entity collection unit from the left, and the third entity collection unit from the left, and record the information of the entity collection unit set and the node information of the detection track, and the second detection result is to start detection from the fifth entity collection unit from the left, since only one legal entity collection unit does not satisfy the continuous number greater than 1, the second detection result is that no entity collection unit set is detected, and only the node information of the detection track is recorded.

The entity collection unit set composed of the first entity collection unit, the second entity collection unit and the third entity collection unit supports intelligent collection of entity words when the entity words are collected, and the fifth entity collection unit (not composed of continuous entity collection unit sets with the number larger than 1) does not support intelligent collection of the entity words. Moreover, the fifth entity collecting unit from the left does not use the entity word information temporarily stored in the dialogue domain memory of the entity collecting unit set formed by the first, the second and the third entity collecting units from the left.

Fig. 2 is a flowchart illustrating a dialog recognition method according to an embodiment of the present application.

The embodiment of the present application is exemplified by the configuration of the dialog recognition method in the dialog recognition device, and the dialog recognition device may be applied to any electronic device, so that the electronic device may perform the dialog recognition function.

The electronic device may be any device with computing capability, for example, a PC (Personal Computer), a mobile terminal, a server, and the like, and the mobile terminal may be a hardware device with various operating systems, touch screens, and/or display screens, such as a mobile phone, a tablet Computer, a Personal digital assistant, a wearable device, an in-vehicle device, and a robot.

As shown in fig. 2, the dialog recognition method may include the steps of:

step 201, obtaining a target dialog in at least one dialog turn, and determining a target question set to be identified from the at least one dialog turn.

In the embodiment of the application, the target problem set is related to the business handled by the user. For example, when a user rents a car, the target question set contains various questions related to the car rental, for example, the target question set may be { "ask which car rental to ask? "," ask for a question about what model of vehicle to rent? "," when to rent a car when asking for a question? "}. Alternatively, when the user queries for weather, the target question set contains various questions related to the queried weather, for example, the target question set may be { "ask where weather is to be queried? "," when to ask for weather? "}. Alternatively, when a user makes a reservation, such as a flight reservation, the target question set may be { "do you reserve a flight on which day? "," your departure time? "," your city of departure? "," your arrival city? "}.

It should be understood that, in order to identify the user's intent and thereby perform an operation matching the user's intent, in the embodiment of the present application, the target dialog may contain individual dialogs that the user has answered in the current session. For example, when the human-computer interaction system presents a question "ask a query about which car rental? ", the user's answer is: "i rent a large car at Beijing bang tomorrow", only one session is obtained, and the above session is "ask for a question about where to rent a car? "the conversation is followed by" I rent large-sized cars on Beijing Help tomorrow ".

And when the target dialog comprises the dialogs which are answered by the user in the current conversation, the target question set is the question set corresponding to the questions in the target dialog.

In one possible implementation manner of the embodiment of the application, a user intention may be identified, and a target problem set may be determined according to the user intention. And according to all questions in the target question set, questions are sequentially asked for the user, and all conversations containing answers of the user are used as target conversations.

An example is that each service handling control, such as controls for car renting, weather inquiry, ticket buying, and the like, can be displayed on the interface of the human-computer interaction system, so that a user can enter a corresponding service handling page by triggering the corresponding control, the human-computer interaction system can determine the intention of the user based on the corresponding service, determine a target question set according to the intention of the user, and ask questions of the user in sequence based on each question in the target question set, so that the user can answer, and each conversation containing the answer of the user is used as a target conversation.

For another example, the user may directly input a service to be transacted (e.g., renting a car, inquiring weather, buying tickets, etc.), wherein the input means includes, but is not limited to, touch input (e.g., sliding, clicking, etc.), keyboard input, voice input, etc., so that the user's intention may be recognized according to the service input by the user, the target question set may be determined according to the user's intention, and the user may be asked questions in order based on the questions in the target question set, so that the user may make an answer, and the respective conversations including the user's answer may be used as the target conversation.

The user intention may be identified based on a semantic analysis technique in an NLP technique in the AI field, or may also be identified based on template matching, text classification, and other methods in the NLP technique, which is not limited in this application.

The AI is a subject for studying a computer to simulate some human mental processes and intelligent behaviors (such as learning, reasoning, thinking, planning, etc.), and has both hardware and software technologies. Artificial intelligence hardware technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing, and the like; the artificial intelligence software technology mainly comprises a computer vision technology, a voice recognition technology, a natural language processing technology, machine learning/deep learning, a big data processing technology, a knowledge map technology and the like.

In another possible implementation manner of the embodiment of the present application, one question set may be extracted from a plurality of question sets according to a set rule to serve as a target question set, questions may be sequentially asked to a user according to the questions in the target question set, and each dialog including the answer of the user may be served as a target dialog.

For example, taking the set rule as an example in sequence, the human-computer interaction system has N question sets in total, and may select one question set first, for example, select the first question set, or may select one question set randomly, and ask a question to the user in sequence according to each question in the selected question set. After the user answers m questions, assuming that the currently selected question set is the tth question set, the tth question set can be used as a target question set, questions can be asked to the user in sequence according to all questions in the target question set, and all conversations containing the answers of the user are used as target conversations.

For another example, a set rule is taken as an example of the correlation between scenes, a previous scene is determined as a ticket booking scene according to the intention of a user, and local tourism, car rental, accommodation and the like may be involved after ticket booking, so that the current scene may be a car rental, accommodation and the like scene, and after the current scene is determined, a target question set corresponding to the current scene may be directly determined, for example, the target question set may be determined according to the intention corresponding to the current scene, so that the user may be asked questions in order according to questions in the target question set, and each conversation including the answer of the user is taken as a target conversation.

Step 202, performing entity word recognition on the target conversation to obtain and store a plurality of entity words in the target conversation.

In the embodiment of the application, entity word recognition can be performed on the target conversation to obtain a plurality of entity words in the target conversation, and the plurality of entity words are stored.

For example, entity word recognition may be performed on the target dialog based on methods such as keyword matching, template matching, and statistical model in NLP technology, so as to obtain a plurality of entity words in the target dialog. It should be noted that the above-mentioned method for recognizing entity words is only an exemplary embodiment, but not limited thereto, and other methods for recognizing entity words known in the art are also included as long as the entity words in the target dialog can be obtained, i.e., the method for recognizing entity words is not limited in the present application.

Step 203, matching each question in the target question set with a plurality of stored entity words respectively to determine the entity words corresponding to each question.

In the embodiment of the application, after a plurality of entity words in the target dialog are obtained, each question in the target question set may be respectively matched with the plurality of entity words in the target dialog, so as to determine the entity word corresponding to each question.

As an application scenario, when the dialog recognition method is applied to a car rental scenario, the set of target questions to be recognized is: { "ask for a question about which rental car to rent? "," ask for a question about what model of vehicle to rent? "," when to rent a car when asking for a question? "}, in the first round of dialogue, the questions sent by the man-machine dialogue system are: "ask for a question about which car rental is to be rented? ", assume that the user's answer is: "i rent a large car at the side of beijing tomorrow", therefore, it can be determined that the entity word corresponding to question 1 ("ask which car to rent.

And step 204, determining answers of the questions according to the entity words corresponding to the questions.

In the embodiment of the application, after the entity word corresponding to each question is determined, the answer of each question can be determined according to the entity word corresponding to each question. For example, the entity words may be filled in the corresponding slots based on a slot filling method to obtain answers to the corresponding questions.

Still as exemplified by the above example, the answer to question 1 can be found to be: "car rental in Beijing", answer to question 2 is "car rental on large scale", and answer to question 3 is "car rental time is tomorrow". Therefore, the man-machine conversation system can execute corresponding business operation according to the answers corresponding to the problems, help the user rent the large-sized vehicle in Beijing, and ensure that the vehicle renting time is tomorrow. Therefore, when a user answers a plurality of questions in one round of conversation, the man-machine interaction system can automatically identify the answers of the questions, so that the unnecessary interaction times can be reduced, the efficiency of transacting services by the user is improved, and the use experience of the user is improved.

The dialog recognition system of the embodiment of the application acquires a target dialog in at least one round of dialog and determines a target problem set to be recognized in the at least one round of dialog; carrying out entity word recognition on the target conversation to obtain and store a plurality of entity words in the target conversation; matching each problem in the target problem set with a plurality of stored entity words respectively to determine the entity words corresponding to each problem; and determining answers of the questions according to the entity words corresponding to the questions. Therefore, when the user answers a plurality of questions in one round of conversation, the system can automatically identify the answers of the questions, reduce the unnecessary interaction times and improve the use experience of the user.

In a possible implementation manner of the embodiment of the present application, the entity words in the target dialog may be identified by the entity collection unit, so as to determine the entity words corresponding to the problems. The above process will be described in detail with reference to the second embodiment.

Fig. 3 is a flowchart illustrating a dialog recognition method according to a second embodiment of the present application.

As shown in fig. 3, the dialog recognition method may include the steps of:

step 301, obtaining a target dialog in at least one round of dialog.

The execution process of step 301 may refer to the execution process of step 201 in the above embodiments, which is not described herein again.

Step 302, according to a set rule, determining a target entity collection unit set from a plurality of entity collection unit sets corresponding to at least one pair of dialogues; the target entity collection unit set comprises one or more entity collection units corresponding to the problems in the target problem set and is used for identifying entity words matched with the corresponding problems.

In the embodiment of the present application, the setting rules are preset, for example, the setting rules may be in sequence, or the setting rules may be other, which is not limited in the present application.

In the embodiment of the present application, the entity words in the dialog process may be collected or identified through a plurality of entity collection unit sets, so that after at least one round of dialog is obtained, a plurality of entity collection unit sets corresponding to the at least one round of dialog may be determined first, and based on a set rule, a target entity collection unit set is determined from the plurality of entity collection unit sets corresponding to the at least one round of dialog, where the target entity collection unit set may include one or more entity collection units corresponding to a problem in the target problem set, and each entity collection unit is used to identify the entity word matching the corresponding problem.

For example, taking the set rule as an example in order, it is assumed that the plurality of entity collection units corresponding to at least one pair of dialogues are entity collection unit set 1, entity collection unit set 2, and entity collection unit set 3, respectively. Assuming that the current scene is booking tickets, the subsequent scenes can discuss local car renting, lodging and tourism, so that entity word recognition can be performed on the conversations in the current scene according to the entity collection units in the entity collection unit set 1, the subsequent entities in the entity collection unit set 2 can be used for performing entity word recognition on the conversations in the car renting scene, and the subsequent entities in the entity collection unit set 3 can be used for performing entity word recognition on the conversations in the lodging scene.

For example, the setting rule may be related to the user intention, for example, the entity collection units in the entity collection unit set 1 may perform entity word recognition on the dialog of the booking scene, the entity collection units in the entity collection unit set 2 may perform entity word recognition on the dialog of the car rental scene, the entity collection units in the entity collection unit set 3 may perform entity word recognition on the dialog of the lodging scene, and the entity collection units in the entity collection unit set 4 may perform entity word recognition on the dialog of querying the weather scene. Assuming that the intention recognition is performed on at least one pair of dialogs, and the intention of the dialog is determined to be car rental, the target entity collection unit set may be the entity collection unit set 2, and thus, the entity word recognition may be performed on the dialog in the car rental scene through each entity collection unit in the entity collection unit set 2.

In the embodiment of the application, one or more entity collecting units corresponding to the problems in the target problem set are adopted to identify the entity words matched with the corresponding problems. Because the entity collecting unit is configured with the nodes required by the entity word collecting process in the conversation process, the accuracy of the entity word recognition result can be improved based on the fact that each entity collecting unit recognizes the entity words matched with the corresponding problems.

And 303, performing entity word recognition on the target conversation to obtain and store a plurality of entity words in the target conversation.

The process of step 303 may refer to the process of step 202 in the above embodiments, which is not described herein again.

And step 304, adopting each entity collecting unit in the target entity collecting unit set to sequentially identify a plurality of entity words in the target conversation so as to obtain the entity words matched with the problems corresponding to each entity collecting unit.

In this embodiment, when the target entity collection unit set includes a plurality of entity collection units, each entity collection unit in the target entity collection unit set may be adopted to sequentially identify a plurality of entity words in the target dialog, so as to obtain the entity words matched by the problem corresponding to each entity collection unit.

Step 305, determining answers to the questions according to the entity words corresponding to the questions.

The process of step 305 may refer to the process of step 204 in the above embodiments, which is not described herein again.

In the embodiment of the application, because the entity collection unit is configured with the nodes required by one entity word collection process in the conversation process, the accuracy of the entity word identification result can be improved based on the fact that each entity collection unit identifies the entity words matched with the corresponding problems.

In a possible implementation manner of the embodiment of the present application, when all the entity collection units in the target entity collection unit set are completely identified, the target entity collection unit set may be stopped from being used for identification, or when an entity collection unit that does not identify a matching entity word exists, the target entity collection unit set may be stopped from being used for identification. Therefore, after the entity words contained in the target dialogue are identified and obtained, the target entity collection unit set is stopped from being adopted to continue identification, and the processing pressure of the system can be reduced.

It can be understood that, when an entity collection unit that does not recognize a matching entity word exists, it indicates that the user has not answered all the questions in the target question set, and at this time, the user cannot be helped to complete the corresponding service, so as to be a possible implementation manner of the embodiment of the present application, in the case that an entity collection unit that does not recognize a matching entity word exists, the following text of the target conversation may be obtained from at least one round of conversation, the entity words in the following text are recognized and stored, and the target entity collection unit set is adopted to continue recognizing each stored entity word until all the entity collection units in the target entity collection unit set recognize the entity word matched with the corresponding question.

The above process is described in detail with reference to example three.

Fig. 4 is a flowchart illustrating a dialog recognition method according to a third embodiment of the present application.

As shown in fig. 4, the dialog recognition method may include the steps of:

step 401, obtaining a target dialog in at least one round of dialog.

Step 402, according to a set rule, determining a target entity collection unit set from a plurality of entity collection unit sets corresponding to at least one pair of dialogues; the target entity collection unit set comprises one or more entity collection units corresponding to the problems in the target problem set and is used for identifying entity words matched with the corresponding problems.

And 403, performing entity word recognition on the target conversation to obtain and store a plurality of entity words in the target conversation.

And step 404, adopting each entity collecting unit in the target entity collecting unit set to sequentially identify a plurality of entity words in the target conversation so as to obtain the entity words matched with the problems corresponding to each entity collecting unit.

The execution process of steps 401 to 404 may refer to the execution process of the above embodiment, which is not described herein again.

And step 405, stopping adopting the target entity collection unit set for recognition under the condition that the entity collection unit without the matching entity word is identified.

For example, the current scenario is a car rental scenario, and the target question set is { "ask for which car rental is to be found? "," ask for a question about what model of vehicle to rent? "," when to rent a car when asking for a question? "when the human-computer interaction system presents a question of asking a question about which car rental? ", the user's answer is: "i rent a large car at beijing bang", it can be determined that the entity word corresponding to question 1 ("ask which car to rent. Accordingly, recognition of the target dialogue using the target entity collecting unit may be stopped.

Step 406, obtaining a context of the target dialog from at least one round of dialog.

It can be understood that, when there is an entity collecting unit that does not recognize a matching entity word, it indicates that the user has not answered all questions in the target question set, at this time, the user cannot be helped to complete the corresponding service, and therefore, as a possible implementation manner of the embodiment of the present application, in the presence of an entity collecting unit that does not recognize a matching entity word, the context of the target conversation may also be obtained from at least one round of conversation.

Step 407, storing the entity words in the following text.

In the embodiment of the application, the entity words in the following text can be identified, and the identified entity words are stored.

And step 408, adopting the target entity collection unit set to continuously identify the stored entity words.

In this embodiment of the application, the target entity collection unit set may be adopted to continuously identify the stored entity words, that is, the entity collection units in the target entity collection unit set are adopted to sequentially identify the stored entity words, so as to obtain the entity words matched with the problems corresponding to the entity collection units. Therefore, the entity words matched with the questions in the target question set can be identified, so that answers of the questions can be determined subsequently, and a user can be helped to handle corresponding services accurately.

Still in the above example, the question presented by the human-computer interaction system in the context of the target dialog is "when asking to rent a car? ", the user's answer is: and in tomorrow, a target entity collection unit set is adopted to continuously identify each stored entity word, so that the entity word corresponding to the problem 3 can be determined to be tomorrow.

And step 409, determining answers of the questions according to the entity words corresponding to the questions.

The execution process of step 409 may refer to the execution process of the above embodiment, and is not described herein again.

In a possible implementation manner of the embodiment of the present application, when each entity collection unit in the target entity collection unit set identifies a matching entity word, a plurality of stored entity words may be deleted. Therefore, when the entity words matched with the problems are identified, the stored entity words are deleted, on one hand, the occupation of storage resources can be reduced, on the other hand, the target entity collecting unit can be used for identifying other conversations, the stored entity words are deleted, and the influence on the identification results of the other conversations can be avoided.

For example, a ticket booking scene may relate to time identification, a lodging scene and a car renting scene may also relate to time identification, assuming that a scene of a target conversation is the ticket booking scene, a stored entity word may include time, and when an entity word matching each problem in a target problem set is identified, if the stored entity words are not deleted, the entity words may occur in the lodging scene or the car renting scene, and the time does not exist in the conversation, the stored time is used as the entity word identified by the target entity collection unit set, which may cause a misjudgment condition, and may seriously reduce the use experience of a user.

For example, the entity words stored in the booking scenario are: booking time, departure city and arrival city, and the entity words identified in the car renting scene are as follows: if the entity words stored in the ticket booking scene are not deleted, the departure time in the ticket booking scene is identified as the car renting time in the car renting scene, obviously, the identification result is incorrect, and therefore, the stored entity words can be deleted under the condition that all the entity collecting units in the target entity collecting unit set identify the matched entity words, on one hand, the accuracy of the identification result can be improved, and on the other hand, the occupation of storage resources can be reduced.

As an example, referring to fig. 5, fig. 5 is a schematic flowchart of a dialog recognition method provided in the fourth embodiment of the present application.

1. A user can input a query sentence (query), wherein the input mode includes but is not limited to touch input, keyboard input, voice input and the like;

2. the man-machine interaction system can perform intention recognition and entity word recognition on the query;

3. entering scenes matched with the user intention, such as a ticket booking scene, a car renting scene and the like;

4. judging whether the entity word recognition result is empty, and under the condition that the entity word recognition result is empty, indicating that the pair of dialogs is not recognized to obtain the entity word, thus directly ending and executing the processing flow corresponding to the current dialogue node; and under the condition that the entity word recognition result is not empty, executing the step 5;

5. judging whether the current conversation node is detected by the entity collecting detector, starting the entity collecting detector under the condition that the current conversation node is not detected by the entity collecting detector, searching whether a legal entity collecting unit set exists in the layer where the current conversation node is located by using the entity collecting detector, temporarily storing a detection result, and continuously executing the step 6; and under the condition that the entity is detected by the entity collecting detector, no processing is carried out, and the step 6 is continuously executed;

6. judging whether the current conversation node belongs to a node in a certain entity collection unit in the entity collection unit set detected by the entity collection detector, and if not, indicating that the current conversation node does not accord with the configuration rule of the entity collection unit and cannot carry out intelligent entity collection, wherein the intelligent collection of the conversation entity words is finished in advance by the current conversation node, and executing a corresponding processing flow of the current conversation node; under the condition that the real words are not in the round of the real words, selectively temporarily storing all recognition results of the round of the real words based on the current scene, for example, when the real words are wound, for example, a mobile phone number and digital real words are wound, at the moment, the wound real words need to be independently collected, the wound real words cannot be temporarily stored, a selective memory temporary storage process is performed, and the step 7 is continuously executed;

7. judging whether an entity collecting unit corresponding to the current conversation node exists or not, and whether the entity words temporarily stored in the step 6 contain the entity words needing to be collected by the currently existing entity collecting unit or not;

and under the condition that the two conditions are met, performing entity word collection of the current entity collection unit, pointing the current entity collection unit to the next entity collection unit in the entity collection unit set detected by the entity collection detector in the step 5, and continuing to perform the step 7 to perform the entity word collection process of the next entity collection unit until the condition of the step 7 is not met. And when the two conditions are not simultaneously satisfied, continuing to execute the step 8;

8. judging whether the entity word collection of the entity collection unit set detected by the entity collection detector in the step 5 is completely finished in the step 7; under the condition that all entity word collection is completed, emptying all the temporarily stored entity words in the step 6, wherein the entity collection unit set detected in the step 5 is completely collected, and the temporarily stored entity word information has no meaning, so that the entity words can be emptied, all the entity word intelligent collection is completed, and answers corresponding to the conversation nodes are replied; and if the entity collection unit set has the entity collection unit which does not finish the entity word collection, the intelligent collection of the dialog entity words is finished, and the answer of the corresponding dialog node is replied.

Therefore, in a certain scene of multi-round conversation, automatic collection of a plurality of continuous entity words can be supported, in the intelligent collection process of the entity words, the short-term memory capacity flexible to the conversation is achieved, more effective conversation information is captured in the conversation, the memory capacity in the conversation process is achieved by the intelligent collection of the entity words, unnecessary interaction times are effectively reduced, the intelligence and the human-simulated performance of the conversation capacity are improved, and meanwhile the efficiency and the experience effect of business handling of a user are improved.

It should be noted that, in the prior art, intelligent collection of entity words can be achieved through a complex configuration, for example, at each collection node, the entity to be collected subsequently is implanted in advance in a "hidden buried point" manner, so as to achieve the purpose by means of pre-collection. For example, at the same time of collecting the place, judging whether collectable vehicle type and time entity words exist, if so, collecting in advance, and analogizing the next collecting node.

However, the above-mentioned physical word collection method introduces additional configuration cost, and too cumbersome and complicated configuration is not conducive to extended maintenance, and even configuration errors may occur, and furthermore, the method cannot be migrated with zero cost because the dialog flow configuration and information collection requirements of each scene are different.

In the method, an additional complex configuration flow is not required, the use threshold is reduced, and the potential uncontrollable risk existing in the traditional configuration mode is effectively and automatically avoided. In addition, the conversation processing method is not a scheme customized and developed based on a certain service scene, is an abstract unified solution for the type of requirements in the conversation system, can be migrated to other scenes with zero cost, and reduces the construction cost for meeting the type of requirements of a multi-turn conversation system.

Corresponding to the dialog recognition method provided in the embodiment of fig. 1 to 4, the present application also provides a dialog recognition device, and since the dialog recognition device provided in the embodiment of the present application corresponds to the dialog recognition method provided in the embodiment of fig. 1 to 4, the embodiment of the dialog recognition method is also applicable to the dialog recognition device provided in the embodiment of the present application, and will not be described in detail in the embodiment of the present application.

Fig. 6 is a schematic structural diagram of a dialog recognition device according to a fifth embodiment of the present application.

As shown in fig. 6, the dialog recognition device 600 includes: a dialog acquisition module 610, a set determination module 620, a recognition module 630, a matching module 640, and an answer determination module 650.

The dialog obtaining module 610 is configured to obtain a target dialog in at least one round of dialog.

A set determination module 620, configured to determine a set of target questions to be identified from at least one round of dialog.

The recognition module 630 is configured to perform entity word recognition on the target dialog, obtain and store a plurality of entity words in the target dialog.

The matching module 640 is configured to match each question in the target question set with a plurality of stored entity words, so as to determine an entity word corresponding to each question.

The answer determining module 650 is configured to determine an answer to each question according to the entity word corresponding to each question.

In a possible implementation manner of the embodiment of the present application, the set determining module 620 is specifically configured to: determining a target entity collection unit set from a plurality of entity collection unit sets corresponding to at least one pair of dialogues according to a set rule; the target entity collection unit set comprises one or more entity collection units corresponding to the problems in the target problem set and is used for identifying entity words matched with the corresponding problems.

In a possible implementation manner of the embodiment of the present application, the target entity collection unit set includes a plurality of entity collection units, and the matching module 640 is specifically configured to: and adopting each entity collecting unit in the target entity collecting unit set to sequentially identify a plurality of entity words in the target conversation so as to obtain the entity words matched with the problems corresponding to each entity collecting unit.

In a possible implementation manner of the embodiment of the present application, the dialog recognition device 600 may further include:

and the stopping module is used for stopping adopting the target entity collecting unit set for recognition if all the entity collecting units in the target entity collecting unit set are completely recognized or an entity collecting unit which does not recognize the matched entity word exists.

and the context acquisition module is used for acquiring the context of the target conversation from at least one round of conversation.

And the storage module is used for storing the entity words in the following text.

The identifying module 630 is further configured to continue identifying the stored entity words by using the target entity collection unit set.

and the deleting module is used for deleting the stored entity words under the condition that all the entity collecting units in the target entity collecting unit set identify the matched entity words.

The dialog recognition device of the embodiment of the application acquires a target dialog in at least one round of dialog and determines a target problem set to be recognized in the at least one round of dialog; carrying out entity word recognition on the target conversation to obtain and store a plurality of entity words in the target conversation; matching each problem in the target problem set with a plurality of stored entity words respectively to determine the entity words corresponding to each problem; and determining answers of the questions according to the entity words corresponding to the questions. Therefore, when the user answers a plurality of questions in one round of conversation, the system can automatically identify the answers of the questions, reduce the unnecessary interaction times and improve the use experience of the user.

There is also provided, in accordance with an embodiment of the present application, an electronic device, a readable storage medium, and a computer program product.

FIG. 7 illustrates a schematic block diagram of an example electronic device 700 that can be used to implement embodiments of the present application. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be examples only, and are not meant to limit implementations of the disclosure described and/or claimed herein.

As shown in fig. 7, the device 700 includes a computing unit 701, which can perform various appropriate actions and processes in accordance with a computer program stored in a ROM (Read-Only Memory) 702 or a computer program loaded from a storage unit 708 into a RAM (Random Access Memory) 703. In the RAM 703, various programs and data required for the operation of the device 700 can also be stored. The computing unit 701, the ROM 702, and the RAM 703 are connected to each other by a bus 704. An I/O (Input/Output) interface 705 is also connected to the bus 704.

Various components in the device 700 are connected to the I/O interface 705, including: an input unit 706 such as a keyboard, a mouse, or the like; an output unit 707 such as various types of displays, speakers, and the like; a storage unit 708 such as a magnetic disk, optical disk, or the like; and a communication unit 709 such as a network card, modem, wireless communication transceiver, etc. The communication unit 709 allows the device 700 to exchange information/data with other devices via a computer network, such as the internet, and/or various telecommunication networks.

Computing unit 701 may be a variety of general purpose and/or special purpose processing components with processing and computing capabilities. Some examples of the computing Unit 701 include, but are not limited to, a CPU (Central Processing Unit), a GPU (graphics Processing Unit), various dedicated AI (Artificial Intelligence) computing chips, various computing Units running machine learning model algorithms, a DSP (Digital Signal Processor), and any suitable Processor, controller, microcontroller, and the like. The calculation unit 701 executes the respective methods and processes described above, such as a dialogue recognition method. For example, in some embodiments, the dialog recognition method may be implemented as a computer software program tangibly embodied in a machine-readable medium, such as storage unit 708. In some embodiments, part or all of a computer program may be loaded onto and/or installed onto device 700 via ROM 702 and/or communications unit 709. When the computer program is loaded into the RAM 703 and executed by the computing unit 701, one or more steps of the dialog recognition method described above may be performed. Alternatively, in other embodiments, the computing unit 701 may be configured to perform the dialog recognition method in any other suitable way (e.g. by means of firmware).

Various implementations of the systems and techniques described here above may be realized in digital electronic circuitry, Integrated circuitry, FPGAs (Field Programmable Gate arrays), ASICs (Application-Specific Integrated circuits), ASSPs (Application Specific Standard products), SOCs (System On Chip, System On a Chip), CPLDs (Complex Programmable Logic devices), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs that are executable and/or interpretable on a programmable system including at least one programmable processor, which may be special or general purpose, receiving data and instructions from, and transmitting data and instructions to, a storage system, at least one input device, and at least one output device.

Program code for implementing the methods of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the program codes, when executed by the processor or controller, cause the functions/operations specified in the flowchart and/or block diagram to be performed. The program code may execute entirely on the machine, partly on the machine, as a stand-alone software package partly on the machine and partly on a remote machine or entirely on the remote machine or server.

In the context of this disclosure, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a RAM, a ROM, an EPROM (Electrically Programmable Read-Only-Memory) or flash Memory, an optical fiber, a CD-ROM (Compact Disc Read-Only-Memory), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having: a Display device (e.g., a CRT (Cathode Ray Tube) or LCD (Liquid Crystal Display) monitor) for displaying information to a user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which a user can provide input to the computer. Other kinds of devices may also be used to provide for interaction with a user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form, including acoustic, speech, or tactile input.

The systems and techniques described here can be implemented in a computing system that includes a back-end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back-end, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: LAN (Local Area Network), WAN (Wide Area Network), and the Internet.

The computer system may include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The Server may be a cloud Server, which is also called a cloud computing Server or a cloud host, and is a host product in a cloud computing service system, so as to solve the defects of high management difficulty and weak service expansibility in a conventional physical host and a VPS (Virtual Private Server). The server may also be a server of a distributed system, or a server incorporating a blockchain.

According to an embodiment of the present application, there is also provided a computer program product, wherein when instructions in the computer program product are executed by a processor, the dialog recognition method provided by the above-mentioned embodiment of the present application is executed.

According to the technical scheme of the embodiment of the application, a target dialog in at least one round of dialog is obtained, and a target problem set to be identified in the at least one round of dialog is determined; carrying out entity word recognition on the target conversation to obtain and store a plurality of entity words in the target conversation; matching each problem in the target problem set with a plurality of stored entity words respectively to determine the entity words corresponding to each problem; and determining answers of the questions according to the entity words corresponding to the questions. Therefore, when the user answers a plurality of questions in one round of conversation, the system can automatically identify the answers of the questions, reduce the unnecessary interaction times and improve the use experience of the user.

It should be understood that various forms of the flows shown above may be used, with steps reordered, added, or deleted. For example, the steps described in the present application may be executed in parallel, sequentially, or in different orders, and the present invention is not limited thereto as long as the desired results of the technical solutions disclosed in the present application can be achieved.

The above-described embodiments should not be construed as limiting the scope of the present application. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may be made in accordance with design requirements and other factors. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A dialog recognition method, the method comprising:

2. The dialog recognition method of claim 1, wherein the determining a set of target questions to be recognized from the at least one round of dialog comprises:

determining a target entity collection unit set from a plurality of entity collection unit sets corresponding to the at least one dialog according to a set rule; the target entity collection unit set comprises one or more entity collection units corresponding to the questions in the target question set, and the entity collection units are used for identifying entity words matched with the corresponding questions.

3. The dialog recognition method of claim 2, wherein the set of target entity collecting units includes a plurality of the entity collecting units, and the matching of each question in the set of target questions with a plurality of stored entity words to determine the entity word corresponding to each question comprises:

and adopting each entity collecting unit in the target entity collecting unit set to sequentially identify a plurality of entity words in the target conversation so as to obtain the entity words matched with the problems corresponding to each entity collecting unit.

4. The dialog recognition method of claim 3, wherein, after the employing each entity collection unit to sequentially recognize the plurality of entity words in the target dialog, further comprises:

and after all the entity collecting units in the target entity collecting unit set are identified, or the entity collecting units which do not identify the matched entity words exist, stopping adopting the target entity collecting unit set for identification.

5. The dialog recognition method of claim 4, wherein the existence of the entity collection unit in which the matching entity word is not recognized further comprises, after stopping recognition with the entity collection unit set:

obtaining the context of the target dialog from the at least one round of dialog;

storing the entity words in the text;

and adopting the target entity collection unit set to continuously identify each stored entity word.

6. The dialog recognition method of claim 4, wherein the method further comprises:

and deleting the stored entity words under the condition that each entity collecting unit in the target entity collecting unit set identifies the matched entity words.

7. A dialog recognition device comprising:

8. The dialog recognition device of claim 7, wherein the set determination module is specifically configured to:

9. The dialog recognition device of claim 8, wherein the set of target entity collection units includes a plurality of the entity collection units, and the matching module is specifically configured to:

10. The dialog recognition device of claim 9, wherein the device further comprises:

and the stopping module is used for stopping adopting the target entity collecting unit set for recognition if all the entity collecting units in the target entity collecting unit set are recognized completely or the entity collecting units which are not recognized to be matched with the entity words exist.

11. The dialog recognition device of claim 10, wherein the device further comprises:

the context acquisition module is used for acquiring the context of the target conversation from the at least one round of conversation;

the storage module is used for storing the entity words in the text;

the identification module is further configured to continuously identify each stored entity word by using the target entity collection unit set.

12. The dialog recognition device of claim 10, wherein the device further comprises:

and the deleting module is used for deleting the stored entity words under the condition that the entity collecting units in the target entity collecting unit set identify the matched entity words.

13. An electronic device, comprising:

at least one processor; and

the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the dialog recognition method of any of claims 1-6.

14. A non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform the dialog recognition method of any of claims 1-6.

15. A computer program product in which instructions, when executed by a processor, perform the dialog recognition method of any of claims 1-6.