CN113132214A

CN113132214A - Conversation method, device, server and storage medium

Info

Publication number: CN113132214A
Application number: CN201911416871.0A
Authority: CN
Inventors: 熊为星; 熊友军
Original assignee: Ubtech Robotics Corp
Current assignee: Ubtech Robotics Corp
Priority date: 2019-12-31
Filing date: 2019-12-31
Publication date: 2021-07-16
Anticipated expiration: 2039-12-31
Also published as: CN113132214B

Abstract

The application discloses an intelligent conversation method, an intelligent conversation device, a server and a storage medium, wherein the conversation server at least comprises task conversation service and non-task conversation service, and the conversation method comprises the following steps: the conversation server receives conversation sentences; obtaining semantic information of a conversation sentence at least by using a task conversation service, wherein the semantic information at least comprises intention information; acquiring a dialogue intention of a dialogue statement in semantic information; calling a dialogue service response dialogue statement corresponding to the dialogue intention and providing a service result; the conversation intents comprise task intents and service intents, the task intents need to provide service results by responding to conversation sentences through task conversation services, and the service intents need to provide service results by responding to the conversation sentences through non-task conversation services. Through the mode, the method and the device can accurately identify the dialogue intention of the user, and the given answer is more effective.

Description

Conversation method, device, server and storage medium

Technical Field

The present application relates to the field of artificial intelligence technologies, and in particular, to a conversation method, device, server, and storage medium.

Background

With the development of artificial intelligence technology, more and more robots are applied in different scenes to realize different functions, such as intelligent question-answering robots, task robots, chatting robots and the like. For example, a question and answer robot is mainly used in the smart customer service, a task robot is mainly used in the smart audio or car robot, and a chatting robot is used for eliminating the time of a client. However, these robots have a single function and hard conversation, and cannot flexibly adjust the response strategy for different conversation sentences, so that the given response is not reasonable enough.

Disclosure of Invention

The technical problem mainly solved by the application is to provide a conversation method, a conversation device, a server and a storage medium, which can accurately identify the conversation intention of a user, and the given answer is more effective.

In order to solve the technical problem, the application adopts a technical scheme that: there is provided a dialog method, the dialog server comprising at least a task dialog service and a non-task dialog service, the dialog method comprising: the conversation server receives conversation sentences; semantic information of a conversation statement is at least obtained by using a task conversation service, the semantic information obtained by the task conversation service is first-class semantic information, and the first-class semantic information at least comprises intention information; acquiring a conversation intention of a conversation sentence in the first type of semantic information; calling a dialogue service response dialogue statement corresponding to the dialogue intention and providing a service result; the conversation intention comprises a task intention and a service intention, the task intention comprises a task identifier, the task identifier is used for identifying a service result which needs to provide a corresponding task intention by using a task conversation service response conversation statement, the service intention comprises a conversation service identifier, and the conversation service identifier is used for identifying a service result which needs to provide a corresponding service intention by using a non-task conversation service response conversation statement of the corresponding conversation service identifier.

Wherein, still include after the dialogue server receives the dialogue statement: distributing the conversation sentences to task conversation services and non-task conversation services; respectively utilizing task conversation service and non-task conversation service to obtain semantic information of conversation sentences to obtain a plurality of semantic information, wherein the semantic information comprises first type semantic information and second type semantic information; the second type of semantic information is semantic information acquired by the non-task conversation service, and at least comprises a service result; invoking a dialog service corresponding to the dialog intention to respond to the dialog statement, and providing a service result comprises: responding to the conversation intention as a task intention, and further judging whether necessary slot position information required by executing the task intention is met; and responding to the service intention, and returning the service result output by the non-task conversation service corresponding to the service intention as a final service result to the user side.

Wherein, judging whether the necessary slot position information required by the task intent satisfies the following steps: in response to the necessary slot location information needed to execute the task intent not being satisfied; returning a question-following statement for clarifying necessary slot position information to the user side; and responding to the condition that necessary slot position information required for executing the task intention is met, and executing the task intention to obtain a service result.

Wherein, still include after the dialogue server receives the dialogue statement: judging whether the current conversation sentence is an associated sentence associated with the previous pair of speaking sentences; responding to the current conversation sentence as an associated sentence associated with the last pair of the speaking sentences, and only distributing the current conversation sentence to the task conversation service to obtain first-class semantic information, wherein the first-class semantic information also comprises slot position information; and continuously judging whether necessary slot position information required by the task intent is met.

And responding to the question tracking turn number larger than the preset threshold value, adjusting the conversation intention to be the preset service intention, wherein the service intention corresponding to the preset service intention is the chatting service.

Wherein, before receiving the dialogue statement, the dialogue server further comprises: receiving corpus information, wherein the corpus information at least comprises intention corpora; and training the service model by utilizing the corpus information to obtain a task conversation service model and/or a non-task conversation service model.

Wherein the non-task dialogue service comprises one or more of an intelligent question and answer service and a chatting service.

In order to solve the above technical problem, another technical solution adopted by the present application is: the conversation device comprises a receiving module, a first obtaining module, a second obtaining module and a processing module, wherein the receiving module is used for receiving conversation sentences; the first acquisition module is used for acquiring semantic information of a conversation sentence at least by using task conversation service, wherein the semantic information acquired by the task conversation service is first-class semantic information which at least comprises intention information; the second acquisition module is used for acquiring the conversation intention of the conversation statement in the first type of semantic information; the processing module is used for calling the dialogue service corresponding to the dialogue intention to respond to the dialogue statement and providing a service result; the conversation intention comprises a task intention and a service intention, the task intention comprises a task identifier, the task identifier is used for identifying a service result which needs to provide a corresponding task intention by using a task conversation service response conversation statement, the service intention comprises a conversation service identifier, and the conversation service identifier is used for identifying a service result which needs to provide a corresponding service intention by using a non-task conversation service response conversation statement of the corresponding conversation service identifier.

In order to solve the above technical problem, another technical solution adopted by the present application is: there is provided a server comprising a processor for executing instructions to implement the above-described conversational method.

In order to solve the above technical problem, another technical solution adopted by the present application is: there is provided a computer-readable storage medium for storing instructions/program data that can be executed to implement the above-described dialogue method.

The beneficial effect of this application is: different from the situation of the prior art, the application provides a conversation method, other non-task conversation services are fused into task conversation services as intents, different conversation intents can be effectively distinguished, corresponding conversation services are called to respond to conversation sentences, services are provided more accurately, more accurate answers are given, and therefore the intelligence of a service platform is enhanced, and man-machine conversation is lubricated.

Drawings

FIG. 1 is a schematic flow chart of a dialog method in an embodiment of the present application;

FIG. 2 is a schematic structural diagram of a session server in an embodiment of the present application;

FIG. 3 is a flow chart illustrating a context management method according to an embodiment of the present disclosure;

FIG. 4 is a schematic structural diagram of a conversation service interaction system in an embodiment of the present application;

FIG. 5 is a schematic structural diagram of a dialogue device according to an embodiment of the present application;

FIG. 6 is a schematic structural diagram of a server according to an embodiment of the present application;

fig. 7 is a schematic structural diagram of a computer-readable storage medium according to an embodiment of the present application.

Detailed Description

In order to make the purpose, technical solution and effect of the present application clearer and clearer, the present application is further described in detail below with reference to the accompanying drawings and examples.

The application provides an intelligent conversation service platform, which integrates various conversation services, can provide the conversation services according to different modes and meets different conversation requirements. In the scheme provided by the application, the conversation service is divided into task conversation service and non-task conversation service.

Task dialog service (TaskBot) refers to a user who, with a clear purpose, wishes to obtain information or services that satisfy certain constraints, such as: ordering a meal, ordering a ticket, looking up music, a movie or a certain commodity, etc. The task dialogue service can carry out semantic analysis on the input dialogue sentences to obtain semantic information of the dialogue sentences; the semantic information at least comprises intention information, so that the dialogue intention of the user can be known, and corresponding task operation is executed according to the dialogue intention so as to provide a result meeting the dialogue intention.

The non-task dialog service may be a service that may give answers or dialog intentions ambiguous directly from questions, smart question answering, chatting, and the like. For example, the intelligent question-answering service (QABot) is a service based on natural language understanding, and is a process of information retrieval focusing on a question-answer, i.e., capable of giving an accurate answer directly according to a question of a user. In the intelligent question-answering service, a knowledge base is required to be prepared in advance, the knowledge base can comprise one or more fields, and when a user asks a question, semantically matched answers can be found from the knowledge base according to sentences asked by the user. ChatBot (ChatBot) is a service that has no explicit purpose, or a rather ambiguous purpose. The function is not to provide answers meeting the user intention quickly, so that the user leaves after obtaining information; but to occupy the user time as much as possible, to prolong the time of chatting and accompanying with the user as much as possible, or to make the user use again as much as possible. Different chat sentences can be stored in the chatting service in advance, and different chat answers are given to the sentence contents according to the user.

According to the task conversation service, the data of the non-task conversation service is embedded into the task conversation service as an intention, so that the task conversation service can recognize the non-task conversation service as an intention, and a result corresponding to the intention is provided. For example, the question-answering service and the chatting service can be embedded into the task dialogue service as two intents, so that the task dialogue service can take the intelligent question-answering service and the chatting service as the intents of the user, and then select the result corresponding to the intents, or call the corresponding dialogue service to give the service. Through the method, different conversation intentions can be effectively distinguished, the corresponding conversation service is called to respond to the conversation sentences, the service is more accurately provided, more accurate answers are given, the intelligence of the service platform is further enhanced, and man-machine conversation is lubricated. The following description will take the non-task dialogue service including the chatting service and the intelligent question-answering service as an example.

Referring to fig. 1, fig. 1 is a schematic flow chart illustrating a dialog method according to an embodiment of the present disclosure. In this embodiment, the dialog server is integrated with at least a task-based dialog service and a non-task-based dialog service, and is capable of providing a plurality of dialog services, and the dialog method includes:

s110: the dialogue server receives the dialogue statement.

The dialog sentence may be information processed by the user side device or information that is not processed by the user side device. The dialogue sentence may be text information or voice information.

S120: semantic information of the dialog statement is acquired at least by using the task dialog service.

The task dialogue service can perform semantic analysis on the spoken sentence to obtain semantic information, the semantic information obtained by the task dialogue service is first-class semantic information, and the first-class semantic information at least comprises intention information and is used for expressing the intention of the dialogue. The semantic information may also include word slot information that is needed to perform the intent.

S130: and acquiring the dialogue intention of the dialogue statement in the first type of semantic information.

Among them, the dialog intention can be divided into a task intention and a service intention. The task intention comprises a task identifier, and the task identifier is used for identifying that a result corresponding to the task intention needs to be provided by a task dialogue service response dialogue statement. For example, the method can be used for ordering dishes, booking tickets, searching music, inquiring weather and the like, and the statements contain instruction semantics so as to drive the task execution operation.

The service intent includes a dialog service identification identifying a result to be provided with the corresponding service intent in response to the dialog statement using a non-task dialog service corresponding to the dialog service identification. If the conversation service identification display is 'chatting', the fact that the chatting conversation service needs to be called to respond to the conversation statement is shown, and a service result is provided; or the result output by the chat session service needs to be provided to the user as the service result.

S140: and calling the dialogue service corresponding to the dialogue intention to respond to the dialogue statement and provide a service result.

In the implementation mode, other non-task conversation services are fused into the task conversation services as intents, different conversation intents can be effectively distinguished, corresponding conversation services are called to respond to conversation sentences, services are provided more accurately, more accurate answers are given, the intelligence of a service platform is further enhanced, and man-machine conversation is lubricated.

Referring to fig. 2, fig. 2 is a schematic structural diagram of a session server according to an embodiment of the present application. In this embodiment, the plurality of session services integrated by the session server may be respectively located in different processing units, and the different processing units may be different processing units in the same server or may be separated in different servers, that is, the session server may be an independent server or a server cluster integrated by a plurality of servers.

In one embodiment, a distributor may be provided in the conversation server, and the distributor may distribute the received conversation statement to different conversation service units.

The distributor can have certain simple judgment capability, can distinguish obvious conversation intentions, and can directly distribute conversation sentences to corresponding conversation service units when judging that the conversation intentions obviously correspond to a certain conversation service. Or when the distributor judges that the conversation intention obviously does not correspond to a certain conversation service, the distributor does not need to distribute the conversation sentence to the corresponding conversation service unit.

When the distributor cannot definitely judge the dialogue intention, the dialogue sentence can be distributed to a plurality of dialogue service units. The distributed multiple dialogue service units at least comprise task dialogue services, so that semantic analysis is carried out by using the task dialogue services, dialogue intents are obtained, and then the dialogue intents are used for judging which dialogue service is required to respond to dialogue sentences. If the dialogue statement can be distributed to the task dialogue service unit and the possible corresponding dialogue service; or distributing the conversation statements to the task conversation service units and conversation services other than the conversation service for which no correspondence is possible.

In this embodiment, by setting up the distributor, simple and rudimentary judgment can be made, and the stress on the session service processing unit can be reduced without calling all session services to participate in the response each time. However, since the distributor only performs the preliminary recognition analysis of the dialogue sentences and has a weak judgment capability, the distributor cannot be completely relied on to perform the differentiated distribution. In other embodiments, the distributor may not be set up, and the conversation statements may be directly sent to all conversation services.

In another embodiment, a decision unit may be provided in the session server, and the decision unit may collect processing results from a plurality of session services, and then comprehensively consider the results to decide which session service to select for providing the processing results.

When the dialogue service returns the processing result, the confidence level can be returned at the same time, the confidence level is used for identifying the confidence level of the dialogue service on the accuracy of the processing result obtained by the dialogue service, and the decision unit can make a decision by referring to the confidence level. The confidence level may be expressed in different ways for different session services.

For example, the result returned by the task conversation service may include conversation intention information, word slot information, and a probability value of the current judgment, which is confidence level, and may indicate how high the probability of the conversation is the conversation intention. The result returned by the intelligent question-answering service can comprise an answer sentence and a matching value of the answer sentence and a dialogue sentence, wherein the matching value is self-credibility, and the probability of the answer being correct can be described. For example, the answer of the corresponding dialogue statement matched with high precision may be provided, three most relevant questions with medium precision matched with the dialogue statement may be provided for the user to click for query, or the answer of the corresponding dialogue statement not queried with low precision matched may be provided. The result returned by the chatting service may be a chatting answer and confidence level matching the dialogue service.

The decision unit can perform decision selection according to the priority, such as the intentions of high-precision matched intelligent question-answer > high confidence (more than 0.5), the intentions of high confidence (0.2-0.5) in word slot > matched intelligent question-answer > and word slot > chatty chat answers.

The decision unit can combine the dialog intention output by the task dialog service during decision making, and if the dialog intention is the service intention, the decision unit can appropriately improve the priority of the processing result of the dialog service corresponding to the service intention. The service result output by the non-task conversation service corresponding to the service intention can be directly returned to the user side as the final service result.

In this embodiment, by setting the decision unit, the processing result can be judged after the spoken sentence is processed, in which case the judgment and selection accuracy is much higher, and the accuracy of the service result is improved.

With the adoption of the embodiment, the conversation server provided by the application can provide various services and improve the service performance by integrating various conversation services. The three-level distinguishing rule of the embedding of the service intention, the distributor and the decision unit is set, so that different conversation intents can be distinguished accurately, the service can be provided more accurately, more accurate answers can be given, the intelligence of the service platform can be further enhanced, and man-machine conversation can be lubricated.

When the man-machine conversation is carried out, if the user requirement is complex, the user requirement may need to be divided into a plurality of rounds of statements, and the requirement intention of the user is continuously modified or improved in the conversation process. In addition, the machine may also help the user find a satisfactory result by asking, clarifying or confirming when the user stated a need not be clear enough.

In order to realize multiple rounds of interaction, the dialogue server provided by the application is also provided with a context management unit, and the context management unit can be used for managing context information. The corresponding context management logic may be set according to different dialog service types. The context management unit is described in the present application by taking the context management unit applied to the task conversation service as an example.

For the purpose of the following description, some noun concepts used when introducing the execution of the method by the context management unit will be explained.

Intent, i.e., classification of the current task, such as booking tickets, looking up weather, may be an intent.

The word slot belongs to an idea under intention, for example, two word slots are arranged in the query weather, one is City, the other is Time, for example, the weather of { City: Shenzhen } { Time: tomorrow } is consulted by the assistant me, wherein the word slot enclosed by { } is the word slot, the former word slot is the City, and the latter word slot is the Time.

The necessary word slot is one type of word slot, namely, the word slot information which is necessary to be filled by the user in the task of completing the current intention is still taken as an example of the intention of checking weather, and the City and the Time are two word slots, and the City can be marked as the necessary word slot, because if the user does not emphasize the specific Time of the query, the weather of the current Time can be inquired by the user by default. Namely: help me to check the weather of (Ctiy: Shenzhen), the user obtains the weather information of today.

The inquiry is an acquisition means for the system to acquire the missing of the necessary word slot of the user. Still taking weather-finding as an example, if we mark City as a necessary word slot, then the user looks up weather when stating the following query? The system asks the user in return, "ask you for the weather of which city you want to look up? ", this question is the question.

The round of hunting, that is, the round of hunting again after the user has not responded to the system after the user is hunted for by the current system.

Input and output context, this concept being derived on a challenge-chasing basis. If the user gives a continuous answer after the system has been consulted, and the user continues to answer "Shenzhen", if the context information is not considered, the system cannot judge what intention the current user expresses "Shenzhen", but can recognize that Shenzhen is word slot information. Therefore, an input and output context is designed for a semantic recognition algorithm without considering context information, and the context with the expression similar to 'Shenzhen' is named as 'inquiring weather lack city'. Namely, the 'checking weather lack city' is used as the output context of the 'checking weather', and the 'checking weather' is the current input context of the 'checking weather lack city'.

The concept of main intention and sub intention is similar to the input and output context, if the weather is the main intention, the supplementary city is the sub intention. The sub-intents here must exist in adherence to the main intention, and if the main intention is not activated, the sub-intents must not be activated.

The context survival turn, that is, the input and output context mentioned above, exists in the context survival period, for example, after the user visits the weather system to chase the city, the system does not respond to the demand of the system, the user says "shenzhen" after interacting 5 times, if the set context survival turn is 3 turns, the "shenzhen" expressed by the user at this time is not recognized as child intention.

The single-round semantic recognition algorithm is used for performing semantic analysis on the input of the current user only, recognizing intention information and word slot information input by the current user, and completely not considering context information. The context management module described herein is taught based on a single round of semantic recognition algorithms.

Referring to fig. 3, fig. 3 is a flowchart illustrating a context management method according to an embodiment of the present disclosure. In this embodiment, the context management method includes:

semantic information is received. The semantic information includes intent information and/or word slot information.

And judging whether the historical previous round of intention is met.

The meaning of the intention satisfaction means that all detail information required by the user to execute the current intention is provided, and the user does not need to be asked to obtain the details. If the intention is satisfied, the weather intention is checked to indicate that the required place and time are available.

The purpose of determining whether the previous round of intention is satisfied is to determine whether the dialog is related to the previous round.

If the previous round of intention is not satisfied, whether the conversation intention is a new intention or chatting can be judged firstly. If not, it is considered as a continuation of the previous round of intent. And continuously judging whether the necessary slot positions meeting the previous round of intentions are successfully extracted, if not, sending a clarification word slot question, outputting the clarification question, and continuously asking the user. If necessary slot position is successfully extracted, judging whether an output context exists, if not, only saving the extracted slot position, and if so, updating the slot position corresponding to the context. And adding the identified word slot information into the context intention and the historical information of the slot position.

And if the last round of intention is met or a new intention is judged, continuously judging whether the main intention is identified, if so, updating the output context corresponding to the current main intention, and storing the extracted slot position. If the main intention is not recognized, but the sub intention information is recognized, whether the name of the corresponding input context of the context name corresponding to the current sub intention is activated or not is judged, if the input context is activated and in the survival period of the effective context, the sub intention is the effective intention, and whether the corresponding context exists or not is continuously judged. And adding the identified word slot information into the context intention and the historical information of the slot position.

And finally, judging whether all the necessary slots are successfully extracted, if so, executing operation to obtain a task result, and if not, continuously asking for the task result to obtain the necessary information of the complete execution intention.

If the task intention is not obtained after multiple times of pursuits, the intention can be identified as the chatting, namely the intention can be adjusted to be the preset service intention, and the preset service intention is the chatting and the chatting service is called to provide the service. By the method, the intelligence of the robot can be enhanced, the man-machine conversation is lubricated, and the conversation interest is increased.

The context management method described above is only an example, and other logical management methods may be designed as needed, and are not limited herein. If an algorithm with semantic recognition of context can be designed and utilized, the input and output context does not need to be set because the algorithm is combined with context information. The context semantic recognition algorithm is that context information of a user is added into the algorithm, and relevant information input by a historical user is considered when the current user is recognized, so that when a similar user inputs Shenzhen, the context semantic recognition algorithm can still recognize that the intention input by the current user is to check weather, and the word slot City is Shenzhen.

By implementing the embodiment, the context management unit can record the context information and provide a plurality of rounds of conversation services.

With continuing reference to fig. 2 and 3, the dialog method provided by the present application includes the following steps:

the dialogue sentences received by the dialogue server firstly pass through the context management unit, the context management unit records the specific sources of the current dialogue sentences, such as Session ID information, App ID information, position information reported by the terminal and the like, and forwards the dialogue sentences to the distributor after recording the information.

After receiving the dialog statement, the distributor may distribute the dialog statement to the dialog service unit according to a preset rule, and for the specific distribution rule, reference is made to the description of the above embodiment, which is not described herein again.

And the service unit receiving the conversation statement processes the conversation statement and returns a processing result to the decision unit.

And the decision unit selects a determination processing result according to a preset decision strategy and returns the determination result to the context management module. For a specific decision strategy, please refer to the description of the above embodiments, which is not repeated herein.

The context management unit processes the semantic information after receiving the processing result (semantic information). Because the context management module provided by the application corresponds to the task dialogue service, when the context management unit identifies that the dialogue intention is non-task intention such as chatting or intelligent question and answer, the corresponding non-task dialogue service is called, and the non-task dialogue service is switched to continue to provide the service. In another embodiment, context management logic corresponding to different service intentions may be provided, and when it is determined that the service intention is a service intention, the corresponding context management unit may perform processing.

When the context management unit receives the task intention, whether the intention is satisfied can be judged according to the above management logic, and when the intention is satisfied, the execution unit is switched to, specifically, the intention operation is executed, and an intention result is obtained.

The context management logic above may be adapted to general task intents, providing general task dialogue services such as weather, music, etc. If the task is the specific task conversation service, a specific service inquiry flow or a context management module can be set, so that after the context management unit corresponding to the task receives the task intention, whether the task is the specific task can be judged firstly, and if the task is the specific task, the corresponding service can be switched to operate to provide the service more accurately. The model built by customer entry can be served as a specific service model. The specific task may be defined by a client, and the corresponding processing mode may be defined by the client.

In order to provide better and more comprehensive services, the dialogue service system can effectively perform services only depending on a large amount of contents, for example, if the intelligent sound asks for weather, the dialogue service system needs to inherit the api of the weather inquiry of a third party to inquire the weather condition of a specific city and report the weather condition to the intelligent sound. Therefore, a set of feasible intelligent conversation platforms needs to integrate a large number of third-party content services, such as weather, music, dining, transportation, cinema and the like, and certainly can build a part of content, such as the contents of jokes, stories, poems and the like.

Referring to fig. 4, fig. 4 is a schematic structural diagram of a dialog service interaction system according to an embodiment of the present application. In the embodiment, the dialogue server (dialogue service platform) can provide services for users and clients respectively, and can also call third-party services compatibly.

The third-party service is used for integrating the online resource information to reply different requests of a user side, such as weather query requests, and the weather conditions of a specific city need to be queried and sent to the conversation server by inheriting an Application Programming Interface (API) of the weather query of the third party, so that the server integrates a large number of third-party services, such as weather, music, catering, traffic, cinema and the like.

The user side equipment facing the user directly communicates with the user, namely receives dialogue sentences from the user and returns response results to the user. After receiving the dialog sentence, the user-side device may perform a preliminary processing on the audio signal of the dialog sentence, such as noise reduction or silence removal of the beginning and the end, and then convert the audio signal into corresponding text information and send the text information to the dialog server, for example, may convert the audio into text information by using a speech recognition technology (ASR). Or directly sending the audio signal to the conversation server to be processed by the conversation server. Similarly, the processed response result may be directly received, or the response result may be further processed. The user side equipment can be a vehicle-mounted intelligent sound box, a mobile phone and the like.

Wherein, the customer service system facing the customer side can provide the input service for the customer. The client side can input question and answer linguistic data on the platform and can also input intention linguistic data, intention expression, word slot information, dictionary content and the like. The platform carries out corresponding client model training according to the corpus information input by the client, and the training can be a task dialogue model, an intelligent question-answering model or both. After the model training is completed, the service can be provided for the client, and the user related inquiry of the client is solved.

The system can multiplex some intentions built in the system while inputting, for example, the system is built with an intention of checking weather, and an intelligent service platform built by a client of a third party also needs the intelligent service of checking weather, so that the intention service built in the system can be completely multiplexed. Even the platform enables those customers to contribute to what they believe is doing to the platform, and other customer merchants can use the intent by paying a corresponding fee when they want to use it.

The conversation service platform provided by the application integrates various conversation service functions, can reasonably forward the request of the user, and can organically and efficiently provide services for the user. The intelligent dialogue platform can provide services of intelligent customer service, intelligent sound equipment, a vehicle-mounted AI robot and the like, and can provide services for third-party customers such as a coffee robot, a customer service robot of a bank website, an airport navigation robot and the like by adding the outward-opened corpus entry platform.

Referring to fig. 5, fig. 5 is a schematic structural diagram of a dialog device according to an embodiment of the present disclosure. The dialog apparatus includes a receiving module 210, a first obtaining module 220, a second obtaining module 230, and a processing module 240.

The receiving module 210 is configured to receive a dialog statement; the first obtaining module 220 is configured to obtain semantic information of a dialog statement by using at least a task dialog service, where the semantic information obtained by the task dialog service is first-class semantic information, and the first-class semantic information at least includes intention information; the second obtaining module 230 is configured to obtain a dialog intention of a dialog statement in the first type of semantic information; the processing module 240 is configured to invoke a dialog service corresponding to the dialog intention to respond to the dialog statement and provide a service result; the conversation intention comprises a task intention and a service intention, the task intention comprises a task identifier, the task identifier is used for identifying a service result which needs to provide a corresponding task intention by using a task conversation service response conversation statement, the service intention comprises a conversation service identifier, and the conversation service identifier is used for identifying a service result which needs to provide a corresponding service intention by using a non-task conversation service response conversation statement of the corresponding conversation service identifier. The dialogue device can effectively distinguish different dialogue intentions by taking other non-task dialogue services as an intention to be fused into the task dialogue services, calls corresponding dialogue service response dialogue sentences, provides services more accurately, gives more accurate answers, further enhances the intelligence of a service platform, and lubricates man-machine dialogue.

Referring to fig. 6, fig. 6 is a schematic structural diagram of a server according to an embodiment of the present disclosure. In this embodiment, the server 50 includes a processor 51.

The processor 51 may also be referred to as a CPU (Central Processing Unit). The processor 51 may be an integrated circuit chip having signal processing capabilities. The processor 51 may also be a general purpose processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components. A general purpose processor may be a microprocessor or the processor 51 may be any conventional processor or the like.

The server 50 may further include a memory (not shown) for storing instructions and data required for the processor 51 to operate.

The processor 51 is configured to execute instructions to implement the methods provided by any of the embodiments of the dialog method of the present application and any non-conflicting combinations thereof.

The server may be an independent server or a server cluster composed of a plurality of servers.

Referring to fig. 7, fig. 7 is a schematic structural diagram of a computer-readable storage medium according to an embodiment of the present application. The computer readable storage medium 60 of the embodiments of the present application stores instructions/program data 61 that when executed enable the methods provided by any of the embodiments of the dialog method of the present application, as well as any non-conflicting combinations. The instructions 61 may form a program file stored in the storage medium 60 in the form of a software product, so as to enable a computer device (which may be a personal computer, a server, or a network device) or a processor (processor) to execute all or part of the steps of the methods according to the embodiments of the present application. And the aforementioned storage medium 60 includes: various media capable of storing program codes, such as a usb disk, a mobile hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, or terminal devices, such as a computer, a server, a mobile phone, and a tablet.

In the several embodiments provided in the present application, it should be understood that the disclosed system, apparatus and method may be implemented in other manners. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the modules or units is only one logical division, and other divisions may be realized in practice, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.

The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be substantially implemented or contributed by the prior art, or all or part of the technical solution may be embodied in a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, a network device, or the like) or a processor (processor) to execute all or part of the steps of the method according to the embodiments of the present application.

The above embodiments are merely examples and are not intended to limit the scope of the present disclosure, and all modifications, equivalents, and flow charts using the contents of the specification and drawings of the present disclosure or those directly or indirectly applied to other related technical fields are intended to be included in the scope of the present disclosure.

Claims

1. A dialog method, characterized in that a dialog server comprises at least a task dialog service and a non-task dialog service, the dialog method comprising:

the conversation server receives conversation sentences;

semantic information of the dialogue sentences is at least obtained by the task dialogue service, the semantic information obtained by the task dialogue service is first type semantic information, and the first type semantic information at least comprises intention information;

acquiring a dialogue intention of the dialogue statement in the first type of semantic information;

calling a dialogue service corresponding to the dialogue intention to respond to the dialogue statement and providing a service result;

the dialog intentions comprise task intentions and service intentions, the task intentions comprise task identifiers, the task identifiers are used for identifying service results corresponding to the task intentions and needing to be provided by responding to the dialog sentences through the task dialog services, the service intentions comprise dialog service identifiers, and the dialog service identifiers are used for identifying service results corresponding to the service intentions and needing to be provided by responding to the dialog sentences through non-task dialog services corresponding to the dialog service identifiers.

2. The dialog method of claim 1,

after receiving the dialog statement, the dialog server further comprises:

distributing the conversation statements to the task conversation service and the non-task conversation service;

respectively utilizing the task dialogue service and the non-task dialogue service to obtain semantic information of the dialogue statement to obtain a plurality of semantic information, wherein the semantic information comprises the first type of semantic information and the second type of semantic information; the second semantic information is semantic information obtained by the non-task dialogue service, and the second semantic information at least comprises a service result;

the calling of the dialogue service corresponding to the dialogue intention responds to the dialogue statement, and the providing of the service result comprises:

responding to the conversation intention as the task intention, and further judging whether necessary slot position information required by executing the task intention is met;

and responding to the conversation intention as the service intention, and returning a service result output by the non-task conversation service corresponding to the service intention as a final service result to the user side.

3. The dialog method of claim 2,

the determining whether the necessary slot position information required for executing the task intent satisfies includes:

in response to the necessary slot location information required to perform the task intent not being satisfied; returning a question-following statement for clarifying the necessary slot position information to the user side;

and executing the task intention acquisition service result in response to the necessary slot position information required for executing the task intention being met.

4. The dialog method of claim 3,

after receiving the dialog statement, the dialog server further comprises:

judging whether the current conversation sentence is an associated sentence associated with the previous pair of speaking sentences;

responding to the current conversation sentence as an associated sentence associated with the last pair of the conversation sentences, only distributing the current conversation sentence to the task conversation service, and acquiring the first type of semantic information, wherein the first type of semantic information also comprises slot position information;

and continuously judging whether necessary slot position information required by executing the task intention is met.

5. The dialog method of claim 3,

and responding to the question-chasing turn number being larger than a preset threshold value, adjusting the conversation intention to be a preset service intention, wherein the service intention corresponding to the preset service intention is a chatting service.

6. The dialog method of claim 1,

before the dialog server receives the dialog statement, the method further comprises the following steps:

receiving corpus information, wherein the corpus information at least comprises intention corpora;

and training a service model by using the corpus information to obtain a task conversation service model and/or a non-task conversation service model.

7. The dialog method of claim 1,

the non-task dialogue service includes one or more of a smart question-answering service and a chat service.

8. A dialogue apparatus, comprising:

the receiving module is used for receiving the dialogue sentences;

the first acquisition module is used for acquiring semantic information of the conversation statement at least by using the task conversation service, wherein the semantic information acquired by the task conversation service is first-class semantic information, and the first-class semantic information at least comprises intention information;

the second acquisition module is used for acquiring the conversation intention of the conversation statement in the first type of semantic information;

the processing module is used for calling the dialogue service corresponding to the dialogue intention to respond to the dialogue statement and providing a service result;

9. A server, characterized in that the server comprises a processor for executing instructions to implement the dialog method according to any of claims 1-7.

10. A computer-readable storage medium for storing instructions/program data executable to implement a dialog method according to any one of claims 1-7.