Question and answer processing method and device

Abstract

The present invention discloses a question-answer processing method and device, which relates to the field of artificial intelligence technology. An embodiment of the method includes: obtaining a voice request; processing the voice request to obtain an intent classification corresponding to the voice request; performing an answer search operation corresponding to the intent classification on the voice request, and feeding back the search answer. In this way, voice requests with multiple intents and incomplete intents can be identified, and an active inquiry function can be provided, thereby improving the intelligence of the dialogue system.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a question-answering processing method and device. Background Art

[0002] The existing dialogue system includes a speech recognition module, a natural language understanding module, a dialogue management module, a language generation module and a speech generation module. The existing language recognition module can only recognize single-intent voice requests, and cannot recognize multi-intent or incomplete-intent voice requests. For example, the user's voice request is "I want to listen", and the answer of the existing dialogue system is "prepare to play "Friends", "play music for you" or "this song is not found", etc. For multi-intent voice requests, for example, the user's voice request is "play Harry Potter", the answer of the existing dialogue system is "prepare to play Harry Potter songs", "play music for you" or "this song is not found", etc.

[0003] For voice requests with multiple or incomplete intents, the existing dialogue system's processing method is relatively simple and crude, directly classifying multiple or incomplete intents into a single intent with a higher probability, and then returning the answer corresponding to the voice request according to the single intent. Although this process completes a round of dialogue, it does not fully reflect the initiative of the dialogue system, thus reducing the accuracy of the dialogue system's response. Summary of the invention

[0004] In view of this, an embodiment of the present invention provides a question-answering processing method and device, which can identify voice requests with multiple intentions and incomplete intentions, and provide an active inquiry function, thereby improving the intelligence of the dialogue system.

[0005] To achieve the above-mentioned purpose, according to the first aspect of an embodiment of the present invention, a question and answer processing method is provided, the method comprising: obtaining a voice request; processing the voice request to obtain an intent classification corresponding to the voice request; performing an answer search operation corresponding to the intent classification on the voice request, and feeding back the search answer.

[0006] Optionally, the processing of the voice request to obtain an intent classification corresponding to the voice request includes: performing semantic analysis on the voice request; using a model to identify the intent of the voice request; determining a voice request whose semantic analysis result is no scenario and whose intent identification result is unclear intent as an incomplete intent; and determining a voice request whose semantic analysis result is a scenario and whose intent identification result is a clear intent as a complete intent.

[0007] Optionally, the performing of an answer search operation corresponding to the intent classification on the voice request includes: if the intent classification is a complete intent, searching for an answer to the voice request in the scenario indicated by the semantic parsing result; if the intent classification is an incomplete intent, predicting multiple pre-selected scenarios corresponding to the voice request, and selecting a pre-selected scenario that meets a first preset condition from the multiple pre-selected scenarios as a candidate scenario; sending an inquiry request corresponding to the candidate scenario, and searching for an answer to the voice request in the candidate scenario indicated by the inquiry request result.

[0008] Optionally, searching for the answer to the voice request in the scenario indicated by the semantic parsing result includes: if there is an entity value in the semantic parsing result, searching for an answer corresponding to the entity value in the scenario indicated by the semantic parsing result, and selecting the answer to the voice request from the searched answers; if there is no entity value in the semantic parsing result, searching for an answer corresponding to the scenario indicated by the semantic parsing result, and selecting the answer to the voice request from the searched answers.

[0009] Optionally, if there is an entity value in the semantic parsing result, searching for an answer corresponding to the entity value in the scenario indicated by the semantic parsing result, including: if there is one entity value, searching for an answer corresponding to the entity value in the scenario indicated by the semantic parsing result; if there are multiple entity values, determining whether the multiple entity values ​​simultaneously satisfy the same semantic slot; if so, searching for answers for each entity value separately in the scenario indicated by the semantic parsing result, and selecting any one answer from the answers corresponding to each entity value; for multiple entity values, concatenating the selected multiple answers to obtain the search answer corresponding to the entity value; if not, searching for answers that simultaneously satisfy multiple entity values ​​in the scenario indicated by the semantic parsing result.

[0010] Optionally, the first preset condition refers to a pre-selected scenario that ranks in the top two in confidence and has a search answer.

[0011] To achieve the above-mentioned purpose, according to the second aspect of an embodiment of the present invention, a question and answer processing device is also provided, which includes: an acquisition module for acquiring a voice request; a processing module for processing the voice request to obtain an intent classification corresponding to the voice request; and a search module for performing an answer search operation corresponding to the intent classification on the voice request and feeding back the search answer.

[0012] Optionally, the processing module includes: a semantic parsing unit, used to perform semantic parsing on the voice request; an intent recognition unit, used to use a model to recognize the intent of the voice request; a determination unit, used to determine a voice request whose semantic parsing result is no scenario and whose intent recognition result is unclear intent as an incomplete intent; and determine a voice request whose semantic parsing result is a scenario and whose intent recognition result is a clear intent as a complete intent.

[0013] Optionally, the search module includes: a first search unit, used to search for the answer to the voice request in the scenario indicated by the semantic parsing result if the intent classification is a complete intent; a second search unit, used to predict multiple pre-selected scenarios corresponding to the voice request if the intent classification is an incomplete intent, select a pre-selected scenario that meets a first preset condition from the multiple pre-selected scenarios as a candidate scenario, send an inquiry request corresponding to the candidate scenario, and search for the answer to the voice request in the candidate scenario indicated by the inquiry request result.

[0014] Optionally, the first search unit includes: an entity value unit, which is used to search for an answer corresponding to the entity value in the scenario indicated by the semantic parsing result if there is an entity value in the semantic parsing result, and select the answer to the voice request from the searched answers; and a scenario unit, which is used to search for an answer corresponding to the scenario indicated by the semantic parsing result if there is no entity value in the semantic parsing result, and select the answer to the voice request from the searched answers.

[0015] To achieve the above-mentioned purpose, according to a third aspect of an embodiment of the present invention, a computer-readable medium is further provided, on which a computer program is stored, and when the program is executed by a processor, the question and answer processing method as described in the first aspect is implemented.

[0016] The embodiment of the present invention processes the acquired voice request to obtain the intent classification corresponding to the voice request, then performs the answer search operation corresponding to the intent classification on the voice request, and feeds back the search answer. Thus, it is possible to identify voice requests with multiple intents and incomplete intents, and provide the function of active inquiry, thereby improving the intelligence of the dialogue system.

[0017] The further effects of the above-mentioned non-conventional optional manner will be described below in conjunction with the specific implementation manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The accompanying drawings are used to better understand the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings, the same or corresponding reference numerals represent the same or corresponding parts.

[0019] Figure 1 A flowchart of a question-and-answer processing method according to an embodiment of the present invention;

[0020] Figure 2 It is a flow chart of a question-answering processing method for an incomplete intended voice request according to an embodiment of the present invention;

[0021] Figure 3 It is a flowchart of a question-answering processing method for a complete intent voice request according to an embodiment of the present invention;

[0022] Figure 4 is a schematic diagram of a question-answering processing device according to an embodiment of the present invention;

[0023] Figure 5 is an exemplary system architecture diagram to which the embodiments of the present invention can be applied;

[0024] Figure 6 It is a schematic diagram of the structure of a computer system of a terminal device or a server suitable for implementing an embodiment of the present invention. DETAILED DESCRIPTION

[0025] The following is a description of exemplary embodiments of the present invention in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0026] like Figure 1 FIG. 1 is a flowchart of a question-answering processing method according to an embodiment of the present invention, and the method at least includes the following operation flow:

[0027] S101, obtaining a voice request.

[0028] Exemplarily, the voice request is obtained through text input or voice input.

[0029] S102: Process the voice request to obtain an intent classification corresponding to the voice request.

[0030] Exemplarily, semantic parsing is performed on voice requests; the model is used to identify the intent of voice requests; voice requests with semantic parsing results of no scenario and intent identification results of unclear intent are determined as incomplete intent; and voice requests with semantic parsing results of scenarios and intent identification results of clear intent are determined as complete intent. When the following two situations occur in the processing results of voice requests, they are both abnormal situations. When encountering such abnormal situations, it is usually necessary to optimize the semantic parsing module or the intent identification module. These two situations are that the semantic parsing result is no scenario and the intent identification result is a clear intent, and the semantic parsing result is a scenario and the intent identification result is an unclear intent.

[0031] S103, performing an answer search operation corresponding to the intent classification on the voice request, and feeding back the search answer.

[0032] Exemplarily, the intent classification includes complete intent and incomplete intent. If the intent classification of the voice request is incomplete intent, multiple candidate answers corresponding to the voice request are searched, and a query request is sent for the candidate answers to determine the answer to the voice request.

[0033] If the intent classification of the voice request is a complete intent, then the answer to the voice request is searched in the scenario indicated by the semantic parsing result; further, if there is an entity value in the semantic parsing result, then the answer corresponding to the entity value is searched in the scenario indicated by the semantic parsing result, and the answer to the voice request is selected from the searched answers; if there is no entity value in the semantic parsing result, then the answer corresponding to the scenario indicated by the semantic parsing result is searched, and the answer to the voice request is selected from the searched answers.

[0034] Therefore, the embodiment of the present invention determines the intent classification of the voice request through the semantic analysis results and the intent recognition results, thereby performing an answer search operation corresponding to the intent classification on the voice request, thereby improving the intelligence of the dialogue system.

[0035] like Figure 2 As shown, it is a flowchart of a question-answering processing method for an incomplete intended voice request according to an embodiment of the present invention; the method at least includes the following operation flow:

[0036] S201, obtaining a voice request.

[0037] S202, performing semantic analysis and intent recognition on the voice request respectively, and classifying the intent corresponding to the voice request as an incomplete intent.

[0038] Exemplarily, the semantic analysis result of the voice request is no scenario, and the intent recognition result of the intent recognition model for the voice request is unclear intent; thus, it is determined that the intent corresponding to the voice request is classified as an incomplete intent.

[0039] S203, predicting multiple pre-selected scenes corresponding to the voice request.

[0040] S204: Select a pre-selected scene that meets a first preset condition from a plurality of pre-selected scenes as a candidate scene.

[0041] Exemplarily, the first preset condition refers to the pre-selected scenes that have the top two confidence levels and have search answers. When the semantic analysis module predicts multiple pre-selected scenes corresponding to the voice request, it outputs the confidence levels of the pre-selected scenes.

[0042] Here, the first preset condition is artificially set according to the actual application scenario.

[0043] S205: Send a query request corresponding to the candidate scenario.

[0044] S206: Search for an answer to the voice request in the candidate scenarios indicated by the inquiry request result.

[0045] Specifically, the voice request is "I want to listen", and the voice request is semantically parsed, and the result of the semantic parsing indicates that the voice request has no scene; the intention recognition model is used to recognize the intention of the voice request, and the result of the intention recognition indicates that the voice request is an unclear intention, thereby determining that the voice request is an incomplete intention. The semantic parsing module predicts multiple pre-selected scenes corresponding to the voice request, and outputs the confidence of each pre-selected scene. The pre-selected scenes with the top two confidences and search answers are selected from the multiple pre-selected scenes. For example, the multiple pre-selected scenes are sorted from large to small according to the confidence of the pre-selected scenes, which are music scenes, poetry scenes, audio book scenes, cross talk scenes, and opera scenes. The top two pre-selected scenes selected are music scenes and poetry scenes respectively; then determine whether the voice request has an answer in the music scene or the poetry scene. If there is a search answer only in the music scene, for example, the search answer is "friend", "same song" and "lover", the music scene is used as a candidate scene for the voice request. If there is a search answer in both the music scene and the poetry scene, for example, the search answer in the poetry scene is "quatrain". A query request is sent, the query request is "Do you want to listen to music or poetry?" If the user feedback content is "I want to listen to music", one is selected from the search answers corresponding to the music scene as the answer to the voice request, for example, "friends" is selected as the answer to "I want to listen" from the three search answers of "friends", "same song" and "lover".

[0046] Here, the search answer ranked first may be selected as the answer to the voice request based on the popularity of the search answers, or a search answer may be randomly selected as the answer to the voice request.

[0047] The embodiment of the present invention can identify incomplete intent and provide a proactive inquiry function for the incomplete intent, thereby achieving the initiative and intelligence of the dialogue system.

[0048] like Figure 3 As shown, it is a flowchart of a question-answering processing method for a complete intention voice request according to an embodiment of the present invention; the method at least includes the following operation flow:

[0049] S301, obtaining a voice request.

[0050] S302, perform semantic analysis and intent recognition on the voice request respectively, and classify the intent corresponding to the voice request as a complete intent; if there is one entity value in the semantic analysis result, execute S303 operation; if there are multiple entity values ​​in the semantic analysis result, execute S304 operation; if there is no entity value in the semantic analysis result, execute S307 operation.

[0051] S303, searching for an answer corresponding to the entity value in the scenario indicated by the semantic parsing result; then executing S308 operation.

[0052] Exemplarily, the search answer may be one or more.

[0053] S304, determine whether multiple entity values ​​simultaneously satisfy the same semantic slot; if so, execute S305 operation, if not, execute S306 operation.

[0054] S305, search for answers for each entity value separately in the scenario indicated by the semantic parsing result, and select any answer from the answers corresponding to each entity value; for multiple entity values, concatenate the selected multiple answers to obtain the search answers corresponding to the entity values; then execute S308 operation.

[0055] S306, searching for answers that simultaneously satisfy multiple entity values ​​in the scenario indicated by the semantic parsing result; then executing S308 operation.

[0056] S307, searching for an answer corresponding to the scenario indicated by the semantic parsing result; then executing S308.

[0057] S308, selecting an answer to the voice request from the search answers.

[0058] Exemplarily, if there is one search answer, the search answer is used as the answer to the voice request. If there are multiple search answers, a search answer that meets the preset conditions is selected from the multiple search answers as the answer to the voice request.

[0059] Specifically, the voice request is "I want to listen to the song "Friends", the semantic analysis result indicates that the scenario of the voice request is "song" and the entity value is "friends", and the intent recognition result is a clear intent, so the intent of the voice request is determined to be classified as a complete intent. Since there is an entity value in the voice request, the answer corresponding to "friends" is searched in the "song" scenario, and three search answers are obtained, such as "Friends" sung by Jay Chou, "Friends" sung by Na Ying, and "Friends" sung by Gao Feng. According to the popularity of "Friends" sung by different singers, the first-ranked one is selected as the answer to the voice request, or a search answer is randomly selected as the answer to the voice request.

[0060] The voice request is "I want to listen to music". The semantic analysis result indicates that the scenario of the voice request is "music" and there is no entity value. The intent recognition result is a clear intent, so the intent of the voice request is determined to be classified as a complete intent. Since there is no entity value for the voice request, the answer is searched in the "music" scenario, and three search answers are obtained, namely "friends", "same song" and "lover". The search answer with the highest playback popularity is selected from the three search answers as the answer to the voice request, for example, "friends" is selected as the answer to "I want to listen to music".

[0061] The voice request is "Help me check the weather in Beijing and Shanghai". The semantic analysis result indicates that the scenario of the voice request is "weather", and the entity values ​​are "Beijing" and "Shanghai" respectively. The intent recognition result is a clear intent, so the intent of the voice request is determined to be classified as a complete intent. Since there are two entity values ​​in the voice request, and the semantic slots corresponding to the two entity values ​​are both "weather location", the two entity values ​​satisfy the same semantic slot. In the "weather" scenario, the answers are searched for the entity values ​​"Beijing" and "Shanghai" respectively. The answers corresponding to the entity value "Beijing" are "sunny, 16-30℃" and "light rain, 25℃", and the answers corresponding to the entity value "Shanghai" are "thunderstorm, 21℃". Select any one answer from the answers corresponding to each entity value; for the two entity values, splice the two selected answers to obtain two groups of answers corresponding to the entity values ​​of the voice request, which are "Beijing is sunny today, 16-30℃; Shanghai is thunderstorm today, 21℃" and "Beijing is light rain tomorrow, 25℃; Shanghai is thunderstorm today, 21℃". Afterwards, a query request can be sent for the two sets of answers, for example, "Which day do you want to know the weather?" Based on the user's answer "I want to know today's weather," "Beijing is sunny today, 16-30℃; Shanghai is thundering today, 21℃" is used as the answer to the voice request and fed back to the user.

[0062] The voice request is "Help me check the weather in Beijing on February 1st". The semantic analysis result indicates that the scenario of the voice request is "weather", and the entity values ​​are "Beijing" and "February 1st". The intent recognition result is a clear intent, so the intent of the voice request is determined to be classified as a complete intent. The voice request has two entity values. The semantic slot corresponding to the entity value "Beijing" is "weather location", and the semantic slot corresponding to the entity value "February 1st" is "weather time". Since the two entity values ​​do not satisfy the same semantic slot. Therefore, in the "weather" scenario, the answer that satisfies both "weather location" and "weather time" is searched. Two answers corresponding to the entity values ​​are obtained, namely "Beijing on February 1st is sunny, 16-30℃" and "Beijing on February 1st is thunderstorm, 16-30℃". The answer with the highest confidence is selected from the two search answers as the answer to the voice request, for example, the selected answer is "Beijing on February 1st is sunny, 16-30℃".

[0063] Therefore, the embodiment of the present invention can recognize multi-intent voice requests, improve the functions of the dialogue system, and enhance the intelligence of the dialogue system.

[0064] It should be understood that in various embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0065] like Figure 4 As shown, it is a schematic diagram of a question and answer processing device according to an embodiment of the present invention; the device 400 includes: an acquisition module 401, used to acquire a voice request; a processing module 402, used to process the voice request to obtain an intent classification corresponding to the voice request; a search module 403, used to perform an answer search operation corresponding to the intent classification on the voice request, and feed back the search answer.

[0066] In an optional embodiment, the processing module 402 includes: a semantic parsing unit, used to perform semantic parsing on a voice request; an intent recognition unit, used to use a model to recognize the intent of a voice request; a determination unit, used to determine a voice request whose semantic parsing result is no scenario and whose intent recognition result is unclear intent as an incomplete intent; and determine a voice request whose semantic parsing result is a scenario and whose intent recognition result is a clear intent as a complete intent.

[0067] In an optional embodiment, the search module 403 includes: a first search unit, which is used to search for an answer to the voice request in a scenario indicated by a semantic parsing result if the intent classification is a complete intent; and a second search unit, which is used to predict multiple pre-selected scenarios corresponding to the voice request if the intent classification is an incomplete intent, select a pre-selected scenario that meets a first preset condition from the multiple pre-selected scenarios as a candidate scenario, send an inquiry request corresponding to the candidate scenario, and search for an answer to the voice request in the candidate scenario indicated by the inquiry request result.

[0068] In an optional embodiment, the first search unit includes: an entity value unit, which is used to search for an answer corresponding to the entity value in the scenario indicated by the semantic parsing result if there is an entity value in the semantic parsing result, and select the answer to the voice request from the search answer; a scenario unit, which is used to search for an answer corresponding to the scenario indicated by the semantic parsing result if there is no entity value in the semantic parsing result, and select the answer to the voice request from the search answer.

[0069] In an optional embodiment, the entity value unit includes: a first entity value unit, which is used to search for an answer corresponding to the entity value in the scenario indicated by the semantic parsing result if there is one entity value; a second entity value unit, which is used to determine whether the multiple entity values ​​simultaneously satisfy the same semantic slot if there are multiple entity values; if so, search for answers for each entity value separately in the scenario indicated by the semantic parsing result, and select any one answer from the answers corresponding to each entity value; for multiple entity values, concatenate the selected multiple answers to obtain the search answer corresponding to the entity value; if not, search for answers that simultaneously satisfy multiple entity values ​​in the scenario indicated by the semantic parsing result.

[0070] In an optional embodiment, the first preset condition refers to a pre-selected scenario that ranks in the top two in confidence and has a search answer.

[0071] The above device can execute the question-answer processing method provided in the embodiment of the present invention, and has the corresponding functional modules and beneficial effects of executing the question-answer processing method. For technical details not described in detail in this embodiment, please refer to the question-answer processing method provided in the embodiment of the present invention.

[0072] like Figure 5 , which is an exemplary system architecture diagram to which the embodiment of the present invention can be applied, the system architecture 500 may include terminal devices 501, 502, 503, a network 504 and a server 505. The network 504 is used to provide a medium for communication links between the terminal devices 501, 502, 503 and the server 505. The network 504 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0073] Users can use terminal devices 501, 502, 503 to interact with server 505 through network 504 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 501, 502, 503, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only as examples).

[0074] The terminal devices 501 , 502 , and 503 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, and the like.

[0075] The server 505 may be a server that provides various services, such as a background management server (only an example) that provides support for click events generated by users using the terminal devices 501, 502, and 503. The background management server may analyze and process the received click data, text content, and other data, and feed back the processing results (such as target push information, product information - only an example) to the terminal device.

[0076] It should be noted that the question and answer processing method provided in the embodiment of the present application is generally executed by the server 505 , and accordingly, the interpretation device is generally arranged in the server 505 .

[0077] It should be understood that Figure 5 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0078] Reference below Figure 6 , which shows a schematic diagram of the structure of a computer system of a terminal device or a server suitable for implementing the embodiment. Figure 6 The terminal device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0079] like Figure 6As shown, the computer system 600 includes a central processing unit (CPU) 601, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 602 or the program loaded from the storage part 608 to the random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the system 600 are also stored. The CPU 601, the ROM 602, and the RAM 603 are connected to each other through the bus 604. The input / output (I / O) interface 605 is also connected to the bus 604. The following components are connected to the I / O interface 605: an input part 606 including a keyboard, a mouse, etc.; an output part 607 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage part 608 including a hard disk, etc.; and a communication part 609 including a network interface card such as a LAN card, a modem, etc. The communication part 609 performs communication processing via a network such as the Internet. A driver 610 is also connected to the I / O interface 605 as needed. A removable medium 611, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is mounted on the drive 610 as needed, so that a computer program read therefrom is installed into the storage section 608 as needed.

[0080] In particular, according to the embodiments disclosed in the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present invention include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 609, and / or installed from the removable medium 611. When the computer program is executed by the central processing unit (CPU) 601, the above-mentioned functions defined in the system of the present invention are executed.

[0081] It should be noted that the computer-readable medium shown in the present invention can be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, system or device, or any combination of the above. More specific examples of computer-readable storage media can include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. In the present invention, a computer-readable signal medium can include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal can take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media that can send, propagate or transmit a program for use by or in conjunction with an instruction execution system, system or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0082] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of the code, and the above-mentioned module, program segment, or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a certain function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0083] The modules involved in the embodiments of the present invention may be implemented in software or hardware. The modules described may also be set in a processor. For example, they may be described as: a processor includes a sending module, an acquisition module, a determination module, and a first processing module. The names of these modules do not, in some cases, constitute a limitation on the unit itself. For example, the sending module may also be described as a "module for sending a picture acquisition request to the connected server."

[0084] As another aspect, the present invention further provides a computer-readable medium, which may be included in the device described in the above embodiment; or it may exist independently without being assembled into the device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by a device, the device includes: S101, obtaining a voice request. S102, processing the voice request to obtain an intent classification corresponding to the voice request. S103, performing an answer search operation corresponding to the intent classification on the voice request, and feeding back the search answer.

[0085] The system of the embodiment of the present invention can not only process single-intent dialogues, but also process incomplete intent and multi-intent dialogues, thereby improving the intelligence of the dialogue system.

[0086] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples, unless they are contradictory.

[0087] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0088] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A question-answering processing method, It is characterized in that include: Get voice request; Processing the voice request to obtain an intent classification corresponding to the voice request; Performing an answer search operation corresponding to the intent classification on the voice request, and feeding back a search answer; The processing of the voice request to obtain an intent classification corresponding to the voice request includes: Performing semantic analysis on the voice request; Using the model to perform intent recognition on the voice request; Determine the voice request with no scene in the semantic analysis result and unclear intent in the intent recognition result as incomplete intent; and determine the voice request with scene in the semantic analysis result and clear intent in the intent recognition result as complete intent; The performing an answer search operation corresponding to the intent classification on the voice request includes: If the intent classification is a complete intent, searching for an answer to the voice request in the scenario indicated by the semantic analysis result; The searching for an answer to the voice request in the scenario indicated by the semantic parsing result includes: If the semantic parsing result has an entity value, searching for an answer corresponding to the entity value in the scenario indicated by the semantic parsing result, and selecting an answer to the voice request from the searched answers; If the semantic parsing result has an entity value, searching for an answer corresponding to the entity value in the scenario indicated by the semantic parsing result includes: If there are multiple entity values, determine whether the multiple entity values ​​satisfy the same semantic slot at the same time; if so, search for answers for each entity value separately in the scenario indicated by the semantic parsing result, and select any one answer from the answers corresponding to each entity value; for multiple entity values, concatenate the selected multiple answers to obtain the search answer corresponding to the entity value; if not, search for answers that simultaneously satisfy multiple entity values ​​in the scenario indicated by the semantic parsing result.

2. The method according to claim 1, It is characterized in that The performing an answer search operation corresponding to the intent classification on the voice request includes: If the intent classification is an incomplete intent, multiple pre-selected scenes corresponding to the voice request are predicted, and a pre-selected scene that meets a first preset condition is selected from the multiple pre-selected scenes as a candidate scene; a query request corresponding to the candidate scene is sent, and an answer to the voice request is searched under the candidate scene indicated by the query request result.

3. The method according to claim 1, It is characterized in that The searching for an answer to the voice request in the scenario indicated by the semantic parsing result includes: If the semantic parsing result does not have an entity value, an answer corresponding to the scenario indicated by the semantic parsing result is searched, and an answer to the voice request is selected from the searched answers.

4. The method according to claim 1, It is characterized in that If the semantic parsing result has an entity value, searching for an answer corresponding to the entity value in the scenario indicated by the semantic parsing result includes: If an entity value exists, an answer corresponding to the entity value is searched in the scenario indicated by the semantic parsing result.

5. The method according to claim 2, It is characterized in that The first preset condition refers to the pre-selected scenarios that rank in the top two in confidence and have search answers.

6. A question-answering processing device, It is characterized in that include: An acquisition module, used to acquire voice requests; A processing module, used to process the voice request to obtain an intent classification corresponding to the voice request; A search module, configured to perform an answer search operation corresponding to the intent classification on the voice request and feed back a search answer; The processing module comprises: A semantic parsing unit, used for performing semantic parsing on the voice request; An intention recognition unit, used to perform intention recognition on the voice request using a model; A determination unit, configured to determine a voice request whose semantic analysis result is that there is no scene and whose intent recognition result is that there is an unclear intent as an incomplete intent; and to determine a voice request whose semantic analysis result is that there is a scene and whose intent recognition result is that there is a clear intent as a complete intent; The search module includes: A first search unit, configured to search for an answer to the voice request in a scenario indicated by the semantic analysis result if the intent classification is a complete intent; The first search unit comprises: An entity value unit, configured to search for an answer corresponding to the entity value in the scenario indicated by the semantic parsing result if the semantic parsing result has an entity value, and select an answer to the voice request from the searched answers; The entity value unit comprises: The second entity value unit is used to determine whether the multiple entity values ​​simultaneously satisfy the same semantic slot if there are multiple entity values; if so, search for answers for each entity value separately in the scenario indicated by the semantic parsing result, and select any one answer from the answers corresponding to each entity value; for the multiple entity values, concatenate the selected multiple answers to obtain the search answer corresponding to the entity value; if not, search for answers that simultaneously satisfy the multiple entity values ​​in the scenario indicated by the semantic parsing result.

7. The device according to claim 6, It is characterized in that The search module includes: The second search unit is used to predict multiple pre-selected scenes corresponding to the voice request if the intent classification is an incomplete intent, select a pre-selected scene that meets a first preset condition from the multiple pre-selected scenes as a candidate scene, send an inquiry request corresponding to the candidate scene, and search for an answer to the voice request under the candidate scene indicated by the inquiry request result.

8. The device according to claim 7, It is characterized in that The first search unit also includes: The scene unit is used to search for an answer corresponding to the scene indicated by the semantic analysis result if there is no entity value in the semantic analysis result, and select an answer to the voice request from the searched answers.

Images