Semantic arbitration method, device, equipment and storage medium

Abstract

The present application discloses a semantic arbitration method, apparatus, device, and storage medium, wherein the method comprises: obtaining visible-to-speak semantics corresponding to a user's voice from an offline recognition engine; obtaining online common semantics corresponding to the user's voice from an online recognition engine; outputting the visible-to-speak semantics when the visible-to-speak semantics exactly matches a visible-to-speak hotword; and outputting the online common semantics or the visible-to-speak semantics according to a preset priority when the visible-to-speak semantics fuzzily matches a visible-to-speak hotword. The technical solution provided by the present application can selectively output online common semantics or visible-to-speak semantics according to a preset priority when the visible-to-speak semantics fuzzily matches a visible-to-speak hotword, thereby avoiding the negative user experience caused by the visible-to-speak semantics mistakenly triggering related instructions.

Description

Technical Field

[0001] The present application belongs to the field of speech processing technology, and in particular relates to a semantic arbitration method, apparatus, device and storage medium. Background Art

[0002] Currently, many voice clients have the see-and-speak function. With the see-and-speak function, users can execute the instructions of the corresponding controls on the voice client interface through voice interaction.

[0003] In the related art, the semantic arbitration strategy of the voice client is to give priority to the visible and audible semantics over other semantics, but such a design often cannot meet the actual needs of the project. For example, in the music playback interface, the application uploaded the visible and audible hot word: play Qilixiang. When the user's voice includes "play", "Qilixiang" or "play Qilixiang", the NLU (Natural Language Understanding) module outputs the visible and audible semantics, and the system will execute the corresponding control for playing Qilixiang in the music playback interface. When the user's voice includes "play", the product design usually expects other semantics to be output and other playback actions to be executed. It can be seen that the current semantic arbitration strategy will mistakenly trigger related instructions due to the output of visible and audible semantics, thereby bringing a bad experience to users. Summary of the Invention

[0004] The embodiments of the present application provide a semantic arbitration method, apparatus, device, and storage medium, thereby at least to some extent resolving the problem that current semantic arbitration strategies may erroneously trigger related instructions due to outputting visible, audible semantics, thereby causing a poor user experience.

[0005] Other features and advantages of the present application will become apparent from the following detailed description, or may be learned in part by practice of the present application.

[0006] According to a first aspect of an embodiment of the present application, a semantic arbitration method is provided, including:

[0007] Obtain the visible and audible semantics corresponding to the user's voice from the offline recognition engine;

[0008] Obtaining online common semantics corresponding to the user speech from an online recognition engine;

[0009] When the visible-can-be-speakable semantics and the visible-can-be-speakable hotwords are exactly matched, outputting the visible-can-be-speakable semantics;

[0010] In the case where the visible-and-speakable semantics and the visible-and-speakable hot words are fuzzy matched, the online common semantics or the visible-and-speakable semantics are output according to a preset priority.

[0011] In some embodiments of the present application, based on the aforementioned solution, the offline recognition engine includes an offline speech recognition module and an offline natural language understanding module, the online recognition engine includes an online speech recognition module and an online natural language understanding module, and the method further includes:

[0012] The visible and utterable hot words are sent to the offline speech recognition module, the offline natural language understanding module and the online speech recognition module respectively, so that the offline speech recognition module outputs an offline speech recognition result according to the user voice and the visible and utterable hot words, the offline natural language understanding module outputs the visible and utterable semantics or offline common semantics according to the user voice and the visible and utterable hot words, the online speech recognition module outputs an online speech recognition result according to the user voice and the visible and utterable hot words, and the online natural language understanding module outputs the online common semantics according to the user voice.

[0013] In some embodiments of the present application, based on the aforementioned solution, the semantic arbitration method further includes:

[0014] If the visible-to-speak semantics are not obtained, obtaining offline common semantics corresponding to the user speech from the offline natural language understanding module;

[0015] In the case that the offline common semantics is an invalid semantics, the online speech recognition result is input into the offline natural language understanding module, so that the offline natural language understanding module generates visible and audible semantics or new offline common semantics according to the online speech recognition result.

[0016] In some embodiments of the present application, based on the aforementioned solution, the semantic arbitration method further includes:

[0017] In the case where the online common semantics is a valid semantics, outputting the online common semantics;

[0018] In a case where the online general semantics is not a valid semantics and the offline general semantics is a valid semantics, the offline general semantics is output.

[0019] In some embodiments of the present application, based on the aforementioned solution, the semantic arbitration method further includes:

[0020] Determining a score of the offline common semantics according to the user speech;

[0021] When the score of the offline common semantics is less than a preset score, it is determined that the offline common semantics is an invalid semantics.

[0022] In some embodiments of the present application, based on the aforementioned solution, the semantic arbitration method further includes:

[0023] In a case where the word slot corresponding to the visible-to-speak semantics is exactly the same as the word slot corresponding to the visible-to-speak hot word, determining that the visible-to-speak semantics is an exact match to the visible-to-speak hot word;

[0024] When the word slot corresponding to the visible-and-speakable semantics is not completely the same as the word slot corresponding to the visible-and-speakable hot word, it is determined that the visible-and-speakable semantics and the visible-and-speakable hot word are fuzzy matched.

[0025] In some embodiments of the present application, based on the aforementioned solution, outputting the online common semantics or the visible-and-speakable semantics according to a preset priority includes:

[0026] In a case where the preset priority is that the priority of the common semantics is higher than the fuzzy-matched visible-that-is-speakable semantics, outputting the online common semantics;

[0027] In a case where the preset priority is that the priority of the common semantics is lower than the fuzzy-matched visible-and-speakable semantics, the visible-and-speakable semantics are output.

[0028] According to a second aspect of an embodiment of the present application, a semantic arbitration device is provided, including: a first semantic acquisition module, used to obtain visible and utterable semantics corresponding to user voice from an offline recognition engine; a second semantic acquisition module, used to obtain online common semantics corresponding to the user voice from an online recognition engine; a semantic arbitration module, used to output the visible and utterable semantics when the visible and utterable semantics and the visible and utterable hot words are exactly matched; the semantic arbitration module is also used to output the online common semantics or the visible and utterable semantics according to a preset priority when the visible and utterable semantics and the visible and utterable hot words are fuzzy matched.

[0029] In some embodiments of the present application, based on the aforementioned scheme, the first semantic acquisition module is further used to obtain offline common semantics corresponding to the user voice from the offline natural language understanding module when the visible and speakable semantics are not obtained; and when the offline common semantics are invalid semantics, the online speech recognition results are input into the offline natural language understanding module, so that the offline natural language understanding module generates visible and speakable semantics or new offline common semantics based on the online speech recognition results.

[0030] In some embodiments of the present application, based on the aforementioned scheme, the semantic arbitration module is further used to output the online common semantics when the online common semantics is a valid semantics; and to output the offline common semantics when the online common semantics is not a valid semantics and the offline common semantics is a valid semantics.

[0031] In some embodiments of the present application, based on the aforementioned scheme, the semantic arbitration module is further used to determine the score of the offline common semantics based on the user voice; when the score of the offline common semantics is less than a preset score, it is determined that the offline common semantics is an invalid semantics.

[0032] In some embodiments of the present application, based on the aforementioned scheme, the semantic arbitration module is further used to determine that the visible and speakable semantics and the visible and speakable hot words are exactly matched when the word slot corresponding to the visible and speakable semantics is exactly the same as the word slot corresponding to the visible and speakable hot words; and to determine that the visible and speakable semantics and the visible and speakable hot words are fuzzy matched when the word slot corresponding to the visible and speakable semantics is not exactly the same as the word slot corresponding to the visible and speakable hot words.

[0033] In some embodiments of the present application, based on the aforementioned scheme, the semantic arbitration module is also used to output the online common semantics when the preset priority is that the priority of common semantics is higher than the visible and speakable semantics of fuzzy matching; and output the visible and speakable semantics when the preset priority is that the priority of common semantics is lower than the visible and speakable semantics of fuzzy matching.

[0034] According to a third aspect of an embodiment of the present application, a semantic arbitration device is provided, comprising a processor and a memory, wherein the memory stores computer program instructions that can be executed by the processor, and when the processor executes the computer program instructions, the steps of the method described in any one of the first aspects above are implemented.

[0035] According to a fourth aspect of an embodiment of the present application, a computer-readable storage medium is provided, in which computer program instructions are stored. When the computer program instructions are executed by a processor, the processor is prompted to implement the steps of the method described in any one of the first aspects above.

[0036] In the present application, the visible-to-speak semantics corresponding to the user's voice are obtained from an offline recognition engine; the online common semantics corresponding to the user's voice are obtained from an online recognition engine; when the visible-to-speak semantics exactly match the visible-to-speak hotword, the visible-to-speak semantics are output; when the visible-to-speak semantics fuzzily match the visible-to-speak hotword, the online common semantics or the visible-to-speak semantics are output according to a preset priority. The technical solution provided by the present application can selectively output the online common semantics or the visible-to-speak semantics according to a preset priority when the visible-to-speak semantics fuzzily match the visible-to-speak hotword, thereby avoiding the bad experience that the visible-to-speak semantics may cause to the user by mistakenly triggering related instructions.

[0037] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The accompanying drawings are incorporated into and constitute a part of the specification, illustrating embodiments consistent with the present application and, together with the specification, explaining the principles of the present application. Obviously, the drawings described below are only some embodiments of the present application, and those skilled in the art can derive other drawings based on these drawings without inventive effort. In the drawings:

[0039] Figure 1 A schematic flow chart of a semantic arbitration method in one embodiment is shown;

[0040] Figure 2 A block diagram of a semantic arbitration device in one embodiment is shown;

[0041] Figure 3 A schematic structural diagram of a semantic arbitration device in one embodiment is shown. DETAILED DESCRIPTION

[0042] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0043] In addition, described feature, structure or characteristic can be combined in one or more embodiments in any suitable manner.In the following description, many specific details are provided so as to provide a full understanding of the embodiments of the present application. However, it will be appreciated by those skilled in the art that the technical scheme of the present application can be put into practice without one or more of the specific details, or other methods, components, devices, steps etc. can be adopted. In other cases, known methods, devices, implementations or operations are not shown or described in detail to avoid blurring the various aspects of the application.

[0044] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically separate entities. That is, these functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0045] The flowcharts shown in the accompanying drawings are for illustrative purposes only and do not necessarily include all contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may be decomposed, while others may be combined or partially combined. Therefore, the actual execution order may vary depending on the actual situation.

[0046] Figure 1 FIG. 1 shows a flow chart of a semantic arbitration method in one embodiment. Figure 1 As shown, a semantic arbitration method is provided, which may include the following steps 101 to 104.

[0047] In step 101 , visible and audible semantics corresponding to user speech are obtained from an offline recognition engine.

[0048] It is understandable that the offline recognition engine is used to recognize the user's voice when the voice client is in an offline state. The offline recognition engine may include an offline voice recognition module (i.e., an offline ASR module) and an offline natural language understanding module (i.e., an offline NLU module). The offline ASR module is used to output offline voice recognition results; the offline NLU module is used to output visible and utterable semantics or offline common semantics, wherein common semantics is relative to visible and utterable semantics. Visible and utterable semantics are semantics that hit visible and utterable hot words, and common semantics are semantics that do not hit visible and utterable hot words. That is to say, if the user's voice hits the visible and utterable hot words in the offline ASR module and the offline NLU module, the offline NLU module will output the visible and utterable semantics, otherwise, the offline NLU module will output the offline common semantics.

[0049] It should be noted that the speech recognition results output by the Automatic Speech Recognition (ASR) module are text converted from the user's voice; while the results output by the NLU module are text obtained by further understanding and intent recognition based on the speech recognition results. The text can be visible and speakable semantics or general semantics.

[0050] In step 102, online common semantics corresponding to the user's speech are obtained from an online recognition engine.

[0051] It is understood that the online recognition engine is used to recognize user speech when the voice client is online. The online recognition engine may include an online speech recognition module (i.e., an online ASR module) and an online natural language understanding module (i.e., an online NLU module). The online ASR module is used to output online speech recognition results, and the online natural language understanding module is used to output online general semantics.

[0052] In some embodiments, visible and audible hot words can be sent to the offline speech recognition module, the offline natural language understanding module and the online speech recognition module respectively, so that the offline speech recognition module outputs offline speech recognition results based on the user voice and the visible and audible hot words, the offline natural language understanding module outputs visible and audible semantics or offline common semantics based on the user voice and the visible and audible hot words, the online speech recognition module outputs online speech recognition results based on the user voice and the visible and audible hot words, and the online natural language understanding module outputs online common semantics based on the user voice.

[0053] It should be noted that since no visible and utterable hotwords have been uploaded to the online NLU module, the online NLU module can only output online general semantics. Since visible and utterable hotwords have been uploaded to the online ASR module, the online ASR module's recognition performance will not be affected when the user's speech is in the visible and utterable language.

[0054] In step 103 , when the "visible and utterable" semantics exactly match the "visible and utterable" hotword, the "visible and utterable" semantics are output.

[0055] It is understandable that if the visible-can-be-spoken semantics and the online common semantics are acquired, arbitration is performed on the visible-can-be-spoken semantics and the online common semantics.

[0056] During the implementation process, the visible-can-be-speaking semantics can be compared with the visible-can-be-speaking hot words. If the two are exactly matched, the visible-can-be-speaking semantics are output.

[0057] In some embodiments, when the word slot corresponding to the visible-to-speak semantics and the word slot corresponding to the visible-to-speak hot word are completely identical, it can be determined that the visible-to-speak semantics and the visible-to-speak hot word are exactly matched.

[0058] In step 104 , when the visible-and-speakable semantics and the visible-and-speakable hot words are fuzzy matched, the online common semantics or the visible-and-speakable semantics are output according to a preset priority.

[0059] During the implementation process, you can first compare the "visible and speakable" semantics with the "visible and speakable" hot words. If the two are not an exact match, it means that they are a fuzzy match. At this time, you need to determine whether to output online general semantics or "visible and speakable" semantics based on the specific scenario or product requirements.

[0060] In some embodiments, when the word slot corresponding to the visible-to-speak semantics and the word slot corresponding to the visible-to-speak hotword are not completely the same, it is determined that the visible-to-speak semantics and the visible-to-speak hotword are fuzzy matched.

[0061] It is understandable that the preset priority can be set according to specific scenarios or product requirements. For some scenarios that require priority output of normal semantics, the preset priority is that normal semantics is higher than fuzzy matching visible and speakable semantics; otherwise, the preset priority is that normal semantics is lower than fuzzy matching visible and speakable semantics.

[0062] When the preset priority is that the priority of common semantics is higher than the visible and utterable semantics of fuzzy matching, the online common semantics is output; when the preset priority is that the priority of common semantics is lower than the visible and utterable semantics of fuzzy matching, the visible and utterable semantics is output.

[0063] In some embodiments, when the visible and speakable semantics are not obtained, the offline common semantics corresponding to the user's voice are obtained from the offline natural language understanding module; when the offline common semantics are invalid semantics, the online speech recognition results are input into the offline natural language understanding module, so that the offline natural language understanding module generates visible and speakable semantics or new offline common semantics based on the online speech recognition results.

[0064] It is understandable that if the visible-to-speak semantics are not obtained, the offline NLU module outputs the offline common semantics. At this time, the offline common semantics can be obtained from the offline NLU module, and then the offline common semantics and the online common semantics are arbitrated.

[0065] In some embodiments, the score of the offline general semantics may be determined based on the user voice; if the score of the offline general semantics is less than a preset score, the offline general semantics is determined to be an invalid semantics.

[0066] Taking the preset score of 60 points as an example, if the score of offline common semantics is 50 points, it means that offline common semantics is invalid semantics. If the score of offline common semantics is 80 points, it means that offline common semantics is valid semantics.

[0067] If the offline common semantics are valid, the output of the common semantics is determined based on whether the online common semantics are valid. If the online common semantics are valid, the online common semantics are output; if the online common semantics are not valid and the offline common semantics are valid, the offline common semantics are output.

[0068] If the offline common semantics are invalid, the flag of the offline common semantics can be set to false. When the online ASR module detects that flag = false, it inputs the online speech recognition result into the offline NLU module. After the offline NLU module generates visible and audible semantics or new offline common semantics based on the online speech recognition result, the semantic arbitration module arbitrates the visible and audible semantics or the new offline common semantics and the online common semantics.

[0069] By inputting the online speech recognition results into the offline NLU module, the offline NLU module's S2S recognition capability can be improved, reducing the impact of the online NLU module's failure to upload S2S hot words on the overall S2S recognition effect.

[0070] The embodiments of the present application obtain the visible-to-speak semantics corresponding to the user's voice from an offline recognition engine; obtain the online common semantics corresponding to the user's voice from an online recognition engine; when the visible-to-speak semantics exactly match the visible-to-speak hotword, the visible-to-speak semantics are output; when the visible-to-speak semantics fuzzily match the visible-to-speak hotword, the online common semantics or the visible-to-speak semantics are output according to a preset priority. The technical solution provided by this application can selectively output the online common semantics or the visible-to-speak semantics according to a preset priority when the visible-to-speak semantics fuzzily match the visible-to-speak hotword, thus avoiding the negative user experience caused by the visible-to-speak semantics mistakenly triggering related instructions.

[0071] The following describes an apparatus embodiment of the present application, which can be used to execute the semantic arbitration method in the above-mentioned embodiment of the present application. For details not disclosed in the apparatus embodiment of the present application, please refer to the above-mentioned embodiment of the semantic arbitration method of the present application.

[0072] See also Figure 2 , shows a block diagram of the semantic arbitration device in the embodiment of the present application. Figure 2 As shown, the semantic arbitration device of the embodiment of the present application includes: a first semantic acquisition module 201, a second semantic acquisition module 202 and a semantic arbitration module 203, wherein the first semantic acquisition module 201 is used to obtain the visible and utterable semantics corresponding to the user voice from the offline recognition engine; the second semantic acquisition module 202 is used to obtain the online common semantics corresponding to the user voice from the online recognition engine; the semantic arbitration module 203 is used to output the visible and utterable semantics when the visible and utterable semantics and the visible and utterable hot words are exactly matched; the semantic arbitration module 203 is also used to output the online common semantics or the visible and utterable semantics according to the preset priority when the visible and utterable semantics and the visible and utterable hot words are fuzzy matched.

[0073] In some embodiments of the present application, based on the aforementioned scheme, the offline recognition engine includes an offline speech recognition module and an offline natural language understanding module, and the online recognition engine includes an online speech recognition module and an online natural language understanding module, and a hot word upload module (not shown in the figure) is used to send the visible and speakable hot words to the offline speech recognition module, the offline natural language understanding module and the online speech recognition module respectively, so that the offline speech recognition module outputs the offline speech recognition result according to the user voice and the visible and speakable hot words, the offline natural language understanding module outputs the visible and speakable semantics or offline common semantics according to the user voice and the visible and speakable hot words, the online speech recognition module outputs the online speech recognition result according to the user voice and the visible and speakable hot words, and the online natural language understanding module outputs the online common semantics according to the user voice.

[0074] In some embodiments of the present application, based on the aforementioned scheme, the first semantic acquisition module 201 is further used to obtain offline common semantics corresponding to the user voice from the offline natural language understanding module when the visible and audible semantics are not obtained; and when the offline common semantics are invalid semantics, input the online speech recognition results into the offline natural language understanding module, so that the offline natural language understanding module generates visible and audible semantics or new offline common semantics based on the online speech recognition results.

[0075] In some embodiments of the present application, based on the aforementioned scheme, the semantic arbitration module 203 is further configured to output the online common semantics when the online common semantics are valid semantics; and output the offline common semantics when the online common semantics are not valid semantics and the offline common semantics are valid semantics.

[0076] In some embodiments of the present application, based on the aforementioned solution, the semantic arbitration module 203 is further used to determine the score of the offline common semantics according to the user voice; when the score of the offline common semantics is less than a preset score, it is determined that the offline common semantics is an invalid semantics.

[0077] In some embodiments of the present application, based on the aforementioned scheme, the semantic arbitration module 203 is further used to determine that the visible and speakable semantics and the visible and speakable hot words are exactly matched when the word slot corresponding to the visible and speakable semantics is exactly the same as the word slot corresponding to the visible and speakable hot words; and to determine that the visible and speakable semantics and the visible and speakable hot words are fuzzy matched when the word slot corresponding to the visible and speakable semantics is not exactly the same as the word slot corresponding to the visible and speakable hot words.

[0078] In some embodiments of the present application, based on the aforementioned scheme, the semantic arbitration module 203 is also used to output the online common semantics when the preset priority is that the common semantics have a higher priority than the fuzzy-matched visible and speakable semantics; and to output the visible and speakable semantics when the preset priority is that the common semantics have a lower priority than the fuzzy-matched visible and speakable semantics.

[0079] Based on the same inventive concept, the present application embodiment also provides a semantic arbitration device, referring to Figure 3 , shows a structural diagram of a semantic arbitration device in an embodiment of the present application, wherein the semantic arbitration device includes one or more memories 304, one or more processors 302, and at least one computer program (computer program instruction) stored in the memories 304 and executable on the processors 302. When the processors 302 execute the computer programs, the method described above is implemented.

[0080] Among them, Figure 3 In the embodiment of the present invention, a bus architecture (represented by bus 300) is shown. Bus 300 may include any number of interconnected buses and bridges, and bus 300 links together various circuits including one or more processors represented by processor 302 and memory represented by memory 304. Bus 300 may also link together various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. Bus interface 305 provides an interface between bus 300 and receiver 301 and transmitter 303. Receiver 301 and transmitter 303 may be the same component, namely a transceiver, which provides a unit for communicating with various other devices over a transmission medium. Processor 302 is responsible for managing bus 300 and general processing, while memory 304 may be used to store data used by processor 302 when performing operations.

[0081] Based on the same inventive concept, an embodiment of the present application provides a computer-readable storage medium, in which computer program instructions are stored. When the computer program instructions are executed by a processor, the processor is prompted to implement the steps of the method as described above.

[0082] The functions described herein may be implemented in hardware, software executed by a processor, firmware, or any combination thereof. If implemented in software executed by a processor, the functions may be stored as one or more instructions or codes on or transmitted via a computer-readable medium. Other examples and implementations are within the scope and spirit of this application and the appended claims. For example, due to the nature of software, the functions described above may be implemented using software executed by a processor, hardware, firmware, hardwiring, or a combination of any of these. Furthermore, the functional units may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit.

[0083] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only exemplary. For example, the division of the units can be a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0084] The units described as separate components may or may not be physically separate, and the components of the control device may or may not be physical units, that is, they may be located in one place or distributed across multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0085] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk, etc. Various media that can store computer program instructions.

[0086] The foregoing is merely an embodiment of the present application and is not intended to limit the present application. Various modifications and variations are possible for those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present application shall be included within the scope of the claims of the present application.

Claims

1. A semantic arbitration method, characterized in that: include: Obtain the visible and audible semantics corresponding to the user's voice from the offline recognition engine; Obtaining online common semantics corresponding to the user speech from an online recognition engine; When the visible-can-be-speakable semantics and the visible-can-be-speakable hotwords are exactly matched, outputting the visible-can-be-speakable semantics; When the visible-to-be-spoken semantics and the visible-to-be-spoken hotwords are fuzzy matched, the online common semantics or the visible-to-be-spoken semantics are output according to a preset priority; wherein the offline recognition engine includes an offline speech recognition module and an offline natural language understanding module, and the online recognition engine includes an online speech recognition module and an online natural language understanding module, and the method further includes: If the visible-to-speak semantics are not obtained, obtaining offline common semantics corresponding to the user speech from the offline natural language understanding module; In the case where the offline common semantics are invalid semantics, the online speech recognition result output by the online speech recognition module is input into the offline natural language understanding module, so that the offline natural language understanding module generates visible and audible semantics or new offline common semantics according to the online speech recognition result.

2. The semantic arbitration method according to claim 1, characterized in that: Also includes: The visible and utterable hot words are sent to the offline speech recognition module, the offline natural language understanding module and the online speech recognition module respectively, so that the offline speech recognition module outputs an offline speech recognition result according to the user voice and the visible and utterable hot words, the offline natural language understanding module outputs the visible and utterable semantics or offline common semantics according to the user voice and the visible and utterable hot words, the online speech recognition module outputs an online speech recognition result according to the user voice and the visible and utterable hot words, and the online natural language understanding module outputs the online common semantics according to the user voice.

3. The semantic arbitration method according to claim 2, characterized in that: Also includes: In the case where the online common semantics is a valid semantics, outputting the online common semantics; In a case where the online general semantics is not a valid semantics and the offline general semantics is a valid semantics, the offline general semantics is output.

4. The semantic arbitration method according to claim 2, characterized in that: Also includes: Determining a score of the offline common semantics according to the user speech; When the score of the offline common semantics is less than a preset score, it is determined that the offline common semantics is an invalid semantics.

5. The semantic arbitration method according to any one of claims 1 to 4, characterized in that: Also includes: In a case where the word slot corresponding to the visible-to-speak semantics is exactly the same as the word slot corresponding to the visible-to-speak hot word, determining that the visible-to-speak semantics is an exact match to the visible-to-speak hot word; When the word slot corresponding to the visible-and-speakable semantics is not completely the same as the word slot corresponding to the visible-and-speakable hot word, it is determined that the visible-and-speakable semantics and the visible-and-speakable hot word are fuzzy matched.

6. The semantic arbitration method according to any one of claims 1 to 4, characterized in that: Outputting the online common semantics or the visible-and-speakable semantics according to the preset priority includes: In a case where the preset priority is that the priority of the common semantics is higher than the fuzzy-matched visible-that-is-speakable semantics, outputting the online common semantics; In a case where the preset priority is that the priority of the common semantics is lower than the fuzzy-matched visible-and-speakable semantics, the visible-and-speakable semantics are output.

7. A semantic arbitration device, characterized in that: include: A first semantics acquisition module is used to obtain the visible and audible semantics corresponding to the user's voice from the offline recognition engine; A second semantics acquisition module is used to acquire online common semantics corresponding to the user speech from an online recognition engine; A semantic arbitration module, configured to output the visible-and-speakable semantics when the visible-and-speakable semantics exactly match the visible-and-speakable hot words; The semantic arbitration module is further used to output the online common semantics or the visible and utterable semantics according to the preset priority when the visible and utterable semantics and the visible and utterable hot words are fuzzy matched; wherein the offline recognition engine includes an offline speech recognition module and an offline natural language understanding module, the online recognition engine includes an online speech recognition module and an online natural language understanding module, and the first semantic acquisition module is also used If the visible-to-speak semantics are not obtained, obtaining offline common semantics corresponding to the user speech from the offline natural language understanding module; In the case where the offline common semantics are invalid semantics, the online speech recognition result output by the online speech recognition module is input into the offline natural language understanding module, so that the offline natural language understanding module generates visible and audible semantics or new offline common semantics according to the online speech recognition result.

8. A semantic arbitration device comprising a processor and a memory, characterized in that: The memory stores computer program instructions that can be executed by the processor, and when the processor executes the computer program instructions, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer program instructions, which, when executed by a processor, prompt the processor to implement the steps of the method according to any one of claims 1 to 6.

Images