TWI780502B - Speech recognition system, command generation system, and speech recognition method thereof - Google Patents

Abstract

A speech recognition system, an instruction generation system and a speech recognition method thereof are provided. The speech recognition system is adapted to communicate with an application system. The application system receives a speech input. The speech recognition system includes a speech recognition module, a natural speech understanding system, and a command generation system. The speech recognition module receives the speech input provided by the application system, and recognizes the speech input to generate speech information. The natural speech understanding system is coupled to the speech recognition module. The natural speech understanding system understands speech information to generate a semantic analysis result. The command generation system utilizes the semantic analysis result to compare a selected item in an interface content of a current user interface, and then outputs a control command to the application system according to a comparison result.

Description

Speech recognition system, command generation system and speech recognition method thereof

The present invention relates to a speech recognition technology, and in particular to a speech recognition system, an instruction generating system and a speech recognition method thereof.

With the evolution of speech recognition technology, various application systems are beginning to try to be equipped with a speech recognition function, so as to improve the operation convenience of the application system. Especially, for a virtual reality (Virtual Reality, VR) system or an augmented reality (Augmented Reality, AR) system, if the voice recognition function can be provided, the performance of the virtual reality system or the augmented reality system can be greatly improved. Ease of operation and user experience. However, since the voice recognition often needs to consume a large amount of system resources, it will increase the construction cost of the system, and even affect the running speed of the system.

Another problem is that since the traditional speech recognition realizes the speech recognition function through the compilation of Automatic Speech Recognition grammar (ASR grammar), the user must compile all the sayings and contents selected by the speech into the automatic speech Among the recognition grammars, only words that fully conform to the automatic speech recognition grammar can be matched. That is to say, the traditional voice recognition method requires a lot of workload for system developers, and the use of voice recognition is not flexible enough. Even, if the voice recognition function needs to be implemented in the virtual reality system or the augmented reality system, it may be necessary to modify the internal system settings of the application system, which will increase the complexity and cost of setting the system.

The invention provides a speech recognition system, an instruction generation system and a speech recognition method thereof, which can provide convenient speech recognition functions.

The speech recognition system of the present invention is adapted to communicate with application systems. The application system is used for receiving voice input. The speech recognition system includes a speech recognition module, a natural speech understanding system, and an instruction generation system. The voice recognition module is used for receiving voice input provided by the application system, and recognizing the voice input to generate voice information. The natural speech understanding system is coupled to the speech recognition module and used for understanding speech information to generate semantic analysis results. The instruction generating system is coupled to the natural speech understanding system, and is used for comparing a selection item in an interface content of a current user interface by using the semantic analysis result, and outputting a control instruction to the application system according to a comparison result .

In an embodiment of the present invention, the above-mentioned command generation system includes a comparison module and a command confirmation module. The comparison module is used to receive the semantic analysis result, and use the semantic analysis result to compare the selected items in the interface content of the current user interface to generate a comparison result. The command confirmation module is coupled to the comparison module, and is used for converting the comparison result according to the command format, and outputting the control command.

In an embodiment of the present invention, the above-mentioned application system is used to display the current user interface. After the application system receives the control command output by the speech recognition system, the application system selects the selection item in the interface content of the current user interface according to the control command, and changes to display the next user interface or perform a specific operation.

In an embodiment of the present invention, the above-mentioned natural speech understanding system includes a natural language processor, a knowledge-aided understanding module, a retrieval system, and an analysis result output module. The natural language processor is coupled to the speech recognition module and is used to receive speech information to generate possible intent grammar data. The knowledge aided comprehension module is coupled to the natural language processor and uses intent data storing possible intent grammar data. The retrieval system is coupled to the knowledge-assisted understanding module, and is used to receive the keywords of the possible intention grammar data provided by the knowledge-assisted understanding module, so as to generate a response result to the knowledge-assisted understanding module according to the keywords, so that the knowledge-assisted understanding module The group generates intent-determined grammar data based on the response results. The analysis result output module is coupled to the knowledge aided comprehension module and the command generation system, and is used to output the semantic analysis result according to the grammar data of the determined intention.

The instruction generating system of the present invention is adapted to communicate with an application system. The application system is used for receiving voice input. The command generation system includes a comparison module and a command confirmation module. The comparison module is used for receiving the semantic analysis result corresponding to the voice input, and using the semantic analysis result to compare the selected items in the interface content of the current user interface to generate a comparison result. The command confirmation module is coupled to the comparison module to convert the comparison result according to the command format, and output the control command to the application system.

The speech recognition method of the present invention is suitable for a speech recognition system. The voice recognition system communicates with the application system, and the application system is used to receive voice input. The speech recognition method includes the following steps. Receive voice input provided by the application system. Recognize voice input to generate voice information. Understand speech information to generate semantic analysis results. Compare the semantic analysis results to output control instructions to the application system.

The speech recognition method of the present invention is suitable for an instruction generation system. The instruction generation system is adapted to communicate with the application system, and the application system is configured to receive speech input. The speech recognition method includes the following steps. Semantic analysis results corresponding to the speech input are received. The semantic analysis result is used to compare the selected items in the interface content of the current user interface, so as to generate a comparison result. The comparison result is confirmed according to the instruction format, and the control instruction is output to the application system.

Based on the above, the speech recognition system, command generation system and speech recognition method thereof of the present invention can recognize the speech input provided by the application system, and return the corresponding control command to the application system, so that the application system can execute the corresponding command according to the control command. corresponding operation. Therefore, the speech recognition system, command generation system and speech recognition method thereof of the present invention can not only provide a convenient speech recognition function, but also reduce system resources required for speech recognition in an application system.

In order to make the above-mentioned features and advantages of the present invention more comprehensible, the following specific embodiments are described in detail together with the accompanying drawings.

In order to make the content of the present invention more comprehensible, the following specific examples are given as examples in which the present invention can indeed be implemented. In addition, wherever possible, elements/components/steps using the same reference numerals in the drawings and embodiments represent the same or similar parts.

FIG. 1 is a schematic diagram of a speech recognition system according to an embodiment of the invention. Referring to FIG. 1 , the speech recognition system 100 is suitable for communicating with the application system 200 , and the communication between the speech recognition system 100 and the application system 200 can be in a wired or wireless manner. In this embodiment, the application system 200 includes a voice receiving module 210 and an instruction executing module 220 . The application system 200 receives the voice input 101 provided by the user through the voice receiving module 210 , and transmits the voice input 101 to the voice recognition system 100 . In this embodiment, the speech recognition system 100 can perform speech recognition on the speech input 101 provided by the application system 200 to generate corresponding instructions, and the speech input 101 returns the instructions to the instruction execution module of the application system 200 220 , so that the instruction execution module 220 executes related operations of the application system 200 . In other words, the speech recognition system 100 of this embodiment can be used with any application system and provide a speech recognition function.

In this embodiment, the application system 200 may be, for example, a game program or an application program running or carried on a virtual reality (Virtual Reality, VR) device or an augmented reality (Augmented Reality, AR) device, and the user The relevant operations in the game program can be controlled by voice. A virtual reality device or an augmented reality device may include hardware circuits such as processing circuits, memories, and voice sensing devices, so that the processing circuits may execute or access relevant modules or programs in the memory, and at least realize the present invention. Invented voice receiving function, command execution function and application program execution function. In this embodiment, the voice recognition system 100 may be built in a cloud server or a local host device, for example, to provide related voice recognition and processing functions. The speech recognition system 100 may also include another processing circuit and another memory, so that another processing circuit executes or accesses a related module or program in another memory, so as to realize at least the speech recognition function of the present invention.

In this embodiment, the speech recognition system 100 includes a speech recognition module 110 , a natural language understanding system 120 , an instruction generation system 130 and a storage device 140 . The voice recognition module 110 is coupled to the voice receiving module 210 of the application system 200 and the natural language understanding system 120 . The instruction generating system 130 is coupled to the natural language understanding system 120 and the storage device 140 . FIG. 2 is a flowchart of a voice recognition method according to an embodiment of the invention. With the voice recognition method in FIG. 2 , the voice recognition system 100 in FIG. 1 can execute steps S210 - S240 in FIG. 2 to realize the voice recognition function. In step S210 , the voice recognition module 110 of the voice recognition system 100 receives the voice input 101 provided by the application system 200 . In step S220 , the voice recognition module 110 recognizes the voice input 101 to generate voice information 102 . In this embodiment, the voice recognition module 110 can convert the voice input 101 signal into voice information 102 (or data) that can be processed and analyzed by a computer.

In step S230 , the natural language understanding system 120 receives the speech information 102 output by the speech recognition module 110 , and understands the speech information 102 to generate a semantic analysis result 103 . In step S240 , the instruction generating system 130 receives the semantic analysis result 103 output by the natural language understanding system 120 , and compares the semantic analysis result 103 to output the control instruction 105 to the application system 200 . In this embodiment, the storage device 140 can provide the interface content 104 of the user interface displayed by the current application system 200 to the instruction generation system 130, so that the instruction generation system 130 can compare the semantic analysis result 103 with the interface of the current user interface content 104 to generate the control command 105 to the command execution module 220 of the application system 200 . In this embodiment, the command execution module 220 can make the application system 200 display the next user interface or execute a specific operation according to the control command 105 . The storage device 140 can be any type of memory in the server or computer system, such as dynamic random access memory (DRAM), static random access memory (SRAM), flash memory (Flash memory), read-only memory (ROM ), etc., the present invention is not limited to this, those skilled in the art can select according to actual needs

In this embodiment, the natural language understanding system 120 may convert the speech information 102 into text information (Text Information), and normalize the text information, so as to generate a semantic analysis result 103 with an intended object. Moreover, the command generation system 130 can generate the control command 105 corresponding to the intended object and provide it to the application system 200, so that the application system 200 can execute the control command 105 to display the next user interface or perform a specific operation. Therefore, the speech recognition method of this embodiment can make the application system 200 not need to spend extra system resources to perform speech recognition, and can effectively save the system resources required for the application system 200 to perform the function of recognizing the user's speech input.

It should be noted that the semantic analysis result 103 output by the natural language understanding system 120 may include one or more possible semantic data, and the semantic data may include keywords and intent data. In other words, the user can express the selection intention in a colloquial manner, such as the full name, abbreviation or alias of the selected item, and the corresponding control command can be generated through the speech recognition system 100 of this embodiment without having to read out The full specific name. Regarding this, regarding the manner in which the natural language understanding system 120 generates the semantic analysis result 103 , the embodiment in FIG. 6 will be used as an example to illustrate below.

FIG. 3 is a schematic diagram of an instruction generation system according to an embodiment of the invention. FIG. 4 is a flowchart of a speech recognition method according to another embodiment of the present invention. Referring to FIG. 1 , FIG. 3 and FIG. 4 , the command generating system 130 in FIG. 1 is an application system interface (Interface). Users can edit the command generating system 130 to make the voice recognition system 100 applicable to the corresponding application system 200 . The instruction generation system 130 may include a system architecture as shown in FIG. 3 . In this embodiment, the instruction generation system 130 includes a comparison module 131 , an instruction confirmation module 132 , a temporary storage device 133 , an access module 134 and an item acquisition module 135 . The comparison module 131 is coupled to the instruction confirmation module 132 and the item acquisition module 135 . The temporary storage device 133 is coupled to the command confirmation module 132 and the access module 134 . The access module 134 is coupled to the item acquisition module 135 and the storage device 140 . In this embodiment, the temporary storage device 133 can be, for example, dynamic random access memory (DRAM), static random access memory (SRAM), flash memory (Flash memory), read-only memory (ROM), etc. This is not limited, and those skilled in the art can select according to actual needs.

With the voice recognition method in FIG. 4 , the command generation system 130 in FIG. 3 can execute steps S410 - S450 in FIG. 4 to realize voice recognition and command generation functions. In step S410 , the temporary storage device 133 receives the interface number 301 of the current user interface provided by the application system 200 . In step S420 , the access module 134 generates the interface content 303 of the current user interface according to the interface number 301 . In this embodiment, the access module 134 can access the interface data preloaded in the storage device 140 according to the interface number 301 to obtain the interface content 303 of the current user interface displayed by the application system 200 .

In step S430, the item acquisition module 135 receives the interface content 303 of the current user interface provided by the access module 134, and the item acquisition module 135 acquires the selected item 304 from the interface content 303 of the current user interface, to The selection item 304 is output to the comparison module 131 . In step S440 , the comparison module 131 uses the semantic analysis result 103 to compare the selection item 304 of the current user interface to generate a comparison result 305 . It should be noted that the selected item 304 may include an item name and a plurality of reference keywords corresponding to the item name. That is to say, the item obtaining module 135 can extract the item name of the selected item 304 and the plurality of reference keywords corresponding to the item name from the interface content 303 of the current user interface. Moreover, the comparison module 131 can compare whether the semantic analysis result 103 matches the item name and one of the plurality of reference keywords to generate a comparison result 305 . In other words, as long as the semantic analysis result 103 generated after the voice input spoken by the user matches the item name and one of the plurality of reference keywords after being understood by the natural language understanding system 120, the comparison module 131 can For example, the comparison result 305 of the corresponding item number is output. The plurality of reference keywords may be, for example, abbreviations or aliases of project names.

In step S450 , the instruction confirmation module 132 converts the comparison result 305 according to the instruction format 307 , and outputs, for example, a control instruction 306 with a corresponding item number. In this embodiment, the instruction format 307 refers to the instruction format that the application system 100 can receive, and the instruction confirmation module 132 outputs the control instruction 306 to the application system 200 through the temporary storage device 133 . Therefore, when the application system 200 displays the current user interface, if the application system 200 receives the control command 306 output by the speech recognition system 100, the application system 200 can select the interface content 303 of the current user interface according to the control command 306 Select the item in the item, so that the application system 200 can, for example, change to display the next user interface or perform a specific operation according to the item number obtained above. Accordingly, the voice recognition method of this embodiment can enable the command generating system 130 to effectively recognize the user's voice input and generate corresponding control commands.

In addition, since the speech recognition system 100 can be applied to various application systems, the user only needs to edit the speech recognition system 100 without changing the application system. For example, the speech recognition system 100 can first be operated in the edit mode (or edited through a software development kit (SDK, Software Development Kit) of the speech recognition system, so that the interface content 104 of the user interface displayed by the application system 200 and The instruction format 302 can be written into the storage device 140 through the temporary storage device 133 and the access module 134 in advance. Therefore, when the speech recognition system 100 is operated in the working mode, the command generation system 130 can receive the current usage through the temporary storage device 133. The interface number 301 of the user interface, and the access module 134 can read the storage device 140 according to the interface number 301 to obtain the interface content 303 of the corresponding current user interface. The command confirmation module 132 can use the access module 134 to Obtain the instruction format 307. That is to say, the voice recognition system 100 of this embodiment can be adapted to match with various application systems to provide effective voice recognition and voice selection functions.

FIG. 5 is a schematic diagram of a user interface of an application system according to an embodiment of the present invention. FIG. 5 is an example of a user interface that may be displayed by the application system 200 of FIG. 1 . Referring to FIG. 1 , FIG. 3 and FIG. 5 , the application system 200 can, for example, execute a virtual reality game program. In this regard, taking a game program as an example, what needs to be explained first is that the game developer can first create one or more corresponding data sets according to each user interface that may be displayed in the game, wherein each user interface Can include one or more project names each. For this, the game developer can create a data set for each item name, including the item label, the number of columns and rows where the item name is located on the interface, and a plurality of reference keywords corresponding to the item name. Therefore, when the speech recognition system 100 is connected with the game program, the game program can input the created multiple data sets into the temporary storage device 133 of the instruction generation system 130 and store them in the temporary storage device 133 . The access module 134 can then read the temporary storage device 133 and store the plurality of data sets in the storage device 140 of the speech recognition system 100 .

Next, assume that the application system 200 first displays the user interface 510 as shown in FIG. 5 . The interface content of the user interface 510 includes an interface name 511 (home page) and a plurality of selection items 512 - 514 . The interface content 303 of the storage device 140 accessed by the access module 134 may, for example, include the data of the plurality of selection items 512 - 514 . It should be noted that the item acquisition module 135 can extract multiple item names of the selected items 512-514 and multiple reference keywords corresponding to the item names from the interface content of the user interface 510, so as to output multiple item names and The corresponding reference keywords are compared with the comparison module 131 . In this example, the comparison module 131 can obtain the following data content from the item acquisition module 135:

{id=0; column=0; line=0; title="Instructions for use"; alias="use", "first", "third from last", "one", "use", "operation",… }

{id=1; column=0; line=1; title="role selection"; alias="casting", "second", "second to last", "role", "character",...}

{id=2; column=0; line=2; title="level selection"; alias="level", "third", "last one", "battle", "war",…}

Among them, "id" is the item number, "column" is the number of columns, "line" is the number of lines, "title" is the name of the project, and "alias" is the reference keyword. In this example, when the user wants to select the selection item 513, he can, for example, say the full name corresponding to the selection item 513 "role selection", "selection" for short, or the number of items "second", "last but one". Two", etc., can make the comparison module 131 compare the selection item 513, and output the corresponding comparison result 305 to the command confirmation module 132, so as to then generate the corresponding control command 306 to the application system 200 for command execution Module 220. Therefore, the application system 200 can then change to display the next user interface 520 .

Then, when the user interface 520 displayed by the system 200 is displayed, the interface content of the user interface 520 includes an interface name 521 (role selection) and a plurality of selection items 522 - 524 . In this example, the comparison module 131 can obtain the following data content from the item acquisition module 135:

{id=3; column=0; line=0; title="Zhao Yun"; alias="Zhao Zilong", "first", "third from last", "one", "Zhao", "Zilong",… }

{id=4; column=0; line=1; title="Guan Yu"; alias="Guan Yunchang", "Second", "Second to last", "Guan", "Yunchang",…}

{id=5; column=0; line=2; title=“Cao Cao”; alias=“Cao Mengde”, “third”, “last one”, “three”, “Cao”, “Meng De”,… }

In this example, the user can, for example, say the full name "Zhao Yun" corresponding to the selection item 522, the abbreviation "Zhao", the alias "Zhao Zilong" or the item number "first", "the third from the bottom", etc., all can be used The comparison module 131 can compare to the selection item 522 selected by the user, and output the corresponding comparison result 305 to the instruction confirmation module 132, so as to then generate the corresponding control instruction 306 to the instruction execution module of the application system 200 Group 220. Therefore, the application system 200 can then change to display the next user interface 530 .

Next, when the user interface 530 displayed by the system 200 is displayed, the interface content of the user interface 530 includes an interface name 531 (weapon selection) and a plurality of selection items 532 - 534 . In this example, the comparison module 131 can obtain the following data content from the item acquisition module 135:

{id=6; column=0; line=0; title="Qinghong Sword"; alias="Sword", "Qinghong", "First", "Third from last", "One",…}

{id=7; column=0; line=1; title="long gun"; alias="gun", "second", "second to last", "two",...}

{id=8; column=0; line=2; title="大刀"; alias="刀", "third", "last one", "three"...}

In this example, the user can, for example, say the full name "Qinghong Sword", the abbreviation "Qinghong" or the item number "No. The user selects the selection item 532 , and outputs the corresponding comparison result 305 to the command confirmation module 132 to then generate the corresponding control command 306 to the command execution module 220 of the application system 200 . Therefore, the application system 200 can then perform subsequent related specific operations in the game program.

However, the voice input provided by the user is not limited to the above-mentioned forms of full name, abbreviation, alias or item number. In one embodiment, the comparison module 131 can also directly extract item number information (may include sequence item number information or reverse order item number information) of multiple selected items in the current user interface from the semantic analysis result 103 Information), directly extract the line number or column number of multiple selection items on the current user interface from the semantic analysis result 103, or directly perform pinyin matching according to the semantic analysis result 103 to find the beginning, end or character of the item name Strings can be matched to select items, etc. Moreover, the comparison module 131 can also output a comparison result 305 corresponding to a plurality of selected items that are successfully matched, so that the application system 200 can also execute a plurality of control instructions simultaneously or sequentially.

FIG. 6 is a schematic diagram of a natural speech understanding system according to an embodiment of the present invention. It should be noted that, in some embodiments of the present invention, the natural language understanding system of the present invention may, for example, apply the framework of a natural language understanding system such as a Chinese invention patent (publication number CN103761242B), but the present invention is not limited to this. In other embodiments of the present invention, the natural language understanding system of the present invention may also adopt other system architectures capable of generating the semantic analysis results described in the various implementations of the present invention. Referring to FIG. 1 and FIG. 6 , the natural language understanding system 620 in FIG. 6 is an implementation example of the natural language understanding system 120 in FIG. 1 , but the natural language understanding system of the present invention is not limited thereto. In this example. The natural language understanding system 620 includes a natural language processor 621 , a knowledge-assisted understanding module 622 , a retrieval system 624 and an analysis result output module 629 . The knowledge aided comprehension module 622 is coupled to the natural language processor 621 and the retrieval system 624 . The knowledge aided comprehension module 622 includes intent data 623 . The retrieval system 624 includes a structured database 625 , a search engine 626 , an instruction data storage device 627 and a retrieval interface unit 628 , wherein the search engine 626 is coupled to the structured database 625 , the instruction data storage device 627 and the retrieval interface unit 628 .

In this embodiment, with reference to the following Table 1, when the natural language understanding system 620 receives the voice information 102 provided by the voice recognition module 110 of FIG. 1 (for example, when the user displays the user interface 520 of FIG. 5 , use verbal input of the request information "I want Zilong"), the natural language processor 621 can analyze the voice information 102 to generate possible intent grammar data 603. The natural language processor 621 can send the possible intention grammar data 603 to the knowledge aided understanding module 622 , wherein the possible intention grammar data 603 includes keywords 604 and intention data 623 . In this regard, since the keywords 604 (such as "Zilong") in the intent grammar data 603 may belong to different fields (such as two fields of role selection (<roleselect>) and movies (<readfilm>), a voice information 102 can be analyzed into multiple possible intent grammar data 603 (such as "<roleselect>,<rolename>=Zilong" or "<watchfilm>,<filmname>=Zilong"), so it is necessary to understand the module 622 through knowledge assistance Do further analysis to confirm the user's intent. In this embodiment, the knowledge aided comprehension module 622 can extract the keyword 604 (such as "zilong") in the possible intent grammar data 603 and send it to the retrieval interface unit 628 of the retrieval system 624 to search for the structure through the search engine 626 Optimize the database 625 to confirm whether there is the character name or movie title of "Zilong". Moreover, the natural language processor 621 stores the intention data 623 in the knowledge-assisted comprehension module 622 . Voice information (request information) possible intent grammar intent data keywords I want Zilong <roleselect>,<rolename>=Zilong <role select> Zi Long <watchfilm>, <filmname>=Zilong ＜watchfilm＞ Zi Long Table 1

In other words, in this embodiment, the natural language understanding system 620 can first extract the keywords 604 in the possible intent grammar data 603, and use the full-text search results of the structured database 625 to determine the domain attributes of the keywords 604, and then Then further analyze and confirm the clear intention of the user. Therefore, users can easily express their intentions or information in a colloquial manner without needing to memorize specific terms, such as the specific terms related to the list of fixed words in the existing practice.

In this embodiment, the structured database 625 in the retrieval system 624 may, for example, store a plurality of records. The search engine 626 in the retrieval system 624 will carry out a full-text search to the structured database 625 according to the keyword 604, and after confirming the user's intention, then the response result 605 of the full-text search (assuming that the structured database 625 stores a certain record , there is a record of "rolenameguid:Zhao Zilong" in its title field, and there is no title field of any record to store the information of "filmnameduid:Zhao Zilong", so the response result 605 will be "rolenameguid") and sent back to the knowledge-aided understanding module 622.

In this embodiment, the retrieval interface unit 628 can obtain the instruction data from the instruction data storage device 627 through the search engine 626, and the retrieval interface unit 628 sequentially outputs the instruction data in the complete match record and the partial match record of the matching keyword 604 As a response result 605 is sent to the knowledge aided understanding system 622 , wherein the priority of the full matching record is higher than that of the partial matching record. Next, the knowledge-aided understanding module 622 can compare the stored intention data 623 according to the response result 605, and obtain the determined intention grammar data 606 (for example, after comparing the response result 605 with the possible intention grammar data 603 , it is determined that the user's intention should be "<roleselect>,<rolename>=Zhao Zilong") and sent to the analysis result output module 629.

However, in another embodiment of the present invention, with reference to Table 2 below, each record stored in the structured database 220 may also include, for example, information such as popularity fields, like fields or dislike fields. For this, it is assumed that the intent grammar data 603 may contain two data (for example, "<roleselect>, <rolename>=Zilong" or "<roleselect>, <rolename>=Zilong"). And, after the search engine 626 of the retrieval system 624 performs a full-text search, if it is judged that there are two records that match the search results (assuming that the structured database 625 stores two records, the title fields in the title fields have "rolenameguid : Zhao Zilong" and "rolenameguid: Zilong" records), then the search engine 626 of the retrieval system 624 can further determine the popularity field, like field and dislike field in these two records. In this regard, the search engine 626 of the retrieval system 624 can, for example, further determine the semantic analysis result 103 according to the value of the popularity column (for example, the popularity value (8) corresponding to "Zhao Zilong" is higher, and the popularity value (8) corresponding to "Zilong" If the value (2) is low, the search engine 626 takes "Zhao Zilong" as the semantic analysis result 103). Or, the search engine 626 of the retrieval system 624 can further determine the semantic analysis result 103 based on the value of the preference field (for example, the preference value (20) corresponding to "Zhao Zilong" is higher, and the preference value corresponding to "Zilong" (5) is lower, then the search engine 626 uses "Zhao Zilong" as the semantic analysis result 103). Alternatively, the search engine 626 of the retrieval system 624 can further determine the semantic analysis result 103 based on the value of the dislike column (for example, the dislike value (1) corresponding to "Zhao Zilong" is lower, and the dislike value corresponding to "Zilong" If the numerical value (20) is higher, then the search engine 626 uses "Zhao Zilong" as the semantic analysis result (103). Moreover, in yet another embodiment of the present invention, the search engine 626 of the retrieval system 624 may also refer to at least one of the above popularity field, favorite field and dislike field, and is not limited to the above single judgment criterion (for example, If the popularity values of "Zhao Zilong" and "Zilong" are the same, the search engine 626 further compares the preference values, or adds the values of the popularity field and the preference field for comparison). Record title content bar popularity bar Favorite column dislike bar 1 rolenameguid:Zhao Zilong character selection 8 20 1 2 rolenameguid: Zilong character selection 2 5 20 Table 2

Therefore, the analysis result output module 629 can output the semantic analysis result 103 with a specific intention object according to the determined intention grammar data 606 . In this regard, since the natural language understanding system 620 can realize the judgment of complete matching and partial matching after the full-text search of the keyword 604, and output an appropriate semantic analysis result 103 (for example, according to the received determined intention grammatical data 606 "<roleselect >, <rolename>=Zhao Zilong" and confirm that the user wants to select Zhao Yun, so the semantic analysis result 103 of "Zhao Yun" is output and sent to the command generation system 105), therefore, in some embodiments of the present invention, the user can A more colloquial or flexible voice input form is provided, and the voice recognition system with the natural language understanding system 620 of this embodiment can effectively and accurately feed back corresponding control commands to the application system, thereby providing an effective voice selection function.

To sum up, the speech recognition system, command generation system and speech recognition method of the present invention can provide speech recognition function through another system externally installed outside the application system, and return corresponding control commands to the application system. Moreover, the speech recognition system, command generation system and speech recognition method thereof of the present invention can also perform effective speech recognition on the colloquial speech input provided by the user. Therefore, the speech recognition system, instruction generation system and speech recognition method thereof of the present invention can effectively reduce system resources required for speech recognition in application systems, and can realize convenient and flexible speech selection functions.

Although the present invention has been disclosed above with the embodiments, it is not intended to limit the present invention. Anyone with ordinary knowledge in the technical field may make some changes and modifications without departing from the spirit and scope of the present invention. The scope of protection of the present invention should be defined by the scope of the appended patent application.

100: Speech Recognition System 101:Voice input 102: Voice information 103: Semantic analysis results 104:Interface content 105: Control command 110:Language recognition module 120, 620: Natural Language Understanding System 130: Command generation system 131: Comparison module 132:Command confirmation module 133: Temporary storage device 134: access module 135: Item acquisition module 140: storage device 200: Application system 210:Voice receiving module 220: Instruction execution module 301: interface number 302, 307: instruction format 303: interface content 304: select item 305: Comparison result 306: Control command 511, 521, 531: interface name 512~514, 522~524, 532~534: select the item 603: Possible intent grammar data 604:Keyword 605: Response result 606: Determine intent grammar data 621: Natural Language Processor 622: Knowledge Assisted Comprehension Module 623: Intent data 624: Retrieval system 625: Structured database 626:Search Engine 627: Indicate data storage device 628: Search interface unit 629: Analysis result output module S210~S240, S410~S450: steps

FIG. 1 is a schematic diagram of a speech recognition system according to an embodiment of the invention. FIG. 2 is a flowchart of a voice recognition method according to an embodiment of the invention. FIG. 3 is a schematic diagram of an instruction generation system according to an embodiment of the invention. FIG. 4 is a flowchart of a speech recognition method according to another embodiment of the present invention. FIG. 5 is a schematic diagram of a user interface of an application system according to an embodiment of the present invention. FIG. 6 is a schematic diagram of a natural speech understanding system according to an embodiment of the present invention.

100: Speech Recognition System

101:Voice input

102: Voice information

103: Semantic analysis results

104:Interface content

105: Control command

110:Language recognition module

120:Natural Language Understanding Systems

130: Command generation system

140: storage device

200: Application system

210:Voice receiving module

220: Instruction execution module

Claims (38)

A voice recognition system, adapted to communicate with an application system, and the application system is used to receive a voice input, wherein the voice recognition system includes: a voice recognition module, used to receive the voice input provided by the application system , and recognize the speech input to generate speech information; a natural speech understanding system, coupled to the speech recognition module, and used to understand the speech information, to generate a semantic analysis result; and an instruction generation system, coupled to The natural speech understanding system is used to use the semantic analysis result to compare a selection item in an interface content of a current user interface displayed by the application system, and output a control command according to a comparison result to the application system. The speech recognition system as described in claim 1, wherein the instruction generation system includes: a comparison module, used to receive the semantic analysis result, and use the semantic analysis result to compare the interface in the current user interface The selection item in the content is used to generate the comparison result; and an instruction confirmation module is coupled to the comparison module and used for converting the comparison result according to an instruction format to output the control instruction. The speech recognition system as described in claim 2, wherein the interface content of the current user interface includes an item name of the selected item and an object number corresponding to the item name and a plurality of reference keywords, the comparison model Group comparison of the language The meaning analysis result matches the item name and one of the reference keywords to generate the comparison result. The speech recognition system as described in claim 3, wherein the command generation system further includes: a temporary storage device for receiving an interface number of the current user interface provided by the application system; an access module coupled to connected to the temporary storage device, and used to generate the interface content of the current user interface according to the interface number; and an item acquisition module, coupled to the access module and the comparison module, and used to obtain from Obtain the item name of the selected item, the item label corresponding to the item name, and the reference keywords from the interface content of the current user interface, so as to output the item name of the selected item and the item name corresponding to the item name The item label and the reference keywords are sent to the comparison module. The speech recognition system according to claim 4, wherein the temporary storage device is further coupled to the command confirmation module, and the command confirmation module outputs the control command to the application system through the temporary storage device. The speech recognition system as described in claim 4, further comprising: a storage device coupled to the access module of the command generation system, and the access module is used to access the storage device to obtain the currently used The interface content of the operator interface. The speech recognition system according to claim 6, wherein the interface content of the current user interface is pre-written into the storage device through the temporary storage device and the access module. The speech recognition system as described in claim 6, wherein the instruction format is written into the storage device through the temporary storage device and the access module in advance, and the instruction confirmation module obtains the instruction through the access module Format. The speech recognition system as described in claim 4, wherein the application system is used to display the current user interface, and when the application system receives the control instruction output by the speech recognition system, the application system will command to select the selection item in the interface content of the current user interface, and change to display the next user interface or perform a specific operation. The speech recognition system as described in claim 9, wherein when the application system changes to display the next user interface, the application system outputs the next interface number of the next user interface to the command generation system, so that the The access module of the command generation system acquires the next interface content of the next user interface through the storage device, and provides it to the item acquisition module. The speech recognition system as described in Claim 1, wherein the natural speech understanding system includes: a natural language processor, coupled to the speech recognition module, and used to receive the speech information to generate a possible intent grammar data; A knowledge-assisted comprehension module, coupled to the natural language processor, and uses an intent data storing the possible intent grammar data; A retrieval system, coupled to the knowledge-aided understanding module, and used to receive a keyword of the possible intent grammar data provided by the knowledge-assisted understanding module, so as to generate a response result to the knowledge-aided understanding module according to the keyword module, so that the knowledge aided comprehension module generates a definite intention grammar data according to the response result; Determine the intent grammar data to output the semantic analysis result. The speech recognition system as described in claim 11, wherein the retrieval system includes: a retrieval interface unit coupled to the knowledge aided comprehension module; a search engine coupled to the retrieval interface unit; an instruction data storage device coupled to the search engine; and a structured database coupled to the search engine and storing a plurality of records, wherein the retrieval interface unit can search the records in the structured database through the search engine according to the keyword , and according to a search result, a corresponding instruction data is acquired through the instruction data storage device as the response result. The speech recognition system as claimed in claim 12, wherein the search engine performs a full-text search on the records, and determines a domain attribute of the keyword according to a full-text search result. The speech recognition system as described in claim 13, wherein the records each include at least one of a popularity field, a favorite field, and a dislike field One, and the search engine determines the search result by comparing the values of at least one of the popularity field, the like field and the dislike field of the respective records. The speech recognition system as claimed in claim 12, wherein the knowledge aided comprehension module compares the response result and the intent data to generate the definite intent grammar data. An instruction generation system is adapted to communicate with an application system, and the application system is used to receive a voice input, wherein the instruction generation system includes: a comparison module, used to receive a semantic analysis corresponding to the voice input result, and use the semantic analysis result to compare a selected item in an interface content of a current user interface displayed by the application system to generate a comparison result; and an instruction confirmation module, coupled to the The comparison module is used for converting the comparison result according to an instruction format, and outputting a control instruction to the application system. The command generating system as described in claim 16, wherein the interface content of the current user interface includes an item name of the selected item and an object number corresponding to the item name and a plurality of reference keywords, the comparison model and comparing whether the semantic analysis result matches the project name and one of the reference keywords to generate the comparison result. The command generation system as described in claim 17 further includes: a temporary storage device for receiving an interface number of the current user interface provided by the application system; An access module, coupled to the temporary storage device, and used to generate the interface content of the current user interface according to the interface number; and an item acquisition module, coupled to the access module and the comparison module, and used to obtain the item name of the selected item, the item label corresponding to the item name, and the reference keywords from the interface content of the current user interface, so as to output the item of the selected item name and the item label corresponding to the item name and the reference keywords to the comparison module. The command generation system as claimed in claim 18, wherein the temporary storage device is further coupled to the command confirmation module, and the command confirmation module outputs the control command to the application system through the temporary storage device. The command generation system as described in claim 18, wherein a storage device is coupled to the access module of the command generation system, and the access module is used to access the storage device to obtain the current user interface The interface content. The command generating system as described in claim 20, wherein the interface content of the current user interface is pre-written into the storage device through the temporary storage device and the access module. The command generating system as described in claim 20, wherein the command format is written into the storage device through the temporary storage device and the access module in advance, and the command confirmation module obtains the command through the access module Format. The command generation system as described in claim 18, wherein the application system is used to display the current user interface, and when the application system receives the control command output by the command generation system, the application system will instruction to select the selection item in the interface content of the current user interface, and change to display the next user interface or perform a specific operation. The command generation system as described in claim 23, wherein when the application system changes to display the next user interface, the application system outputs the next interface number of the next user interface to the command generation system, so that the The access module of the command generation system acquires the next interface content of the next user interface through the storage device, and provides it to the item acquisition module. A speech recognition method suitable for a speech recognition system, wherein the speech recognition system communicates with an application system, and the application system is used to receive a speech input, wherein the speech recognition method includes: receiving the Voice input; identifying the voice input to generate voice information; understanding the voice information to generate a semantic analysis result; and using the semantic analysis result to compare an interface of a current user interface displayed on the application system An item is selected in the content, and a control command is output to the application system according to a comparison result. The speech recognition method according to claim 25, wherein the step of outputting the control instruction to the application system includes: converting the comparison result according to an instruction format, and outputting the control instruction. The speech recognition method as described in claim 26, wherein the interface content of the current user interface includes an item name of the selected item and an object number corresponding to the item name and a plurality of reference keywords, and the semantic Analyzing the result to compare the selection item of the current user interface to generate the comparison result includes: comparing the semantic analysis result with the item name and one of the reference keywords to generate The comparison result. The voice recognition method as described in claim 27, further comprising: receiving an interface number of the current user interface provided by the application system; generating the interface content of the current user interface according to the interface number; and from the The item name of the selected item, the item label corresponding to the item name, and the reference keywords are obtained from the interface content of the current user interface. The speech recognition method as described in claim 28, wherein the application system is used to display the current user interface, and when the application system receives the control instruction output by the speech recognition system, the application system will command to select the selection item in the interface content of the current user interface, and change to display the next user interface or perform a specific operation. The speech recognition method as described in claim 25, wherein the step of understanding the speech information to generate the semantic analysis result includes: generating a possible intention grammar data according to the speech information; generating a keyword according to the possible intention grammar data A response result; generating a definite intention grammar data according to the response result; and outputting the semantic analysis result according to the definite intention grammar data. The speech recognition method as described in claim item 30, wherein the step of generating the response result according to the keyword includes: searching for a plurality of records according to the keyword, and obtaining a corresponding instruction data according to a search result as the Respond to the result. The speech recognition method as described in claim item 31, wherein the step of searching these records according to the keyword comprises: performing a full-text search on these records, and judging a field attribute of the keyword according to a full-text search result . The speech recognition method as described in claim 32, wherein the step of searching the records according to the keyword further includes: comparing at least One of the values to determine the search results. The speech recognition method as claimed in claim 31, wherein the step of outputting the semantic analysis result according to the definite intention grammar data includes: comparing the response result and the possible intention grammar data to generate the definite intention grammar data. A voice recognition method suitable for an instruction generating system, wherein the instruction generating system is adapted to communicate with an application system, and the application system is used to receive a voice input, wherein the voice recognition method includes: receiving a voice corresponding to the voice input a semantic analysis result; using the semantic analysis result to compare a selected item in an interface content of a current user interface displayed by the application system to generate a comparison result; and confirm the comparison result according to an instruction format, and output a control instruction to the application system. The speech recognition method as described in claim 35, wherein the interface content of the current user interface includes an item name of the selected item and a plurality of reference keywords corresponding to the item name, and the semantic analysis result is used to compare The step of generating the comparison result for the selected item in the current user interface includes: comparing whether the semantic analysis result matches the item name and one of the reference keywords to generate the comparison result . The voice recognition method as described in claim 35, further comprising: receiving an interface number of the current user interface provided by the application system; generating the interface content of the current user interface according to the interface number; and from the Get the selected item from the interface content of the current user interface. The speech recognition method as described in claim item 37, wherein the application system is used to display the current user interface, and when the application system receives the control instruction output by the instruction generation system, the application system according to the control command to select the selection item in the interface content of the current user interface, and change to display the next user interface or perform a specific operation.

Images