Method for operating game via voice, and related apparatus

Abstract

A method for operating a game via voice, and a related apparatus. The method comprises: recognizing a pending voice command in a target game so as to obtain a pending text command; identifying an intent of the pending text command so as to obtain a pending intent, and extracting entities of the pending text command so as to obtain a set of entities to be operated; on the basis of a correspondence between a plurality of game intents in the target game, a plurality of game functions, and a plurality of function parameter sets, determining a function to be called and a set of parameters to be called corresponding to the pending intent; and when the set of entities to be operated matches the set of parameters to be called, on the basis of the set of entities to be operated, operating the target game by calling the function to be called, so as to obtain a voice operation result of the target game.

Description

Voice operation game method and related device

[0001] The present application claims priority to the Chinese patent application No. 2024106561249, filed on May 24, 2024, and entitled "Voice operation game method and related device", the whole content of which is incorporated herein by reference. TECHNICAL FIELD

[0002] The present application relates to the technical field of computer, in particular to voice operation game. BACKGROUND

[0003] At present, the game needs to be operated through multiple levels of interfaces, and the game player needs to be familiar with the interface levels. In addition, at the critical moment of the game, the extra operation of multiple levels of interfaces affects the game experience. Therefore, the voice operation game can be used to replace the operation of multiple levels of interfaces.

[0004] In the related art, the voice operation game method is to define the corresponding relationship between the operation text and the game function in advance, recognize the input voice of the game player to obtain the input text, and if the input text matches the corresponding relationship between the operation text and the game function, call the game function to realize the voice operation game.

[0005] However, in the above method, the game player needs to learn and remember a large number of operation voice corresponding to a large number of operation text, so that the input voice is the operation voice, the input text is the operation text, the requirement for voice input and voice recognition is high, it is difficult to realize the voice operation game, and the game experience is seriously affected. SUMMARY

[0006] In order to solve the above technical problems, the present application provides a voice operation game method and related device, which can simply, quickly and effectively realize the target game to be operated by voice, thereby improving the game experience.

[0007] The present application embodiment discloses the following technical scheme:

[0008] On the one hand, the present application embodiment provides a voice operation game method, which comprises:

[0009] Voice recognition is performed on the target game to be operated by voice to obtain the target text to be operated corresponding to the target voice to be operated;

[0010] Intention recognition and entity extraction are performed on the target text to be operated to obtain the target intention to be operated and the target entity set corresponding to the target text to be operated;

[0011] determine, according to a correspondence among the plurality of game intents, the plurality of game functions, and the plurality of function parameter sets in the target game, a to-be-called function and a to-be-called parameter set corresponding to the to-be-operated intent among the plurality of game functions and the plurality of function parameter sets;

[0012] If the to-be-operated entity set matches the to-be-called parameter set, the to-be-called function is called according to the to-be-operated entity set to perform game operation on the target game, and a voice operation result of the target game is obtained.

[0013] In another aspect, an embodiment of the present application provides a voice operation game device, the device comprising: an identification unit, an extraction unit, a determination unit, and a calling unit;

[0014] The identification unit is configured to perform voice recognition on a to-be-operated voice in a target game, and obtain to-be-operated text corresponding to the to-be-operated voice.

[0015] The extraction unit is configured to perform intent recognition and entity extraction on the to-be-operated text, and obtain a to-be-operated intent and a to-be-operated entity set corresponding to the to-be-operated text.

[0016] The determination unit is configured to determine, according to a correspondence among a plurality of game intents, a plurality of game functions, and a plurality of function parameter sets in the target game, a to-be-called function and a to-be-called parameter set corresponding to the to-be-operated intent among the plurality of game functions and the plurality of function parameter sets.

[0017] The calling unit is configured to, if the to-be-operated entity set matches the to-be-called parameter set, call the to-be-called function according to the to-be-operated entity set to perform game operation on the target game, and obtain a voice operation result of the target game.

[0018] In another aspect, an embodiment of the present application provides a computer device, the computer device comprising a processor and a memory:

[0019] The memory is configured to store a computer program and transmit the computer program to the processor.

[0020] The processor is configured to execute the method according to the instructions in the computer program.

[0021] In another aspect, an embodiment of the present application provides a computer readable storage medium for storing a computer program, when the computer program is run on a computer device, the computer device executes the method according to any one of the preceding aspects.

[0022] In another aspect, an embodiment of the present application provides a computer program product, comprising a computer program which, when running on a computer device, causes the computer device to execute the method of any one of the preceding aspects.

[0023] It can be seen from the above technical solution that, first, the target game to be operated text is obtained by recognizing the target game to be operated voice, which provides text data for subsequent implementation of the voice operation game. Second, the target game to be operated intent is obtained by recognizing the intent of the target game to be operated text, and the target game to be operated entity set is obtained by extracting the entity of the target game to be operated text, which realizes understanding of the natural language target game to be operated text as the target game to be operated intent, and extracts the target game to be operated entity set corresponding to the target game to be operated intent, thereby providing input data for subsequent implementation of the voice operation game. Then, based on the corresponding relationship between the plurality of game intents, the plurality of game functions and the plurality of function parameter sets in the target game, the target game to be called function and the target game to be called parameter set are determined from the plurality of game functions and the plurality of function parameter sets, which can map the target game to be operated intent to the target game to be called function in the plurality of game functions of the target game, and determine the target game to be called parameter set in the plurality of function parameter sets of the target game, thereby providing calling data for subsequent implementation of the voice operation game. Finally, when the target game to be operated entity set matches the target game to be called parameter set, the target game to be called function is called based on the target game to be operated entity set to operate the target game, thereby obtaining the voice operation result of the target game; and the voice operation game is completed when the input data of the voice operation game matches the calling data of the voice operation game. Based on this, based on the corresponding relationship between the plurality of game intents, the plurality of game functions and the plurality of function parameter sets in the target game, the method does not need to learn and memorize a large number of operation texts corresponding to a large number of operation voices, has a lower requirement for voice input and voice recognition, and after recognizing the target game to be operated text from the target game to be operated voice, the target game to be operated text is first understood as the target game to be operated intent, and the target game to be operated entity set corresponding to the target game to be operated intent is extracted, and then mapped to the target game to be called function and the target game to be called parameter set in the target game, and the target game to be called function is called when the target game to be operated entity set matches the target game to be called parameter set, thereby simply, quickly and effectively operating the target game from the target game to be operated voice, and thereby improving the game experience. BRIEF DESCRIPTION OF DRAWINGS

[0024] FIG. 1 is a system schematic diagram of a voice operation game method provided by an embodiment of the present application;

[0025] FIG. 2 is a flowchart of a voice operation game method provided by an embodiment of the present application;

[0026] FIG. 3 is a partial schematic diagram of a target game to be operated intent, a target game to be operated entity set, and a corresponding relationship between a plurality of game intents, a plurality of game functions and a plurality of function parameter sets provided by an embodiment of the present application;

[0027] FIG. 4 is an intent recognition schematic diagram of a target game to be operated text provided by an embodiment of the present application;

[0028] FIG. 5 is a schematic diagram of training an intent recognition model according to an embodiment of the present application;

[0029] FIG. 6 is a schematic diagram of calling a game function according to an embodiment of the present application;

[0030] FIG. 7 is a schematic diagram of correcting a text to be operated according to an embodiment of the present application;

[0031] FIG. 8 is a schematic diagram of an interface of a voice-operated game according to an embodiment of the present application;

[0032] FIG. 9 is a structural diagram of a voice-operated game device according to an embodiment of the present application;

[0033] FIG. 10 is a structural diagram of a server according to an embodiment of the present application;

[0034] FIG. 11 is a structural diagram of a terminal according to an embodiment of the present application. DETAILED DESCRIPTION

[0035] Embodiments of the present application will be described below with reference to the accompanying drawings.

[0036] At present, a game player can operate a game by voice instead of multi-level interface, thereby improving the game experience. That is, on the basis of predefining the correspondence between an operation text and a game function, the game player inputs a voice for the game to obtain an input voice, recognizes the input voice as an input text, and if the input text matches the correspondence between the operation text and the game function, calls the game function to implement the voice-operated game.

[0037] However, it is found through research that the game player needs to learn and memorize a large number of operation voices corresponding to a large number of operation texts in the above voice-operated game process, which is relatively complicated for the game player and requires a higher voice input and voice recognition, resulting in difficulty in implementing the voice-operated game.

[0038] Embodiments of the present application provide a voice-operated game method, which is based on the correspondence among a plurality of game intents, a plurality of game functions and a plurality of function parameter sets in a target game, does not need to learn and memorize a large number of operation voices corresponding to a large number of operation texts, has a lower requirement for voice input and voice recognition, after recognizing a to-be-operated voice as a to-be-operated text, understands the to-be-operated text as a to-be-operated intent, extracts a to-be-operated entity set corresponding to the to-be-operated intent, and maps the to-be-operated entity set as a to-be-called function and a to-be-called parameter set in the target game, calls the to-be-called function when the to-be-operated entity set matches the to-be-called parameter set, thereby simply, quickly and effectively operating the target game by the to-be-operated voice and improving the game experience.

[0039] Next, the system architecture of the voice operation game method will be introduced. Referring to FIG. 1, FIG. 1 is a system schematic diagram of a voice operation game method provided by an embodiment of the present application, the system comprising a computer device 100 configured to execute the voice operation game method.

[0040] The computer device 100 performs voice recognition on the target game voice to be operated to obtain the target game text to be operated corresponding to the voice to be operated. The voice to be operated is a voice input by a game object (i.e., a game player) to the target game for the purpose of operating the target game; the voice recognition refers to recognizing the voice to be operated as a text; and the text to be operated is a text obtained by recognizing the voice to be operated.

[0041] As an example, the target game is a basketball game, and the voice to be operated is "voice: pass the ball to Xiaoming". The computer device 100 performs voice recognition on "voice: pass the ball to Xiaoming" in the basketball game to obtain the text to be operated corresponding to "voice: pass the ball to Xiaoming" as "text: pass the ball to Xiaoming".

[0042] As another example, the target game is a basketball game, and the voice to be operated is "voice: pass the ball to Xiaoming". The computer device 100 performs voice recognition on "voice: pass the ball to Xiaoming" in the basketball game to obtain the text to be operated corresponding to "voice: pass the ball to Xiaoming" as "text: pass the ball to Xiaoming".

[0043] As another example, the target game is a basketball game, and the voice to be operated is "voice: pass the ball to Xiaoming". The computer device 100 performs voice recognition on "voice: pass the ball to Xiaoming" in the basketball game to obtain the text to be operated corresponding to "voice: pass the ball to Xiaoming" as "text: pass the ball to Xiaoming".

[0044] The computer device 100 performs intent recognition and entity extraction on the text to be operated to obtain the target game intent to be operated and the set of target game entities to be operated corresponding to the text to be operated. The intent recognition refers to recognizing the text to be operated as a target game intent; the target game intent to be operated is a target game intent recognized from the text to be operated; the entity extraction refers to extracting a target game entity from the text to be operated; and the set of target game entities to be operated is a set of target game entities extracted from the text to be operated.

[0045] As an example, on the basis of the above example, the computer device 100 performs intent recognition and entity extraction on "text: pass the ball to Xiaoming" to obtain the target game intent to be operated and the set of target game entities to be operated corresponding to "text: pass the ball to Xiaoming" as "intent: pass the ball to a specified player" and "entity: Xiaoming".

[0046] As another example, on the basis of the above example, the computer device 100 performs intent recognition and entity extraction on the text "pass the ball to Xiaoming", and obtains the to-be-operated intent and the to-be-operated entity set corresponding to the text "pass the ball to Xiaoming" as "intent: pass the ball to a specified player" and "entity: Xiaoming".

[0047] As another example, on the basis of the above example, the computer device 100 performs intent recognition and entity extraction on the text "pass the ball to Xiaoming", and obtains the to-be-operated intent and the to-be-operated entity set corresponding to the text "pass the ball to Xiaoming" as "intent: pass the ball to a specified player" and "entity: Xiaoming".

[0048] The computer device 100 determines the to-be-called function and the to-be-called parameter set corresponding to the to-be-operated intent in the plurality of game functions and the plurality of function parameter sets according to the correspondence relationship between the plurality of game intents, the plurality of game functions and the plurality of function parameter sets. Wherein, the correspondence relationship between the plurality of game intents, the plurality of game functions and the plurality of function parameter sets is a mapping relationship between the plurality of game intents, the plurality of game functions and the plurality of function parameter sets constructed in advance; the plurality of game intents are a plurality of intents corresponding to a plurality of game functions implemented in the target game; the plurality of game functions are a plurality of functions corresponding to a plurality of game functions implemented in the target game; the plurality of function parameter sets are a plurality of parameter sets corresponding to a plurality of game functions implemented in the target game; the to-be-called function is a game function corresponding to the to-be-operated intent in the correspondence relationship between the plurality of game intents, the plurality of game functions and the plurality of function parameter sets; and the to-be-called parameter set is a function parameter set corresponding to the to-be-called function corresponding to the to-be-operated intent in the correspondence relationship between the plurality of game intents, the plurality of game functions and the plurality of function parameter sets.

[0049] As an example, on the basis of the above example, the computer device 100 determines the to-be-called function and the to-be-called parameter set corresponding to "intent: pass the ball to a specified player" in the plurality of game functions and the plurality of function parameter sets according to the correspondence relationship between the plurality of game intents, the plurality of game functions and the plurality of function parameter sets in the basketball game, as "function: pass the ball to a specified player PassBallByPlayername" and "parameter: player".

[0050] The computer device 100 calls the to-be-called function according to the to-be-operated entity set to perform game operation on the target game, and obtains a voice operation result of the target game, if the to-be-operated entity set matches the to-be-called parameter set. The to-be-operated entity set matches the to-be-called parameter set means that the to-be-operated entity in the to-be-operated entity set matches the to-be-called parameter in the to-be-called parameter set. The game operation means that the to-be-operated voice operation target game is realized by calling the to-be-operated entity set to call the to-be-called function. The voice operation result of the target game is the operation result of the to-be-operated voice operation target game realized by calling the to-be-operated entity set to call the to-be-called function.

[0051] As an example, on the basis of the above example, if "entity: Xiaoming" matches "parameter: player", the computer device 100 calls "function: PassBallByPlayername" to perform game operation on the basketball game according to "entity: Xiaoming", and obtains a voice operation result of the basketball game.

[0052] That is, the to-be-operated text is obtained by recognizing the to-be-operated voice in the target game, to provide text data for subsequent implementation of voice operation game; the to-be-operated intent is obtained by recognizing the intent of the to-be-operated text, and the to-be-operated entity set is obtained by extracting the entity of the to-be-operated text, to realize understanding of the to-be-operated text in natural language as the to-be-operated intent, and extraction of the to-be-operated entity set corresponding to the to-be-operated intent, to provide input data for subsequent implementation of voice operation game; based on the corresponding relationship among the plurality of game intents, the plurality of game functions and the plurality of function parameter sets in the target game, the to-be-called function corresponding to the to-be-operated intent and the to-be-called parameter set are determined from the plurality of game functions and the plurality of function parameter sets, to be able to map the to-be-operated intent to the to-be-called function in the plurality of game functions of the target game, and determine the to-be-called parameter set in the plurality of function parameter sets of the target game, to provide calling data for subsequent implementation of voice operation game; when the to-be-operated entity set matches the to-be-called parameter set, the to-be-called function is called based on the to-be-operated entity set to operate the target game, to obtain a voice operation result of the target game; the voice operation game is implemented when the input data of the voice operation game matches the calling data of the voice operation game.

[0053] It should be noted that, in the embodiments of the present application, the computer device can be a server or a terminal, and the method provided by the embodiments of the present application can be executed by the terminal or the server alone, or by the terminal and the server together. When the method provided by the embodiments of the present application is executed by the terminal or the server alone, the execution method is similar to the corresponding embodiment of FIG. 1, and the computer device is replaced by the terminal or the server. In addition, when the method provided by the embodiments of the present application is executed by the terminal and the server together, the steps that need to be reflected on the front-end interface can be executed by the terminal, and some steps that need to be calculated in the background and do not need to be reflected on the front-end interface can be executed by the server.

[0054] The terminal can be a smartphone, a tablet computer, a notebook computer, a desktop computer, a smart voice interaction device, a vehicle-mounted terminal, an extended reality device, an aircraft, or the like, but is not limited thereto. The server can be a standalone physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server providing cloud computing services, but is not limited thereto. The terminal and the server can be directly or indirectly connected through wired or wireless communication, which is not limited in the present application. For example, the terminal and the server can be connected through a network, which can be a wired or wireless network.

[0055] In addition, the embodiments of the present application can be applied to various scenarios, including but not limited to cloud technology, artificial intelligence, intelligent transportation, assisted driving, autonomous driving, digital people, virtual people, virtual reality, augmented reality, mixed reality, audio and video, and games.

[0056] The game involved in the present application includes but is not limited to role-playing game (RPG), action game (ACT), adventure game (AVG), strategy game (SLG), simulation game (SIM), shooting game, sports game (SPG), puzzle game (PUZ), music game, fighting game, and racing game. Specific introductions are as follows:

[0057] Role-playing game: game players play one or more roles in the game, and promote the development of the plot by completing tasks, exploring the world, and interacting with non-player characters (NPCs). RPG usually contains rich plot and character growth system, allowing game players to deeply experience the game world.

[0058] Action game: a game characterized by fast, precise, and continuous action responses controlled by game players. Action games usually contain elements such as jumping, attacking, and dodging. Action games can include multiplayer online battle arena (MOBA).

[0059] Adventure game: a game characterized by exploration, puzzle solving, and story. Game players need to play game characters, solve puzzles, complete tasks, and interact with NPCs to promote the progress of the game.

[0060] Strategy game: a game in which players use strategies to compete against the computer or other players to achieve various forms of victory. Strategy games usually require players to develop plans, allocate resources, and use tactics to achieve goals. Strategy games can include turn-based strategy games and real-time strategy games.

[0061] Simulation game: a game that simulates a certain situation or role in real life. For example, games such as city simulation, business simulation, and driving simulation belong to the simulation game category.

[0062] Shooting game: a shooting game played in first-person or third-person perspective, i.e., shooting games include first-person shooting games (FPS) and third-person shooting games (TPS). Players need to control game characters to shoot, aim, and dodge to defeat other players and complete tasks.

[0063] Sports game: a game that can simulate various sports. Sports games can include soccer games, basketball games, racing games, etc.

[0064] Puzzle game: a game that mainly features solving puzzles and intellectual challenges.

[0065] The presentation of the game includes but is not limited to two-dimensional (2-dimension, 2D) form, three-dimensional (3-dimension, 3D) form, virtual reality (virtual reality, VR) form, augmented reality (augmented reality, AR) form and mixed reality (mixed reality, MR) form. The game can be a player versus player (player versus player, PVP) game or a player versus environment (player versus enviroment, PVE) game. Next, the speech operation game method provided by the embodiments of the present application will be described in detail with the computer device executing the method provided by the embodiments of the present application as an example and in combination with the accompanying drawings.

[0066] Referring to FIG. 2, which is a flowchart of a speech operation game method provided by an embodiment of the present application, the method comprises:

[0067] S201: performing speech recognition on the speech to be operated in the target game to obtain the text to be operated corresponding to the speech to be operated.

[0068] In order to realize speech operation game, first, the speech issued by the game player for operating the game needs to be converted into text in the target game, so as to call the game function in the game through the text to operate the game subsequently.

[0069] Based on this, in the embodiment of the application, any game is taken as a target game, the voice of the game object (i.e. the game player) inputting an operation target game for the target game is taken as a to-be-operated voice, and the text recognized from the to-be-operated voice is taken as to-be-operated text, so as to realize subsequent voice operation of the game.

[0070] Since the to-be-operated voice and the to-be-operated text are processed in the application, for example, the to-be-operated voice is an audio form "pass the ball to Xiaoming", and the to-be-operated text is a text form "pass the ball to Xiaoming". In order to accurately realize the distinguishing description, in subsequent embodiments, the writing method of "A: B" is adopted, wherein A is used to identify the expression type of the content, such as voice and text, and B is used to identify the text information corresponding to the content, such as passing the ball to Xiaoming. For example, the expression meaning of "voice: pass the ball to Xiaoming" is that the content (such as the to-be-operated voice) of the voice type corresponds to the text of passing the ball to Xiaoming.

[0071] The S201 recognizes the to-be-operated voice input by the game player for operating the target game into to-be-operated text, provides text data for subsequent operation of the target game through the to-be-operated voice, so as to realize subsequent voice operation of the game.

[0072] As an example of S201, the target game is a basketball game, and the to-be-operated voice is "voice: pass the ball to Xiaoming". The voice recognition is performed on "voice: pass the ball to Xiaoming" in the basketball game, and the to-be-operated text corresponding to "voice: pass the ball to Xiaoming" is "text: pass the ball to Xiaoming".

[0073] As another example of S201, the target game is a basketball game, and the to-be-operated voice is "voice: use q four". The voice recognition is performed on "voice: use q four" in the basketball game, and the to-be-operated text corresponding to "voice: use q four" is "text: use q four".

[0074] S202: Intention recognition and entity extraction are performed on the to-be-operated text to obtain a to-be-operated intention and a to-be-operated entity set corresponding to the to-be-operated text.

[0075] S203: According to the corresponding relationship between a plurality of game intentions and a plurality of game functions in the target game, a to-be-called function corresponding to the to-be-operated intention in the plurality of game functions is determined.

[0076] In the related art, a corresponding relationship between operation text and a game function is predefined, a game player needs to learn and memorize a large number of operation voices corresponding to a large number of operation texts, input text is obtained by recognizing an input voice of the game player, if the input text matches the corresponding relationship between the operation text and the game function, that is, the input voice is the operation voice and the input text is the operation text, the game function is called to implement voice operation game; the method is relatively complicated to learn and memorize a large number of operation voices corresponding to a large number of operation texts, and has a high requirement for voice input and voice recognition, and it is difficult to implement voice operation game, which seriously affects the game experience.

[0077] Therefore, in the embodiments of the present application, in order to avoid the game player from learning and memorizing a large number of operation voices corresponding to a large number of operation texts, a corresponding relationship between a plurality of game functions and a plurality of intents, a plurality of functions and a plurality of parameter sets corresponding to the plurality of game functions can be constructed in advance, that is, a corresponding relationship between a plurality of game intents and a plurality of game functions is constructed in advance.

[0078] That is, in the present application, through the understanding of natural language, as long as the corresponding game intent (such as the operation intent) can be recognized from the operation text, the game operation of the target game can be implemented through the voice. Therefore, the player object does not need to memorize the operation password completely consistent with the game operation, but can express the game intent through any voice content. Taking the game operation of passing the ball as an example, in the related art, if the game operation of passing the ball to Xiaoming is implemented through the voice, the corresponding predefined operation text is "make a pass action to the player Xiaoming", in this case, the player object must say the content of the operation text word by word to implement the game operation, in the present application, as long as the player object says the content that can express the pass to Xiaoming, the game operation can be implemented, for example, the player object can say "pass the ball to Xiaoming", "pass the ball to Xiaoming", "pass the ball to Xiaoming", and other voices expressing the same game intent can implement the game operation of passing the ball to Xiaoming which can only be implemented based on the voice "make a pass action to the player Xiaoming".

[0079] Based on this, the game player inputs the natural language voice of the operation target game as the operation voice, recognizes the operation voice as the natural language text, recognizes the natural language text obtained by recognizing the operation voice as the operation text, and then recognizes the operation text as the game intent in the target game, extracts the game entity in the operation text, recognizes the game intent in the target game obtained by recognizing the operation text as the operation intent, and extracts the set of game entities obtained by recognizing the operation text as the operation entity set.

[0080] On the basis of the correspondence among the multiple game intents, the multiple game functions, and the multiple function parameter sets in the target game, the game function corresponding to the to-be-operated intent is taken as the to-be-called function, and the function parameter set corresponding to the determined game function is taken as the to-be-called parameter set, so that a large amount of operation texts corresponding to a large amount of operation voices without the need of memorization can be simply, quickly, and effectively realized, the to-be-operated text in natural language corresponding to the to-be-operated voice in natural language is understood as the to-be-operated intent, and the to-be-operated entity set corresponding to the to-be-operated intent is extracted, and the to-be-operated intent is mapped to the to-be-called function in the multiple game functions of the target game, and the to-be-called parameter set in the multiple function parameter sets of the target game is determined.

[0081] The S202 and the S203 identify the to-be-operated text as the to-be-operated intent of the game intent in the target game, and extract the game entity in the to-be-operated text to form the to-be-operated entity set, so as to understand the to-be-operated text in natural language as the to-be-operated intent, and extract the to-be-operated entity set corresponding to the to-be-operated intent, and provide input data for subsequent implementation of voice operation of the game; through the correspondence among the multiple game intents, the multiple game functions, and the multiple function parameter sets in the target game, the game function corresponding to the to-be-operated intent is determined as the to-be-called function, and the function parameter set corresponding to the game function corresponding to the to-be-operated intent is determined as the to-be-called parameter set, so as to map the to-be-operated intent to the to-be-called function in the multiple game functions of the target game, and determine the to-be-called parameter set in the multiple function parameter sets of the target game, and provide calling data for subsequent implementation of voice operation of the game, so as to smoothly implement voice operation of the game subsequently.

[0082] As an example of the S202 and the S203, on the basis of the example of the S201 described above, the intent recognition and entity extraction are performed on the text “pass the ball to Xiaoming”, and the to-be-operated intent and the to-be-operated entity set corresponding to the text “pass the ball to Xiaoming” are obtained as “intent: pass the ball to a specified player” and “entity: Xiaoming”; considering that the to-be-operated entity set includes the name of the to-be-operated entity and the type of the to-be-operated entity, the name of the to-be-operated entity is the complete expression of the game entity extracted from the to-be-operated text, and the type of the to-be-operated entity is the category to which the game entity extracted from the to-be-operated text belongs, the to-be-operated entity set can also be “entity: name: Xiaoming, type: player”. According to the correspondence among the multiple game intents, the multiple game functions, and the multiple function parameter sets in the basketball game, the to-be-called function and the to-be-called parameter set corresponding to “intent: pass the ball to a specified player” in the multiple game functions and the multiple function parameter sets are determined as “function: pass the ball to a specified player PassBallByPlayername” and “parameter: player”.

[0083] As another example of S202 and S203, based on another example of S201, the text "use q four" is subjected to intent recognition and entity extraction, and the operation-intent and operation-entity set corresponding to the text "use q four" are obtained as "intent: call tactics" and "entity: name: QUICK 4HORNS FLARE, type: tactic name", wherein "QUICK 4HORNS FLARE" is the complete expression of "q four". According to the correspondence between the multiple game intents, multiple game functions, and multiple function parameter sets, the function to be called and the parameter set to be called corresponding to the intent "call tactics" in the multiple game functions and multiple function parameter sets are determined as "function: call tactics CallStrategy" and "parameter: tactic name".

[0084] Referring to FIG. 3, FIG. 3 is a partial schematic diagram of the correspondence between the operation-intent, operation-entity set, and multiple game intents, multiple game functions, and multiple function parameter sets provided by an embodiment of the present application. FIG. 3 illustrates that the operation-intent is "intent: pass the ball to a specified player", and the operation-entity set is "entity: name: Xiaoming, type: player". In the correspondence between the multiple game intents, multiple game functions, and multiple function parameter sets, the game intent corresponding to the intent "pass the ball to a specified player" is "intent: pass the ball to a specified player", the game function corresponding to the intent "pass the ball to a specified player" is "function: pass the ball to a specified player PassBallByPlayername", and the function parameter set corresponding to the function "pass the ball to a specified player PassBallByPlayername" is "parameter: player".

[0085] FIG. 3 illustrates that the operation-intent is "intent: call tactics", and the operation-entity set is "entity: name: QUICK 4HORNS FLARE, type: tactic name". In the correspondence between the multiple game intents, multiple game functions, and multiple function parameter sets, the game intent corresponding to the intent "call tactics" is "intent: call tactics", the game function corresponding to the intent "call tactics" is "function: call tactics CallStrategy", and the function parameter set corresponding to the function "call tactics CallStrategy" is "parameter: tactic name".

[0086] S204: calling the function to be called according to the operation-entity set to perform game operation on the target game, and obtaining the voice operation result of the target game.

[0087] In some possible implementation manners, a correspondence relationship between the game function and the function parameter set can also be set. The function parameter set can be used to match whether the to-be-operated entity set meets the to-be-called function, so as to avoid an abnormal execution of a game operation caused by calling the game function using the to-be-operated entity set that does not meet the requirement of the game function.

[0088] Correspondingly, S203 can include: determining, according to the correspondence relationship between the multiple game intents, the multiple game functions, and the multiple function parameter sets in the target game, the to-be-called function and the to-be-called parameter set corresponding to the to-be-operated intent in the multiple game functions and the multiple function parameter sets.

[0089] S204 can include: if the to-be-operated entity set matches the to-be-called parameter set, calling the to-be-called function according to the to-be-operated entity set to perform a game operation on the target game, and obtaining a voice operation result of the target game.

[0090] In the embodiments of the present application, after the to-be-operated intent is obtained by performing the above-mentioned S202-S203 to identify the intent of the to-be-operated text, and the to-be-operated entity set is obtained by extracting the entity of the to-be-operated text, and the to-be-called function and the to-be-called parameter set corresponding to the to-be-operated intent are determined from the multiple game functions and the multiple function parameter sets based on the correspondence relationship between the multiple game intents, the multiple game functions, and the multiple function parameter sets in the target game, it is further needed to consider whether the to-be-operated entity set matches the to-be-called parameter set, that is, to judge whether the to-be-operated entity in the to-be-operated entity set matches the to-be-called parameter in the to-be-called parameter set. If yes, it indicates that the to-be-operated entity in the to-be-operated entity set can be transmitted into the to-be-called parameter in the to-be-called parameter set to call the to-be-called function, and then the to-be-operated voice operation target game is implemented by calling the to-be-called function through the to-be-operated entity set, and an operation result of implementing the to-be-operated voice operation target game by calling the to-be-called function through the to-be-operated entity set is obtained as the voice operation result of the target game.

[0091] It should be noted that the above is only one implementation manner. When the game function does not have the corresponding function parameter set, S204 can be directly implemented after the to-be-called function is determined.

[0092] The S204 takes the to-be-operated intent and the to-be-operated entity set corresponding to the to-be-operated text as input data of the voice operation game, and takes the to-be-called function and the to-be-called parameter set corresponding to the to-be-operated intent as calling data of the voice operation game. When the input data of the voice operation game matches the calling data of the voice operation game, the to-be-called function is called through the to-be-operated entity set, so that the to-be-operated voice operation target game can be simply, quickly, and effectively implemented.

[0093] As an example of S204, based on the example of S201-S203 described above, if "entity: Xiaoming" matches "parameter: player", a game operation is performed on the basketball game according to "entity: Xiaoming" calling "function: PassBallByPlayername" to obtain the voice operation result of the basketball game. As another example of S204, based on another example of S201-S203 described above, if "entity: QUICK 4HORNS FLARE" matches "parameter: strategy name", a game operation is performed on the basketball game according to "entity: QUICK 4HORNS FLARE" calling "function: CallStrategy" to obtain the voice operation result of the basketball game.

[0094] As can be seen from the above technical solutions, the text to be operated is obtained by recognizing the target game voice to be operated, which provides text data for subsequent implementation of voice operation games. The intent to be operated is obtained by recognizing the intent to be operated, and the entity set to be operated is obtained by extracting the entity to be operated, which realizes the understanding of the natural language text to be operated as the intent to be operated, and extracts the entity set to be operated corresponding to the intent to be operated, which provides input data for subsequent implementation of voice operation games. Based on the corresponding relationship between the multiple game intents, multiple game functions and multiple function parameter sets in the target game, the calling function to be operated and the parameter set to be called corresponding to the intent to be operated are determined from the multiple game functions and the multiple function parameter sets, which can map the intent to be operated to the calling function to be operated in the multiple game functions of the target game, and determine the calling parameter set to be operated in the multiple function parameter sets of the target game, which provides calling data for subsequent implementation of voice operation games. When the entity set to be operated matches the parameter set to be called, the target game is operated based on the entity set to be operated calling the calling function to be operated, and the voice operation result of the target game is obtained; the voice operation game is implemented when the input data of the voice operation game matches the calling data of the voice operation game. Based on this, based on the corresponding relationship between the multiple game intents, multiple game functions and multiple function parameter sets in the target game, there is no need to maintain a large number of corresponding relationships between operation texts and game functions, and there is no need to learn and remember a large number of operation texts corresponding to a large number of operation voices. The requirement for voice input and voice recognition is low. After recognizing the voice to be operated as the text to be operated, the text to be operated is first understood as the intent to be operated, and the entity set to be operated corresponding to the intent to be operated is extracted, and then mapped to the calling function to be operated and the parameter set to be called in the target game. When the entity set to be operated matches the parameter set to be called, the calling function to be operated is called, the target game is operated, and the voice operation result of the target game is obtained, thereby simply, quickly and effectively implementing the voice operation of the target game, and improving the game experience.

[0095] In the above embodiments, when S202 is specifically implemented, the intention of the to-be-operated text is identified to obtain a to-be-operated intention, and the entity of the to-be-operated text is extracted to obtain a to-be-operated entity set, which provides input data for subsequent implementation of the voice operation game, is used to determine the calling data of the voice operation game and implement the voice operation game, therefore, it is necessary to guarantee the recognition accuracy of the to-be-operated intention and the extraction accuracy of the to-be-operated entity set. Based on this, the following implementation manner can be used to accurately identify the intention of the to-be-operated text to obtain a to-be-operated intention, and extract the entity of the to-be-operated text to obtain a to-be-operated entity set.

[0096] One implementation manner is as follows: first, in order to understand the to-be-operated text in natural language as a to-be-operated intention, so as to subsequently map the to-be-operated intention to a to-be-called function in a plurality of game functions of a target game, the to-be-operated text and a plurality of game intentions can be matched through an intention matching strategy, the matching degree between the to-be-operated text and the plurality of game intentions is calculated, and a plurality of first confidence degrees of the to-be-operated text corresponding to the plurality of game intentions are obtained; since the greater the first confidence degree is, the more matched the to-be-operated text is with the corresponding game intention, therefore, the game intention corresponding to the maximum confidence degree in the plurality of first confidence degrees can be taken as the to-be-operated intention. Then, in order to extract a to-be-operated entity set corresponding to the to-be-operated intention, so as to subsequently call the to-be-called function by inputting a to-be-called parameter set in a plurality of function parameter sets of the target game, the to-be-operated text can be processed through a natural language processing technology, and entity words in the to-be-operated text are extracted to obtain a first entity set corresponding to the to-be-operated text; since the first entity set is obtained by directly performing natural language processing on the to-be-operated text, the first entity in the first entity set can not belong to a game entity set of the target game, therefore, the first entity set can be subjected to entity screening based on the game entity set of the target game, and entity words in the first entity set that belong to the game entity set of the target game are screened to obtain the to-be-operated entity set. Based on this, the present application provides a possible implementation manner, and S202 includes the following S202a-S202d (not shown in the figure). In the present embodiment, the present application embodiment also provides a plurality of manners of obtaining the to-be-operated entity set, and in the present embodiment, the manner one of obtaining the to-be-operated entity set is explained through S202c and S202d.

[0097] S202a: The to-be-operated text and a plurality of game intentions are matched according to an intention matching strategy, and a first confidence degree of the to-be-operated text corresponding to the plurality of game intentions is obtained. One first confidence degree corresponds to one game intention, and is used to identify the degree to which the to-be-operated text conforms to the game intention.

[0098] S202b: The to-be-operated intention is determined according to the game intention corresponding to the maximum confidence degree in the plurality of first confidence degrees.

[0099] S202c: performing natural language processing on the to-be-operated text to obtain a first entity set corresponding to the to-be-operated text.

[0100] S202d: performing entity screening on the first entity set according to a game entity set of the target game to obtain a to-be-operated entity set.

[0101] The intent matching strategy is a pre-defined strategy for calculating whether the to-be-operated text matches the multiple game intents. The intent matching strategy can include a regular expression strategy or a text content similarity strategy. The regular expression strategy refers to calculating whether the to-be-operated text matches the multiple game intents by using a regular expression. The text content similarity strategy refers to calculating whether the to-be-operated text matches the multiple game intents by using a text content similarity. The matching processing refers to calculating the matching degree between the to-be-operated text and the multiple game intents. The first confidence degree refers to the matching degree between the to-be-operated text and the corresponding game intent. The natural language processing refers to extracting entities in the to-be-operated text. The first entity set is a set of entities extracted from the to-be-operated text. The game entity set of the target game refers to a set of multiple game entities associated with the target game. The entity screening refers to screening entities in the first entity set that belong to the game entity set of the target game.

[0102] The S202a-S202d calculate multiple first confidence degrees of the to-be-operated text corresponding to the multiple game intents by using the pre-defined intent matching strategy. The game intent corresponding to the maximum confidence degree in the multiple first confidence degrees best reflects the game intent of the to-be-operated text. The game intent corresponding to the maximum confidence degree is determined as the to-be-operated intent of the to-be-operated text, thereby improving the accuracy of the determined to-be-operated intent. The first entity set corresponding to the to-be-operated text is obtained by performing natural language processing on the to-be-operated text. The first entity set is screened by using the pre-obtained game entity set of the target game, thereby obtaining a relatively accurate to-be-operated entity set. That is, the to-be-operated intent and the to-be-operated entity set are determined by using the pre-defined intent matching strategy, natural language processing, and entity screening. The intent recognition and entity extraction are realized, thereby guaranteeing the accuracy of the intent recognition and entity extraction and improving the accuracy of the to-be-operated intent and the to-be-operated entity set in the case that the text data associated with the target game is less.

[0103] As an example of S202a-S202b, the text to be operated is "Text: Move the position of Little Red", the regular expression of the plurality of game intents "Intent: Move the character" includes "move", "Move" and "walk", and the intent matching strategy is the regular expression strategy; according to the regular expression strategy, the "Text: Move the position of Little Red" and the plurality of game intents are matched to obtain a plurality of first confidence degrees of the "Text: Move the position of Little Red" corresponding to the plurality of game intents, wherein the first confidence degree of the "Text: Move the position of Little Red" corresponding to the "Intent: Move the character" is "0.8"; the maximum confidence degree in the plurality of first confidence degrees is "0.8", and according to the game intent corresponding to "0.8" being "Intent: Move the character", it is determined that the "Text: Move the position of Little Red" corresponds to the "Intent: Move the character".

[0104] As an example of S202c-S202d, the game entity set includes "Entity: QUICK 4HORNS FLARE", and the text to be operated is "Text: Use q four"; the "Text: Use q four" is processed by natural language processing to obtain a first entity set corresponding to the "Text: Use q four" as "Entity: q four"; according to the game entity set of the target game, the "Entity: q four" is filtered to obtain a to-be-operated entity set as "Entity: QUICK 4HORNS FLARE".

[0105] Another implementation manner is that: first, in order to understand the natural language to-be-operated text as a to-be-operated intention, so as to subsequently map the to-be-operated intention to a to-be-called function in a plurality of game functions of a target game, the to-be-operated text can be input into a pre-trained intention recognition model for intention recognition, that is, to recognize the matching degree between the to-be-operated text and a plurality of game intentions, and output a plurality of second confidences of the plurality of game intentions corresponding to the to-be-operated text; since the greater the second confidence is, the more matched the to-be-operated text and the corresponding game intention are, therefore, the game intention corresponding to the maximum confidence in the plurality of second confidences can be taken as the to-be-operated intention. Then, in order to extract a to-be-operated entity set corresponding to the to-be-operated intention, so as to subsequently transmit the to-be-operated entity set into a to-be-called parameter set in a plurality of function parameter sets of the target game to call the to-be-called function, the to-be-operated text can be input into a pre-trained entity extraction model for entity extraction, to extract entity words in the to-be-operated text, and output a second entity set corresponding to the to-be-operated text; since the second entity set is obtained by using the entity extraction model for entity extraction, a second entity in the second entity set can not belong to a game entity set of the target game, therefore, the second entity set can be subjected to entity screening based on the game entity set of the target game, to screen the second entity in the second entity set that belongs to the game entity set of the target game, and obtain the to-be-operated entity set. Based on this, the present application provides another possible implementation manner, and S202 includes the following S202e-S202h (not shown in the figure). Wherein, the present application embodiment also provides a plurality of manners of obtaining the to-be-operated entity set, and in the present embodiment, the manner two of obtaining the to-be-operated entity set is explained by S202g and S202h.

[0106] S202e: performing intention recognition on the to-be-operated text by using the intention recognition model, to obtain second confidences of the to-be-operated text corresponding to a plurality of game intentions respectively. Wherein, one second confidence corresponds to one game intention, and is used to identify the degree to which the to-be-operated text conforms to the game intention.

[0107] S202f: determining the to-be-operated intention according to the game intention corresponding to the maximum confidence in the second confidences.

[0108] S202g: performing entity extraction on the to-be-operated text by using the entity extraction model, to obtain a second entity set corresponding to the to-be-operated text.

[0109] S202h: performing entity screening on the second entity set according to the game entity set of the target game, to obtain a to-be-operated entity set.

[0110] The intent recognition model is a pre-trained model for calculating whether the to-be-operated text matches a plurality of game intents, and the second confidence degree refers to the matching degree between the to-be-operated text and the corresponding game intent. The entity extraction model is a pre-trained model for extracting entities in the to-be-operated text. The second entity set is a set of second entities extracted from the to-be-operated text. The entity screening refers to screening entities in the second entity set that belong to the game entity set of the target game.

[0111] The intent recognition model and the entity extraction model can be a recurrent neural network or other neural network. Specifically, the intent recognition model and the entity extraction model can be a Bi-directional Long Short-Term Memory (BiLSTM), an Attention Bi-directional Long Short-Term Memory (AttBiLSTM), a Transformer, or a Text Convolutional Neural Network (TextCNN).

[0112] The S202e-S202h obtains a plurality of second confidence degrees corresponding to a plurality of game intents of the to-be-operated text by using the pre-trained intent recognition model. The game intent corresponding to the maximum confidence degree in the plurality of second confidence degrees best reflects the game intent of the to-be-operated text. The game intent corresponding to the maximum confidence degree is determined as the to-be-operated intent of the to-be-operated text, thereby improving the accuracy of the determined to-be-operated intent. The second entity set corresponding to the to-be-operated text is obtained by using the pre-trained entity extraction model, and the second entity set is screened by using the pre-obtained game entity set of the target game, thereby obtaining a relatively accurate to-be-operated entity set. That is, the to-be-operated intent and the to-be-operated entity set are determined by using the pre-trained intent recognition model and the entity extraction model, intent recognition and entity extraction are realized, thereby quickly realizing intent recognition and entity extraction in the case of a large number of game intents and a large number of operation texts, and the accuracy of the to-be-operated intent and the to-be-operated entity set is improved.

[0113] As an example of S202e-S202f: the text to be operated is "text: pass the ball to Xiaoming", the intent recognition model is used to perform intent recognition on "text: pass the ball to Xiaoming", and a plurality of second confidences corresponding to a plurality of game intents of "text: pass the ball to Xiaoming" are obtained, wherein the second confidence of "text: pass the ball to Xiaoming" corresponding to "intent: pass the ball to the specified player" is "0.9"; the maximum confidence in the plurality of second confidences is "0.9", and according to the game intent corresponding to "0.9" being "intent: pass the ball to the specified player", it is determined that the text to be operated "text: pass the ball to Xiaoming" corresponds to the intent to be operated "intent: pass the ball to the specified player".

[0114] As an example of S202g-S202h, the game entity set includes "entity: Xiaoming", and the text to be operated is "text: pass the ball to Xiaoming"; the entity extraction model is used to perform entity extraction on "text: pass the ball to Xiaoming", and the second entity set corresponding to "text: pass the ball to Xiaoming" is obtained as "entity: Xiaoming"; according to the game entity set of the target game, the entity "entity: Xiaoming" is screened, and the entity set to be operated is obtained as "entity: Xiaoming".

[0115] In another implementation, first, in order to understand the natural language to-be-operated text as a to-be-operated intent, so as to subsequently map the to-be-operated intent to a to-be-called function in a plurality of game functions of a target game; on one hand, the to-be-operated text and the plurality of game intents can be matched through an intent matching strategy, the matching degree between the to-be-operated text and the plurality of game intents is calculated, and a plurality of first confidences of the to-be-operated text corresponding to the plurality of game intents are obtained; on the other hand, the to-be-operated text can also be input into a pre-trained intent recognition model for intent recognition, that is, the matching degree between the to-be-operated text and the plurality of game intents is recognized, and a plurality of second confidences of the to-be-operated text corresponding to the plurality of game intents are output; since the maximum confidence in the first confidence and the second confidence indicates that the to-be-operated text is most matched with the corresponding game intent, the game intent corresponding to the maximum confidence in the plurality of first confidences and the plurality of second confidences can be taken as the to-be-operated intent. Then, in order to extract a to-be-operated entity set corresponding to the to-be-operated intent, so as to subsequently call the to-be-called function by inputting a to-be-called parameter set in a plurality of function parameter sets of the target game; on one hand, the to-be-operated text can be processed through a natural language processing technology, and entity words in the to-be-operated text are extracted, to obtain a first entity set corresponding to the to-be-operated text; on the other hand, the to-be-operated text can also be input into a pre-trained entity extraction model for entity extraction, and entity words in the to-be-operated text are extracted, to output a second entity set corresponding to the to-be-operated text; since the first entity set is obtained by directly performing natural language processing on the to-be-operated text, a first entity in the first entity set can not belong to a game entity set of the target game, and correspondingly, the second entity set is obtained by performing entity extraction on the to-be-operated text by using the entity extraction model, and a second entity in the second entity set can also not belong to the game entity set of the target game, therefore, the first entity set and the second entity set can be subjected to entity screening based on the game entity set of the target game, to screen entity words in the first entity set that belong to the game entity set of the target game, and to screen the second entity in the second entity set that belongs to the game entity set of the target game, to obtain the to-be-operated entity set. Based on this, the present application provides another possible implementation, and S202 includes the following S202i-S202n (not shown in the figure). In the present embodiment, the present application provides a plurality of ways to obtain the to-be-operated entity set, and in the present embodiment, the way one of obtaining the to-be-operated entity set is explained through S202i-S202n.

[0116] S202i: The to-be-operated text and the plurality of game intents are matched according to an intent matching strategy, to obtain a first confidence of the to-be-operated text corresponding to the plurality of game intents. One first confidence corresponds to one game intent, and is used to identify the degree to which the to-be-operated text conforms to the game intent.

[0117] S202j: performing intent recognition on the to-be-operated text by using an intent recognition model to obtain second confidences of the to-be-operated text corresponding to multiple game intents respectively. One second confidence corresponds to one game intent, and is used to identify a degree to which the to-be-operated text meets the game intent.

[0118] S202k: determining the to-be-operated intent according to a game intent corresponding to a maximum confidence in the first confidences and the second confidences.

[0119] S202l: performing natural language processing on the to-be-operated text to obtain a first entity set corresponding to the to-be-operated text.

[0120] S202m: performing entity extraction on the to-be-operated text by using an entity extraction model to obtain a second entity set corresponding to the to-be-operated text.

[0121] S202n: performing entity screening on the first entity set and the second entity set according to a game entity set of the target game to obtain a to-be-operated entity set.

[0122] The S202i-S202n obtain multiple first confidences and multiple second confidences of the to-be-operated text corresponding to multiple game intents by using a pre-defined intent matching strategy and a pre-trained intent recognition model. A game intent corresponding to a maximum confidence in the multiple first confidences and the multiple second confidences can best reflect a game intent of the to-be-operated text. The game intent corresponding to the maximum confidence is determined as the to-be-operated intent of the to-be-operated text, thereby further improving the accuracy of the determined to-be-operated intent. The first entity set and the second entity set corresponding to the to-be-operated text are obtained by using natural language processing and a pre-trained entity extraction model. The first entity set and the second entity set are jointly screened by using a pre-obtained game entity set of the target game, thereby obtaining a relatively accurate to-be-operated entity set. That is, the pre-defined intent matching strategy and the pre-trained intent recognition model can further improve the speed and accuracy of intent recognition. The natural language processing and the pre-trained entity extraction model are used to jointly perform entity extraction, thereby further improving the accuracy of entity extraction.

[0123] Referring to FIG. 4, FIG. 4 is a schematic diagram of intent recognition of a text to be operated according to an embodiment of the present application. As shown in FIG. 4, the intent recognition is performed by using an intent matching strategy and an intent recognition model respectively. The game intents that can be matched by the intent matching strategy include intent 1 to intent n, and the game intents that can be recognized by the intent recognition model also include intent 1 to intent n. The intents 1 to intent n can be distinguished by using the intent identifier, and n is a positive integer greater than 1. The n first confidences of the text to be operated and the intents 1 to intent n are obtained by using the intent matching strategy, and the n second confidences of the text to be operated and the intents 1 to intent n are obtained by using the intent recognition model. The maximum confidence in the n first confidences and the n second confidences is the second confidence 0.95 of the intent n corresponding to the text to be operated, and therefore, the intent n corresponding to the second confidence 0.95 is determined as the operation intent.

[0124] In the above embodiment, the obtaining step of the intent recognition model in S202e and S202j refers to: first, obtaining the training sample of the trained intent recognition model, that is, obtaining the text to be trained of the target game and the training intent corresponding to the text to be trained; then, the intent recognition of the text to be trained can be performed by using the pre-trained language model, the prediction intent corresponding to the text to be trained is obtained, and whether the training intent corresponding to the text to be trained is correct is measured. When the prediction intent does not match the training intent, it indicates that the training intent does not correctly correspond to the game intent of the text to be trained. The training intent can be adjusted, and the text to be trained is used as the input of the first recognition model, and the adjusted training intent is used as the training target of the first recognition model, so as to train the first recognition model by using the text to be trained and the adjusted training intent, and finally obtain the intent recognition model with high accuracy.

[0125] The specific training process refers to: obtaining the text to be trained of the target game and the training intent corresponding to the text to be trained; inputting the text to be trained into the pre-trained language model for intent recognition, and outputting the prediction intent corresponding to the text to be trained; when the prediction intent does not match the training intent, adjusting the training intent to obtain the adjusted training intent; inputting the text to be trained into the first recognition model for intent recognition, and outputting the intent recognition result of the text to be trained; calculating the loss between the intent recognition result and the adjusted training intent by using the loss function of the first recognition model, adjusting the model parameters of the first recognition model by using the loss, until the loss no longer decreases or reaches the preset iteration number, determining that the training of the first recognition model is completed, and taking the trained first recognition model as the intent recognition model. Therefore, the present application provides a possible implementation manner of the obtaining step of the intent recognition model in S202e and S202j, which may, for example, include S1-S4 (not shown in the figure).

[0126] S1: Obtain the to-be-trained text of a target game and a to-be-trained intent corresponding to the to-be-trained text.

[0127] S2: Perform intent recognition on the to-be-trained text according to a pre-trained language model to obtain a predicted intent corresponding to the to-be-trained text.

[0128] S3: If the predicted intent does not match the to-be-trained intent, perform intent adjustment on the to-be-trained intent to obtain an adjusted to-be-trained intent.

[0129] S4: Perform model training on the first recognition model according to the to-be-trained text and the adjusted to-be-trained intent to obtain an intent recognition model.

[0130] The S1-S4 perform intent recognition on the to-be-trained text by using the pre-trained language model to obtain the predicted intent, which is the correct game intent corresponding to the to-be-trained text. If the predicted intent does not match the to-be-trained intent, it means that the to-be-trained intent is not the correct game intent corresponding to the to-be-trained text, and the to-be-trained intent needs to be adjusted to be the correct game intent corresponding to the to-be-trained text. Then, the model training is performed by using the adjusted to-be-trained intent and the to-be-trained text. Since the adjusted to-be-trained intent and the to-be-trained text are accurate training samples, the model training is performed by using the accurate training samples, thereby improving the accuracy of the intent recognition model.

[0131] Referring to FIG. 5, which is a training diagram of an intent recognition model provided by an embodiment of the present application. The to-be-trained text and the to-be-trained intent corresponding to the to-be-trained text are stored in an intent training database as training samples of the intent recognition model. The to-be-trained text and the to-be-trained intent corresponding to the to-be-trained text can be determined according to the historical operation text of a target game and the historical operation intent corresponding to the historical operation text. The pre-trained language model is a large language model. The large language model is used to perform intent recognition on the to-be-trained text to obtain the predicted intent corresponding to the to-be-trained text. If it is determined that the predicted intent does not match the to-be-trained intent, the to-be-trained intent is adjusted. The adjusted to-be-trained intent and the to-be-trained text are updated and stored in the intent training database. The model training is performed by using the to-be-trained text and the adjusted to-be-trained intent to obtain the intent recognition model.

[0132] In the above embodiments, in order to perform entity screening on the first entity set or the second entity set by using the game entity set of the target game to determine the to-be-operated entity set related to the target game and correctly and completely expressed, the game entity set of the target game is constructed by extracting entities in the game text associated with the target game. Based on this, the present application provides a possible implementation manner. The obtaining step of the game entity set of the target game in S202d, S202h and S202n includes the following S5 (not shown in the figure):

[0133] S5: performing entity extraction on the game text associated with the target game to obtain a game entity set.

[0134] In the method, the entity extraction on the game text associated with the target game can be performed by using a pre-trained model, i.e., the game text is input into the pre-trained model, the pre-trained model is used to extract entities included in the game text and relationship information between the entities, and the entities and the relationship information between the entities are used to form the game entity set. The entities and the relationship information between the entities are stored in a graph database, the entities are used as nodes of the graph database, and the relationship information between the entities is used as edges of the graph database.

[0135] The game entity set obtained by performing the entity extraction on the game text associated with the target game is associated with the target game, has high accuracy, and is convenient for subsequent accurate screening of entities in the first entity set or the second entity set by using the game entity set, so as to determine the target game and the entity set with the correct name to be operated, and improve the accuracy of the entity set to be operated.

[0136] As an example, the game text associated with the target game is one or more of game popular science, game hot post and game announcement.

[0137] In the above embodiment, when S204 is specifically implemented, the target game is implemented by calling the to-be-called function by using the entity set to be operated. When the to-be-called function is called, the entity to be operated in the entity set to be operated is not directly used as parameter data of a to-be-called parameter in a to-be-called parameter set corresponding to the to-be-called function. The entity to be operated in the entity set to be operated needs to be converted into executable parameter data based on the to-be-called function, so as to obtain an executable function parameter data set corresponding to the entity set to be operated. After the entity set to be operated is converted into the executable function parameter data set, whether the to-be-called function can be called to implement the target game in the voice operation according to the function parameter data set also needs to be determined. If the function parameter data set matches the game scene of the target game in the voice operation, it indicates that the to-be-called function can be called to operate the target game in the game scene in the voice operation according to the function parameter data set, that is, the voice operation result of the target game is obtained. Based on this, the present application provides a possible implementation manner, and S204 includes the following S2041-S2042 (not shown in the figure).

[0138] S2041: performing data conversion on the entity set to be operated according to the to-be-called function to obtain a function parameter data set corresponding to the entity set to be operated.

[0139] S2042: If the function parameter data set matches the game scene of the target game under the to-be-operated voice, the to-be-called function is called according to the function parameter data set to perform game operation on the target game, and a voice operation result is obtained.

[0140] The function parameter data set can be transmitted to the to-be-called parameter set to call the to-be-called function, so as to call the to-be-called function through the function parameter data set, thereby realizing the operation of the target game by the to-be-operated voice.

[0141] The S2041-S2042 can perform data conversion on the to-be-operated entity set according to the to-be-called function, so as to obtain the executable function parameter data set corresponding to the to-be-operated entity set, and determine whether the function parameter data set matches the game scene of the target game under the to-be-operated voice. After determining that the function parameter data set matches the game scene of the target game under the to-be-operated voice, it is determined that the function parameter data set is executable and reasonable under the current game scene. Then, the to-be-called function can be called according to the function parameter data set to perform game operation on the target game. The operation result of the target game by the to-be-operated voice through the function parameter data set calling the to-be-called function is used as the voice operation result of the target game.

[0142] As an example of S2041-S2042, on the basis of the example of S204, the entity "Xiaoming" is data-converted according to the function "PassBallByPlayername", and the function parameter data set corresponding to the entity "Xiaoming" is obtained. "Parameter data: 38", wherein 38 is the player ID of Xiaoming. If the basketball game has a player Xiaoming under the voice "pass the ball to Xiaoming", that is, "parameter data: 38" matches the game scene of the basketball game under the voice "pass the ball to Xiaoming", the function "PassBallByPlayername" is called according to "parameter data: 38" to perform game operation on the basketball game, and the voice operation result of the basketball game is obtained.

[0143] As another example of S2041-S2042, based on the example of S204 described above, the data conversion is performed on the entity QUICK 4HORNS FLARE according to the function: CallStrategy, and the function parameter data set corresponding to the entity QUICK 4HORNS FLARE is obtained as parameter data: A, where A is the enumeration value of the tactic name QUICK 4HORNS FLARE. If the basketball game in the game scene under the voice: use q four can trigger QUICK 4HORNS FLARE, that is, the parameter data: A matches the game scene of the basketball game under the voice: use q four, the function: CallStrategy is called according to the parameter data: A to perform game operation on the basketball game, and the voice operation result of the basketball game is obtained.

[0144] In the above embodiment, when S204 is implemented, considering the case that the set of entities to be operated does not match the set of parameters to be called, and the case that the set of function parameter data does not match the game scene of the target game under the voice to be operated, at this time, the function to be called cannot be called for voice operation, and it is determined that the voice operation to be operated fails. Based on this, the present application provides a possible implementation manner, S204 can also include S6-S7 (not shown in the figure).

[0145] S6: If the set of entities to be operated does not match the set of parameters to be called, it is determined that the voice operation to be operated fails.

[0146] S7: If the set of function parameter data does not match the game scene of the target game under the voice to be operated, it is determined that the voice operation to be operated fails.

[0147] Wherein, the set of entities to be operated and the set of parameters to be called do not match, that is, the entities to be operated in the set of entities to be operated do not match the parameters to be called in the set of parameters to be called, which means that the entities to be operated in the set of entities to be operated cannot be passed into the parameters to be called in the set of parameters to be called to call the function to be called, and the set of entities to be operated cannot be used to call the function to be called, and the target game cannot be operated by the voice, and it is determined that the voice operation to be operated fails.

[0148] The S6 and S7 are implemented when the set of entities to be operated does not match the set of parameters to be called and the set of function parameter data does not match the game scene of the target game under the voice to be operated, and the result of the voice operation to be operated is fed back to the game player, thereby realizing the result feedback of the voice operation game.

[0149] Referring to FIG. 6, FIG. 6 is a schematic diagram of a game function call provided by an embodiment of the present application. Based on the above examples of S2041-S2042, after obtaining the to-be-operated intent "intent: pass the ball to the specified player" and the to-be-operated entity set "entity: Xiaoming", the to-be-called function "function: pass the ball to the specified player PassBallByPlayername" and the to-be-called parameter set "parameter: player" are determined according to the correspondence between the plurality of game intents, the plurality of game functions and the plurality of function parameter sets; "entity: Xiaoming" is passed into "parameter: player" to call "function: pass the ball to the specified player PassBallByPlayername"; the function parameter data set "parameter data: 38" corresponding to "entity: Xiaoming" is obtained by data conversion of "entity: Xiaoming" according to "function: pass the ball to the specified player PassBallByPlayername", wherein 38 is the player ID of Xiaoming; if the player Xiaoming exists in the game scene of the basketball game under the voice "pass the ball to Xiaoming", that is, "parameter data: 38" matches the game scene of the basketball game under the voice "pass the ball to Xiaoming", the game operation of the basketball game is performed according to "parameter data: 38" calling "function: pass the ball to the specified player PassBallByPlayername", and the voice operation result of the basketball game is obtained. If the player Xiaoming does not exist in the game scene of the basketball game under the voice "pass the ball to Xiaoming", that is, "parameter data: 38" does not match the game scene of the basketball game under the voice "pass the ball to Xiaoming", it is determined that the voice "pass the ball to Xiaoming" operation fails.

[0150] In the above embodiment, when S202 is specifically implemented, considering that the plurality of to-be-operated words in the to-be-operated text obtained by voice recognition of the to-be-operated voice may have string confusion or may not be the exclusive words of the target game, in order to improve the accuracy of intent recognition and entity extraction of the to-be-operated text, a plurality of corresponding relationships between confused strings and an exclusive word set of the target game can also be constructed in advance. Based on this, first, the to-be-operated text is segmented according to the lengths of the plurality of words to obtain a plurality of to-be-operated word sets corresponding to the to-be-operated text; second, the plurality of to-be-operated word sets are corrected to obtain a plurality of corrected word sets corresponding to the to-be-operated text through the plurality of corresponding relationships between the confused strings and the exclusive word set of the target game; third, the corrected text corresponding to the to-be-operated text is determined based on the plurality of corrected word sets and the plurality of word lengths corresponding to the plurality of corrected word sets; and finally, the game intent in the target game is recognized from the corrected text, and the game entity in the corrected text is extracted, the game intent in the target game recognized from the corrected text is taken as the to-be-operated intent, and the set of game entities extracted from the corrected text is taken as the to-be-operated entity set.

[0151] In the present application, the confused strings are strings that are easily misidentified, for example, string a and string b are easily misidentified as each other, such as string a is misidentified as string b, or string b is misidentified as string a, so string a and string b can be determined as confused strings. Confused strings with a corresponding relationship generally belong to strings with the same or similar pronunciation, for example, the aforementioned string a and string b can be confused strings with a corresponding relationship.

[0152] Therefore, the present application provides a possible implementation manner, S202 includes the following S2021-S2024 (not shown in the figure).

[0153] S2021: performing segmentation processing on the to-be-operated text according to a plurality of segmentation lengths, to obtain a plurality of to-be-operated segmentation sets corresponding to the to-be-operated text.

[0154] S2022: performing error correction processing on the plurality of to-be-operated segmentation sets according to a corresponding relationship between the plurality of confused strings and an exclusive vocabulary set of a target game, to obtain a plurality of corrected segmentation sets respectively corresponding to the plurality of to-be-operated segmentation sets.

[0155] S2023: determining a corrected text corresponding to the to-be-operated text according to the corrected segmentation sets and segmentation lengths of the corresponding to-be-operated segmentation sets.

[0156] S2024: performing intent recognition and entity extraction on the corrected text, to obtain a to-be-operated intent and a to-be-operated entity set corresponding to the corrected text.

[0157] Among them, the plurality of segmentation lengths are determined according to a multiple n-gram model (nGram) manner; the plurality of to-be-operated segmentation sets correspond to the plurality of segmentation lengths, and each to-be-operated segmentation set refers to a set formed by a plurality of to-be-operated segments with the same segmentation length; the corresponding relationship between the plurality of confused strings refers to a mapping relationship between one string that is confused to another string; the exclusive vocabulary set of the target game refers to a set of exclusive vocabularies associated with the target game; the corrected segmentation set refers to a to-be-operated segmentation set after correction; and the corrected text refers to a to-be-operated text after correction.

[0158] The S2021-S2024 respectively perform word segmentation processing on the to-be-operated text through multiple word segmentation lengths to obtain multiple to-be-operated word segmentation sets, thereby providing a data basis for subsequent comprehensive error correction on the to-be-operated text; the error correction processing of the multiple to-be-operated word segmentation sets is implemented through the corresponding relationship between the multiple confused strings and the exclusive vocabulary set of the target game, thereby avoiding the case that the multiple corrected word segmentation sets contain confused strings and exclusive vocabulary of non-target games; since the corrected word segmentation in the multiple corrected word segmentation sets has different word segmentation lengths, the to-be-operated text corresponding to the corrected text is re-determined through the multiple corrected word segmentation sets and the word segmentation lengths corresponding to the multiple corrected word segmentation sets, so that the text accuracy of the corrected text is relatively high; based on this, when the corrected text is used for intent recognition and entity extraction, the accuracy of the to-be-operated intent and the to-be-operated entity set is also relatively high.

[0159] As an example of S2021-S2023, the word segmentation lengths are 1 and 2; the corresponding relationship between the multiple confused strings includes the corresponding relationship between "confused string: channel" and "confused string: pindown"; and the to-be-operated text is "text: execute channel". According to 1 and 2, the "text: execute channel" is respectively subjected to word segmentation processing to obtain multiple to-be-operated word segmentation sets including to-be-operated word segmentation set 1 and to-be-operated word segmentation set 2; wherein the to-be-operated word segmentation set 1 is "word segmentation: execute, channel", and the to-be-operated word segmentation set 2 is "word segmentation: execute, channel". According to the corresponding relationship between the multiple confused strings and the exclusive vocabulary set of the target game, the "word segmentation: execute, channel" and "word segmentation: execute, channel" are subjected to error correction processing to obtain multiple corrected word segmentation sets including corrected word segmentation set 1 and corrected word segmentation set 2; wherein the corrected word segmentation set 1 is "word segmentation: execute, channel", and the corrected word segmentation set 2 is "word segmentation: execute, channel". According to "word segmentation: execute, channel" and the word segmentation length 1, and "word segmentation: execute, channel" and the word segmentation length 2, the corrected text is determined to be "text: execute pindown".

[0160] As another example of S2021-S2023, the tokenization length is 1 and 2; the correspondence between the plurality of confused strings includes the correspondence between the confused string "I saw" and the confused string "ISO"; and the text to be operated is "text: I saw". According to the tokenization processing of "text: I saw" with 1 and 2 respectively, the plurality of to-be-operated token sets includes to-be-operated token set 3 and to-be-operated token set 4; wherein the to-be-operated token set 3 is "token: I, saw", and the to-be-operated token set 4 is "token: I saw". According to the correspondence between the plurality of confused strings and the exclusive vocabulary set of the target game, the error correction processing is performed on "token: I, saw" and "token: I saw", and the plurality of corrected token sets includes corrected token set 3 and corrected token set 4; wherein the corrected token set 3 is "token: I, saw", and the corrected token set 4 is "token: ISO". According to "token: I, saw" and the tokenization length 1, "token: ISO" and the tokenization length 2, it is determined that the corrected text is "text: ISO".

[0161] In the above embodiment, when S2022 is specifically implemented, considering that the error correction priority of the correspondence between the plurality of confused strings is higher than the error correction priority of the exclusive vocabulary set of the target game, for each to-be-operated token in the plurality of to-be-operated token sets, it is first judged whether the to-be-operated token matches the correspondence between the plurality of confused strings. If yes, the to-be-operated token corresponding corrected token can be directly determined through the correspondence between the plurality of confused strings; if not, further considering that the to-be-operated token may also exist in English and Chinese mixed form, the plurality of corresponding relationships between non-pinyin and pinyin can be constructed in advance, the to-be-operated token is converted into a pinyin string, and the to-be-operated token is corrected through the pinyin string corresponding to each exclusive vocabulary in the exclusive vocabulary set of the target game to obtain the corrected token, so as to improve the error correction efficiency and accuracy of the to-be-operated token. Based on this, the present application provides a possible implementation manner, and S2022 includes S2022a-S2022c (not shown in the figure).

[0162] S2022a: For each to-be-operated token in the plurality of to-be-operated token sets, if the to-be-operated token matches the correspondence between the plurality of confused strings, the to-be-operated token corresponding corrected token is determined according to the correspondence between the plurality of confused strings.

[0163] S2022b: If the to-be-operated token does not match the correspondence between the plurality of confused strings, the to-be-operated token is mapped to a pinyin string according to the plurality of corresponding relationships between non-pinyin and pinyin, and the pinyin string corresponding to the to-be-operated token is obtained.

[0164] S2022c: correcting the pinyin string corresponding to the to-be-operated segmented word according to the pinyin string corresponding to each exclusive term in the exclusive term set, to obtain the corrected segmented word corresponding to the to-be-operated segmented word.

[0165] The corresponding relationship between the plurality of non-pinyin and pinyin refers to a pre-set mapping relationship between the plurality of non-pinyin and pinyin, and the non-pinyin refers to a plurality of languages other than pinyin. In addition, it is also necessary to pre-map each exclusive term in the exclusive term set of the target game according to the corresponding relationship between the plurality of non-pinyin and pinyin to obtain the pinyin string corresponding to each exclusive term in the exclusive term set.

[0166] The S2022a-S2022c can not only improve the efficiency of error correction processing and realize efficient error correction of the to-be-operated text, but also improve the accuracy of the to-be-operated text by first performing error correction processing on the confused string and then using the corresponding relationship between the non-pinyin and the pinyin for error correction processing.

[0167] In the above embodiment, when S2022c is specifically implemented, the pinyin string corresponding to the to-be-operated segmented word is corrected using the pinyin string corresponding to the exclusive term. It is necessary to first consider whether the pinyin string corresponding to the to-be-operated segmented word has confused pinyin, and then consider whether the pinyin string corresponding to the to-be-operated segmented word and the pinyin string corresponding to the exclusive term have similarity. Based on this, the corresponding relationship between a plurality of confused pinyin can be pre-constructed. When the length of the pinyin string corresponding to the to-be-operated segmented word is the same as the length of the pinyin string corresponding to the exclusive term, and the pinyin string corresponding to the to-be-operated segmented word matches the corresponding relationship between the plurality of confused pinyin, the confused pinyin string corresponding to the to-be-operated segmented word is determined through the corresponding relationship between the plurality of confused pinyin, the first similarity between the confused pinyin string corresponding to the to-be-operated segmented word and the pinyin string corresponding to the exclusive term is calculated, and when the first similarity is greater than the similarity threshold, it indicates that the similarity between the confused pinyin string corresponding to the to-be-operated segmented word and the pinyin string corresponding to the exclusive term is high, and the to-be-operated segmented word needs to be corrected. The corrected segmented word corresponding to the to-be-operated segmented word can be determined through the exclusive term. When the length of the pinyin string corresponding to the to-be-operated segmented word is different from the length of the pinyin string corresponding to the exclusive term, or the pinyin string corresponding to the to-be-operated segmented word does not match the corresponding relationship between the plurality of confused pinyin, the second similarity between the pinyin string corresponding to the to-be-operated segmented word and the pinyin string corresponding to the exclusive term is directly calculated, and when the second similarity is greater than the similarity threshold, it indicates that the similarity between the pinyin string corresponding to the to-be-operated segmented word and the pinyin string corresponding to the exclusive term is high, and the to-be-operated segmented word needs to be corrected. The corrected segmented word corresponding to the to-be-operated segmented word is determined through the exclusive term. Based on this, the present application provides a possible implementation manner, and S2022c includes the following S8-S12 (not shown in the figure).

[0168] S8: If the length of the pinyin string corresponding to the to-be-operated word is the same as the length of the pinyin string corresponding to the exclusive word, and the pinyin string corresponding to the to-be-operated word matches the corresponding relationship between the plurality of confused pinyins, the corresponding confused pinyin string of the to-be-operated word is determined according to the corresponding relationship between the plurality of confused pinyins.

[0169] S9: Similarity calculation is performed on the confused pinyin string corresponding to the to-be-operated word and the pinyin string corresponding to the exclusive word to obtain a first similarity.

[0170] S10: If the first similarity is greater than a similarity threshold, the to-be-operated word corresponding to the exclusive word is determined.

[0171] S11: If the length of the pinyin string corresponding to the to-be-operated word is different from the length of the pinyin string corresponding to the exclusive word, or the pinyin string corresponding to the to-be-operated word does not match the corresponding relationship between the plurality of confused pinyins, similarity calculation is performed on the pinyin string corresponding to the to-be-operated word and the pinyin string corresponding to the exclusive word to obtain a second similarity.

[0172] S12: If the second similarity is greater than the similarity threshold, the to-be-operated word corresponding to the exclusive word is determined.

[0173] The confused pinyin can be an initial or a final, and the corresponding relationship between the plurality of confused pinyins is a mapping relationship between the plurality of confused initials or finals that are easy to confuse.

[0174] The S8-S12, when performing error correction processing on the pinyin string corresponding to the to-be-operated word by using the pinyin string corresponding to the exclusive word, when the length of the pinyin string corresponding to the to-be-operated word is the same as the length of the pinyin string corresponding to the exclusive word, first determines whether the pinyin string corresponding to the to-be-operated word matches the corresponding relationship between the plurality of confused pinyins to determine whether the pinyin string corresponding to the to-be-operated word needs to be converted into the confused pinyin string corresponding to the to-be-operated word, if so, similarity between the confused pinyin string corresponding to the to-be-operated word and the pinyin string corresponding to the exclusive word is calculated to determine whether to perform error correction processing on the to-be-operated word; when the length of the pinyin string corresponding to the to-be-operated word is different from the length of the pinyin string corresponding to the exclusive word, or the pinyin string corresponding to the to-be-operated word does not match the corresponding relationship between the plurality of confused pinyins, similarity between the pinyin string corresponding to the to-be-operated word and the pinyin string corresponding to the exclusive word is directly calculated to determine whether to perform error correction processing on the to-be-operated word, which can improve the error correction accuracy of the to-be-operated text.

[0175] As an example, the corresponding relationship between the plurality of confused pinyins includes one or more of the corresponding relationship between q and x, and the corresponding relationship between zh and z.

[0176] In the above embodiment, the correspondence between the plurality of confused strings, the correspondence between the plurality of non-pinyin and pinyin, and the correspondence between the plurality of confused pinyins are used to measure whether the to-be-operated word segmentation needs to be corrected, and therefore, the correspondence needs to be obtained from the speech recognition error data of the target game. Based on this, the present application provides a possible implementation manner, and the obtaining steps of the correspondence between the plurality of confused strings, the correspondence between the plurality of non-pinyin and pinyin, and the correspondence between the plurality of confused pinyins include the following S13-S14 (not shown in the figure).

[0177] S13: obtaining the plurality of confused strings, the plurality of non-pinyin and pinyin, and the plurality of confused pinyins from the speech recognition error data of the target game.

[0178] S14: determining the correspondence between the plurality of confused strings, the correspondence between the plurality of non-pinyin and pinyin, and the correspondence between the plurality of confused pinyins according to the plurality of confused strings, the plurality of non-pinyin and pinyin, and the plurality of confused pinyins.

[0179] The S13-S14 obtains the plurality of confused strings, the plurality of non-pinyin and pinyin, and the plurality of confused pinyins from the speech recognition error data of the target game, and determines the correspondence between the plurality of confused strings by the plurality of confused strings, determines the correspondence between the plurality of non-pinyin and pinyin by the plurality of non-pinyin and pinyin, and determines the correspondence between the plurality of confused pinyins by the plurality of confused pinyins; which facilitates subsequent implementation of accurately correcting the to-be-operated word segmentation in the to-be-operated text by using the correspondence between the plurality of confused strings, the correspondence between the plurality of non-pinyin and pinyin, and the correspondence between the plurality of confused pinyins, so as to reduce the probability of the to-be-operated text appearing the same error, thereby improving the correction accuracy of the to-be-operated text.

[0180] In the above embodiment, when S2023 is specifically implemented, considering that the lengths of the corrected word segmentation in the plurality of corrected word segmentation sets are different, and there is a containing relationship between different corrected word segmentations in different corrected word segmentation sets, the corrected word segmentation with longer length further considers the context content compared with the corrected word segmentation with shorter length, in order to avoid the conflict situation when the different corrected word segmentation sets with the containing relationship determine the corrected text, for the case that the first word segmentation belongs to the second word segmentation in the plurality of corrected word segmentation sets, and the length of the first word segmentation is less than the length of the second word segmentation, the corrected text is determined by the second word segmentation. Based on this, the present application provides a possible implementation manner, and S2023 includes S2023a (not shown in the figure): if the first word segmentation belongs to the second word segmentation in the plurality of corrected word segmentation sets, and the length of the first word segmentation is less than the length of the second word segmentation, the corrected text is determined according to the second word segmentation.

[0181] wherein the first segmented word and the second segmented word belong to different corrected segmented word sets, and the first segmented word belongs to the second segmented word, i.e., the second segmented word includes the first segmented word.

[0182] The S2023a selects the second segmented word with a longer length to determine the corrected text when the length of the first segmented word is less than the length of the second segmented word and the first segmented word belongs to the second segmented word, thereby improving the accuracy of the corrected text and further improving the accuracy of the corrected text.

[0183] Referring to FIG. 7, which is a schematic diagram of text error correction, the text error correction is divided into two stages, i.e., an error detection stage and a text correction stage. In the error detection stage, the text to be operated is segmented according to the lengths of multiple segmented words to obtain multiple segmented word sets corresponding to the text to be operated, and whether the segmented words to be operated need to be corrected is detected according to the corresponding relationship between multiple confused strings, the exclusive word set of the target game, the corresponding relationship between multiple non-pinyin and pinyin, and the corresponding relationship between multiple confused pinyin. In the correction stage, the segmented words to be operated are corrected according to the corresponding relationship between multiple confused strings, the exclusive word set of the target game, the corresponding relationship between multiple non-pinyin and pinyin, and the corresponding relationship between multiple confused pinyin, and multiple corrected segmented word sets are obtained to determine the corrected text.

[0184] In the above embodiment, when S201 is specifically implemented, it is considered that the voice operation target game usually uses the exclusive voice corresponding to the exclusive text of the target game. Therefore, in order to improve the voice recognition accuracy of the voice to be operated, a voice recognition model can be obtained by training a second recognition model according to the training voice associated with the target game and the training text corresponding to the training voice, and then the voice to be operated is input into the voice recognition model for voice recognition, and the text to be operated corresponding to the voice to be operated is output. Based on this, the present application provides a possible implementation manner, S201 includes S2011 (not shown in the figure): S2011: voice recognition of the voice to be operated is performed by a voice recognition model to obtain the text to be operated; the voice recognition model is obtained by training a second recognition model according to the training voice associated with the target game and the training text corresponding to the training voice.

[0185] The voice recognition model in the above S2011 is trained by using the training voice associated with the target game and the training text corresponding to the training voice, i.e., there are enough exclusive texts and exclusive voices corresponding to the target game as the training basis when training the voice recognition model, and the recognition accuracy of the exclusive texts and exclusive voices corresponding to the target game is high when subsequently performing voice recognition on the voice to be recognized, thereby finally improving the game experience of the voice operation game.

[0186] Referring to FIG. 8, FIG. 8 is a game interface schematic diagram of a voice operation game provided by an embodiment of the present application. The target game is a basketball game, and the game interface includes a game field, multiple game characters, game items, and multiple game controls. The multiple game controls include a shooting control, a passing control, a tactic control, a moving control, and a voice control. A game player can trigger a voice operation game process by clicking the voice control.

[0187] It should be noted that, on the basis of the implementation manners provided by the above aspects, further combinations can be made to provide more implementation manners.

[0188] Based on the voice operation game method provided by the corresponding embodiment of FIG. 2, an embodiment of the present application further provides a voice operation game device. Referring to FIG. 9, FIG. 9 is a structural diagram of a voice operation game device provided by an embodiment of the present application. The voice operation game device 900 includes an identification unit 910, an extraction unit 920, a determination unit 930, and a calling unit 940.

[0189] The identification unit 910 is configured to perform voice recognition on a voice to be operated in a target game, and obtain a text to be operated corresponding to the voice to be operated.

[0190] The extraction unit 920 is configured to perform intent recognition and entity extraction on the text to be operated, and obtain a set of entities to be operated corresponding to the text to be operated.

[0191] The determination unit 930 is configured to determine, according to a corresponding relationship between multiple game intents and multiple game functions in the target game, a function to be called corresponding to the intent to be operated in the multiple game functions.

[0192] The calling unit 940 is configured to call the function to be called according to the set of entities to be operated to perform game operation on the target game, and obtain a voice operation result of the target game.

[0193] In a possible implementation manner, the extraction unit 920 is configured to:

[0194] perform matching processing on the text to be operated and the multiple game intents according to an intent matching strategy, and obtain a first confidence degree corresponding to the multiple game intents respectively.

[0195] determine the intent to be operated according to a game intent corresponding to a maximum confidence degree in the first confidence degree.

[0196] In a possible implementation manner, the extraction unit 920 is configured to:

[0197] perform intent recognition on the text to be operated by using an intent recognition model, and obtain a second confidence degree corresponding to the multiple game intents respectively.

[0198] According to the game intent corresponding to the maximum confidence in the second confidence, the operation intent to be operated is determined.

[0199] In a possible implementation, the extraction unit 920 is configured to:

[0200] According to the intent matching strategy, the text to be operated and the plurality of game intents are matched to obtain a first confidence of the text to be operated corresponding to the plurality of game intents respectively;

[0201] The text to be operated is subjected to intent recognition by using the intent recognition model to obtain a second confidence of the text to be operated corresponding to the plurality of game intents respectively;

[0202] According to the game intent corresponding to the maximum confidence in the first confidence and the second confidence, the operation intent to be operated is determined.

[0203] In a possible implementation, the voice operation game device 900 further includes a training unit.

[0204] The training unit is configured to:

[0205] Obtain the text to be trained of the target game and the operation intent to be trained corresponding to the text to be trained;

[0206] According to the pre-trained language model, the text to be trained is subjected to intent recognition to obtain a predicted intent corresponding to the text to be trained;

[0207] If the predicted intent does not match the operation intent to be trained, the operation intent to be trained is subjected to intent adjustment to obtain an adjusted operation intent to be trained;

[0208] According to the text to be trained and the adjusted operation intent to be trained, the first recognition model is subjected to model training to obtain the intent recognition model.

[0209] In a possible implementation, the determination unit 930 is further configured to:

[0210] According to the correspondence among the plurality of game intents, the plurality of game functions, and the plurality of function parameter sets in the target game, a function to be called and a parameter set to be called corresponding to the operation intent to be operated in the plurality of game functions and the plurality of function parameter sets are determined;

[0211] The game operation on the target game by using the function to be called according to the set of entities to be operated to obtain a voice operation result of the target game, includes:

[0212] If the set of entities to be operated matches the parameter set to be called, the game operation on the target game by using the function to be called according to the set of entities to be operated is performed to obtain a voice operation result of the target game.

[0213] In a possible implementation, the extraction unit 920 is further configured to obtain the entity set to be operated in any of the following manners:

[0214] Manner one:

[0215] performing natural language processing on the text to be operated to obtain a first entity set corresponding to the text to be operated;

[0216] performing entity screening on the first entity set according to the game entity set of the target game to obtain the entity set to be operated;

[0217] Manner two:

[0218] performing entity extraction on the text to be operated by using an entity extraction model to obtain a second entity set corresponding to the text to be operated;

[0219] performing entity screening on the second entity set according to the game entity set of the target game to obtain the entity set to be operated;

[0220] Manner three:

[0221] performing natural language processing on the text to be operated to obtain a first entity set corresponding to the text to be operated;

[0222] performing entity extraction on the text to be operated by using an entity extraction model to obtain a second entity set corresponding to the text to be operated;

[0223] performing entity screening on the first entity set and the second entity set according to the game entity set of the target game to obtain the entity set to be operated.

[0224] In a possible implementation, the extraction unit 920 is further configured to:

[0225] perform entity extraction on a game text associated with the target game to obtain a game entity set.

[0226] In a possible implementation, the calling unit 940 is configured to:

[0227] perform data conversion on the entity set to be operated according to the function to be called to obtain a function parameter data set corresponding to the entity set to be operated;

[0228] if the function parameter data set matches a game scene of the target game in the voice to be operated, calling the function to be called according to the function parameter data set to perform game operation on the target game, and obtaining a voice operation result.

[0229] In a possible implementation, the determination unit 930 is further configured to:

[0230] If the set of entities to be operated on does not match the set of parameters to be invoked, the voice operation to be operated is determined to have failed; or,

[0231] If the function parameter data set does not match the game scene of the target game under the voice command to be performed, the voice command to be performed is determined to have failed.

[0232] In one possible implementation, the extraction unit 920 is used for:

[0233] The text to be operated on is segmented according to multiple segmentation lengths to obtain multiple sets of segmented words corresponding to the text to be operated on.

[0234] Based on the correspondence between multiple obfuscated strings and the target game's specific word set, multiple word segmentation sets to be operated on are corrected to obtain the corrected word segmentation sets corresponding to each of the multiple word segmentation sets to be operated on.

[0235] Based on the segmentation lengths of multiple corrected word segmentation sets and the corresponding word segmentation sets to be operated on, determine the corrected text corresponding to the text to be operated on;

[0236] The intent recognition and entity extraction are performed on the corrected text to obtain the set of intents and entities to be operated on corresponding to the corrected text.

[0237] In one possible implementation, the extraction unit 920 is used for:

[0238] For each word segment to be operated on in multiple sets of word segments to be operated on, if the word segment to be operated on matches the correspondence between multiple obfuscated strings, the corrected word segment corresponding to the word segment to be operated on is determined according to the correspondence between the multiple obfuscated strings;

[0239] If the word segment to be operated on does not match the correspondence between multiple obfuscated strings, the word segment to be operated on is mapped to pinyin according to the correspondence between multiple non-pinyin and pinyin, and the pinyin string corresponding to the word segment to be operated on is obtained;

[0240] Based on the pinyin string corresponding to each exclusive word in the exclusive word set, the pinyin string corresponding to the word to be segmented is corrected to obtain the corrected word segment.

[0241] In one possible implementation, the extraction unit 920 is used for:

[0242] If the length of the pinyin string corresponding to the word to be segmented is the same as the length of the pinyin string corresponding to the exclusive word, and the pinyin string corresponding to the word to be segmented matches the correspondence between multiple confused pinyin, the confused pinyin string corresponding to the word to be segmented is determined according to the correspondence between the multiple confused pinyin;

[0243] The similarity between the to-be-operated segmented word corresponding to the confused pinyin string and the exclusive word corresponding to the pinyin string is calculated to obtain a first similarity;

[0244] If the first similarity is greater than a similarity threshold, the to-be-operated segmented word corresponding to the exclusive word is determined;

[0245] If the length of the pinyin string corresponding to the to-be-operated segmented word is different from the length of the pinyin string corresponding to the exclusive word, or the pinyin string corresponding to the to-be-operated segmented word does not match the corresponding relationship between the multiple confused pinyins, the similarity between the pinyin string corresponding to the to-be-operated segmented word and the pinyin string corresponding to the exclusive word is calculated to obtain a second similarity;

[0246] If the second similarity is greater than the similarity threshold, the to-be-operated segmented word corresponding to the exclusive word is determined.

[0247] In a possible implementation, the voice operation game device 900 further includes an obtaining unit;

[0248] The obtaining unit is configured to obtain the multiple confused strings, the multiple non-pinyin and pinyins, and the multiple confused pinyins from the voice recognition error data of the target game.

[0249] The determining unit 930 is further configured to determine the corresponding relationship between the multiple confused strings, the corresponding relationship between the multiple non-pinyin and pinyins, and the corresponding relationship between the multiple confused pinyins according to the multiple confused strings, the multiple non-pinyin and pinyins, and the multiple confused pinyins.

[0250] In a possible implementation, the extracting unit 920 is configured to:

[0251] If the first segmented word in the multiple corrected segmented word sets belongs to the second segmented word, and the length of the segmented word corresponding to the first segmented word is less than the length of the segmented word corresponding to the second segmented word, the corrected text is determined according to the second segmented word.

[0252] In a possible implementation, the identifying unit 910 is configured to:

[0253] The to-be-operated voice is subjected to voice recognition through a voice recognition model to obtain to-be-operated text. The voice recognition model is obtained by training a second recognition model according to the to-be-trained voice associated with the target game and the to-be-trained text corresponding to the to-be-trained voice.

[0254] It can be seen from the above technical solution that the voice operation game device comprises an identification unit, an extraction unit, a determination unit and a calling unit. The identification unit identifies the to-be-operated voice in the target game to obtain to-be-operated text, thereby providing text data for subsequent voice operation game implementation. The extraction unit identifies the intention of the to-be-operated text to obtain a to-be-operated intention, and extracts the entity of the to-be-operated text to obtain a to-be-operated entity set, thereby implementing understanding of the to-be-operated text in natural language as the to-be-operated intention, and extracting the to-be-operated entity set corresponding to the to-be-operated intention, thereby providing input data for subsequent voice operation game implementation. The determination unit determines the to-be-called function and the to-be-called parameter set corresponding to the to-be-operated intention from the plurality of game functions and the plurality of function parameter sets based on the corresponding relationship between the plurality of game intentions, the plurality of game functions and the plurality of function parameter sets, thereby being capable of mapping the to-be-operated intention to the to-be-called function in the plurality of game functions of the target game, and determining the to-be-called parameter set in the plurality of function parameter sets of the target game, thereby providing calling data for subsequent voice operation game implementation. The calling unit calls the to-be-called function to operate the target game based on the to-be-operated entity set when the to-be-operated entity set matches the to-be-called parameter set, thereby obtaining a voice operation result of the target game. The voice operation game is implemented when the input data of the voice operation game matches the calling data of the voice operation game. Based on this, the voice operation game device does not need to learn and memorize a large number of operation texts corresponding to a large number of operation voices based on the corresponding relationship between the plurality of game intentions, the plurality of game functions and the plurality of function parameter sets in the target game, and has a lower requirement for voice input and voice recognition. After the identification unit identifies the to-be-operated voice as the to-be-operated text, the extraction unit first understands the to-be-operated text as the to-be-operated intention, and extracts the to-be-operated entity set corresponding to the to-be-operated intention. The determination unit then maps it to the to-be-called function and the to-be-called parameter set in the target game. The calling unit calls the to-be-called function when the to-be-operated entity set matches the to-be-called parameter set, thereby simply, quickly and effectively implementing the to-be-operated voice operation on the target game, and thereby improving the game experience.

[0255] The embodiments of the present application further provide a computer device, which can be a server. Referring to FIG. 10, which is a structural diagram of a server provided by the embodiments of the present application, the server 1000 can have great differences due to different configurations or performances, and can include one or more processors, such as a central processing unit CPU 1022, and a memory 1032, one or more storage media 1030 (such as one or more mass storage devices) for storing application programs 1042 or data 1044. The memory 1032 and the storage media 1030 can be temporary storage or persistent storage. The programs stored in the storage media 1030 can include one or more modules (not shown in the diagram), each of which can include a series of instruction operations in the server. Furthermore, the central processing unit 1022 can be configured to communicate with the storage media 1030 and execute the series of instruction operations in the storage media 1030 on the server 1000.

[0256] The server 1000 can further include one or more power supplies 1026, one or more wired or wireless network interfaces 1050, one or more input and output interfaces 1058, and / or one or more operating systems 1041, such as Windows Server TM , Mac OS X TM , Unix TM , Linux TM , FreeBSD TM , and the like.

[0257] In the embodiments, the central processing unit 1022 in the server 1000 can execute the methods provided in the various optional implementations of the above embodiments.

[0258] The computer device provided by the embodiments of the present application can also be a terminal. Referring to FIG. 11, which is a structural diagram of a terminal provided by the embodiments of the present application. Taking a smart phone as an example, the smart phone includes: radio frequency (RF) circuit 1110, memory 1120, input unit 1130, display unit 1140, sensor 1150, audio circuit 1160, wireless fidelity (WiFi) module 1170, processor 1180, and power supply 1190, and the like. The input unit 1130 can include a touch panel 1131 and other input devices 1132, and the display unit 1140 can include a display panel 1141. The audio circuit 1160 can include a speaker 1161 and a microphone 1162. Those skilled in the art can understand that the structure of the smart phone shown in FIG. 11 does not constitute a limitation on the smart phone, and can include more or fewer components than the diagram, or combine certain components, or different component arrangements.

[0259] The memory 1120 can be used to store software programs and modules, and the processor 1180 executes various function applications and data processing of the smart phone by running the software programs and modules stored in the memory 1120. The memory 1120 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application program required by a function (such as a sound playing function, an image playing function, etc.), and the like; and the data storage area can store data created according to the use of the smart phone (such as audio data, a phone book, etc.), and the like. In addition, the memory 1120 can include a high-speed random access memory, and can also include a non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other volatile solid-state memory device.

[0260] The processor 1180 is a control center of the smart phone, connects all parts of the smart phone through various interfaces and lines, and executes various functions and processes data of the smart phone by running or executing the software programs and / or modules stored in the memory 1120 and calling the data stored in the memory 1120. Optionally, the processor 1180 can include one or more processing units; preferably, the processor 1180 can integrate an application processor and a modem processor, wherein the application processor mainly processes an operating system, a user interface, and an application program, and the modem processor mainly processes wireless communication. It can be understood that the above-mentioned modem processor can also not be integrated into the processor 1180.

[0261] In the embodiment, the processor 1180 in the smart phone can execute the method provided in various optional implementation manners of the above-mentioned embodiments.

[0262] According to an aspect of the present application, a computer readable storage medium is provided, the computer readable storage medium is used to store a computer program, when the computer program runs on a computer device, the computer program causes the computer device to execute the method provided in various optional implementation manners of the above-mentioned embodiments.

[0263] According to an aspect of the present application, a computer program product is provided, the computer program product includes a computer program stored in a computer readable storage medium. The processor of a computer device reads the computer program from the computer readable storage medium, and the processor executes the computer program, so that the computer device executes the method provided in various optional implementation manners of the above-mentioned embodiments.

[0264] The description of the corresponding flow or structure of each of the above-mentioned figures has its own emphasis, and the part not described in detail in a certain flow or structure can refer to the related description of other flows or structures.

[0265] The terms "first", "second", and the like in the description and in the claims of this application are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the terms so used are interchangeable under appropriate circumstances such that the embodiments of the application described herein are, for example, capable of practical implementation irrespective of the particular sequential or chronological order of the steps contained therein. Moreover, the term "comprising" and variations thereof, as used in enlisting the elements of this application, are intended to mean that the installation contain the recited elements, but not excluding others. In other words, it is meant to be an "inclusive inclusive" term and not an "exclusive exclusive" term.

[0266] In several embodiments provided in the present application, it should be understood that the disclosed system, apparatus and method can be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, for example, the division of the units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed units can be indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0267] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, that is, they can be located in one place, or can be distributed on a plurality of network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0268] In addition, each functional unit in each embodiment of the present application can be integrated into a processing unit, or each unit can exist physically, or two or more units can be integrated into one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0269] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for causing a computer device to perform all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a RAM, a magnetic disk or an optical disk, and various computer program storage media that can store computer programs.

[0270] In the embodiments of the present application, the term "module" or "unit" refers to a computer program or a part of a computer program with a predetermined function, and works together with other related parts to achieve a predetermined target, and can be implemented entirely or partially by using software, hardware (such as a processing circuit or a memory) or a combination thereof. Similarly, one processor (or multiple processors or memories) can be used to implement one or more modules or units. In addition, each module or unit can be a part of an integral module or unit that includes the functions of the module or unit.

[0271] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements for some technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A voice operation game method, the method being executed by a computer device, the method comprising: performing voice recognition on a voice to be operated in a target game to obtain a text to be operated corresponding to the voice to be operated; performing intent recognition and entity extraction on the text to be operated to obtain an operation intent corresponding to the text to be operated and a set of operation entities corresponding to the text to be operated; determining a calling function corresponding to the operation intent from a plurality of game functions according to a correspondence between a plurality of game intents and the plurality of game functions in the target game; and performing game operation on the target game by calling the calling function according to the set of operation entities to obtain a voice operation result of the target game.

2. The method of claim 1, wherein the intent recognition on the text to be operated to obtain the operation intent corresponding to the text to be operated comprises: matching the text to be operated with the plurality of game intents according to an intent matching strategy to obtain a first confidence of the text to be operated corresponding to each of the plurality of game intents; and determining the operation intent according to a game intent corresponding to a maximum confidence in the first confidence.

3. The method of claim 1, wherein the intent recognition on the text to be operated to obtain the operation intent corresponding to the text to be operated comprises: performing intent recognition on the text to be operated by an intent recognition model to obtain a second confidence of the text to be operated corresponding to each of the plurality of game intents; and determining the operation intent according to a game intent corresponding to a maximum confidence in the second confidence.

4. The method of claim 1, wherein the intent recognition on the text to be operated to obtain the operation intent corresponding to the text to be operated comprises: matching the text to be operated with the plurality of game intents according to an intent matching strategy to obtain a first confidence of the text to be operated corresponding to each of the plurality of game intents; performing intent recognition on the text to be operated by an intent recognition model to obtain a second confidence of the text to be operated corresponding to each of the plurality of game intents; and determining the operation intent according to a game intent corresponding to a maximum confidence in the first confidence and the second confidence.

5. The method of claim 3 or 4, wherein the obtaining of the intent recognition model comprises: obtaining a training text of the target game and a training intent corresponding to the training text; performing intent recognition on the training text according to a pre-trained language model to obtain a predicted intent corresponding to the training text; adjusting the training intent according to the predicted intent if the predicted intent does not match the training intent to obtain an adjusted training intent; and training a first recognition model according to the training text and the adjusted training intent to obtain the intent recognition model.

6. The method of any one of claims 1-5, wherein the determining of the calling function corresponding to the operation intent from the plurality of game functions according to the correspondence between the plurality of game intents and the plurality of game functions in the target game comprises: ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ According to a correspondence relationship among the multiple game intents, the multiple game functions, and the multiple function parameter sets in the target game, a function to be called and a parameter set to be called corresponding to the intent to be operated are determined from the multiple game functions and the multiple function parameter sets; The game operation on the target game according to the function to be called and the parameter set to be called is performed by using the entity set to be operated, and a voice operation result of the target game is obtained. If the entity set to be operated matches the parameter set to be called, the game operation on the target game is performed by using the entity set to be operated according to the function to be called, and the voice operation result of the target game is obtained.

7. The method of claim 6, wherein the entity set to be operated is obtained in any of the following manners: Manner one: The natural language processing is performed on the text to be operated, and a first entity set corresponding to the text to be operated is obtained; The entity set to be operated is obtained by performing entity screening on the first entity set according to the game entity set of the target game; Manner two: The entity extraction model is used to perform entity extraction on the text to be operated, and a second entity set corresponding to the text to be operated is obtained; The entity set to be operated is obtained by performing entity screening on the second entity set according to the game entity set of the target game; Manner three: The natural language processing is performed on the text to be operated, and a first entity set corresponding to the text to be operated is obtained; The entity extraction model is used to perform entity extraction on the text to be operated, and a second entity set corresponding to the text to be operated is obtained; The entity set to be operated is obtained by performing entity screening on the first entity set and the second entity set according to the game entity set of the target game.

8. The method of claim 7, wherein the game entity set is obtained by: performing entity extraction on the game text associated with the target game, and obtaining the game entity set.

9. The method of claim 6, wherein the game operation on the target game according to the function to be called and the parameter set to be called is performed by using the entity set to be operated, and the voice operation result of the target game is obtained, including: data conversion is performed on the entity set to be operated according to the function to be called, and a function parameter data set corresponding to the entity set to be operated is obtained; If the function parameter data set matches a game scene of the target game under the voice to be operated, the game operation on the target game is performed according to the function parameter data set, and the voice operation result is obtained.

10. The method of claim 9, further comprising: If the entity set to be operated does not match the parameter set to be called, it is determined that the voice operation fails. Or, If the function parameter data set does not match the game scene of the target game under the voice to be operated, it is determined that the voice operation fails. 11.The method of any one of claims 1-10, wherein the intent recognition and entity extraction on the to-be-operated text to obtain a to-be-operated intent and a to-be-operated entity set corresponding to the to-be-operated text comprises: performing segmentation processing on the to-be-operated text according to a plurality of segmentation lengths to obtain a plurality of to-be-operated segmentation sets corresponding to the to-be-operated text; performing error correction processing on the plurality of to-be-operated segmentation sets according to a correspondence between a plurality of confused strings and a set of exclusive terms of the target game to obtain a plurality of error-corrected segmentation sets corresponding to the plurality of to-be-operated segmentation sets, respectively; determining an error-corrected text corresponding to the to-be-operated text according to the error-corrected segmentation sets and segmentation lengths of the corresponding to-be-operated segmentation sets; and performing intent recognition and entity extraction on the error-corrected text to obtain a to-be-operated intent and a to-be-operated entity set corresponding to the error-corrected text. 12.The method of claim 11, wherein the error correction processing on the plurality of to-be-operated segmentation sets according to a correspondence between a plurality of confused strings and a set of exclusive terms of the target game to obtain a plurality of error-corrected segmentation sets corresponding to the plurality of to-be-operated segmentation sets, respectively, comprises: for each to-be-operated segmentation in the plurality of to-be-operated segmentation sets, if the to-be-operated segmentation matches the correspondence between the plurality of confused strings, determining an error-corrected segmentation corresponding to the to-be-operated segmentation according to the correspondence between the plurality of confused strings; if the to-be-operated segmentation does not match the correspondence between the plurality of confused strings, performing pinyin mapping on the to-be-operated segmentation according to a correspondence between a plurality of non-pinyin and pinyin to obtain a pinyin string corresponding to the to-be-operated segmentation; and performing error correction processing on the pinyin string corresponding to the to-be-operated segmentation according to a pinyin string corresponding to each exclusive term in the set of exclusive terms to obtain an error-corrected segmentation corresponding to the to-be-operated segmentation. 13.The method of claim 12, wherein the error correction processing on the pinyin string corresponding to the to-be-operated segmentation according to a pinyin string corresponding to each exclusive term in the set of exclusive terms to obtain an error-corrected segmentation corresponding to the to-be-operated segmentation comprises: if a length of the pinyin string corresponding to the to-be-operated segmentation is the same as a length of the pinyin string corresponding to the exclusive term, and the pinyin string corresponding to the to-be-operated segmentation matches a correspondence between a plurality of confused pinyins, determining a confused pinyin string corresponding to the to-be-operated segmentation according to the correspondence between the plurality of confused pinyins; performing similarity calculation on the confused pinyin string corresponding to the to-be-operated segmentation and the pinyin string corresponding to the exclusive term to obtain a first similarity; and if the first similarity is greater than a similarity threshold, determining the error-corrected segmentation corresponding to the to-be-operated segmentation according to the exclusive term. ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ If the length of the pinyin string corresponding to the to-be-operated word is different from the length of the pinyin string corresponding to the exclusive word, or the pinyin string corresponding to the to-be-operated word does not match the corresponding relationship between the plurality of confused pinyins, the similarity between the pinyin string corresponding to the to-be-operated word and the pinyin string corresponding to the exclusive word is calculated to obtain a second similarity; If the second similarity is greater than the similarity threshold, the to-be-operated word is determined as the corrected word corresponding to the to-be-operated word according to the exclusive word.

14. The method of claim 13, wherein the obtaining of the corresponding relationship between the plurality of confused strings, the corresponding relationship between the plurality of non-pinyin and pinyin, and the corresponding relationship between the plurality of confused pinyins comprises: obtaining a plurality of confused strings, a plurality of non-pinyin and pinyin, and a plurality of confused pinyins from the speech recognition error data of the target game; determining the corresponding relationship between the plurality of confused strings, the corresponding relationship between the plurality of non-pinyin and pinyin, and the corresponding relationship between the plurality of confused pinyins according to the plurality of confused strings, the plurality of non-pinyin and pinyin, and the plurality of confused pinyins.

15. The method of claim 11, wherein the determining of the corrected text corresponding to the to-be-operated text according to the plurality of corrected word sets and the plurality of word lengths corresponding to the plurality of corrected word sets comprises: if a first word in the plurality of corrected word sets belongs to a second word, and the word length corresponding to the first word is less than the word length corresponding to the second word, the second word is determined as the corrected text according to the second word.

16. A voice operated game device, the device comprising: an identifying unit, an extracting unit, a determining unit, and a calling unit; the identifying unit is configured to perform speech recognition on a to-be-operated speech in a target game to obtain a to-be-operated text corresponding to the to-be-operated speech; the extracting unit is configured to perform intent recognition and entity extraction on the to-be-operated text to obtain a to-be-operated intent and a to-be-operated entity set corresponding to the to-be-operated text; the determining unit is configured to determine a to-be-called function and a to-be-called parameter set corresponding to the to-be-operated intent according to a corresponding relationship between a plurality of game intents, a plurality of game functions, and a plurality of function parameter sets in the target game; the calling unit is configured to call the to-be-called function according to the to-be-operated entity set to perform game operation on the target game if the to-be-operated entity set matches the to-be-called parameter set, and obtain a speech operation result of the target game.

17. A computer device, comprising a processor and a memory: the memory is configured to store a computer program and transmit the computer program to the processor; the processor is configured to execute the method of any one of claims 1-15 according to instructions in the computer program.

18. A computer readable storage medium for storing a computer program, which, when executed on a computer device, causes the computer device to execute the method of any one of claims 1-15.

19. A computer program product comprising a computer program which, when run on a computer device, causes the computer device to perform the method of any one of claims 1-15.

Images