CN117610539A

CN117610539A - Intention execution method, device, electronic equipment and storage medium

Info

Publication number: CN117610539A
Application number: CN202311643403.3A
Authority: CN
Inventors: 常欣宇
Original assignee: Guangdong Oppo Mobile Telecommunications Corp Ltd
Current assignee: Guangdong Oppo Mobile Telecommunications Corp Ltd
Priority date: 2023-12-01
Filing date: 2023-12-01
Publication date: 2024-02-27

Abstract

The application discloses an intention execution method, an apparatus, an electronic device and a storage medium, wherein the intention execution method is applied to the electronic device and comprises the following steps: acquiring input information to be processed; generating execution flow information corresponding to the information to be processed through a large language model, wherein the execution flow information is used for realizing user intention corresponding to the information to be processed; and determining a target operation corresponding to the execution flow information through the large language model according to a knowledge database and an operation database, and executing the target operation, wherein the knowledge database is constructed according to the equipment information of the electronic equipment, and the operation database is constructed according to functions realized by input operations in different application interfaces. The method can identify and realize complex user intention according to the information input by the user.

Description

Intention execution method, device, electronic equipment and storage medium

Technical Field

The present disclosure relates to the technical field of electronic devices, and more particularly, to an intended execution method, an apparatus, an electronic device, and a storage medium.

Background

With rapid progress in the technological level and the living standard, electronic devices (such as smartphones, tablet computers, etc.) have become one of the commonly used electronic products in people's lives. Many electronic devices today are able to implement a user's intent by identifying the user's intent to provide a corresponding service. However, in the related art, only a simple user intention is generally recognized and executed, and when the user intention is complex, the electronic device cannot better execute the user intention.

Disclosure of Invention

The application provides an intention executing method, an intention executing device, electronic equipment and a storage medium, which can identify and realize complex user intention.

In a first aspect, an embodiment of the present application provides an intent execution method, applied to an electronic device, where the method includes: acquiring input information to be processed; generating execution flow information corresponding to the information to be processed through a large language model, wherein the execution flow information is used for realizing user intention corresponding to the information to be processed; and determining a target operation corresponding to the execution flow information through the large language model according to a knowledge database and an operation database, and executing the target operation, wherein the knowledge database is constructed according to the equipment information of the electronic equipment, and the operation database is constructed according to functions realized by input operations in different application interfaces.

In a second aspect, an embodiment of the present application provides an intent execution apparatus applied to an electronic device, the apparatus including: the device comprises an information acquisition module, an intention recognition module and an operation execution module, wherein the information acquisition module is used for acquiring input information to be processed; the intention recognition module is used for generating execution flow information corresponding to the information to be processed through a large language model, and the execution flow information is used for realizing user intention corresponding to the information to be processed; the operation execution module is used for determining a target operation corresponding to the execution flow information through the large language model according to a knowledge database and an operation database, and executing the target operation, wherein the knowledge database is constructed according to the equipment information of the electronic equipment, and the operation database is constructed according to functions realized by input operations in different application interfaces.

In a third aspect, an embodiment of the present application provides an electronic device, including: one or more processors; a memory; one or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, the one or more applications configured to perform the intended execution method provided in the first aspect above.

In a fourth aspect, embodiments of the present application provide a computer readable storage medium having stored therein program code that is callable by a processor to perform the intended execution method provided in the first aspect described above.

According to the scheme, the input information to be processed is obtained, the execution flow information corresponding to the information to be processed is generated through the large language model, the execution flow information is used for achieving user intention corresponding to the information to be processed, target operation corresponding to the execution flow information is determined and performed through the large language model according to the knowledge database and the operation database, the knowledge database is built according to equipment information of the electronic equipment, and the operation database is built according to functions achieved by input operations in different application interfaces. Therefore, the input information with complex user intention can be identified as the execution flow information, and then the target operation corresponding to the execution flow information is determined and executed by utilizing the large language model, the knowledge database and the operation database, so that the complex user intention can be understood and realized.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the description of the embodiments will be briefly introduced below, it being obvious that the drawings in the following description are only some embodiments of the present application, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

FIG. 1 shows a flow diagram of a method intended to be performed according to one embodiment of the present application.

Fig. 2 shows a flow diagram of a method intended to be performed according to another embodiment of the present application.

Fig. 3 shows a flow diagram of a method intended to be performed according to a further embodiment of the present application.

Fig. 4 shows a flow diagram of a method intended to be performed according to yet another embodiment of the present application.

Fig. 5 shows a block diagram of an intent execution device in accordance with one embodiment of the present application.

Fig. 6 is a block diagram of an electronic device for performing the method of intent to perform in accordance with an embodiment of the present application.

Fig. 7 is a storage unit for holding or carrying program code for implementing the method of intent to execute in accordance with an embodiment of the present application.

Detailed Description

In order to enable those skilled in the art to better understand the present application, the following description will make clear and complete descriptions of the technical solutions in the embodiments of the present application with reference to the accompanying drawings in the embodiments of the present application.

Intent recognition is an important task in natural language processing. Currently, most electronic devices are capable of recognizing a user's intention and then performing a corresponding operation according to the recognized intention to implement the user's intention. For example, a voice assistant may be provided in the electronic device to assist the user in controlling the terminal to perform tasks such as playing music, opening a call record, etc., and in controlling the terminal using the voice assistant, the user inputs a voice signal, the terminal needs to recognize intention information of the voice signal, and then perform a corresponding task based on the intention information.

In the related art, when the electronic device learns the intention of the user, the electronic device mainly depends on preset keywords and depends on the instruction template to execute actions, so that the electronic device cannot fully learn the intention of the user and cannot automatically execute the complex intention of the user under the condition that the intention of the user is complex, and further the use experience of the user when using the electronic device can be influenced. For example, in the related art, when the user inputs "open APP1 search for ice cream a", the electronic device may recognize that the APP target keyword is "APP1", the operation action keyword is "search", the search target keyword is "ice cream a", and then may execute the above user intention according to the above keywords in combination with the instruction template; however, after searching the result of the ice cream, if the user inputs the input information "help me find a section with the highest cost performance and order to my company" with a relatively complex intention, the electronic device generally cannot understand the intention of the user or cannot automatically execute the intention of the user after understanding the intention of the user because the input information does not have a preset keyword or the instruction template cannot realize the intention.

In order to solve the above problems, the inventor proposes an intention execution method, an apparatus, an electronic device and a storage medium provided by the embodiments of the present application, which can implement input information with complex user intention, identify the input information as execution flow information, and then determine and execute a target operation corresponding to the execution flow information by using a large language model, a knowledge database and an operation database, thereby implementing understanding and implementing the complex user intention. Among them, a specific intended execution method is described in detail in the following examples.

The method for executing the program provided in the embodiment of the present application will be described in detail with reference to the accompanying drawings.

Referring to fig. 1, fig. 1 is a schematic flow chart of an implementation method according to an embodiment of the present application. In a specific embodiment, the intention execution method is applied to the intention execution apparatus 400 shown in fig. 5 and the electronic device 100 (fig. 6) configured with the intention execution apparatus 400. The specific flow of the embodiment will be described below by taking an electronic device as an example, and it will be understood that the electronic device applied in the embodiment may be a smart phone, a tablet computer, a smart watch, an electronic book, etc., which is not limited herein. The following will describe the flow shown in fig. 1 in detail, and the method for performing intent may specifically include the following steps:

Step S110: and acquiring the input information to be processed.

In the embodiment of the application, the electronic device can acquire the input information to be processed so as to identify and execute the intention of the user according to the input information to be processed.

In some embodiments, the information to be processed may be input by text input, or may be input by voice input, which is not limited herein. As an implementation manner, if the information to be processed is input by means of text, the input text may be determined as the information to be processed. As yet another implementation manner, if the information to be processed is input by means of voice, the input voice may be subjected to text conversion operation, and the text obtained after text conversion may be determined as the information to be processed, where the input voice may be subjected to text conversion operation by means of an automatic voice recognition technology (automatic speech recognition, ASR).

In some embodiments, the electronic device may display an input interface of the information to be processed in response to an operation input by the user, so that the user can input the information to be processed in the input interface. Optionally, for example, the electronic device may display an information input interface corresponding to the voice assistant, and the electronic device may detect information input by the user to obtain information to be processed when the information input interface is displayed. Optionally, the electronic device may display an application interface corresponding to the voice assistant, and when detecting an operation for triggering chat with the virtual session object in the application interface, display a chat interface with the virtual session object, and obtain the information to be processed by detecting chat information input by a user in the chat interface. The electronic equipment can display an application interface corresponding to the voice assistant according to voice information input by a user and used for waking up the voice assistant; the triggering operation of the application icon for the voice assistant can also be responded, the application interface corresponding to the voice assistant is displayed, and the triggering mode of the application interface for triggering the voice assistant specifically is not limited.

Of course, the specific manner in which the electronic device obtains the information to be processed may not be limited, for example, the electronic device may obtain an input image, then perform content identification on the image, and determine the information to be processed according to the identified content; for another example, the electronic device may acquire an input video, then perform content recognition on the video image, perform speech-to-text processing according to audio in the video, and then determine the information to be processed according to the content recognized in the video image and the content obtained by speech-to-text processing.

Step S120: and generating execution flow information corresponding to the information to be processed through a large language model, wherein the execution flow information is used for realizing user intention corresponding to the information to be processed.

In the embodiment of the application, after the electronic device acquires the information to be processed, the intention recognition can be performed on the information to be processed, so that corresponding operation is performed according to the recognized intention of the user, the intention of the user is realized, and the task required to be executed by the user is completed. When the electronic equipment performs intention recognition on the information to be processed, the execution flow information corresponding to the information to be processed can be generated through the large language model, and the execution flow information is used for realizing the user intention corresponding to the information to be processed. The execution flow information may include step information corresponding to a step of realizing the user intention, that is, when the user intention corresponding to the information to be processed is identified, the identified user intention is broken down into steps, and the execution flow information is formed. The large language model (Large Language Model, LLM) refers to a complex artificial neural network trained based on a large amount of data and computing resources, and can learn rich language modes and knowledge, so that accurate response is generated to natural language input, and the main purpose of the large language model is to enable a machine to better understand and generate natural language texts of human beings, such as seals, dialogues, searches and the like.

It can be appreciated that, since the data size of the training data required by the large language model is very large, and the data size for training the large language model is even more than billions of words, the large language model can understand the intention of the user more accurately and can understand the intention of the complex user, so that the intention understanding of the execution flow information can be performed through the large language model, and the corresponding execution flow information can be generated.

In some embodiments, input information for inputting a large language model may be generated according to information to be processed, and then the input information is input into the large language model, so that the large language model outputs execution flow information corresponding to the information to be processed according to the input information. The input information includes prompt information, which may be used to prompt the large language model to understand intention of the information to be processed, and generate execution flow information for achieving (or implementing) understanding of user intention, where the prompt information may be understood as prompt (prompt word).

For example, if the above information to be processed is "help me find an ice cream with highest cost performance and order to my company", then the above prompt may be "please analyze the intention expressed by the user and give a possible execution flow information for achieving the intention of the user", and after the information to be processed is spliced with the prompt, the above input information may be obtained. The large language model may output execution flow information according to the input information input: if ' help me find an ice cream with highest cost performance and order to my company ', the required execution flow information is ' 1; 2. acquiring an address of a company; 3. selecting one with larger preferential strength, higher score and lower unit price; 4. the selected commodity is placed in order.

In one possible implementation manner, a first prompt word template may be created in advance, where the first prompt word template is used to obtain input information for inputting the large language model after being spliced with the information to be processed, so that the large language model can output execution flow information corresponding to the information to be processed according to the input information.

In one possible implementation manner, the large language model may be deployed in a server, and when execution flow information corresponding to information to be processed needs to be generated through the large language model, the electronic device may generate the input information according to the information to be processed, and then send the input information to the server; correspondingly, the server can receive the input information and input the input information into the large language model, so that execution flow information corresponding to the information to be processed output by the large language model is obtained, and the execution flow information is returned to the electronic equipment; correspondingly, the electronic device can receive the execution flow information returned by the server.

In a possible implementation manner, under the condition that the hardware configuration level of the electronic device is relatively high, the large language model can be deployed in the electronic device, when execution flow information corresponding to the information to be processed needs to be generated through the large language model, the electronic device can directly call the local large language model, and after the input information is generated according to the information to be processed, the input information is input into the large language model, so that the execution flow information corresponding to the information to be processed, which is output by the large language model, is obtained.

In some embodiments, since the large language model in the embodiments of the present application is to implement the task customization model for the to-be-processed information, after understanding the user intention, and generating the execution flow information corresponding to the to-be-processed information, there may be a difference between the task of daily processing of the large language model and the task of daily processing of the large language model, the large language model in the embodiments of the present application may be a model customized according to the task of generating the execution flow information, that is, the large language model is obtained by adjusting and optimizing the task of generating the execution flow information according to training data. Thus, the accuracy of the large language model in generating the execution flow information corresponding to the information to be processed can be ensured.

Step S130: and determining a target operation corresponding to the execution flow information through the large language model according to a knowledge database and an operation database, and executing the target operation, wherein the knowledge database is constructed according to the equipment information of the electronic equipment, and the operation database is constructed according to functions realized by input operations in different application interfaces.

The above execution flow information may include step information corresponding to steps for realizing the user's intention, so that to realize the user's intention, the steps corresponding to the respective step information in the above execution flow information are completed, so that the user's intention can be achieved. Therefore, after the execution flow information is obtained, the target operation corresponding to the execution flow information can be determined through the large language model according to the knowledge database and the operation database, and the target operation is executed, so that each step in the execution flow information is completed, and the intention of the user is realized.

In the embodiment of the application, the knowledge database is constructed according to the equipment information of the electronic equipment, and the operation database is constructed according to functions realized by input operations in different application interfaces. The device information may refer to basic information of the electronic device, the device information may include application information of an application program installed in the electronic device, preference setting of the electronic device, personal information of a user corresponding to the electronic device, configuration information of the electronic device, and the like, the application information may include a name of the installed application program, a type of the application program, and the like, the preference setting of the electronic device may include a set sound output volume, a network used, brightness of a screen, a duration of automatic screen locking, and the like, the personal information of the user corresponding to the electronic device may include a region, an age, a income, a home address, a company address, and the like, the configuration information of the electronic device may include parameters of a processor of the electronic device, parameters of a memory of the electronic device, a screen size of the electronic device, a maximum battery power of the electronic device, and the specific device information may not be limited, of course; the different application interfaces may at least include respective application interfaces in different application programs installed in the electronic device, and functions implemented by input operations in the application interfaces may be understood as functions that may be implemented in the application interfaces. As can be appreciated, since the operations database is built according to functions implemented by input operations in different application interfaces, the operations database may provide a priori knowledge for the large language model to determine the target operations to which the above execution flow information corresponds; the knowledge database is constructed according to the device information of the electronic device, so that the knowledge database can provide information for executing the target operation corresponding to the execution flow information, all operations for realizing the execution flow information (serving as the target operation corresponding to the execution flow information) can be determined through a large language model according to the knowledge database and the operation database, and the determined target operation can be automatically executed, so that the intention of a user is achieved.

In some embodiments, the electronic device may generate input information for inputting the large language model according to the above execution flow information and information in the operation database, and then input the input information into the large language model, so that the large language model outputs operation information of the target operation corresponding to the execution flow information according to the input information. The input information can also comprise prompt information, and the prompt information can be used for prompting the large language model to determine operation information required for realizing the execution flow information according to the execution flow information and the information in the operation database. For example, the prompt message may be "for the execution flow information a, a specific operation for implementing the execution flow information a is given; the operations to be implemented to implement the execution flow information a may refer to the following information: { abc } ", in the hint information, abc refers to the information in the above operation database, that is, the information in the actual operation database needs to be replaced with the abc.

It will be appreciated that in large language models, reAct refers to a technique that solves complex problems using both reasoning and action operations in a single interaction cycle. Specifically, the REAct utilizes the semantic understanding capability (i.e. reasoning) of the large language model to analyze and understand the input text, so that deep knowledge and information of the input text are obtained; the parsed information is then further processed and utilized by other capabilities (i.e., actions) outside of the model to perform corresponding operations, such as computing, searching for up-to-date messages, user-defined actions, etc. In this way, the REAct can efficiently realize various complex interactive interfaces and support various types of input and output, so that support is provided for more flexible and intelligent application programs, and meanwhile, the REAct also has good expandability and adaptability, and can be integrated and fused with other technologies to meet the requirements and applications of different fields. Therefore, based on the REAct of the large language model, the large language model can determine the target operation corresponding to the execution flow information according to the input information.

In addition, after the target operations determined by the large language model are obtained, in consideration of the fact that when a lot of operations are realized, equipment information of the electronic equipment is needed to be used, so that the determined target operations can be executed according to the knowledge database, and the realization of the execution flow information is completed, namely, the intention of a user is achieved. For example, for the foregoing example execution flow information "1. Query shopping software for information of ice cream merchandise; 2. acquiring an address of a company; 3. selecting one with larger preferential strength, higher score and lower unit price; 4. and ordering the selected commodity, wherein when the operation for realizing the corresponding operation of inquiring the information of the ice cream commodity in the shopping software is executed, the application information of the installed application program can be inquired from the knowledge database, and the shopping software to be used is determined according to the application information.

Similarly, in the case that the large language model is deployed in the server, the electronic device may generate input information according to the above information to be processed and information in the operation database and send the generated input information to the server, so that the server determines the above target operation according to the received input information through the large language model and feeds back to the electronic device; under the condition that the large language model is deployed in the electronic equipment, the electronic equipment can directly input the input information into the local large language model, so that the determined target operation is obtained.

In some embodiments, when the knowledge database is constructed, a prompt interface may be displayed, and the prompt interface may include prompt information for prompting the user to collect device information; under the condition that the confirmation operation input in the prompt interface is detected, the equipment information of the electronic equipment can be scanned, and a knowledge database is constructed according to the equipment information. When a knowledge database is constructed, each piece of original information in the equipment information can be processed into a set size; then using text to vector (text to vector) method, such as hash algorithm, cosine similarity, PCA (principal component analysis), transform architecture Embedding model, etc., to encode each piece of original information into vector; the encoded vector may be stored in a database, for example, a database such as MySql, SQLite, mongoDB, redis, where the data is in the form of a key value pair of < vector encoding, original file path >, and original information corresponding to the vector may be found through the original file path.

In some embodiments, the above operation database may be locally configured by the electronic device, or may be downloaded from a server. When the operation database is constructed, the functions realized by the input operations in different application interfaces can be generated through an automatic script or can be generated through manual summarization. The automatic script generation simulates human clicking and input, and simultaneously analyzes response conditions of an interface, mainly by mining out operable buttons in an application interface, and effects after operation, such as an interface displayed after operation, processing executed after operation and the like. According to the function information corresponding to the function realized by the input operation in the application interface, the function information can be stored into a database according to the processing mode of the equipment information, so as to obtain an operation database.

According to the intention executing method provided by the embodiment of the application, the input information to be processed is obtained, the understanding capability of the large language model is utilized for the information to be processed, the intention is understood through the large language model, the information to be processed is generated, then the operation corresponding to the execution flow information is determined through the large language model, the knowledge database and the operation database which are built in advance, and the determined operation is executed, so that the intention of a user can be achieved, the complex user intention can be further realized, the use experience of the user when the user uses the relevant service for intention understanding of the electronic equipment is improved, and the usability of the electronic equipment is enhanced.

Referring to fig. 2, fig. 2 is a schematic flow chart of an intended execution method according to another embodiment of the present application. The method for executing intention is applied to the electronic device, and will be described in detail with respect to the flow shown in fig. 2, and the method for executing intention may specifically include the following steps:

step S210: and acquiring the input information to be processed.

Step S220: and generating execution flow information corresponding to the information to be processed through a large language model, wherein the execution flow information is used for realizing user intention corresponding to the information to be processed.

In the embodiment of the present application, the step S210 and the step S220 may refer to the content of the foregoing embodiment, which is not described herein.

Step S230: and sequentially determining each operation of a plurality of operations sequentially executed through the large language model according to a knowledge database and an operation database, and sequentially executing each operation, wherein the knowledge database is constructed according to equipment information of the electronic equipment, and the operation database is constructed according to functions realized by input operations in different application interfaces.

In the embodiment of the present application, generally when the intention of the current user is complex, a plurality of operations with an execution sequence relationship are needed to implement the above generated execution flow information, so that each operation in the plurality of operations can be determined in turn and each operation is executed in turn through a large language model according to a knowledge database and an operation database, and thus, the complex intention of the user can be better achieved.

In some embodiments, when each operation of the plurality of operations is determined in turn and each operation is executed in turn by the large language model according to the knowledge database and the operation database, the first operation is determined by the large language model according to the operation database and the first operation is executed according to the knowledge database, and an operation result of the first operation is obtained; for the N-th operation in the plurality of operations, N is a positive integer greater than 1, namely, for any operation other than the first operation, the N-th operation can be determined through a large language model according to the operation result of the N-1-th operation and the operation database, and the N-th operation is executed according to the knowledge database, so that the operation result of the N-th operation is obtained. That is, for any operation other than the first operation, it is necessary to rely on the execution result of the last operation (i.e., the above operation result) in determining the specific operation, and thus the operation is determined by a large language model and based on the operation result of the last operation of the operation and the operation database.

Illustratively, the information of ice cream commodity in shopping software is inquired according to the execution flow information '1' exemplified in the previous embodiment; 2. acquiring an address of a company; 3. selecting one with larger preferential strength, higher score and lower unit price; 4. when determining and executing the operation corresponding to the information for inquiring ice cream commodities in shopping software, firstly determining that a first operation is to acquire shopping application, and after executing the first operation according to the knowledge database, obtaining the result (namely the operation result of the first operation) inquired in the knowledge database, wherein the shopping application is to be provided with application j and application k; then determining a second operation 'to ask a question to the user' through a large language model according to the operation result of the first operation 'shopping application has application j and application k' and an operation database: whether the application j or the application k is used for purchase is carried out, and after the second operation is carried out, the result of the user feedback aiming at the prompt (namely, the operation result of the second operation) can be obtained, namely, the application k is used for purchase; then determining a third operation application through the large language model according to the operation result of the second operation, namely 'using application k to purchase', and the operation database: applying k, searching commodities and keywords: ice cream ", after this third operation is performed, an operation result" commodity information list of ice cream searched by applying k "may be obtained.

In a possible implementation manner, when the first operation is determined through the large language model and according to the operation database, the first operation is executed according to the knowledge database, and an operation result of the first operation is obtained, since when the first operation is determined and executed, an operation result of the last operation does not exist, first input information for inputting the large language model can be generated according to target prompt information, the operation database and execution flow information, and the target prompt information is used for prompting the large language model to output an operation for realizing execution flow information according to data in the operation database; and inputting the first input information into the large language model to obtain first operation information output by the large language model, and determining an operation corresponding to the first operation information as a first operation. The above target prompt message may be understood as a prompt (prompt) as described in the foregoing embodiment for prompting the processing required for the large language model.

Optionally, the target prompt information may be a predetermined prompt word template, and the electronic device may fill information in the operation database into the prompt word template, and splice the execution flow information with the prompt word target, so as to obtain the first input information. For example, the hint word template may be: "specific operation for realizing the execution flow information A is given for the execution flow information A; if the information in the operation database is needed to be used in realizing the flow information A, the following format is used: { operation number: parameter list }; the supportable operation is selected from the operation database, and the information of the operation database is as follows: { abc }; if the step contains an unsupported operation, please attempt to replace with a supported operation; if the intent is still not resolved, the reply is temporarily unable to support the intent.

In a possible implementation manner, when the nth operation is determined according to the operation result of the nth operation and the operation database through the large language model for the nth operation, third input information for inputting the large language model can be generated according to the second input information input to the large language model when the nth-1 operation is determined and the operation result of the nth-1 operation; and inputting the third input information into the large language model to obtain second operation information output by the large language model, and determining the operation corresponding to the second operation information as an Nth operation. That is, the input information to be input to the large language model at the time of determining and executing the nth operation may be generated based on the operation result of the last operation and the input information to be input to the large language model at the time of determining the last operation. Alternatively, the operation result of the N-1 th operation may be spliced with the above second input information to obtain the above third input information. It will be appreciated that the input information input to the large language model at the time of determining the last operation already includes the execution flow information, the related prompt information (which may be the target prompt information described above) and the information in the operation database, so that the input information input to the large language model at the time of determining the last operation and the operation result of the last operation can be directly used to generate the input information that needs to be input to the large language model at the present time.

In one possible implementation manner, when the first operation is performed according to the knowledge database and the operation result of the first operation is obtained, the first operation may be performed according to the knowledge database; if the operation result is not obtained after the first operation is executed, outputting first prompt information, wherein the first prompt information is used for prompting the input of the operation result corresponding to the first operation; and acquiring the input first result information as an operation result corresponding to the first operation. It can be understood that, in some cases, when the electronic device automatically performs the operation, the required operation cannot be completed, for example, for the operation of "obtaining the address of the company", the electronic device cannot obtain the address of the company "after querying the knowledge database, so that the user can input corresponding result information according to the prompt information by outputting the prompt information as the operation result.

Likewise, for the nth operation, if no operation result is obtained after the nth operation is performed, outputting second prompting information, where the second prompting information is used to prompt for inputting an operation result corresponding to the nth operation; and acquiring the input second result information as an operation result corresponding to the Nth operation.

According to the intention executing method provided by the embodiment of the application, the input information to be processed is obtained, the understanding capability of the large language model is utilized for the information to be processed, the intention is understood through the large language model, the information to be processed is generated, then each operation in a plurality of operations which need to be sequentially executed is sequentially determined through the large language model, the knowledge database and the operation database which are built in advance, and each operation is sequentially executed, so that the intention of a user with complex understanding can be realized, and the complex intention of the user can be better achieved.

Referring to fig. 3, fig. 3 is a schematic flow chart of a method for executing the method according to another embodiment of the present application. The method for executing intention is applied to the electronic device, and will be described in detail with respect to the flowchart shown in fig. 3, where the method for executing intention may specifically include the following steps:

step S310: and acquiring the input information to be processed.

Step S320: and generating execution flow information corresponding to the information to be processed through a large language model, wherein the execution flow information is used for realizing user intention corresponding to the information to be processed, and the execution flow information comprises step information corresponding to a plurality of execution steps.

In the embodiment of the present application, the step S310 and the step S320 may refer to the content of the foregoing embodiment, which is not described herein.

Step S330: and determining the operation corresponding to each step information through the large language model according to a knowledge database and an operation database, and executing the operation corresponding to each step information, wherein the knowledge database is constructed according to the equipment information of the electronic equipment, and the operation database is constructed according to functions realized by input operations in different application interfaces.

In this embodiment of the present application, when the intention of the current user is generally complex, a plurality of execution steps are required to be performed when the intention of the user is implemented, so the above execution flow information may include step information corresponding to the plurality of execution steps, and when the above generated execution flow information is implemented, an operation corresponding to each step information in the step information corresponding to the plurality of execution steps may be determined through a large language model and according to a knowledge database and an operation database, and an operation corresponding to each step information is performed, thereby implementing the above plurality of execution steps, that is, implementing the above execution flow information, so that the complex intention of the user may be better achieved.

In some embodiments, the method for executing intent provided in the embodiments of the present application may be implemented in combination with the method for executing intent provided in the previous embodiment, for example, to implement a plurality of operations that need to be executed sequentially in the above executing steps, when determining and executing the operation corresponding to the executing step, each operation in the plurality of operations corresponding to the executing step may be determined sequentially through a large language model and the knowledge database and the operation database, and each operation is executed sequentially.

Illustratively, the information of ice cream commodity in shopping software is inquired according to the execution flow information '1' exemplified in the previous embodiment; 2. acquiring an address of a company; 3. selecting one with larger preferential strength, higher score and lower unit price; 4. and ordering the selected commodity, wherein the execution flow information comprises 4 steps, and the operations corresponding to the information of each step can be determined through a large language model according to a knowledge database and an operation database and executed.

When determining and executing the step 1, inquiring the information of ice cream commodities in shopping software, firstly determining that a first operation is to acquire shopping application, and after executing the first operation according to the knowledge database, obtaining an inquired result (namely an operation result of the first operation) in the knowledge database, wherein the shopping application is to acquire application j and application k; then determining a second operation 'to ask a question to the user' through a large language model according to the operation result of the first operation 'shopping application has application j and application k' and an operation database: whether the application j or the application k is used for purchase is carried out, and after the second operation is carried out, the result of the user feedback aiming at the prompt (namely, the operation result of the second operation) can be obtained, namely, the application k is used for purchase; then determining a third operation application through the large language model according to the operation result of the second operation, namely 'using application k to purchase', and the operation database: applying k, searching commodities and keywords: the ice cream can obtain an operation result of 'the commodity information list of the ice cream searched by applying k' after the third operation is executed, and then the step '1' is completed.

When determining and executing step "2. Obtain company address", it may be determined first that "query knowledge database: the first operation of the company address ' can obtain the result (namely the operation result of the first operation) which is inquired in the knowledge database ' none ' after the first operation is executed according to the knowledge database; in this case, according to the operation result "none" of the first operation and the operation database, it is determined through the large language model that the second operation "ask a question to the user: what the company's shipping address is, "after the second operation is performed, the result of the user feedback for the prompt (i.e., the operation result of the second operation) may be obtained, where the company's shipping address is: cell D "of street C of zone B of city a, and thus step" 2 "is completed, the address of the company is obtained.

When the step 3 is determined and executed, the operation of determining the commodity with larger preference and lower unit price from the commodity information list obtained in the step 1 can be determined, and after the operation is executed, the operation result of determining the commodity can be obtained: ice cream f).

When determining and executing step "4. Order the selected merchandise", it may be determined that "operation application: k, purchasing commodity: after the operation of the ice cream f and the receiving address is performed, an operation result of 'completed user intention' can be obtained, and thus, the user intention contained in the information to be processed is realized.

In some embodiments, considering that the electronic device may have accuracy problems in understanding intention and automatically executing intention for the input information to be processed, the electronic device may further display a scoring interface after determining, through the large language model and according to the knowledge database and the operation database, a target operation corresponding to the execution flow information and executing the target operation; acquiring data input in a scoring interface as scoring data aiming at the current intention execution; the scoring data is then fed back to the server. Therefore, the server can collect the scores of the user feedback of the electronic equipment after intention understanding and automatic intention execution, and further can determine the improvement direction of the function of the electronic equipment according to the collected scores, so that the complicated intention of the user can be better understood and automatically executed after the function of the electronic equipment is improved.

According to the intention executing method provided by the embodiment of the application, through acquiring the input information to be processed, aiming at the information to be processed, the understanding capability of the large language model is utilized, the intention is understood through the large language model, the information to be processed is generated, then the operation corresponding to each step information in the plurality of step information of executing the flow information is determined according to the knowledge database and the operation database through the large language model, and the operation corresponding to each step information is executed, so that the intention of a user with complex understanding can be realized, and the complex intention of the user can be better achieved.

Referring to fig. 4, fig. 4 is a schematic flow chart of an intended execution method according to still another embodiment of the present application. The method for executing intention is applied to the electronic device, and will be described in detail with respect to the flowchart shown in fig. 4, where the method for executing intention may specifically include the following steps:

step S410: and acquiring the input information to be processed.

Step S420: and generating execution flow information corresponding to the information to be processed through a large language model, wherein the execution flow information is used for realizing user intention corresponding to the information to be processed.

In the embodiment of the present application, step S410 and step S420 may refer to the content of the foregoing embodiment, and are not described herein.

Step S430: and determining the target operation corresponding to the execution flow information through the large language model according to an operation database, wherein the operation database is constructed according to functions realized by input operations in different application interfaces.

Step S440: and displaying operation prompt information corresponding to the target operation, wherein the operation prompt information is used for prompting whether the target operation is executed.

Step S450: and responding to the determining operation aiming at the operation prompt information, and executing the target operation according to the knowledge database, wherein the knowledge database is constructed according to the equipment information of the electronic equipment.

In the embodiment of the application, considering that the electronic device may have misunderstanding intention and related operations related to information such as property and privacy of the user in the process of understanding and executing the intention of the user, the electronic device determines, through a large language model and according to a knowledge database and an operation database, a target operation corresponding to the execution flow information, and when executing the target operation, the electronic device may display operation prompt information corresponding to the target operation after determining, through the large language model and according to the operation database, the target operation corresponding to the execution flow information, where the operation prompt information is used for prompting whether to execute the target operation; after the operation prompt information corresponding to the target operation is displayed, under the condition that the determination operation aiming at the operation prompt information is detected, the target operation can be executed according to the knowledge database in response to the determination operation; if the determining operation for the above operation prompt information is not detected within the target duration, the above target operation may not be executed, and the flow is ended, that is, the process of performing intention understanding and achieving user intention according to the above information to be processed is ended. Therefore, the above target operation can be executed only when the user determines to execute the above target operation, so that the accuracy of the intended execution can be ensured, and the user can be prevented from suffering related loss.

According to the intention execution method provided by the embodiment of the application, the input information to be processed is obtained, the understanding capability of the large language model is utilized for the information to be processed, the intention is understood through the large language model, the information to be processed is generated, then the operation corresponding to the execution flow information is determined through the large language model, the knowledge database and the operation database which are built in advance, and the determined operation is executed, so that the intention of a user can be achieved, the complex user intention can be further realized, the use experience of the user when the user uses relevant service for intention understanding of the electronic equipment is improved, and the usability of the electronic equipment is enhanced; in addition, before the operation corresponding to the execution flow information is executed, the electronic equipment also outputs related prompt information, and the operation is executed only when the determination operation input by the user is detected, so that the accuracy of the intention execution can be ensured, and the user can be prevented from suffering related loss.

Referring to fig. 5, a block diagram of an apparatus 500 for executing an intent according to an embodiment of the present application is shown. The intention execution apparatus 500 is applied to an upper electronic device, and the intention execution apparatus 500 includes: an information acquisition module 510, an intention recognition module 520, and an operation execution module 530. The information obtaining module 510 is configured to obtain input information to be processed; the intention recognition module 520 is configured to generate, through a large language model, execution flow information corresponding to the information to be processed, where the execution flow information is used to implement a user intention corresponding to the information to be processed; the operation execution module 530 is configured to determine, through the large language model, a target operation corresponding to the execution flow information according to a knowledge database and an operation database, and execute the target operation, where the knowledge database is constructed according to device information of the electronic device, and the operation database is constructed according to functions implemented by input operations in different application interfaces.

In some embodiments, the operation execution module 530 may be specifically configured to sequentially determine each operation of the plurality of operations through the large language model according to the knowledge database and the operation database, and sequentially execute each operation.

In a possible implementation manner, the operation execution module 530 may be further configured to determine, for a first operation of the plurality of operations, through the large language model and according to the operation database, the first operation, and execute the first operation according to the knowledge database, and obtain an operation result of the first operation; and aiming at an Nth operation in the plurality of operations, determining the Nth operation according to the operation result of the N-1 th operation and the operation database through the large language model, executing the Nth operation according to the knowledge database, and obtaining the operation result of the Nth operation.

Optionally, the operation execution module 530 may be further configured to generate, for a first operation of the plurality of operations, first input information for inputting the large language model according to target prompt information, the operation database, and the execution flow information, where the target prompt information is used to prompt the large language model to output an operation for implementing the execution flow information according to data in the operation database; and inputting the first input information into the large language model to obtain first operation information output by the large language model, and determining an operation corresponding to the first operation information as the first operation.

Alternatively, the operation execution module 530 may be further configured to generate, for an nth operation of the plurality of operations, third input information for inputting the large language model according to second input information input to the large language model when the nth-1 operation is determined, and an operation result of the nth-1 operation; and inputting the third input information into the large language model to obtain second operation information output by the large language model, and determining an operation corresponding to the second operation information as the Nth operation, wherein N is a positive integer greater than 1.

Optionally, the operation execution module 530 may be further configured to execute the first operation according to the knowledge database; if the operation result is not obtained after the first operation is executed, outputting first prompt information, wherein the first prompt information is used for prompting the input of the operation result corresponding to the first operation; and acquiring input first result information as an operation result corresponding to the first operation.

In some embodiments, the execution flow information includes step information corresponding to a plurality of execution steps, and the operation execution module 530 may be specifically configured to determine, according to the knowledge database and the operation database, an operation corresponding to each step information, and execute the operation corresponding to each step information.

In some embodiments, the operation execution module 530 may be specifically configured to determine, through the large language model and according to the operation database, a target operation corresponding to the execution flow information; displaying operation prompt information corresponding to the target operation, wherein the operation prompt information is used for prompting whether the target operation is executed or not; and responding to the determining operation aiming at the operation prompt information, and executing the target operation according to the knowledge database.

In some embodiments, the intent execution device 500 may further include an interface display module, a data acquisition module, and a data feedback module. The interface display module is used for displaying a scoring interface after the target operation corresponding to the execution flow information is determined and the target operation is executed according to the knowledge database and the operation database through the large language model; the data acquisition module is used for acquiring data input in the evaluation interface and taking the data as the evaluation data aiming at the current intention execution; the data feedback module is used for feeding back the evaluation data to the server.

It will be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working process of the apparatus and modules described above may refer to the corresponding process in the foregoing method embodiment, which is not repeated herein.

In several embodiments provided herein, the coupling of the modules to each other may be electrical, mechanical, or other.

In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist alone physically, or two or more modules may be integrated into one module. The integrated modules may be implemented in hardware or in software functional modules.

In summary, according to the scheme provided by the application, by acquiring the input information to be processed, generating the execution flow information corresponding to the information to be processed through the large language model, wherein the execution flow information is used for realizing the user intention corresponding to the information to be processed, determining the target operation corresponding to the execution flow information and executing the target operation according to the knowledge database and the operation database, wherein the knowledge database is constructed according to the equipment information of the electronic equipment, and the operation database is constructed according to the functions realized by the input operation in different application interfaces. Therefore, the input information with complex user intention can be identified as the execution flow information, and then the target operation corresponding to the execution flow information is determined and executed by utilizing the large language model, the knowledge database and the operation database, so that the complex user intention can be understood and realized.

Referring to fig. 6, a block diagram of an electronic device according to an embodiment of the present application is shown. The electronic device 100 may be an electronic device capable of running an application program, such as a smart phone, a tablet computer, a smart watch, an electronic book, etc. The electronic device 100 in this application may include one or more of the following components: a processor 110, a memory 120, and one or more applications, wherein the one or more applications may be stored in the memory 120 and configured to be executed by the one or more processors 110, the one or more applications configured to perform the method as described in the foregoing method embodiments.

Processor 110 may include one or more processing cores. The processor 110 utilizes various interfaces and lines to connect various portions of the overall electronic device 100, perform various functions of the electronic device 100, and process data by executing or executing instructions, programs, code sets, or instruction sets stored in the memory 120, and invoking data stored in the memory 120. Alternatively, the processor 110 may be implemented in hardware in at least one of digital signal processing (Digital Signal Processing, DSP), field programmable gate array (Field-Programmable Gate Array, FPGA), programmable logic array (Programmable Logic Array, PLA). The processor 110 may integrate one or a combination of several of a central processing unit (Central Processing Unit, CPU), a graphics processor (Graphics Processing Unit, GPU), and a modem, etc. The CPU mainly processes an operating system, a user interface, an application program and the like; the GPU is used for being responsible for rendering and drawing of display content; the modem is used to handle wireless communications. It will be appreciated that the modem may not be integrated into the processor 110 and may be implemented solely by a single communication chip.

The Memory 120 may include a random access Memory (Random Access Memory, RAM) or a Read-Only Memory (Read-Only Memory). Memory 120 may be used to store instructions, programs, code, sets of codes, or sets of instructions. The memory 120 may include a stored program area and a stored data area, wherein the stored program area may store instructions for implementing an operating system, instructions for implementing at least one function (such as a touch function, a sound playing function, an image playing function, etc.), instructions for implementing the various method embodiments described below, etc. The storage data area may also store data created by the electronic device 100 in use (e.g., phonebook, audiovisual data, chat log data), and the like.

Referring to fig. 7, a block diagram of a computer readable storage medium according to an embodiment of the present application is shown. The computer readable medium 800 has stored therein program code which can be invoked by a processor to perform the methods described in the method embodiments described above.

The computer readable storage medium 800 may be an electronic memory such as a flash memory, an EEPROM (electrically erasable programmable read only memory), an EPROM, a hard disk, or a ROM. Optionally, the computer readable storage medium 800 comprises a non-volatile computer readable medium (non-transitory computer-readable storage medium). The computer readable storage medium 800 has storage space for program code 810 that performs any of the method steps described above. The program code can be read from or written to one or more computer program products. Program code 810 may be compressed, for example, in a suitable form.

Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present application, and are not limiting thereof; although the present application has been described in detail with reference to the foregoing embodiments, one of ordinary skill in the art will appreciate that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not drive the essence of the corresponding technical solutions to depart from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An intended execution method, applied to an electronic device, comprising:

acquiring input information to be processed;

generating execution flow information corresponding to the information to be processed through a large language model, wherein the execution flow information is used for realizing user intention corresponding to the information to be processed;

and determining a target operation corresponding to the execution flow information through the large language model according to a knowledge database and an operation database, and executing the target operation, wherein the knowledge database is constructed according to the equipment information of the electronic equipment, and the operation database is constructed according to functions realized by input operations in different application interfaces.

2. The method according to claim 1, wherein the target operation includes a plurality of operations sequentially performed, the determining, by the large language model and according to a knowledge database and an operation database, the target operation corresponding to the execution flow information, and performing the target operation includes:

and sequentially determining each operation in the plurality of operations through the large language model according to the knowledge database and the operation database, and sequentially executing each operation.

3. The method of claim 2, wherein said sequentially determining each of said plurality of operations by said large language model and from said knowledge database and operations database and sequentially performing said each operation comprises:

determining a first operation in the plurality of operations through the large language model according to the operation database, executing the first operation according to the knowledge database, and obtaining an operation result of the first operation;

and aiming at an Nth operation in the plurality of operations, determining the Nth operation through the large language model according to an operation result of the N-1 th operation and the operation database, and executing the Nth operation according to the knowledge database to obtain an operation result of the Nth operation, wherein N is a positive integer greater than 1.

4. A method according to claim 3, wherein said determining a first operation of said plurality of operations, through said large language model, and from said operation database, comprises:

for a first operation in the plurality of operations, generating first input information for inputting the large language model according to target prompt information, the operation database and the execution flow information, wherein the target prompt information is used for prompting the large language model to output operations for realizing the execution flow information according to data in the operation database;

and inputting the first input information into the large language model to obtain first operation information output by the large language model, and determining an operation corresponding to the first operation information as the first operation.

5. The method of claim 3, wherein the determining, for an nth operation of the plurality of operations, the nth operation by the large language model and based on the operation result of the nth-1 operation and the operation database comprises:

generating third input information for inputting the large language model according to second input information input to the large language model when determining the (N-1) -th operation and operation results of the (N-1) -th operation for the (N) -th operation of the plurality of operations;

And inputting the third input information into the large language model to obtain second operation information output by the large language model, and determining an operation corresponding to the second operation information as the Nth operation.

6. A method according to claim 3, wherein said performing said first operation based on said knowledge database and obtaining an operation result of said first operation comprises:

executing the first operation according to the knowledge database;

if the operation result is not obtained after the first operation is executed, outputting first prompt information, wherein the first prompt information is used for prompting the input of the operation result corresponding to the first operation;

and acquiring input first result information as an operation result corresponding to the first operation.

7. The method according to any one of claims 1-6, wherein the execution flow information includes step information corresponding to a plurality of execution steps, the determining, by the large language model and according to a knowledge database and an operation database, a target operation corresponding to the execution flow information, and executing the target operation, includes:

and determining the operation corresponding to each step information through the large language model according to the knowledge database and the operation database, and executing the operation corresponding to each step information.

8. The method according to any one of claims 1-6, wherein determining, by the large language model and according to a knowledge database and an operation database, a target operation corresponding to the execution flow information, and executing the target operation, includes:

determining target operation corresponding to the execution flow information through the large language model according to the operation database;

displaying operation prompt information corresponding to the target operation, wherein the operation prompt information is used for prompting whether the target operation is executed or not;

and responding to the determining operation aiming at the operation prompt information, and executing the target operation according to the knowledge database.

9. The method according to any one of claims 1-6, wherein after said determining the target operation corresponding to the execution flow information through the large language model and according to a knowledge database and an operation database, and performing the target operation, the method further comprises:

displaying a scoring interface;

acquiring data input in the evaluation interface as evaluation data aiming at the current intention execution;

and feeding the evaluation data back to a server.

10. An intended execution apparatus, characterized by being applied to an electronic device, comprising: an information acquisition module, an intention recognition module and an operation execution module, wherein,

the information acquisition module is used for acquiring input information to be processed;

the intention recognition module is used for generating execution flow information corresponding to the information to be processed through a large language model, and the execution flow information is used for realizing user intention corresponding to the information to be processed;

the operation execution module is used for determining a target operation corresponding to the execution flow information through the large language model according to a knowledge database and an operation database, and executing the target operation, wherein the knowledge database is constructed according to the equipment information of the electronic equipment, and the operation database is constructed according to functions realized by input operations in different application interfaces.

11. An electronic device, comprising:

one or more processors;

a memory;

one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs configured to perform the method of any of claims 1-9.

12. A computer readable storage medium, characterized in that the computer readable storage medium has stored therein a program code, which is callable by a processor for executing the method according to any one of claims 1-9.