Intention recognition method, electronic equipment and computer readable storage medium

Abstract

The invention discloses an intention recognition method, electronic equipment and a computer readable storage medium. The method comprises the steps of obtaining to-be-recognized data; performing intention recognition processing on the to-be-recognized data to obtain an initial intention corresponding to the to-be-recognized data; and in response to the situation that the intention number of the initial intentions is greater than a preset number threshold, performing verification processing on the initial intentions according to the to-be-identified data to obtain a target intention. Therefore, the accuracy and reliability of intention recognition can be ensured.

Description

Technical Field

[0001] This invention relates to the field of dialogue processing technology, and more particularly to an intent recognition method, an electronic device, and a computer-readable storage medium. Background Technology

[0002] Intent recognition aims to understand user needs and provide corresponding services. By accurately understanding user intent, user experience and system interaction quality can be significantly improved.

[0003] In practical applications, due to the diversity and complexity of natural language, systems may identify multiple related or conflicting intents when recognizing intents. Current methods for resolving intent conflicts often rely on manual verification, thus impacting the efficiency and automation of intent recognition. Summary of the Invention

[0004] The main technical problem addressed by this application is to provide an intent recognition method, electronic device, and computer-readable storage medium that can improve the accuracy and efficiency of intent recognition.

[0005] To solve the above-mentioned technical problems, one technical solution adopted in this application is: to provide an intent recognition method, the intent recognition method comprising: acquiring data to be recognized; performing intent recognition processing on the data to be recognized to obtain an initial intent corresponding to the data to be recognized; and, in response to the number of intents of the initial intent being greater than a preset number threshold, performing verification processing on each initial intent according to the data to be recognized to obtain a target intent.

[0006] To solve the above-mentioned technical problems, another technical solution adopted in this application is: to provide an electronic device, including a memory and a processor, wherein the memory stores program instructions, and the processor retrieves the program instructions from the memory to execute the above-mentioned intent recognition method.

[0007] To solve the above-mentioned technical problems, another technical solution adopted in this application is to provide a computer-readable storage medium including program data, which is used to implement the above-mentioned intent recognition method when executed by a processor.

[0008] The intent recognition method of this application acquires data to be recognized; performs intent recognition processing on the data to be recognized to obtain the initial intent corresponding to the data to be recognized; in response to the number of initial intents exceeding a preset threshold, verifies each initial intent based on the data to be recognized to obtain the target intent. Thus, when multiple initial intents are recognized, verification is performed using the data to be recognized, ensuring the accuracy and reliability of intent recognition. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly described below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort, wherein: Figure 1 This is a flowchart illustrating an exemplary embodiment of the intent recognition method shown in this application; Figure 2 yes Figure 1 A flowchart illustrating an exemplary embodiment of the intent recognition method after step S130 is shown. Figure 3 yes Figure 1 A flowchart illustrating an exemplary embodiment of step S130 in the intent recognition method is shown. Figure 4 This is a schematic flowchart illustrating a specific embodiment of the intent recognition method shown in this application; Figure 5 This is a schematic diagram of an exemplary embodiment of the intent recognition device shown in this application; Figure 6 This is a schematic diagram of the structure of an embodiment of the electronic device provided in this application; Figure 7 This is a schematic diagram of an embodiment of the computer-readable storage medium provided in this application. Detailed Implementation

[0010] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It is understood that the specific embodiments described herein are only for explaining this application and not for limiting it. Furthermore, it should be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all structures. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0011] First, it should be noted that intent recognition is a key task in natural language processing, aiming to analyze and determine the purpose of input in order to provide more intelligent and efficient services. Based on this, embodiments of this application propose an intent recognition method, electronic device, and computer-readable storage medium. When multiple initial intents are output, the data to be recognized is used to verify these initial intents, resulting in a more accurate target intent. For details, please refer to [reference needed]. Figure 1 , Figure 1 This is a flowchart illustrating an exemplary embodiment of the intent recognition method shown in this application.

[0012] The execution subject of the intent recognition method can be a terminal device, a server, or other processing devices. Among them, the terminal device can be a user equipment (UE), a computer, a mobile device, a user terminal, a terminal, a cellular phone, a cordless phone, a personal digital assistant (PDA), a handheld device, a computing device, a vehicle-mounted device, a wearable device, etc. The execution subject of the intent recognition method can also be an intent recognition device. In some possible implementation manners, the intent recognition method can be implemented by a processor calling computer-readable instructions stored in a memory.

[0013] Specifically, the intent recognition method of this embodiment includes the following steps: S110: Obtain the data to be recognized.

[0014] The data to be recognized refers to the data that needs to be subjected to intent recognition. Exemplarily, the intent recognition device receives the user's input from the client to obtain the data to be recognized; the intent recognition device can also obtain the data to be recognized from the internal storage. The data type of the data to be recognized can include but is not limited to text, voice, or image, etc. In some embodiments, the intent recognition device uses the acquired raw data as the data to be recognized. In other embodiments, the intent recognition device can preprocess the acquired raw data to obtain the data to be recognized.

[0015] As an example, when the data type being processed is text, the preprocessing includes but is not limited to removing special characters, stop words, word segmentation, key information processing, etc. Among them, stop words are function words such as "de", "di", "de" that have little impact on semantics. Perform lexical analysis on the raw data and perform word segmentation processing on the raw data according to lexical logic to obtain at least one word segment. For example, segment the sentence "I want to book a flight ticket to somewhere tomorrow" to get "I", "want", "book", "tomorrow", "go", "somewhere", "flight ticket", thereby converting the raw data into a form that is easier to process. Key information includes time information and location information, etc. Processing the time information in the raw data can extract the corresponding time point from the events with strong time correlation in the raw data. For example, National Day in 24 years can be processed as October 1, 2024; the information such as the area, location, or place where the raw data appears can be processed into the corresponding coding information. For example, convert a certain street into the corresponding coding of that street.

[0016] S120: Perform intent recognition processing on the data to be recognized to obtain the initial intent corresponding to the data to be recognized.

[0017] The initial intent is the intent obtained from the initial identification of the data to be identified. In some embodiments, the intent recognition device can process the data to be identified using a machine learning model to obtain the initial intent. In other embodiments, the intent recognition device can process the data to be identified using a deep learning model to obtain the initial intent. In still other embodiments, the intent recognition device can also process the data to be identified using a large model to obtain the initial intent.

[0018] The intent recognition device can pre-set a preset intent library, which stores several preset intents. When the data to be recognized is processed for intent recognition using the above method, an initial intent corresponding to the data to be recognized can be matched from these preset intents. The preset intents may include other intents. When performing intent recognition on the data to be recognized, if a corresponding intent cannot be matched from the preset intents (excluding other intents), then the other intent is used as the initial intent of the data to be recognized. For example, before using a machine learning model to process the data for intent recognition, the machine learning model can be pre-trained using a preset corpus to obtain a pre-trained machine learning model; then, the pre-trained machine learning model is used to process the data for intent recognition to obtain the initial intent. The preset corpus includes the preset intent library.

[0019] It should be noted that, due to the complexity of natural language, multiple initial intentions may be identified when performing intent recognition on the data to be identified. When multiple initial intentions are identified, step S130 can be executed to verify the intent in order to obtain a more accurate intent.

[0020] S130: In response to the number of initial intents exceeding a preset threshold, each initial intent is verified based on the data to be identified to obtain the target intent.

[0021] The intent count refers to the number of initial intents output when performing intent recognition processing on the data to be recognized. For example, after acquiring the initial intent, the intent recognition device performs statistical processing on the initial intent to obtain the intent count of the initial intent.

[0022] A preset quantity threshold is used to limit the maximum number of initial intents. In some embodiments, the preset quantity threshold corresponding to the initial intent can be manually set in advance. For example, the preset quantity threshold can be 1. When the number of initial intents is greater than 1, each initial intent is verified based on the data to be identified to obtain the target intent.

[0023] The target intent is the final intent of the data to be identified. For example, when the number of initial intents is less than or equal to a preset threshold, the initial intent is taken as the target intent of the data to be identified; when the number of initial intents is greater than the preset threshold, each initial intent can be verified based on the data to be identified to obtain the target intent.

[0024] As can be seen, the intent recognition method in this application acquires data to be recognized; performs intent recognition processing on the data to be recognized to obtain the initial intent corresponding to the data to be recognized; in response to the number of initial intents exceeding a preset threshold, verifies each initial intent based on the data to be recognized to obtain the target intent. Therefore, when multiple initial intents are recognized, verification is performed using the data to be recognized, ensuring the accuracy and reliability of intent recognition.

[0025] Furthermore, during initial intent recognition, a large model can be used to process the intent of the data to be recognized. Specifically, the steps for processing the data to be recognized to obtain the initial intent corresponding to the data include: inputting the data to be recognized into the large model for semantic parsing to obtain at least one intent output by the large model; and determining the at least one intent output by the large model as the initial intent corresponding to the data to be recognized. This fully utilizes the natural language processing capabilities of the large model and improves the efficiency of intent output.

[0026] In this embodiment, the large model used can be a pre-trained large model. The intent recognition device fine-tunes the large model using a pre-organized preset corpus to obtain a pre-trained large model, thereby giving the large model a strong intent recognition capability. However, the output of the large model has a certain degree of uncertainty. Therefore, this embodiment further proposes that when the number of initial intents output by the large model exceeds a preset threshold, the initial intent is verified based on the data to be recognized to obtain the target intent, thus ensuring the accuracy of the target intent.

[0027] In other embodiments, the intent recognition device may also use machine learning models or other methods to perform intent recognition processing on the data to be recognized, so as to obtain the initial intent corresponding to the data to be recognized.

[0028] Based on the above embodiments, the embodiments of this application adopt... Figure 2 The flowchart illustrates in detail how to select the appropriate task execution tool based on the target intent. Please refer to [link / reference]. Figure 2 , Figure 2 yes Figure 1 The illustrated flowchart shows an exemplary embodiment of the intent recognition method after step S130. Specifically, after step S130, the method further includes the following steps: S210: Match the target task execution tool from the preset mapping relationship library according to the target intent. The preset mapping relationship library includes the correspondence between preset intents and preset task execution tools.

[0029] The preset mapping relationship library includes the correspondence between preset intents and preset task execution tools. The preset intents in the preset mapping relationship library can be preset intents from the preset intent library. For example, each preset intent in the preset intent library is pre-set with a corresponding preset task execution tool, and this correspondence is stored in the preset mapping relationship library. As an example, the table structure of the preset mapping relationship library can be found in Table 1:

[0030] Table 1 As can be seen, when establishing the correspondence between preset intents and preset task execution tools, the intent name of the preset intent and the tool name, tool input parameter structure, and tool output parameter structure of the preset task execution tool can be used as table fields. The field attributes include a data type of varchar, a data length of 255, and non-null values. The correspondence between preset intents and preset task execution tools can also be considered as the correspondence between the preset intent and the tool name, tool input parameter structure, and tool output parameter structure.

[0031] Furthermore, the preset mapping relationship library can be stored in a relational database and supports dynamic adjustment. New preset intents and corresponding preset task execution tools can be added according to business needs, or existing correspondences in the preset mapping relationship library can be modified.

[0032] The target task execution tool is used to realize the target intent corresponding to the data to be identified. For example, after identifying the target intent, the intent recognition device calculates the similarity between the target intent and preset intents in a preset mapping database, obtaining several similarity scores. Preset intents with similarity scores greater than a similarity threshold are selected as target preset intents, and the preset task execution tool corresponding to the target preset intent is selected as the target execution tool. For example, the preset intent with the highest similarity score can be selected as the target preset intent.

[0033] S220: Execute the target intent of the data to be identified through the target task execution tool to obtain the target execution result.

[0034] The target execution result is the desired outcome of the data to be identified. For example, the intent recognition device can execute the target intent of the data to be identified using a target task execution tool to obtain the target execution result. Specifically, the target task execution tool includes a tool input parameter structure, which adaptively fills in the parameters required by the target intent based on the tool input parameter structure. The corresponding task processing module is dynamically invoked through a reflection mechanism to execute the task and obtain the target execution result.

[0035] As can be seen, the intent recognition method in this application embodiment matches the target task execution tool from a preset mapping relationship library based on the target intent. The preset mapping relationship library includes the correspondence between preset intents and preset task execution tools. The target task execution tool executes the target intent of the data to be identified to obtain the target execution result. Thus, by querying the preset mapping relationship library to obtain the target task execution tool corresponding to the target intent, the correspondence can be flexibly added or modified.

[0036] Based on the above embodiments, the embodiments of this application adopt... Figure 3 The flowchart details how the initial intent is validated using the data to be identified. Please refer to [link / reference]. Figure 3 , Figure 3 yes Figure 1 The illustrated flowchart shows an exemplary embodiment of step S130 in the intent recognition method. Specifically, step S130 further includes the following steps: S310: Obtain at least one initial intent class data by matching from the preset intent class database according to each initial intent.

[0037] A preset intent class database is used to store several preset intent class data. Each preset intent class data corresponds to a preset intent; that is, in the preset intent class database, one preset intent can correspond to one or more preset intent class data, and the intents of one or more preset intent class data all point to that preset intent. For example, the preset intent class database can be obtained by training with several historical data sets. Specifically, the database is trained on each historical data set to obtain the preset intent corresponding to each historical data set, and the historical data is identified as the preset intent class data under the corresponding preset intent. In other embodiments, the preset intent class database can also store several data vectors corresponding to the preset intent class data. Specifically, after the intent recognition device acquires the preset intent class data under each preset intent, it performs vectorization processing on each preset intent class data set to obtain the data vector corresponding to each preset intent class data set; and saves each preset intent class data set and its corresponding data vector to the preset intent class database. As an example: { "intents": { "Preset Intent Data 1": { "feature_vector": [0.1, 0.2, 0.3, 0.4, 0.5] }, "Preset Intent Data 2": { "feature_vector": [0.2, 0.3, 0.4, 0.5, 0.6] }, "Preset Intent Data 3": { "feature_vector": [0.3, 0.4, 0.5, 0.6, 0.7] } } } For ease of querying, the preset intent category database can be categorized and stored according to preset intents, and preset intent category data belonging to the same preset intent can be saved in the same area.

[0038] The initial intent data is data obtained by matching the initial intent from a preset intent class database. For example, after acquiring the initial intent, the intent recognition device matches the initial intent with preset intents in the preset intent class database, and uses the preset intent class data under the matched preset intents as the initial intent class data for each initial intent. As an example, the preset intent class database includes vehicle query data (preset intent is vehicle query), traffic violation query data (preset intent is traffic violation query), and accident query data (preset intent is accident query), etc. When each initial intent includes both vehicle query and traffic violation query, the vehicle query data and traffic violation query data can be used as the initial intent class data. Therefore, filtering by initial intent can reduce the computational burden.

[0039] S320: Obtain the target intent class data by matching it with each initial intent class data based on the data to be identified.

[0040] The target intent class data is intent class data obtained by filtering the initial intent class data. For example, after the intent recognition device obtains at least one initial intent class data from a preset intent class database based on the initial intent, it matches the data to be recognized with each initial intent class data to obtain the target intent class data. As an example, the intent recognition device can calculate the vector similarity between the data to be recognized and each initial intent class data to obtain at least one second vector similarity; the initial intent class data whose second vector similarity is greater than a second preset similarity threshold is determined as the target intent class data. As another example, the intent recognition device can also determine the similarity between the data to be recognized and each initial intent class data through a machine learning model or a large model, and determine the target intent class data based on the similarity.

[0041] Specifically, the intent recognition device acquires the data vector of the data to be recognized and the data vectors of each initial intent class data. Then, it calculates a second vector similarity between the data vector of the data to be recognized and the data vectors of each initial intent class data, and identifies the initial intent class data whose second vector similarity is greater than a second preset similarity threshold as the target intent class data. For example, the initial intent class data corresponding to the maximum second vector similarity can be used as the target intent class data. The data vectors of the data to be recognized and / or each preset intent class data in the preset intent class database can be acquired through machine learning models or large models.

[0042] S330: In response to the target intent class data meeting the preset requirements, the target intent is selected from each initial intent based on the target intent class data.

[0043] The preset requirements can be set according to actual needs. In some embodiments, the target intent data is selected by calculating the similarity between the data to be identified and the initial intent data; therefore, a similarity threshold can be set as a preset requirement. Specifically, the intent recognition device obtains the third vector similarity between the target intent data and the data to be identified; in response to the third vector similarity being greater than the third preset similarity threshold, it confirms that the target intent data meets the preset requirements; in response to the third vector similarity being less than or equal to the third preset similarity threshold, it confirms that the target intent data does not meet the preset requirements. Thus, determining the target intent data through the third vector similarity can improve the accuracy of the target intent data. In other embodiments, the intent recognition device can also use other matching methods to obtain the matching value between the target intent data and the data to be identified; if the matching value is greater than a preset matching value, it is determined that the target intent data meets the preset requirements.

[0044] When the target intent data meets preset requirements, the target intent can be selected from the initial intents based on the target intent data. Specifically, the target intent to which the target intent data belongs is matched with each initial intent to obtain the target intent. When the target intent data does not meet the preset requirements, the intent recognition is confirmed to have failed, and a recognition failure message is returned directly. In some other embodiments, when the target intent data does not meet the preset requirements, the target intent of the data to be recognized can be determined manually.

[0045] In other embodiments, the intent recognition device obtains at least one initial intent class data by matching each initial intent from a preset intent class database; it then obtains target intent class data by matching the data to be recognized from each initial intent class data; and in response to the target intent class data not meeting preset requirements, it performs matching processing between the data to be recognized and each preset intent class data in the preset intent class database to obtain the target intent. Thus, when the target intent class data does not meet preset requirements, matching is performed through the preset intent class database, ensuring the stability of the target intent output.

[0046] It should be noted that when the target intent data does not meet the preset requirements, the initial intent result is considered incorrect. In this case, the data to be identified can be matched with each preset intent data in the preset intent data database. The target intent data is selected from the preset intent data in the preset intent data database, and the intent corresponding to the target intent data is taken as the target intent. In some embodiments, the intent recognition device can match the data to be identified with each preset intent data in the preset intent data database using keyword matching. In other embodiments, the intent recognition device can calculate the first vector similarity between the data to be identified and each preset intent data in the preset intent data database. In response to the presence of preset intent data with a first vector similarity greater than a first preset similarity threshold, the preset intent corresponding to the preset intent data is determined as the target intent.

[0047] The first vector similarity is the vector similarity between the data to be identified and each preset intent class data. For example, the data vector of the data to be identified and the data vectors of each preset intent class data can be obtained respectively; the vector similarity between the data vector of the data to be identified and the data vectors of each preset intent class data can be calculated respectively to obtain several first vector similarities.

[0048] The first preset similarity threshold can be preset to ensure the accuracy of the selected preset intent class data. In some embodiments, if the preset intent class data with a first vector similarity greater than the first preset similarity threshold is greater than 1, the preset intent class data corresponding to the maximum first vector similarity can be selected, and the preset intent corresponding to this preset intent class data is determined as the target intent. In other embodiments, after obtaining the first vector similarity between the data to be identified and each preset intent class data, the intent recognition device uses the preset intent class data corresponding to the maximum first vector similarity as the target preset intent class data; if the first vector similarity corresponding to the target preset intent class data is greater than the first preset similarity threshold, the preset intent corresponding to the target preset intent class data is used as the target intent. Furthermore, if the first vector similarity corresponding to the target preset intent class data is less than or equal to the first preset similarity threshold, it is considered that the data to be identified cannot select a corresponding intent. In this case, other intents can be defaulted to the target intent of the data to be identified, and the selection of the task execution tool is no longer performed; a query failure message is directly returned.

[0049] The first, second, and third preset similarity thresholds can be set to the same or different. It should be noted that after calculating the second vector similarity between the data to be identified and the initial intent class data, the intent recognition device can determine the third vector similarity between the data to be identified and the target intent class data, and the first vector similarity between the data to be identified and the preset intent class data, through the corresponding second vector similarity. If the similarity cannot be obtained from the second vector similarity, it is calculated separately.

[0050] As can be seen, the intent recognition device in this application first filters out initial intent class data from a preset intent class database based on the initial intent, which can reduce the computational burden for the subsequent process. Then, it obtains target intent class data by matching the data to be recognized, thus verifying the initial intent in a simple and effective way and improving the accuracy of the target intent.

[0051] In other embodiments, the intent recognition device of this application may also perform matching processing on the data to be recognized and each initial intent, and obtain the target intent from each initial intent.

[0052] To elaborate on the data synchronization method of this application, Figure 4 The flowchart shown below provides further explanation, as detailed below: Receive initial data input from the user. The data type of the initial data can be text, audio, image, etc. The following example uses text as the data type.

[0053] The initial data is preprocessed to obtain the data to be identified. The preprocessing methods include, but are not limited to: removing special characters, segmenting the initial data into words, and processing the time and location information in the initial data. The large model is used to perform initial intent recognition on the data to be identified, and at least one initial intent is obtained. It is then determined whether to enter the intent conflict handling process. The determination process includes: counting the number of initial intents output by the large model, and determining whether to enter the intent conflict handling process based on the number of initial intents. If the number of initial intents is less than or equal to a preset threshold, it is determined that the intent conflict handling process will not be entered, and the initial intent is directly used as the target intent of the text to be identified. If the number of initial intents is greater than the preset threshold, it is determined that the intent conflict handling process will be entered. The intent conflict handling process includes: inputting the data to be identified into a large model for vectorization to obtain the data vector of the data to be identified; and calculating similarity using a preset intent class database to obtain the target intent. Specifically, based on the initial intent, initial intent class data corresponding to the preset intent is obtained from the preset intent class database; the data vector of the data to be identified is compared with the data vectors of each initial intent class data to calculate vector similarity, and the initial intent class data whose second vector similarity is greater than the second preset similarity threshold is selected as the target intent class data; it is determined whether the third vector similarity corresponding to the target intent class data is greater than the third preset similarity threshold; if the third vector similarity corresponding to the target intent class data is greater than the third preset similarity threshold, the target intent is selected from each initial intent according to the preset intent corresponding to the target intent class data; if the target intent class data corresponds to... If the third vector similarity is less than or equal to the third preset similarity threshold, then the vector similarity of the data to be identified and each preset intent class in the preset intent class database is recalculated to obtain the first vector similarity, and the preset intent class with the largest first vector similarity is selected as the target preset intent class; if the first vector similarity of the target preset intent class is greater than the first preset similarity threshold, then the preset intent corresponding to the target preset intent class is taken as the target intent; if the first vector similarity of the target preset intent class is less than or equal to the first preset similarity threshold, it means that no intent has been selected, and other intents can be set as the target intent by default. After obtaining the target intent, the target task execution tool is matched from the preset mapping relationship library according to the target intent. Before this, it can be determined whether the target intent is another intent. If it is, a tool selection is performed and a recognition failure message is returned directly; otherwise, a tool selection is performed and the target task execution tool is obtained. Then, the corresponding task processing module is dynamically called through the reflection mechanism to execute the specific task and obtain the target execution result.

[0054] Please see Figure 5 , Figure 5This is a schematic diagram of an exemplary embodiment of the intent recognition device shown in this application. The intent recognition device 500 includes an acquisition module 510, an identification module 520, and a verification module 530. The acquisition module 510 is used to acquire data to be identified; the identification module 520 is used to perform intent recognition processing on the data to be identified to obtain the initial intent corresponding to the data to be identified; the verification module 530 is used to verify each initial intent according to the data to be identified in response to the number of intents of the initial intents being greater than a preset number threshold, to obtain the target intent.

[0055] In the above scheme, the intent recognition device acquires the data to be recognized; performs intent recognition processing on the data to be recognized to obtain the initial intent corresponding to the data to be recognized; in response to the number of initial intents exceeding a preset threshold, each initial intent is verified based on the data to be recognized to obtain the target intent. Thus, when multiple initial intents are recognized, verification is performed using the data to be recognized, ensuring the accuracy and reliability of intent recognition.

[0056] The functions of each module can be found in the implementation examples of the intent recognition method, and will not be repeated here.

[0057] To implement the intent recognition method of the above embodiments, this application proposes another electronic device, please refer to [link / reference needed]. Figure 6 , Figure 6 This is a schematic diagram of the structure of an embodiment of the electronic device provided in this application.

[0058] Electronic device 600 includes memory 610 and processor 620, wherein memory 610 and processor 620 are coupled together.

[0059] The memory 610 is used to store program data, and the processor 620 is used to execute the program data to implement the intent recognition method of the above embodiments.

[0060] In this embodiment, processor 620 can also be referred to as a CPU (Central Processing Unit). Processor 620 may be an integrated circuit chip with signal processing capabilities. Processor 620 can also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The general-purpose processor can be a microprocessor, or processor 620 can be any conventional processor.

[0061] This application also provides a computer-readable storage medium, such as Figure 7 As shown, the computer-readable storage medium 700 is used to store program data 710, which, when executed by a processor, is used to implement the intent recognition method as described in the method embodiments of this application.

[0062] This application intends to identify the methods involved in the method embodiments. When implemented as software functional units and sold or used as independent products, these methods can be stored in a device, such as a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0063] The above description is merely an embodiment of this application and does not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. An intent recognition method, characterized in that, The intent recognition method includes: Obtain the data to be identified; The data to be identified is subjected to intent recognition processing to obtain the initial intent corresponding to the data to be identified; If the number of initial intentions exceeds a preset threshold, the initial intentions are verified based on the data to be identified to obtain the target intention.

2. The intent recognition method according to claim 1, characterized in that, After the step of verifying each initial intent based on the data to be identified to obtain the target intent when the number of intents responding to the initial intent is greater than a preset threshold, the method further includes: The target task execution tool is obtained by matching the target intent from a preset mapping relationship library, wherein the preset mapping relationship library includes the correspondence between preset intents and preset task execution tools; The target intent of the data to be identified is executed by the target task execution tool to obtain the target execution result.

3. The intent recognition method according to claim 1, characterized in that, The step of verifying each initial intent based on the data to be identified to obtain the target intent includes: At least one initial intent class data is obtained by matching each initial intent from the preset intent class database; The target intent class data is obtained by matching the data to be identified from each initial intent class data; In response to the target intent data meeting preset requirements, the target intent is selected from each initial intent based on the target intent data.

4. The intent recognition method according to claim 1, characterized in that, The step of verifying each initial intent based on the data to be identified to obtain the target intent includes: At least one initial intent class data is obtained by matching each initial intent from the preset intent class database; The target intent class data is obtained by matching the data to be identified from each initial intent class data; If the target intent data does not meet the preset requirements, the data to be identified is matched with each preset intent data in the preset intent database to obtain the target intent.

5. The intent recognition method according to claim 4, characterized in that, The step of matching the data to be identified with each preset intent class in the preset intent class database to obtain the target intent includes: Calculate the first vector similarity between the data to be identified and each preset intent class data in the preset intent class database; If there is a preset intent class data whose similarity to each first vector is greater than a first preset similarity threshold, then the preset intent corresponding to the preset intent class data is determined as the target intent.

6. The intent recognition method according to claim 3 or 4, characterized in that, The step of matching the target intent class data from each initial intent class data based on the data to be identified includes: The vector similarity between the data to be identified and each initial intent class data is calculated to obtain at least one second vector similarity. The initial intent class data whose similarity to the second vector is greater than the second preset similarity threshold is determined as the target intent class data.

7. The intent recognition method according to claim 3 or 4, characterized in that, The method further includes: Obtain the third vector similarity between the target intent class data and the data to be identified; If the similarity of the third vector is greater than a third preset similarity threshold, then the target intent class data is confirmed to meet the preset requirements. If the similarity of the third vector is less than or equal to the third preset similarity threshold, then it is confirmed that the target intent class data does not meet the preset requirements.

8. The intent recognition method according to claim 1, characterized in that, The step of performing intent recognition processing on the data to be identified to obtain the initial intent corresponding to the data to be identified includes: The data to be identified is input into a large model for semantic parsing to obtain at least one intent output by the large model; At least one intent output by the large model is determined as the initial intent corresponding to the data to be identified.

9. An electronic device, characterized in that, include: A memory and a processor, wherein the memory stores program instructions, and the processor retrieves the program instructions from the memory to perform the method as claimed in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, include: The system stores program data, which, when executed by a processor, is used to implement the method as described in any one of claims 1-8.

Images