Data processing method, apparatus and device

Abstract

Embodiments of the present specification provide a data processing method, device and equipment, wherein the data processing method comprises: in the processing process of user demand, obtaining target demand data of at least one data modality and generating target demand description information corresponding to the target demand data based on the data modality, if a plurality of first demand description information stored includes first target description information matched with the target demand description information, obtaining an intent category corresponding to the first target description information, processing the target demand data according to the intent category corresponding to the first target description information and the target demand description information, and if the plurality of first demand description information does not include the first target description information, determining suspected intent category information of the target demand description information, and processing the target demand data according to the suspected intent category information, so as to realize accurate processing of user demand.

Description

Technical Field

[0001] This document relates to the field of data processing technology, and in particular to a data processing method, apparatus and equipment. Background Technology

[0002] With the development of internet technology, applications can automatically process user requests based on intelligent systems. For example, if a user asks the application's intelligent system to "modify registration information," the system can automatically reply with the specific process for modifying the registration information or automatically execute the required modification procedure. However, as user needs become more complex and their expressions more diverse in real-world scenarios, higher demands are being placed on intelligent systems. Summary of the Invention

[0003] The purpose of the embodiments in this specification is to provide a data processing method, apparatus, and device that can improve the accuracy of processing user needs.

[0004] To achieve the above technical solution, the embodiments in this specification are implemented as follows: This specification provides one or more embodiments of a data processing method, the method comprising: acquiring target demand data to be processed, the target demand data corresponding to at least one data modality; generating target demand description information corresponding to the target demand data based on the data modality corresponding to the target demand data; if a plurality of stored first demand description information includes first target description information matching the target demand description information, then acquiring an intent category corresponding to the first target description information, and processing the target demand data according to the intent category corresponding to the first target description information and the target demand description information; if the plurality of first demand description information does not include the first target description information, then determining suspected intent category information of the target demand description information, and processing the target demand data according to the suspected intent category information; the suspected intent category information is used to represent the suspected intent category corresponding to the target demand description information and the credibility of the suspected intent category.

[0005] This specification provides one or more embodiments of a data processing apparatus, the apparatus comprising: a data acquisition unit for acquiring target demand data to be processed, the target demand data corresponding to at least one data modality; an information generation unit for generating target demand description information corresponding to the target demand data based on the data modality corresponding to the target demand data; a first data processing unit for acquiring an intent category corresponding to the first target description information if a plurality of stored first demand description information includes first target description information matching the target demand description information, and processing the target demand data according to the intent category corresponding to the first target description information and the target demand description information; and a second data processing unit for determining a suspected intent category information of the target demand description information if the plurality of first demand description information does not include the first target description information, and processing the target demand data according to the suspected intent category information; the suspected intent category information is used to represent the suspected intent category corresponding to the target demand description information and the credibility of the suspected intent category.

[0006] This specification provides one or more embodiments of a data processing device, including: a processor; and a memory configured to store computer-executable instructions, which, when executed, cause the processor to: acquire target demand data to be processed, the target demand data corresponding to at least one data modality; generate target demand description information corresponding to the target demand data based on the data modality corresponding to the target demand data; if a plurality of stored first demand description information includes first target description information matching the target demand description information, then acquire an intent category corresponding to the first target description information, and process the target demand data according to the intent category corresponding to the first target description information and the target demand description information; if the plurality of first demand description information does not include the first target description information, then determine suspected intent category information of the target demand description information, and process the target demand data according to the suspected intent category information; the suspected intent category information is used to indicate the suspected intent category corresponding to the target demand description information and the credibility of the suspected intent category.

[0007] This specification provides one or more embodiments of a computer-readable storage medium for storing computer-executable instructions, which, when executed, perform the following process: acquiring target requirement data to be processed, the target requirement data corresponding to at least one data modality; generating target requirement description information corresponding to the target requirement data based on the data modality corresponding to the target requirement data; if a plurality of stored first requirement description information includes first target description information matching the target requirement description information, then acquiring the intent category corresponding to the first target description information, and processing the target requirement data according to the intent category corresponding to the first target description information and the target requirement description information; if the plurality of first requirement description information does not include the first target description information, then determining suspected intent category information of the target requirement description information, and processing the target requirement data according to the suspected intent category information; the suspected intent category information is used to represent the suspected intent category corresponding to the target requirement description information and the credibility of the suspected intent category.

[0008] This specification also provides a computer program product, including a computer program that, when executed by a processor, implements the following process: acquiring target requirement data to be processed, the target requirement data corresponding to at least one data modality; generating target requirement description information corresponding to the target requirement data based on the data modality corresponding to the target requirement data; if a plurality of stored first requirement description information includes first target description information matching the target requirement description information, then acquiring the intent category corresponding to the first target description information, and processing the target requirement data according to the intent category corresponding to the first target description information and the target requirement description information; if the plurality of first requirement description information does not include the first target description information, then determining suspected intent category information of the target requirement description information, and processing the target requirement data according to the suspected intent category information; the suspected intent category information is used to represent the suspected intent category corresponding to the target requirement description information and the credibility of the suspected intent category. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in one or more embodiments of this specification or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Figure 1 A flowchart illustrating a data processing method provided in one or more embodiments of this specification; Figure 2 A flowchart illustrating another data processing method provided in one or more embodiments of this specification; Figure 3 A flowchart illustrating yet another data processing method provided in one or more embodiments of this specification; Figure 4 A schematic diagram of the structure of a data processing apparatus provided in one or more embodiments of this specification; Figure 5 This is a schematic diagram of the structure of a data processing device provided for one or more embodiments of this specification. Detailed Implementation

[0010] To enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this specification, and not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of this document.

[0011] This specification provides one or more embodiments of a data processing method, the application scenarios of which may include servers.

[0012] The server can be one or more servers, a server cluster consisting of several servers, or a cloud server of a cloud computing platform. During the processing of user requests, the server acquires target request data in at least one data modality and generates target request description information corresponding to the target request data based on the data modality. If the stored multiple first request descriptions include first target description information that matches the target request description information, then the intent category corresponding to the first target description information is acquired. The target request data is processed according to the intent category corresponding to the first target description information and the target request description information. If the multiple first request descriptions do not include the first target description information, then the suspected intent category information of the target request description information is determined, and the target request data is processed according to the suspected intent category information, thereby improving the accuracy of user request processing.

[0013] The server can obtain target demand data sent by users through terminal devices such as computers, mobile phones, tablets, and laptops. When processing the target demand data, the server can retrieve the corresponding response data and send the retrieved response data to the terminal device, which then displays the response data to the user. Alternatively, the server can execute the operation process corresponding to the target demand data, send the execution result of the operation process to the terminal device, and the terminal device displays the execution result to the user. The execution result includes whether the execution was successful or failed.

[0014] Figure 1 This is a flowchart illustrating a data processing method provided in one or more embodiments of this specification. This method can be applied to a server and executed by the server. Figure 1 As shown, the method includes: Step S102: Obtain the target requirement data to be processed, which corresponds to at least one data modality.

[0015] Step S104: Based on the data modality corresponding to the target requirement data, generate target requirement description information corresponding to the target requirement data.

[0016] Step S106: If the stored multiple first demand description information includes first target description information that matches the target demand description information, then obtain the intent category corresponding to the first target description information, and process the target demand data according to the intent category corresponding to the first target description information and the target demand description information.

[0017] Step S108: If the first target description information is not included among the multiple first demand description information, then the suspected intent category information of the target demand description information is determined, and the target demand data is processed according to the suspected intent category information; the suspected intent category information is used to represent the suspected intent category corresponding to the target demand description information and the credibility of the suspected intent category.

[0018] In step S102 above, the server acquires the target demand data to be processed, which corresponds to at least one data modality. The target demand data refers to the data provided by the user when submitting a request through a specific application, and includes, but is not limited to, at least one of text data, image data, and video data. A data modality refers to the form, carrier, or type of data; for example, a data modality includes, but is not limited to, at least one of text modality, image modality, and video modality.

[0019] Taking the security field as an example, the target requirement data in this embodiment is illustrated by way of example. The target requirement data in other fields can be referred to the above example about the security field, and will not be listed one by one in this specification.

[0020] In one example, the target requirement data is divided into single-modal data and multimodal mixed data based on the data modality.

[0021] Single-modal data refers to inputs containing only one data modality. For example, the target required data may be one of the following: plain text modal data, plain image modal data, or plain video modal data. Plain text modal data could be fault description text, consultation request text, etc. Plain image modal data could be screenshots of system errors, screenshots of maintenance interfaces, log screenshots, etc. Plain video modal data could be error videos recording fault phenomena, screen recordings of operations, etc.

[0022] Multimodal mixed data: refers to mixed input that includes at least two of the following modalities: text, image, and video. For example, the target required data is a mixed input that simultaneously contains fault description text, system error screenshots, and system error videos.

[0023] In practice, the server acquires the target requirement data to be processed, which can be achieved by receiving target requirement data uploaded by the client. The client can run a specific application, and the target requirement data can be collected by the client through the interactive page of that application. For example, the client obtains the question text and a screenshot associated with the question text through the question-and-answer interactive page of application 1, and uploads the question text and the screenshot to the server. The server receives the question text and the screenshot, and uses them as the target requirement data, which corresponds to both text and image modalities.

[0024] The server acquires the target requirement data to be processed, or it receives a processing instruction uploaded by the client. Based on this instruction, the server searches locally for the data corresponding to that instruction and identifies the searched data as the target requirement data. For example, the client receives input for a preset control through the video interaction page of application 2, generates a video fault diagnosis instruction in response to the input, and uploads it to the server. The server receives the video fault diagnosis instruction, searches locally for the video data indicated by the video identifier carried in the instruction, and uses the searched video data as the target requirement data, which corresponds to the video modality.

[0025] The server can obtain target requirement data to be processed, or it can obtain tasks to be processed from the list of tasks to be processed and read the relevant data of the tasks to be processed as target requirement data.

[0026] In step S104 above, the server generates target requirement description information corresponding to the target requirement data based on the data modality corresponding to the target requirement data. The target requirement description information refers to the structured semantic representation of the target requirement data. In one optional implementation, this structured semantic representation can be represented in vector form or other non-vector forms.

[0027] In practice, during the process of generating target requirement description information corresponding to the target requirement data based on the data modality corresponding to the target requirement data, the data conversion module corresponding to the data modality can be called to perform data conversion processing on the target requirement data to obtain the target requirement description information corresponding to the target requirement data.

[0028] For example, a text structuring module is pre-configured for text modalities, an image-to-text module is pre-configured for image modalities, and a video-to-text module is pre-configured for video modalities. Target requirement data 1 includes question text 1. The server calls the text structuring module to perform data conversion processing on question text 1, obtaining the target requirement description information corresponding to target requirement data 1. Target requirement data 2 includes question text 2, image 1, and video 1. The server calls the text structuring module to perform data conversion processing on question text 2, obtaining subtext 1; calls the image-to-text module to perform data conversion processing on image 1, obtaining subtext 2; and calls the video-to-text module to perform data conversion processing on video 1, obtaining subtext 3. Subtext 1, subtext 2, and subtext 3 are concatenated according to a preset assembly method to obtain the target requirement description information.

[0029] In one embodiment, generating target requirement description information based on the data modality corresponding to the target requirement data includes: dividing the target requirement data into at least one sub-data based on the data modality corresponding to the target requirement data; each sub-data having a corresponding data modality; extracting the data content of the sub-data based on the data modality corresponding to the sub-data; and performing structured extraction and fusion of the data content of each sub-data using a large language model to obtain the target requirement description information.

[0030] Based on the data modality corresponding to the target demand data, the target demand data is divided into at least one sub-data; each sub-data has a corresponding data modality. For example, if the target demand data corresponds to two data modalities: text modality and image modality, the target demand data is divided into two sub-data: sub-data 1 corresponds to the text modality, and sub-data 2 corresponds to the image modality.

[0031] Optionally, the data modality corresponding to the sub-data includes a text modality. Based on the data modality corresponding to the sub-data, the data content of the sub-data is extracted, including: performing word segmentation and part-of-speech tagging on the sub-data; extracting nouns and verbs from the sub-data as keywords based on the part-of-speech tagging results; identifying entity elements in the sub-data; entity elements include at least one of time, location, and people; and constructing the data content of the sub-data based on the keywords and entity elements. Optionally, if the data modality corresponding to the sub-data includes a text modality, the sub-data can also be used as its data content.

[0032] Optionally, the data modality corresponding to the sub-data includes an image modality; based on the data modality corresponding to the sub-data, the data content of the sub-data is extracted, including: identifying the text content and graphic elements other than text content in the sub-data; determining the semantic information of the graphic elements, and determining the layout relationship between the text content and the graphic elements; generating text data to represent the semantic information and layout relationship of the text content and graphic elements, and using the text data as the data content of the sub-data.

[0033] In practice, before recognizing the text content in the sub-data, image quality enhancement can be performed on the sub-data to optimize recognition conditions. Image quality enhancement methods include, but are not limited to, denoising, sharpening, and angle correction. During the recognition of the text content in the sub-data, OCR (Optical Character Recognition) can be used to identify the visible text in the sub-data, obtaining initial recognized text. This initial recognized text is then corrected and structured to obtain intermediate recognized text, and a confidence score is generated for this intermediate recognized text. If the confidence score is greater than or equal to a preset confidence threshold, the intermediate recognized text is identified as the text content in the sub-data. If the confidence score is less than a preset confidence threshold, the OCR recognition process can be repeated or manual processing can be performed. For example, the data modality corresponding to sub-data 1 includes image modalities. An OCR model is used to extract visible text from sub-data 1, such as error codes, user interface text, and log fragments.

[0034] Graphical elements refer to non-textual visual components in user interfaces, design works, or any visual presentation. These graphic elements include, but are not limited to, interface elements, tooltips, and visual icons. Identifying graphic elements other than text content in sub-data can be achieved by inputting the sub-data into a graphic detection model for graphic detection processing, obtaining a detection result for at least one graphic element, including its bounding box. Furthermore, the semantic information of the graphic element is determined, representing its meaning; for example, semantic information might indicate that the graphic element means "please note."

[0035] In one example, graphic elements can be identified and their semantic information determined simultaneously. For instance, sub-data is input into a visual understanding model, which extracts and semantically identifies graphic elements from the sub-data to obtain identification information for at least one graphic element. This identification information describes the location of the graphic element and expresses its semantic information in natural language.

[0036] Layout relationships refer to the relative positions, arrangement, hierarchical structure, and functional connections between components within a system, design, or organization. Specifically, in this embodiment, the layout relationship between text content and graphic elements refers to the spatial arrangement of text content and graphic elements in the user interface, such as position, alignment, spacing, stacking order, and responsiveness. Determining the layout relationship between text content and graphic elements can be based on the position of the text content within sub-data and the position of each graphic element within the sub-data. For example, a layout relationship might indicate that a graphic element is located to the right of the text content.

[0037] In one scenario, generating text data representing the semantic information and layout relationships of text content and graphic elements can be achieved by concatenating these elements to obtain structured text data. This structured text data is then used to represent the semantic information and layout relationships of the text content and graphic elements. In another scenario, generating text data representing the semantic information and layout relationships of text content and graphic elements can also be achieved by inputting these elements into a large language model for information fusion to obtain structured text data. This structured text data is then used to represent the semantic information and layout relationships of the text content and graphic elements. Finally, the text data obtained above is used as the data content of the sub-data.

[0038] Optionally, the data modality corresponding to the sub-data includes a video modality; based on the data modality corresponding to the sub-data, the data content of the sub-data is extracted, including: identifying the text content in the sub-data, and determining the semantic information of the images in the sub-data; generating text data to represent the semantic information of the text content and images, and using the text data as the data content of the sub-data.

[0039] Sub-data can include multiple frames. In practice, before recognizing the text content in the sub-data, image quality enhancement can be performed on each frame in the sub-data to optimize recognition conditions. Image quality enhancement methods include, but are not limited to, denoising, sharpening, and angle correction. During the recognition of the text content in the sub-data, OCR can be used to recognize the visible text in each frame of the sub-data to obtain initial recognized text. The initial recognized text is then corrected and structured to obtain intermediate recognized text, and a confidence score is generated for this intermediate recognized text. If the confidence score is greater than or equal to a preset confidence threshold, the intermediate recognized text is determined as the text content in the sub-data. If the confidence score is less than a preset confidence threshold, the OCR recognition process can be repeated or manual processing can be performed. For example, the data modality corresponding to sub-data 2 includes a video modality, and sub-data 2 includes N frames, where N is an integer greater than 1. The OCR model is used to extract visible text from each frame of sub-data 2, such as error codes, operation interface text, log fragments, etc.

[0040] Then, the semantic information of the images in the sub-data can be determined through a visual understanding model. For example, the semantic information can represent the fault detection process of the system; that is, the sub-data of the video modality mentioned above represents the process of fault detection for the system.

[0041] In one scenario, generating text data to represent the semantic information of text content and images can be achieved by concatenating the semantic information of the text content and images to obtain structured text data, which is then used to represent the semantic information of the text content and images. In another scenario, generating text data to represent the semantic information of text content and images can also involve inputting the semantic information of the text content and images into a large language model for information fusion to obtain structured text data, which is then used to represent the semantic information of the text content and images. Finally, the resulting text data is used as the data content of the sub-data.

[0042] The server uses a large language model to perform structured extraction and fusion of the data content from each sub-data set to obtain the target requirement description information. In practice, the data content of each sub-data set is input into the large language model, which extracts the corresponding sub-texts. Then, the large language model fuses these sub-texts according to a preset order to obtain the target requirement description information. Of course, preprocessing operations may or may not be performed before fusing the sub-texts. Preprocessing operations include, but are not limited to, data cleaning, data transformation, feature engineering, and data augmentation. Data cleaning may involve removing noise and handling missing values; data transformation may involve vectorizing the text; feature engineering may involve extracting keywords and entities from the text; and data augmentation may involve synonym / near-synonym replacement.

[0043] As can be seen, this embodiment utilizes data modalities to divide target requirement data into at least one sub-data and extracts the data content of each sub-data. The data content of each sub-data is then structurally extracted and fused to obtain target requirement description information. This enables the accurate conversion of target requirement data involving multiple data modalities into a structured form, opening up the transformation link from raw, unstructured customer input to structured, executable product requirements. Furthermore, by fully utilizing the reasoning capabilities of the large language model, it effectively solves the problems of semantic alignment and information loss in multimodal data fusion.

[0044] In one embodiment, a large language model is used to perform structured extraction and fusion of the data content of each sub-data to obtain target requirement description information. This includes: obtaining an information template corresponding to the requirement description information; the information template includes multiple sub-structures; each sub-structure has corresponding data content requirements; using the large language model, extracting the data content corresponding to each sub-data from the data content of each sub-data according to the data content requirements corresponding to each sub-structure; and using the large language model, fusing the data content corresponding to each sub-structure according to the fusion relationship between each sub-structure in the information template to obtain the target requirement description information.

[0045] The system retrieves the information template corresponding to the requirement description information. This template can be a pre-built general template. The user-input target requirement data is processed into structured target requirement description information based on this template. The information template includes multiple substructures; each substructure has corresponding data content requirements, and each substructure corresponds to a data modality. For example, information template 1 includes: substructure 1 and substructure 2. Substructure 1 corresponds to the text modality, and its data content requirements are: the total number of characters is less than a first character count threshold; if the data content contains English characters, they are translated into corresponding Chinese characters, and the final text uses only Chinese characters. Substructure 2 corresponds to the image modality, and its data content requirements are: the total number of characters is less than a second character count threshold; only Chinese characters, numbers, and common punctuation are retained, while special symbols and garbled characters are filtered out. In one example, the information template corresponding to the requirement description information is pre-created and stored on the server. When the server retrieves the information template corresponding to the requirement description information, it can directly read the information template from its local storage.

[0046] In the process of extracting the data content corresponding to the substructure from the data content of each sub-data using a large language model based on the data content requirements of the substructure, the data content requirements corresponding to the substructure can be used as prompt words. The data content of the sub-data and the prompt words are input into the large language model, and the large language model performs content extraction to extract the data content corresponding to the substructure from the data content of each sub-data.

[0047] Using a large language model, the data content corresponding to each substructure in the information template is merged according to the fusion relationship between them to obtain the target requirement description information. The fusion relationship between the substructures in the information template refers to the logical combination or organization rules among these substructure contents when combining the data content corresponding to multiple substructures into a complete target requirement description information.

[0048] Integration relationships can be pre-categorized into various types, including but not limited to: temporal relationships, causal relationships, master-slave relationships, and parallel relationships. Among these, a temporal relationship indicates that the data content corresponding to each substructure is arranged in chronological order. A causal relationship indicates that the data content corresponding to some substructures is the cause, while the data content corresponding to other substructures is the result. A master-slave relationship indicates that the data content corresponding to some substructures is the main description, while the data content corresponding to other substructures supplements or modifies the main description. A parallel relationship indicates that the data content corresponding to multiple substructures is of equal status and collectively describes a scenario.

[0049] By using a large language model, the data content corresponding to each substructure in the information template is fused according to the fusion relationship between the substructures to obtain the target requirement description information. Alternatively, the information template and the data content corresponding to each substructure can be input into the large language model, and the data content can be fused through the large language model to obtain the target requirement description information.

[0050] As can be seen, this embodiment can fully utilize the natural language understanding capabilities of the large language model to transform potentially irregular data content into structured target requirement description information, thereby providing uniformly formatted input data for the matching operation in subsequent steps.

[0051] In step S106 above, if the stored multiple first demand description information includes first target description information that matches the target demand description information, the server obtains the intent category corresponding to the first target description information and processes the target demand data according to the intent category corresponding to the first target description information and the target demand description information.

[0052] The server can store multiple pre-created first requirement descriptions and the intent categories corresponding to each first requirement description. The first requirement description refers to a structured semantic representation of the requirement data set for each intent category. In one optional implementation, this structured semantic representation can be in vector form or other non-vector form; a vector library can be constructed based on the multiple first requirement descriptions represented in vector form.

[0053] For example, a question library can be pre-built and stored on a server. This question library includes a set of "question-intent-answer / processing flow" sequences that have been established in the security field, used to handle scenarios in the security field where "the expression is diverse, but the essential problem is the same." In this question library, each question (i.e., the initial requirement description information) is represented in vector form.

[0054] Optionally, matching the first target description information with the target requirement description information can be achieved by using the first requirement description information as the first target description information if the keywords in the first requirement description information match the keywords in the target requirement description information. For example, the keyword matching condition can be pre-set to require at least two keywords to be the same. The keyword set of the target requirement description information includes: word 1, word 2, and word 5. The keyword set of the first requirement description information 1 includes: word 1, word 2, word 3, and word 4. If word 1 and word 2 are the same in the keyword sets of the first requirement description information and the target requirement description information, the keyword matching condition is satisfied. Therefore, the keywords in the first requirement description information 1 match the keywords in the target requirement description information, and the first requirement description information 1 is used as the first target description information.

[0055] Optionally, the matching of the first target description information with the target requirement description information can be achieved by using the first requirement description information as the first target description information if the semantics of the first requirement description information matches the semantics of the target requirement description information. For example, a semantic matching condition can be pre-set as a vector similarity greater than a first similarity threshold. The target requirement description information can be represented as vector 1. The first requirement description information 2 can be represented as vector 2. If the vector similarity between vector 1 and vector 2 is greater than the first similarity threshold, the semantic matching condition is satisfied. Therefore, the semantics of the first requirement description information 2 matches the semantics of the target requirement description information, and the first requirement description information 2 is used as the first target description information.

[0056] Optionally, the first target description information and the target requirement description information can be matched if the following two conditions are met simultaneously: (1) the keywords of the target requirement description information match the keywords of the first requirement description information, and (2) the semantics of the target requirement description information match the semantics of the first requirement description information.

[0057] There may be one or more first target descriptions that match the target requirement description information. If a first target description is matched, the server obtains its corresponding intent category; if multiple first target descriptions are matched, the server selects the one with the most identical keywords or the highest semantic similarity to the target requirement description from the multiple first target descriptions and obtains its corresponding intent category.

[0058] During the process of processing the target requirement data based on the intent category and target requirement description information corresponding to the first target description information, the answer text corresponding to the target requirement data can be retrieved from the knowledge base based on the intent category and target requirement description information corresponding to the first target description information.

[0059] For example, target requirement data 1 includes text data "error 004" and image data "screenshot". Figure 1 The server acquires target requirement data 1, and based on the data modality corresponding to target requirement data 1, generates target requirement description information corresponding to target requirement data 1, which can be represented as vector X. The server stores multiple first requirement descriptions, and the i-th first requirement description can be represented as vector i, where i is an integer greater than 0. Among the multiple stored first requirement descriptions is vector 1 that matches vector X. The server acquires the intent category "system error" corresponding to vector 1, and retrieves the answer text from the knowledge base based on the intent category "system error" and vector X. This answer text is an operation guidance text used to guide the user to correct the system error indicated by target requirement data 1.

[0060] During the process of processing the target requirement data based on the intent category and target requirement description information corresponding to the first target description information, the target tool that matches the intent category can also be invoked to execute the data processing flow corresponding to the target requirement data based on the intent category and target requirement description information corresponding to the first target description information. The working parameters of the target tool are set based on the target requirement description information.

[0061] For example, target requirement data 1 includes text data "error 004" and image data "screenshot". Figure 1 The server acquires target requirement data 1, and based on the data modality corresponding to target requirement data 1, generates target requirement description information corresponding to target requirement data 1, which can be represented as vector X. The server stores multiple first requirement descriptions, and the i-th first requirement description can be represented as vector i, where i is an integer greater than 0. Among the multiple stored first requirement descriptions is vector 1 that matches vector X. The server acquires the intent category "system error" corresponding to vector 1, and based on the intent category "system error" and vector X, calls tool 1 that matches the intent category "system error" to execute the error correction process corresponding to target requirement data 1. The working parameters of tool 1 are set based on vector X.

[0062] In one embodiment, the data processing method further includes: acquiring multiple first known intent categories, determining the structured requirement description method corresponding to the first known intent category; creating structured requirement description information corresponding to the first known intent category based on the structured requirement description method; and using the structured requirement description information as the first requirement description information.

[0063] First, the server obtains multiple known intent categories. For example, known intent categories include, but are not limited to, system error reporting, data querying, and transaction operation guidance. The intent category corresponding to the aforementioned first target description information can be at least one of these multiple known intent categories.

[0064] The first known intent category can be constructed from historical records or manually. The first known intent category can be any of the commonly defined intent categories. After obtaining the first known intent category, a structured requirement description method corresponding to that category is determined. Based on this method, structured requirement description information corresponding to the first known intent category is created, and this information is used as the first requirement description information. The structured requirement description method can be a set of structured methods used to generate the first requirement description information for a specific intent category. This first requirement description information is used to match the target requirement description information, which is generated from unstructured target requirement data provided by the user.

[0065] The server determines the structured requirement description method corresponding to the first known intent category by determining the target description method corresponding to the first known intent category based on the first known intent category and the pre-set correspondence between intent categories and description methods, and then uses the target description method as the structured requirement description method corresponding to the first known intent category.

[0066] In one approach, the structured requirement description method is represented by a structured requirement description template, which can be considered a standardized description template for the first known intent category. During the process of the server creating structured requirement description information corresponding to the first known intent category based on the structured requirement description method, the required field values ​​in the structured requirement description template can be obtained. The template can be filled based on these field values ​​to obtain structured requirement description text. This structured requirement description text is then encoded to obtain a description text vector representing the text. This description text vector is then identified as the structured requirement description information corresponding to the first known intent category, and the structured requirement description information is used as the first requirement description information.

[0067] For example, for the intent category "System Error", its structured requirement description template is represented as: [Page Name x1] [Configuration Item x2] prompts a configuration error, how should it be handled? Obtain the field value "Page Name 1" of field x1 and the field values ​​"Configuration Item 1", "Configuration Item 2", and "Configuration Item 3" of field x2. Based on the field values ​​"Page Name 1" and "Configuration Item 1", "Configuration Item 2", and "Configuration Item 3", populate the structured requirement description template to obtain the structured requirement description text: Page Name 1 [Configuration Item 1 / Configuration Item 2 / Configuration Item 3] prompts a configuration error, how should it be handled? Encode this structured requirement description text to obtain vector 3 representing this structured requirement description text, and determine this vector 3 as the structured requirement description information corresponding to the intent category "System Error".

[0068] For example, for the intent category "Data Query", its structured requirement description template is represented as: I need to query order information related to [Product Type x4] in [Store x3]. In one example, the field value "Store 1" of field x3 and the field value "Product Type 1" of field x4 are obtained. Based on the field values ​​"Store 1" and "Product Type 1", the structured requirement description template is populated to obtain the structured requirement description text: I need to query order information related to Product Type 1 in Store 1. This structured requirement description text is encoded to obtain vector 4 representing the structured requirement description text, and this vector 4 is determined as the structured requirement description information corresponding to the intent category "Data Query".

[0069] As can be seen, through this embodiment, the structured requirement description method is first determined, and then the structured requirement description information corresponding to the first known intent category is created using the structured requirement description method. The structured requirement description information is used as the first requirement description information, which realizes the rapid creation of the first requirement description information and ensures that the first requirement description information is structured information.

[0070] In one embodiment, determining the structured requirement description method corresponding to the first known intent category includes: determining at least one key intent information associated with the first known intent category and the order of these key intent information; the key intent information includes intent keywords and / or key semantic information; and determining the structured requirement description method based on the order of the key intent information.

[0071] The server determines at least one key intent information associated with a first known intent category and the order of these key intent information. Key intent information may include intent keywords; alternatively, key intent information may include key semantic information; or alternatively, key intent information may include both intent keywords and key semantic information. Intent keywords are words used to identify intent categories, while key semantic information consists of semantic elements and structural information expressing the deeper meaning of the intent. For example, at least one key intent information associated with the intent category "system error" includes: intent keywords such as "configuration error," "system exception," and "configuration failure," and key semantic information such as "occurrence object: page / configuration item," "error type: configuration error," and "status: failure message, unable to save, interface error."

[0072] In addition, determine the order of the key intent information. For example, the key intent information associated with intent category 1, from first to last, is: intent keyword 1, intent keyword 2, key semantic information 1, and key semantic information 2.

[0073] Based on the order of key intent information, a structured requirement description method is determined. This structured requirement description method represents the order of key intent information. For example, based on the order of key intent information, the aforementioned structured requirement description template is generated. This template includes various fields, each with a value to be filled in. The values ​​of each field represent the aforementioned key intent information.

[0074] In this approach, the field values ​​corresponding to each field can be obtained. When the key intent information includes intent keywords, the field value can include the aforementioned intent keywords. When the key intent information includes key semantic information, the field value can include text used to represent that key semantic information. Based on the field values, the structured requirement description template is filled to obtain structured requirement description text. The structured requirement description text or a vector of structured requirement description text is used as structured requirement description information, and the structured requirement description information is used as the first requirement description information.

[0075] As can be seen, by determining the structured requirement description method based on the key intent information and the order of key intent information in this embodiment, unstructured requirement data that may be relatively complex in expression can be sorted into a structured expression.

[0076] In one embodiment, processing the target demand data according to the intent category corresponding to the first target description information and the target demand description information includes: taking the intent category corresponding to the first target description information as the first target intent category corresponding to the target demand data, and determining the processing method corresponding to the target demand data; the processing method includes generating response data corresponding to the target demand data or executing the operation process corresponding to the target demand data; and processing the target demand data according to the processing method corresponding to the target demand data based on the first target intent category and the target demand description information.

[0077] The server uses the intent category corresponding to the first target description information as the first target intent category corresponding to the target demand data, and determines the processing method corresponding to the target demand data. The processing method includes generating response data corresponding to the target demand data or executing the operation process corresponding to the target demand data. Finally, the server processes the target demand data according to the processing method corresponding to the target demand data. The server can determine the processing method based on the type of the target demand data. For example, when the target demand data is question-type data including question semantics, its corresponding processing method includes generating corresponding response data. Similarly, when the target demand data is operation-type data including request operation semantics, its corresponding processing method includes executing the corresponding operation process.

[0078] In one example, problem-type data could be "My network connection suddenly dropped, and when I tried to reconnect, a connection error message popped up on the page. Why is this happening?" The corresponding response data for the target requirement data could be "The server node connection may have failed. Please try a different available server node and reconnect." In another example, operation-type data could be "I keep getting a login error message after opening page 1 of application A. Please open page 1 for me," along with a screenshot of page 1. The corresponding operation flow for the target requirement data could be: call the login error detection tool to check page 1, and then perform corresponding processing measures based on the detection results.

[0079] As can be seen, by using the intent category to determine the processing method corresponding to the target demand data in this embodiment, different intents can be flexibly adapted to achieve differentiated processing of the target demand data.

[0080] In one embodiment, the processing method includes generating response data corresponding to the target demand data; processing the target demand data according to the processing method corresponding to the target demand data based on the first target intent category and the target demand description information, including: retrieving knowledge point information corresponding to the target demand description information from the question-answering knowledge base based on the first target intent category and the target demand description information through a question-answering model, and generating response data corresponding to the target demand data based on the knowledge point information.

[0081] The server uses a question-answering model to retrieve knowledge points corresponding to the target requirement description from a question-answering knowledge base, based on the first target intent category and the target requirement description information. The question-answering model is an artificial intelligence model capable of understanding user questions and automatically providing corresponding answers. The question-answering knowledge base can include multiple question-answer pairs, each consisting of a question text and an answer text. The question-answering knowledge base can also include a knowledge graph pre-built based on multiple question-answer pairs.

[0082] In practice, the server can store multiple question-and-answer knowledge bases, each corresponding to a first known intent category. The definition and examples of the first known intent category can be found in the relevant explanations above.

[0083] In the process of retrieving knowledge point information corresponding to the target requirement description information from the question-and-answer knowledge base based on the first target intent category and the target requirement description information, the target question-and-answer knowledge base corresponding to the first target intent category can be determined from multiple question-and-answer knowledge bases based on the first target intent category; and information retrieval can be performed in the target question-and-answer knowledge base based on the target requirement description information to obtain knowledge point information.

[0084] When there are multiple pieces of knowledge information, the server can combine the knowledge information to obtain the response data corresponding to the target requirement data during the process of generating response data based on the knowledge information.

[0085] As can be seen, through this embodiment, using the first target intent category to locate the question-and-answer knowledge base can narrow the scope of information retrieval, and using the target demand description information to perform information detection in the question-and-answer knowledge base can improve the accuracy of the response data.

[0086] In one embodiment, the processing method includes executing the operation process corresponding to the target demand data; processing the target demand data according to the processing method corresponding to the target demand data based on the first target intent category and the target demand description information, including: determining the process execution tool and the operation process corresponding to the target demand data based on the first target intent category and the target demand description information through the process execution model, and executing the operation process through the process execution tool.

[0087] The server, through the process execution model, determines the process execution tool and the corresponding operation flow for the target requirement data based on the first target intent category and target requirement description information. The process execution model is also used to execute the operation flow through the process execution tool. The process execution tool can be a tool or module that performs a specific task. For example, when the current requirement is to perform a word count check, the process execution tool can be a word count check tool.

[0088] In practice, the server can be configured with multiple candidate tools that can be invoked, and stores the correspondence between intent categories and candidate tools. During the process of determining the process execution tool and operation flow corresponding to the target requirement data based on the first target intent category and target requirement description information, the server can select one of the multiple candidate tools as the process execution tool based on the first target intent category, and determine the operation flow corresponding to the target requirement data based on the target requirement description information. Then, the operation flow is executed through the process execution tool to process the target requirement data.

[0089] In one example, the operational flow corresponding to the target requirement data can be determined based on the target requirement description information and the baseline operation information of the process execution tool. The aforementioned baseline operation information can be a set of metadata describing the smallest granularity, atomic functions that a candidate tool can perform. It defines the basic framework of "what the tool can do" and "how it does it," and is a prefabricated component for the server to dynamically generate the operational flow. For example, if Tool 1 is a configuration and debugging tool, its baseline operations include: reading the configuration file, backing up the current configuration, modifying configuration items, restarting the service, and rolling back the configuration. The baseline operation information of Tool 1 defines the basic framework of how each baseline operation is implemented.

[0090] For example, based on the intent category "system error", the server selects Tool 1, which corresponds to the intent category "system error", as the process execution tool from multiple candidate tools. The target requirement description information can be represented as a vector X. The server generates an error correction process adapted to vector X based on vector X and the baseline operation information of Tool 1, and uses the semantic information contained in vector X to fill in the specific execution parameters of the error correction process to obtain the operation process corresponding to the target requirement data.

[0091] As can be seen, through this embodiment, by using the first target intent category to determine the process execution tool and using the target requirement description information to determine the operation process, personalized operation processes can be dynamically generated, solving the diverse problems that users may encounter in actual application scenarios.

[0092] In one embodiment, the data processing method further includes: if the stored plurality of second demand description information includes second target description information that matches the target demand description information, then obtaining the response data or operation process corresponding to the second target description information, and processing the target demand data based on the response data or operation process.

[0093] The second requirement description information can be constructed from historical records. This second requirement description information represents user requirements that occur more frequently than a frequency threshold. Matching the second target requirement description information with the target requirement description information can be achieved by using the second requirement description information as the second target description information if the semantics of the second requirement description information matches the semantics of the target requirement description information. For example, a semantic matching condition can be pre-set as a vector similarity greater than a second similarity threshold. The target requirement description information can be represented as vector 1. The second requirement description information 1 can be represented as vector 2. If the vector similarity between vector 1 and vector 2 is greater than the second similarity threshold, the semantic matching condition is satisfied. Therefore, the semantics of the second requirement description information 1 matches the semantics of the target requirement description information, and the second requirement description information 1 is used as the second target description information.

[0094] It is important to note that the second similarity threshold is greater than the first similarity threshold. For example, if the first similarity threshold is set to 70% and the second similarity threshold is set to 95%, then the second similarity threshold is greater than the first similarity threshold.

[0095] For example, two question databases can be pre-built and stored on a server. Question database 1 includes a set of "question-intent-answer / processing flow" sequences already compiled in the security domain, used to handle scenarios in the security domain where "the expression forms are diverse, but the essential problem is the same." In this database, each question (i.e., the first requirement description information) is represented in vector form. Question database 2 includes a set of "question-answer / processing flow" sequences already compiled in the security domain. In this database, each question (i.e., the second requirement description information) is represented in vector form.

[0096] Considering that the second similarity threshold is greater than the first similarity threshold, in order to reduce unnecessary redundant calculations, the target requirement description information can be matched with multiple second requirement description information first. If all second requirement description information fails to match, the target requirement description information can then be matched with multiple first requirement description information.

[0097] Since the second target description information matches the target requirement description information, it can be assumed that the second target description information and the target requirement description information represent the same user requirement. Therefore, after determining the second target description information, the corresponding response data or operation process can be obtained, and the target requirement data can be processed based on the response data or operation process. For example, the response data corresponding to the second target description information can be output to the user, or the operation process corresponding to the second target description information can be executed.

[0098] As can be seen, this embodiment utilizes multiple stored second requirement descriptions to match the target requirement description, skips the intent matching process, and quickly and accurately reuses some common requirement response data or operation processes, thereby improving requirement processing efficiency.

[0099] In step S108 above, if the stored multiple first demand description information does not include first target description information that matches the target demand description information, then the suspected intent category information of the target demand description information is determined, and the target demand data is processed according to the suspected intent category information; the suspected intent category information is used to indicate the suspected intent category corresponding to the target demand description information and the credibility of the suspected intent category.

[0100] Optionally, if the keywords of the target requirement description information do not match the keywords of each of the first requirement description information, it can be determined that the first target description information that matches the target requirement description information is not included among the multiple first requirement description information.

[0101] Optionally, if the semantics of the target requirement description information does not match the semantics of each of the first requirement description information, it can be determined that the first target description information that matches the target requirement description information is not included among the multiple first requirement description information.

[0102] Optionally, the server may set the following two conditions: (1) the keywords of the target requirement description information match the keywords of the first requirement description information, and (2) the semantics of the target requirement description information match the semantics of the first requirement description information. If there is no first requirement description information that satisfies both conditions, it can be determined that the multiple first requirement description information does not include the first target description information that matches the target requirement description information.

[0103] If none of the multiple first demand descriptions match the target demand description, the server can determine that the intent category corresponding to the target demand description is outside the aforementioned known first intent categories. In this case, the intent category corresponding to the target demand description can be considered a long-tail intent. A long-tail intent refers to an intent that occurs infrequently and is not covered by the system's preset known first intent categories. In this situation, the server can determine the suspected intent category information of the target demand description and process the target demand data based on the suspected intent category information.

[0104] Suspected intent category information is used to represent the suspected intent category corresponding to the target demand description information and the confidence level of the suspected intent category. In one example, each suspected intent category information includes a category label representing the suspected intent category and a confidence parameter representing the confidence level of the suspected intent category. The confidence parameter is a quantitative indicator used to represent how confident the model or system is in a certain identification result (such as a suspected intent category). It is usually presented in numerical form, such as 0.95, 85%, etc. The higher the value, the more confident the system is in the accuracy of the result.

[0105] In practice, during the process of determining the suspected intent category information of the target demand description information, the server inputs the target demand description information into a pre-trained text classifier for text classification to obtain the suspected intent category information.

[0106] In the process of determining the suspected intent category information of the target demand description information, the server can also determine the key intent information corresponding to the target demand description information, determine the target long-tail intent that matches the key intent information from multiple pre-set candidate long-tail intents, and obtain the suspected intent category information based on the target long-tail intent.

[0107] The server processes the target demand data based on the suspected intent category information.

[0108] When the credibility of the suspected intent category meets the credibility condition, the server processes the target demand data based on the suspected intent category information. Refer to the corresponding explanation in step S106 above, which states, "Process the target demand data based on the intent category corresponding to the first target description information and the target demand description information." The credibility condition can be flexibly set according to specific scenarios. For example, for security scenarios, the credibility condition can be set to a confidence level parameter greater than a parameter threshold of 50%.

[0109] When the credibility of a suspected intent category does not meet the credibility criteria, the server processes the target requirement data based on the suspected intent category information. This can be done by obtaining the corresponding key intent information based on the suspected intent category information, generating data processing reference opinions based on the key intent information, and then using an artificial intelligence model to process the data based on the data processing reference opinions and the target requirement data.

[0110] When the credibility of the suspected intent category does not meet the credibility criteria, and considering that the target demand data may deviate significantly from the suspected intent category and cannot be used directly, data processing reference opinions can be generated by utilizing the suspected intent category information. This allows for the reuse of existing knowledge and reduces the operational difficulty of the artificial intelligence model.

[0111] In addition, if the credibility of a suspected intent category does not meet the credibility criteria, the server can also transfer the target demand data to a human agent for processing.

[0112] If the credibility of a suspected intent category does not meet the credibility criteria, the target demand data can be labeled as long-tail intent data and recorded. This long-tail intent data can be used to provide more training samples for subsequent optimization of the text classifier.

[0113] In one embodiment, determining the suspected intent category information of the target demand description information includes: using an intent classification model, determining the second target intent category corresponding to the target demand data in each second known intent category based on the target demand description information, and using the second target intent category as the suspected intent category; obtaining the credibility information of the target demand description information corresponding to the second target intent category, and generating the suspected intent category information based on the second target intent category and the credibility information.

[0114] The granularity of the second known intent category can be smaller than that of the first known intent category. Specifically, the second known intent category can be a subcategory of the first known intent category.

[0115] For example, each of the second known intent categories can be a subcategory of the first known intent category "system error". Each of the second known intent categories includes, but is not limited to, VPN (Virtual Private Network) failure, system configuration error, account and permission issues, etc.

[0116] The server uses an intent classification model to determine the second target intent category corresponding to the target demand data among various second known intent categories based on the target demand description information, and then uses this second target intent category as the suspected intent category. The intent classification model can be a machine learning model trained using training data from each of the second known intent categories.

[0117] For example, the target demand description information is represented as vector X. Vector X is input into an intent classification model for intent classification. The intent recognition results for vector X are the predicted probabilities of each second known intent category: the predicted probability for intent category 1 is 0, the predicted probability for intent category 2 is 5%, the predicted probability for intent category 3 is 20%, and so on. Intent category 3, with the highest predicted probability, is taken as the suspected intent category.

[0118] In practice, the server can train multiple machine learning models with different architectures using training data for the second known intent category, resulting in multiple intent classification models. The target demand description information is then input into each intent classification model for intent classification, yielding multiple intent recognition results. Each result includes the predicted probability for each second known intent category. The consistency of the multiple intent recognition results is statistically analyzed, and a confidence parameter for each second known intent category is determined based on the statistical results. The second target intent category with the highest confidence parameter is identified as the suspected intent category.

[0119] It should be noted that since the classification scope of the intent classification model is limited to the preset second known intent categories, when the user's demand belongs to the long tail intent, the suspected intent category output by the model (i.e. the intent category with the highest prediction probability) may only be the known category that is semantically closest to the real intent, rather than an accurate identification of the user's real intent.

[0120] The server obtains the credibility information of the target demand description information corresponding to the second target intent category. This credibility information can be represented by a confidence parameter. This confidence parameter can be obtained together with the suspected intent category through the aforementioned consciousness classification model, or it can be obtained through a confidence calibrator set outside the intent classification model.

[0121] The server generates suspected intent category information based on the second target intent category and confidence level information. This can be achieved by combining the intent category and confidence level parameters together.

[0122] As can be seen, by using the fine-grained second known intent category to determine the suspected intent category and obtain the credibility information of the suspected intent category, even when the target demand data does not match any of the known intents, the suspected intent category with the highest degree of relevance to the target demand data can be found as much as possible, thereby using the suspected intent category to provide reusable information for subsequent processing.

[0123] In one embodiment, processing the target requirement data based on the suspected intent category information includes: if the credibility of the suspected intent category is determined to be greater than a credibility threshold based on the suspected intent category information, then processing the target requirement data based on the suspected intent category and the target requirement description information; if the credibility of the suspected intent category is determined to be no greater than a credibility threshold based on the suspected intent category information, then processing the target requirement description information through an artificial intelligence model.

[0124] The confidence threshold can be a pre-set threshold for the confidence parameter. For example, the confidence threshold could be 30%.

[0125] The server compares the confidence parameter of the suspected intent category in the suspected intent category information with the credibility threshold, and obtains the comparison result: the credibility of the suspected intent category is greater than the credibility threshold, or the credibility of the suspected intent category is less than or equal to the credibility threshold.

[0126] The server processes the target demand data based on the suspected intent category and the target demand description information. This can be done by retrieving the answer text from the knowledge base based on the suspected intent category and the target demand description information. This answer text is the answer text corresponding to the target demand data.

[0127] The server processes the target requirement data based on the suspected intent category and the target requirement description information. Alternatively, it can call a target tool that matches the suspected intent category to execute the data processing flow corresponding to the target requirement data, with the working parameters of the target tool set based on the target requirement description information.

[0128] The server processes the target requirement description information through an artificial intelligence model. This can be done by inputting the target requirement description information into an artificial intelligence model, such as a large language model or an intelligent agent based on a large language model. The artificial intelligence model analyzes the target requirement description information to obtain the corresponding processing method for the target requirement data. The processing method includes outputting a response text, executing the corresponding process, or transferring it to a human.

[0129] As can be seen, this embodiment allows for different processing measures to be taken depending on the degree of credibility of the suspected intent category, thereby reasonably reducing the number of calls to the artificial intelligence model and reducing the consumption of computing resources while ensuring the accuracy of processing user needs.

[0130] In one embodiment, the target demand data is processed according to the suspected intent category and the target demand description information, including: determining the processing method corresponding to the target demand data; the processing method includes generating response data corresponding to the target demand data or executing the operation process corresponding to the target demand data; and processing the target demand data according to the processing method corresponding to the target demand data based on the suspected intent category and the target demand description information.

[0131] The server determines the processing method corresponding to the target demand data; the processing method includes generating response data corresponding to the target demand data or executing the operation process corresponding to the target demand data.

[0132] Optionally, the processing method includes generating response data corresponding to the target demand data; processing the target demand data according to the processing method corresponding to the target demand data based on the suspected intent category and the target demand description information, including: retrieving knowledge point information corresponding to the target demand description information in the question-answering knowledge base through a question-answering model, based on the suspected intent category and the target demand description information, and generating response data corresponding to the target demand data based on the knowledge point information.

[0133] Optionally, the processing method includes executing the operation process corresponding to the target demand data; processing the target demand data according to the processing method corresponding to the target demand data based on the suspected intent category and the target demand description information, including: determining the process execution tool and the operation process corresponding to the target demand data based on the suspected intent category and the target demand description information through the process execution model, and executing the operation process through the process execution tool.

[0134] As can be seen, this embodiment can determine the processing method corresponding to the target demand data. Based on the suspected intent category and the target demand description information, the target demand data is processed according to the processing method corresponding to the target demand data, thereby accurately processing user needs based on the suspected intent category.

[0135] It is worth noting that since the target requirement description information is generated based on the target requirement data, it is essentially a structured representation of the target requirement data. Therefore, in the various embodiments of this specification, the processing method / response data / operation flow corresponding to the target requirement data can also be referred to as the processing method / response data / operation flow corresponding to the target requirement description information. Processing the target requirement data is equivalent to processing the target requirement description information.

[0136] In summary, through the various embodiments of the data processing methods described above, the pre-stored first requirement description information is first matched with the target requirement description information. Then, in the case of a failed match, the suspected intent category and the credibility of the suspected intent category are determined. Different processing methods can be adopted in the case of successful and failed matches to achieve accurate processing of user needs.

[0137] Figure 2 This is a flowchart illustrating another data processing method provided in one or more embodiments of this specification. The process includes: Step S202: Obtain the target requirement data to be processed, wherein the target requirement data corresponds to at least one data modality; Step S204: Based on the data modality corresponding to the target requirement data, generate target requirement description information corresponding to the target requirement data; Step S206: Determine whether the stored multiple first requirement description information includes first target description information that matches the target requirement description information.

[0138] If yes, proceed to step S208; otherwise, proceed to step S210.

[0139] Step S208: Obtain the intent category corresponding to the first target description information, and process the target demand data according to the intent category corresponding to the first target description information and the target demand description information; Step S210: Determine the suspected intent category information of the target requirement description information, and process the target requirement data according to the suspected intent category information; the suspected intent category information is used to represent the suspected intent category corresponding to the target requirement description information and the credibility of the suspected intent category.

[0140] It should be noted that any one or more steps in steps S202 to S210 can be combined with any one or more steps in steps S102 to S108 to form a new implementation method according to the needs of implementation and deployment. In addition, any one or more technical features in steps S202 to S210 can be selected and combined with any one or more technical features provided in steps S102 to S108 to form a new implementation method according to the actual deployment needs. Alternatively, any one or more technical features in steps S202 to S210 can be replaced with any one or more technical features provided in steps S102 to S108 to form a new implementation method according to the actual deployment needs. These will not be elaborated on here.

[0141] In summary, this embodiment first matches the pre-stored first requirement description information with the target requirement description information, and then determines the suspected intent category and the credibility of the suspected intent category in the case of matching failure. Different processing methods can be adopted in the case of matching success and matching failure to achieve accurate processing of user needs.

[0142] Figure 3 This is a flowchart illustrating another data processing method provided in one or more embodiments of this specification. The process includes: Step S302: Obtain the target requirement data to be processed, wherein the target requirement data corresponds to at least one data modality; Step S304: Based on the data modality corresponding to the target requirement data, generate target requirement description information corresponding to the target requirement data; Step S306: Determine whether the stored multiple second requirement description information includes second target description information that matches the target requirement description information; If yes, proceed to step S308; otherwise, proceed to step S310.

[0143] Step S308: Obtain the response data or operation process corresponding to the second target description information, and process the target requirement data based on the response data or operation process; Step S310: Determine whether the stored multiple first requirement description information includes first target description information that matches the target requirement description information; If yes, proceed to step S208; otherwise, proceed to step S210.

[0144] Step S312: Obtain the intent category corresponding to the first target description information, and process the target demand data according to the intent category corresponding to the first target description information and the target demand description information; Step S314: Determine the suspected intent category information of the target requirement description information, and process the target requirement data according to the suspected intent category information; the suspected intent category information is used to represent the suspected intent category corresponding to the target requirement description information and the credibility of the suspected intent category.

[0145] It should be noted that any one or more steps in steps S302 to S314 can be combined with any one or more steps in steps S102 to S108 to form a new implementation method according to the needs of implementation and deployment. In addition, any one or more technical features in steps S302 to S314 can be selected and combined with any one or more technical features provided in steps S102 to S108 to form a new implementation method according to the actual deployment needs. Alternatively, any one or more technical features in steps S302 to S314 can be replaced with any one or more technical features provided in steps S102 to S108 to form a new implementation method according to the actual deployment needs. These will not be elaborated on here.

[0146] In summary, this embodiment first matches the pre-stored second requirement descriptions with the target requirement descriptions. If the matching fails, it matches the pre-stored first requirement descriptions with the target requirement descriptions. If the matching fails again, it determines the suspected intent category and the credibility of the suspected intent category. This reduces redundant calculations. Different processing methods are adopted in cases where the target requirement descriptions and second requirement descriptions match successfully, where the target requirement descriptions and second requirement descriptions fail to match but the target requirement descriptions and first requirement descriptions match successfully, and where the target requirement descriptions and second requirement descriptions fail to match and the target requirement descriptions and first requirement descriptions fail to match, in order to achieve accurate processing of user needs.

[0147] Figure 4 This is a schematic diagram of the structure of a data processing apparatus provided in one or more embodiments of this specification, such as... Figure 4 As shown, this device can run on a server and includes: Data acquisition unit 42 acquires target demand data to be processed, wherein the target demand data corresponds to at least one data modality; Information generation unit 44 generates target requirement description information corresponding to the target requirement data based on the data modality corresponding to the target requirement data; The first data processing unit 46, if the stored plurality of first demand description information includes first target description information that matches the target demand description information, then obtains the intent category corresponding to the first target description information, and processes the target demand data according to the intent category corresponding to the first target description information and the target demand description information; The second data processing unit 48, if the first target description information is not included in the plurality of first demand description information, determines the suspected intent category information of the target demand description information, and processes the target demand data according to the suspected intent category information; the suspected intent category information is used to represent the suspected intent category corresponding to the target demand description information and the credibility of the suspected intent category.

[0148] Optionally, the information generation unit 44 divides the target demand data into at least one sub-data based on the data modality corresponding to the target demand data; the sub-data has a corresponding data modality; the data content of the sub-data is extracted based on the data modality corresponding to the sub-data; and the data content of each sub-data is structurally extracted and fused through a large language model to obtain the target demand description information.

[0149] Optionally, the information generation unit 44 obtains an information template corresponding to the requirement description information; the information template includes multiple substructures; each substructure has corresponding data content requirements; using the large language model, the data content corresponding to each substructure is extracted from the data content of each substructure according to the data content requirements of each substructure; using the large language model, the data content corresponding to each substructure is fused according to the fusion relationship between each substructure in the information template to obtain the target requirement description information.

[0150] Optionally, the data processing device further includes a requirement generation unit, which acquires multiple first known intent categories, determines the structured requirement description method corresponding to the first known intent category, creates structured requirement description information corresponding to the first known intent category based on the structured requirement description method, and uses the structured requirement description information as the first requirement description information.

[0151] Optionally, the requirement generation unit determines at least one key intent information associated with the first known intent category and the order of each key intent information; the key intent information includes intent keywords and / or key semantic information; and determines the structured requirement description method based on the order of the key intent information.

[0152] Optionally, the first data processing unit 46 takes the intent category corresponding to the first target description information as the first target intent category corresponding to the target demand data, and determines the processing method corresponding to the target demand data; the processing method includes generating response data corresponding to the target demand data or executing the operation process corresponding to the target demand data; and processes the target demand data according to the processing method corresponding to the target demand data based on the first target intent category and the target demand description information.

[0153] Optionally, the processing method includes generating response data corresponding to the target demand data; the first data processing unit 46, through a question-and-answer model, retrieves knowledge point information corresponding to the target demand description information in a question-and-answer knowledge base based on the first target intent category and the target demand description information, and generates response data corresponding to the target demand data based on the knowledge point information.

[0154] Optionally, the processing method includes executing the operation process corresponding to the target demand data; the first data processing unit 46, through the process execution model, determines the process execution tool corresponding to the target demand data and the operation process corresponding to the target demand data according to the first target intent category and the target demand description information, and executes the operation process through the process execution tool.

[0155] Optionally, the second data processing unit 48, through an intent classification model, determines the second target intent category corresponding to the target demand data in each of the second known intent categories based on the target demand description information, and uses the second target intent category as the suspected intent category; obtains the credibility information of the target demand description information corresponding to the second target intent category, and generates the suspected intent category information based on the second target intent category and the credibility information.

[0156] Optionally, the second data processing unit 48, if it determines that the credibility of the suspected intent category is greater than the credibility threshold based on the suspected intent category information, then processes the target demand data based on the suspected intent category and the target demand description information; if it determines that the credibility of the suspected intent category is not greater than the credibility threshold based on the suspected intent category information, then processes the target demand description information through an artificial intelligence model.

[0157] Optionally, the second data processing unit 48 determines the processing method corresponding to the target demand data; the processing method includes generating response data corresponding to the target demand data or executing the operation process corresponding to the target demand data; and processing the target demand data according to the processing method corresponding to the target demand data based on the suspected intent category and the target demand description information.

[0158] Optionally, the data processing device further includes: a third data processing unit, which, if the stored plurality of second demand description information includes second target description information that matches the target demand description information, acquires response data or operation process corresponding to the second target description information, and processes the target demand data based on the response data or the operation process.

[0159] As can be seen, through this embodiment, the pre-stored first requirement description information is first matched with the target requirement description information, and then the suspected intent category and the credibility of the suspected intent category are determined in the case of matching failure. Different processing methods can be adopted in the case of matching success and matching failure, so as to achieve accurate processing of user needs.

[0160] For ease of description, the above devices are described by dividing them into various modules or units based on their functions. Of course, when implementing one or more of these specifications, the functions of each module or unit can be implemented in the same or different software and / or hardware, or a module that performs the same function can be implemented by a combination of multiple sub-modules or sub-units, etc. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0161] Since the apparatus embodiment corresponds to the method embodiment, the description is relatively simple. Figure 4 For a more detailed description of the apparatus, please refer to the description of the method embodiments, which will not be repeated here.

[0162] Corresponding to the data processing method described above, and based on the same technical concept, one or more embodiments of this specification also provide a data processing apparatus for performing the data processing method provided above. Figure 5 This is a schematic diagram of the structure of a data processing device provided for one or more embodiments of this specification.

[0163] like Figure 5As shown, device 500 mainly consists of a communication interface 502, a user interface 504, a processor 506, and a data storage 508. These components are interconnected and communicate with each other via a system bus, network, or other connection mechanism 510. The communication interface 502 enables device 500 to communicate with other devices, access networks, and transmission networks via analog or digital modulation. For example, the communication interface 502 may include a chipset and antenna for wireless communication with a radio access network or access point. Furthermore, the communication interface 502 can be a wired interface such as Ethernet, Token Ring, or a USB port, or a wireless interface such as Wi-Fi, Bluetooth, Global Positioning System (GPS), or a wide-area wireless interface (e.g., WiMAX or LTE). Of course, the communication interface 502 can also support other forms of physical layer interfaces and standard or proprietary communication protocols. The communication interface 502 may also include multiple physical communication interfaces, such as Wi-Fi interfaces, Bluetooth interfaces, and wide-area wireless interfaces.

[0164] User interface 504 includes receiving user input and providing output to the user. Therefore, user interface 504 may include input components such as a keypad, keyboard, touch-sensitive or presence-sensitive panel, computer mouse, trackball, joystick, microphone, still camera, and video camera, and output components such as a display screen (which may be combined with a touch-sensitive panel), CRT, LCD, LED, display using DLP technology, printer, and other similar devices known or developed in the future. User interface 504 may also generate auditory output via speakers, speaker jacks, audio output ports, audio output devices, headphones, and other similar devices known or developed in the future. In some embodiments, user interface 504 may include software, circuitry, or other forms of logic capable of transmitting and receiving data from external user input / output devices. Additionally or alternatively, device 500 may support remote access from other devices via communication interface 502 or another physical interface (not shown). User interface 504 may be configured to receive user input, the position and movement of which may be indicated by indicators or cursors described herein. User interface 504 may also be configured as a display device for rendering or displaying text fragments.

[0165] Processor 506 may include one or more general-purpose processors and / or special-purpose processors. Data storage 508 may include one or more volatile and / or non-volatile storage components and may be integrated wholly or partially with processor 506. Data storage 508 may include removable and non-removable components.

[0166] Processor 506 is capable of executing program instructions 518 (e.g., compiled or uncompiled program logic and / or machine code) stored in data storage 508 to perform the various functions described herein. Data storage 508 may contain a non-transitory computer-readable medium on which program instructions are stored, which, when executed by device 500, enable device 500 to perform any methods, processes, or functions disclosed in this specification and / or the accompanying drawings. Execution of program instructions 518 by processor 506 may result in processor 506 using data 512. For example, program instructions 518 may include an operating system 522 (e.g., an operating system kernel, device drivers, and / or other modules) installed on device 500 and one or more application programs 520 (e.g., a browser, social application, or game application).

[0167] Similarly, data 512 may contain operating system data 516 and application data 514. Operating system data 516 is primarily accessible to the operating system 522, while application data 514 is primarily accessible to one or more applications 520. Application data 514 may reside in a file system visible or hidden from the user of device 500. Applications 520 may communicate with the operating system 522 through one or more application programming interfaces (APIs). These APIs facilitate applications 520 in reading and / or writing application data 514, transmitting or receiving information via communication interface 502, receiving or displaying information on user interface 504, etc. In some terms, application 520 may be simply referred to as an "app". Furthermore, application 520 may be downloaded to device 500 through one or more online app stores or app markets. However, applications may also be installed on device 500 in other ways, such as through a web browser or a physical interface on device 500 (e.g., a USB port).

[0168] In one specific embodiment, the data processing includes a memory and one or more programs, wherein the one or more programs are stored in the memory, and the one or more programs may include one or more modules, and each module may include a series of computer-executable instructions for the data processing device, and is configured to be executed by one or more processors. The one or more programs include computer-executable instructions for performing the following: Acquire target demand data to be processed, wherein the target demand data corresponds to at least one data modality; Based on the data modality corresponding to the target demand data, generate target demand description information corresponding to the target demand data; If the stored multiple first demand description information includes first target description information that matches the target demand description information, then the intent category corresponding to the first target description information is obtained, and the target demand data is processed according to the intent category corresponding to the first target description information and the target demand description information. If the first target description information is not included among the plurality of first demand description information, then the suspected intent category information of the target demand description information is determined, and the target demand data is processed according to the suspected intent category information; the suspected intent category information is used to represent the suspected intent category corresponding to the target demand description information and the credibility of the suspected intent category.

[0169] As can be seen, through this embodiment, the pre-stored first requirement description information is first matched with the target requirement description information, and then the suspected intent category and the credibility of the suspected intent category are determined in the case of matching failure. Different processing methods can be adopted in the case of matching success and matching failure, so as to achieve accurate processing of user needs.

[0170] Since the device embodiment corresponds to the method embodiment, the description is relatively simple. For more information about the device, please refer to the description of the method embodiment, which will not be repeated here.

[0171] This specification provides an embodiment of a computer-readable storage medium as follows: Corresponding to the data processing method described above, and based on the same technical concept, one or more embodiments of this specification also provide a computer-readable storage medium.

[0172] The computer-readable storage medium provided in this embodiment is used to store computer-executable instructions, which, when executed, implement the following process: Acquire target demand data to be processed, wherein the target demand data corresponds to at least one data modality; Based on the data modality corresponding to the target demand data, generate target demand description information corresponding to the target demand data; If the stored multiple first demand description information includes first target description information that matches the target demand description information, then the intent category corresponding to the first target description information is obtained, and the target demand data is processed according to the intent category corresponding to the first target description information and the target demand description information. If the first target description information is not included among the plurality of first demand description information, then the suspected intent category information of the target demand description information is determined, and the target demand data is processed according to the suspected intent category information; the suspected intent category information is used to represent the suspected intent category corresponding to the target demand description information and the credibility of the suspected intent category.

[0173] As can be seen, through this embodiment, the pre-stored first requirement description information is first matched with the target requirement description information, and then the suspected intent category and the credibility of the suspected intent category are determined in the case of matching failure. Different processing methods can be adopted in the case of matching success and matching failure, so as to achieve accurate processing of user needs.

[0174] It should be noted that the embodiments of a computer-readable storage medium described in this specification and the embodiments of a data processing method described in this specification are based on the same inventive concept. Therefore, the specific implementation of this embodiment can be referred to the implementation of the corresponding method described above, and the repeated parts will not be described again.

[0175] This specification provides an example of a computer program product as follows: Corresponding to the data processing method described above, and based on the same technical concept, one or more embodiments of this specification also provide a computer program product.

[0176] A computer program product includes a computer program / instructions that, when executed by a processor, perform the following steps: Acquire target demand data to be processed, wherein the target demand data corresponds to at least one data modality; Based on the data modality corresponding to the target demand data, generate target demand description information corresponding to the target demand data; If the stored multiple first demand description information includes first target description information that matches the target demand description information, then the intent category corresponding to the first target description information is obtained, and the target demand data is processed according to the intent category corresponding to the first target description information and the target demand description information. If the first target description information is not included among the plurality of first demand description information, then the suspected intent category information of the target demand description information is determined, and the target demand data is processed according to the suspected intent category information; the suspected intent category information is used to represent the suspected intent category corresponding to the target demand description information and the credibility of the suspected intent category.

[0177] As can be seen, through this embodiment, the pre-stored first requirement description information is first matched with the target requirement description information, and then the suspected intent category and the credibility of the suspected intent category are determined in the case of matching failure. Different processing methods can be adopted in the case of matching success and matching failure, so as to achieve accurate processing of user needs.

[0178] It should be noted that the embodiments of a computer program product described in this specification and the embodiments of a data processing method described in this specification are based on the same inventive concept. Therefore, the specific implementation of this embodiment can be referred to the implementation of the corresponding method described above, and the repeated parts will not be described again.

[0179] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments. For example, the device embodiment, equipment embodiment and computer-readable storage medium embodiment are all similar to the method embodiment, so the description is relatively simple. When reading the relevant content of the device embodiment, equipment embodiment and computer-readable storage medium embodiment, please refer to the description of the method embodiment.

[0180] While one or more embodiments of this specification provide method steps as described in the embodiments or flowcharts, it is understood that the order of steps listed in the embodiments or flowcharts is merely one possible execution order among many steps, and does not represent the only execution order. Therefore, when the claims involve method steps, any changes or adjustments to the order of such steps, or the parallelism between steps, are also within the scope of protection of the claims. This specification uses specific terms to describe embodiments of this specification. For example, "an embodiment," "one embodiment," and / or "some embodiments" refer to a particular feature, structure, or characteristic related to at least one embodiment of this specification. Therefore, it should be emphasized and noted that "an embodiment," "one embodiment," or "an alternative embodiment" mentioned twice or more in different locations in this specification do not necessarily refer to the same embodiment. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0181] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0182] In the 1930s, improvements to a technology could be clearly distinguished as either hardware improvements (e.g., improvements to the circuit structure of diodes, transistors, switches, etc.) or software improvements (improvements to the methodology). However, with technological advancements, many improvements to the methodology today can be considered direct improvements to the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved methodology into the hardware circuit. Therefore, it cannot be said that an improvement to the methodology cannot be implemented using hardware physical modules. For example, a Programmable Logic Device (PLD) (such as a Field Programmable Gate Array (FPGA)) is such an integrated circuit whose logic function is determined by the user programming the device. Designers can program and "integrate" a digital system onto a PLD themselves, without needing chip manufacturers to design and manufacture dedicated integrated circuit chips. Furthermore, nowadays, instead of manually manufacturing integrated circuit chips, this programming is mostly implemented using "logic compiler" software. Similar to the software compiler used in program development, the original code before compilation must also be written in a specific programming language, called a Hardware Description Language (HDL). There are many HDLs, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, and RHDL (Ruby Hardware Description Language). Currently, the most commonly used are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also understand that by simply performing some logic programming on the method flow using one of these hardware description languages ​​and programming it into an integrated circuit, the hardware circuit implementing the logical method flow can be easily obtained.

[0183] The controller can be implemented in any suitable manner. For example, it can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicon Labs C8051F320. A memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also recognize that, in addition to implementing the controller in purely computer-readable program code form, the same functionality can be achieved by logically programming the method steps to make the controller take the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the means included therein for implementing various functions can also be considered as structures within the hardware component. Alternatively, the means for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0184] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0185] For ease of description, the above apparatus is described by dividing it into various functional units. Of course, when implementing the embodiments of this specification, the functions of each unit can be implemented in one or more software and / or hardware.

[0186] Those skilled in the art will understand that one or more embodiments of this specification can be provided as a method, system, or computer program product. Therefore, one or more embodiments of this specification may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this specification may take the form of a computer program product embodied on one or more computer-readable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0187] This specification is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this specification. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0188] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0189] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0190] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0191] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0192] Computer-readable media include both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer-readable storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0193] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising at least one…" does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0194] It should also be noted that the terms "one," "an," and "the" do not specifically refer to the singular; they can also include the plural. Ordinal numbers such as "first," "second," etc., do not necessarily indicate order; often they are used to distinguish objects. For example, "first server" and "second server" usually refer to two servers. To differentiate between these two servers, they are described as "first server" and "second server." Of course, sometimes these two servers may be the same server. Unless explicitly stated, "receiving and sending data" does not necessarily mean direct receipt and transmission; it can be indirect. For example, A receiving data sent by B can be understood as A directly receiving data sent by B, or it can be understood as A indirectly receiving data sent by B through other entities such as C. Similarly, B sending data to A can be understood as B sending data directly to A, or it can be understood as B indirectly sending data to A through other entities such as C. Here, C can be one entity, or it can be two or more entities.

[0195] Unless explicitly stated otherwise, the relationships between structures can be direct or indirect. For example, when describing "A is connected to B," unless it is explicitly stated that A and B are directly connected, it should be understood that A can be directly connected to B or indirectly connected to B. Similarly, when describing "A is above B," unless it is explicitly stated that A is directly above B (AB is adjacent and A is above B), it should be understood that A can be directly above B or indirectly above B (AB is separated by other elements, and A is above B). And so on.

[0196] One or more embodiments of this specification can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a particular task or implement a particular abstract data type. One or more embodiments of this specification can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0197] The above description is merely an embodiment of this document and is not intended to limit the scope of this document. Various modifications and variations can be made to this document by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this document should be included within the scope of the claims of this document.

Claims

1. A data processing method, comprising: Acquire target demand data to be processed, wherein the target demand data corresponds to at least one data modality; Based on the data modality corresponding to the target demand data, generate target demand description information corresponding to the target demand data; If the stored multiple first demand description information includes first target description information that matches the target demand description information, then the intent category corresponding to the first target description information is obtained, and the target demand data is processed according to the intent category corresponding to the first target description information and the target demand description information. If the first target description information is not included among the plurality of first demand description information, then the suspected intent category information of the target demand description information is determined, and the target demand data is processed according to the suspected intent category information; the suspected intent category information is used to represent the suspected intent category corresponding to the target demand description information and the credibility of the suspected intent category.

2. The method according to claim 1, wherein generating target requirement description information corresponding to the target requirement data based on the data modality corresponding to the target requirement data includes: Based on the data modality corresponding to the target demand data, the target demand data is divided into at least one sub-data. The sub-data has a corresponding data modality; Based on the data modality corresponding to the sub-data, extract the data content of the sub-data; By using a large language model, the data content of each sub-data is structurally extracted and fused to obtain the target requirement description information.

3. The method according to claim 2, wherein the step of extracting and fusing the data content of each of the sub-data using a large language model to obtain the target requirement description information includes: Obtain the information template corresponding to the requirement description information; The information template includes multiple substructures; each substructure has corresponding data content requirements. Using the large language model, the data content corresponding to the substructure is extracted from the data content of each sub-data according to the data content requirements corresponding to the substructure; Using the large language model, the data content corresponding to each substructure is fused according to the fusion relationship between the substructures in the information template to obtain the target requirement description information.

4. The method according to claim 1, further comprising: Obtain multiple known intent categories and determine the structured requirement description method corresponding to each known intent category; Based on the structured requirement description method, create structured requirement description information corresponding to the first known intent category; The structured requirement description information is used as the first requirement description information.

5. The method according to claim 4, wherein determining the structured requirement description method corresponding to the first known intent category includes: Determine at least one key intent information associated with the first known intent category and the order of the key intent information; The key intent information includes intent keywords and / or key semantic information; The structured requirement description method is determined based on the key intent information and the order in which the key intent information is presented.

6. The method according to claim 1, wherein processing the target demand data based on the intent category corresponding to the first target description information and the target demand description information comprises: The intent category corresponding to the first target description information is taken as the first target intent category corresponding to the target demand data, and the processing method corresponding to the target demand data is determined. The processing method includes generating response data corresponding to the target demand data or executing the operation process corresponding to the target demand data; Based on the first target intent category and the target requirement description information, the target requirement data is processed according to the processing method corresponding to the target requirement data.

7. The method according to claim 6, wherein the processing method includes generating response data corresponding to the target demand data; and the step of processing the target demand data according to the processing method corresponding to the target demand data based on the first target intent category and the target demand description information includes: Using a question-and-answer model, based on the first target intent category and the target demand description information, knowledge point information corresponding to the target demand description information is retrieved from the question-and-answer knowledge base, and response data corresponding to the target demand data is generated based on the knowledge point information.

8. The method according to claim 6, wherein the processing method includes executing the operation flow corresponding to the target demand data; the step of processing the target demand data according to the processing method corresponding to the target demand data based on the first target intent category and the target demand description information includes: Based on the first target intent category and the target requirement description information, the process execution tool corresponding to the target requirement data and the operation process corresponding to the target requirement data are determined by the process execution model, and the operation process is executed by the process execution tool.

9. The method according to claim 1, wherein determining the suspected intent category information of the target demand description information includes: Using the intent classification model, based on the target demand description information, the second target intent category corresponding to the target demand data is determined among each second known intent category, and the second target intent category is used as the suspected intent category; Obtain the credibility information corresponding to the second target intent category from the target demand description information, and generate the suspected intent category information based on the second target intent category and the credibility information.

10. The method according to claim 1, wherein processing the target demand data based on the suspected intent category information includes: If the credibility of the suspected intent category is determined to be greater than the credibility threshold based on the suspected intent category information, then the target demand data is processed based on the suspected intent category and the target demand description information. If the credibility of the suspected intent category is determined to be no greater than the credibility threshold based on the suspected intent category information, then the target demand description information is processed using an artificial intelligence model.

11. The method according to claim 10, wherein processing the target demand data based on the suspected intent category and the target demand description information includes: Determine the processing method corresponding to the target demand data; The processing method includes generating response data corresponding to the target demand data or executing the operation process corresponding to the target demand data; Based on the suspected intent category and the target demand description information, the target demand data is processed according to the processing method corresponding to the target demand data.

12. The method according to claim 1, further comprising: If the stored multiple second requirement description information includes second target description information that matches the target requirement description information, then the response data or operation process corresponding to the second target description information is obtained, and the target requirement data is processed based on the response data or the operation process.

13. A data processing apparatus, comprising: A data acquisition unit acquires target demand data to be processed, wherein the target demand data corresponds to at least one data modality; The information generation unit generates target requirement description information corresponding to the target requirement data based on the data modality corresponding to the target requirement data. The first data processing unit, if the stored plurality of first demand description information includes first target description information that matches the target demand description information, then obtains the intent category corresponding to the first target description information, and processes the target demand data according to the intent category corresponding to the first target description information and the target demand description information; The second data processing unit, if the first target description information is not included among the plurality of first demand description information, determines the suspected intent category information of the target demand description information, and processes the target demand data according to the suspected intent category information; the suspected intent category information is used to represent the suspected intent category corresponding to the target demand description information and the credibility of the suspected intent category.

14. A data processing apparatus, comprising: processor; And, a memory configured to store computer-executable instructions, which, when executed, cause the processor to: Acquire target demand data to be processed, wherein the target demand data corresponds to at least one data modality; Based on the data modality corresponding to the target demand data, generate target demand description information corresponding to the target demand data; If the stored multiple first demand description information includes first target description information that matches the target demand description information, then the intent category corresponding to the first target description information is obtained, and the target demand data is processed according to the intent category corresponding to the first target description information and the target demand description information. If the first target description information is not included among the plurality of first demand description information, then the suspected intent category information of the target demand description information is determined, and the target demand data is processed according to the suspected intent category information; the suspected intent category information is used to represent the suspected intent category corresponding to the target demand description information and the credibility of the suspected intent category.

Images