Data processing method and electronic equipment

Abstract

The invention provides a data processing method. The data processing method comprises the following steps: determining first input information of a large model; based on the first input information, at least one target file is determined, and the target file represents a user reference object in the first input information. And supplementing the first input information based on the at least one target file to obtain second input information. Based on the second input information, an output of the large model is determined.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and in particular to a data processing method and an electronic device. Background Art

[0002] Knowledge Base Question Answering (KB-QA) is a question-answering system based on a knowledge base. When given a natural language question, the system semantically understands and parses the question, and then uses the knowledge base to query and infer the answer.

[0003] Currently, when using a local large model for knowledge base question and answer, you need to manually filter the knowledge base before entering the question, and the intelligent experience is not good. Summary of the invention

[0004] One aspect of the present disclosure provides a data processing method, including: determining first input information of a large model. Based on the first input information, determining at least one target file, the target file representing a user-referenced object in the first input information. Based on the at least one target file, supplementing the first input information to obtain second input information. Based on the second input information, determining the output of the large model.

[0005] Optionally, based on the first input information, determining at least one target file includes at least one of the following: in response to a user inputting a first text into the macro model, searching multiple knowledge bases to obtain at least one target file, and the multiple knowledge bases are associated with the macro model. In response to a user operation on the current application, determining a first associated object of a page displayed by the current application. Based on the first associated object, searching multiple knowledge bases to obtain at least one target file.

[0006] Optionally, in response to the user inputting the first text into the big model, searching multiple knowledge bases to obtain at least one target file includes at least one of the following: in response to the user inputting instruction information into the big model, searching multiple knowledge bases to obtain at least one target file, the instruction information represents the user's action instruction for the big model. In response to the user inputting reference information into the big model, searching multiple knowledge bases to obtain at least one target file, the reference information represents the target reference object stored in the multiple knowledge bases.

[0007] Optionally, in response to a user operation on the current application, determining a first associated object of a page displayed by the current application includes at least one of the following: extracting information from a displayed page of the current application to obtain the first associated object. Obtaining the first associated object based on an association between the current application and the large model, wherein the current application and the large model are respectively associated with the same user account.

[0008] Optionally, determining at least one target file based on the first input information further includes: determining a second associated object of the page displayed by the current application in response to a user operation on the current application. Determining the second associated object to be the at least one target file.

[0009] Optionally, the first input information includes at least one of a first text, an associated object, and a second text, and the associated object is determined based on a user's operation on the current application. Based on at least one target file, the first input information is supplemented to obtain the second input information, including at least one of the following: adding the first target file to the first text to obtain the second input information, and the first target file is determined based on the first text. Adding the second target file to the second text to obtain the second input information, and the second target file is determined based on the associated object.

[0010] Optionally, the data processing method further comprises: searching multiple knowledge bases based on the associated objects to obtain at least one candidate file, displaying the at least one candidate file, and determining a second target file in response to a user selecting the at least one candidate file.

[0011] Optionally, determining the first input information of the large model includes: obtaining copy information of the user, displaying the copy information, and determining the first input information in response to the user selecting the copy information.

[0012] Optionally, determining the output of the large model based on the second input information includes: extracting information from the second input information to obtain index information, searching at least one target file based on the index information, and integrating the search results to obtain the output of the large model.

[0013] Another aspect of the present disclosure provides an electronic device, comprising: an input part, for inputting first input information and second input information into a large model; a storage part, for storing at least one target file; one or more processors, for determining at least one target file based on the first input information, the target file representing the user-referenced object in the first input information; based on the at least one target file, updating the input content of the large model to obtain the second input information; and determining the output of the large model based on the second input information. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The above and other objects, features and advantages of the present disclosure will become more apparent through the following description of the embodiments of the present disclosure with reference to the accompanying drawings, in which:

[0015] Figure 1 The application scenario diagram of the data processing method according to the embodiment of the present disclosure is schematically shown;

[0016] Figure 2 A flowchart schematically shows a data processing method according to an embodiment of the present disclosure;

[0017] Figure 3 A flowchart of a method for determining a target file according to an embodiment of the present disclosure is schematically shown;

[0018] Figure 4 A flowchart of a method for determining a target file according to another embodiment of the present disclosure is schematically shown;

[0019] Figure 5 A flowchart of a method for determining an associated object according to an embodiment of the present disclosure is schematically shown;

[0020] Figure 6 A flowchart of a method for determining a target file according to another embodiment of the present disclosure is schematically shown;

[0021] Figure 7 A flowchart of a method for obtaining second input information according to an embodiment of the present disclosure is schematically shown;

[0022] Figure 8 A flowchart schematically shows a data processing method according to another embodiment of the present disclosure;

[0023] Fig. 9 A flowchart of a method for determining first input information according to an embodiment of the present disclosure is schematically shown;

[0024] Fig.10 A flowchart of a method for determining a large model output according to an embodiment of the present disclosure is schematically shown;

[0025] Fig.11 The structure block diagram of the data processing device according to the embodiment of the present disclosure is schematically shown;

[0026] Fig.12 A block diagram of an electronic device suitable for implementing a data processing method according to an embodiment of the present disclosure is schematically shown. DETAILED DESCRIPTION

[0027] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In addition, in the following description, descriptions of well-known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0028] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0029] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0030] Some block diagrams and / or flow charts are shown in the accompanying drawings. It should be understood that some blocks or combinations thereof in the block diagrams and / or flow charts may be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that these instructions, when executed by the processor, may create a device for implementing the functions / operations described in these block diagrams and / or flow charts.

[0031] Therefore, the technology of the present disclosure can be implemented in the form of hardware and / or software (including firmware, microcode, etc.). In addition, the technology of the present disclosure can take the form of a computer program product on a computer-readable medium storing instructions, which can be used by an instruction execution system or in combination with an instruction execution system. In the context of the present disclosure, a computer-readable medium can be any medium that can contain, store, transmit, propagate, or transmit instructions. For example, a computer-readable medium can include, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, device, or propagation medium. Specific examples of computer-readable media include: magnetic storage devices, such as magnetic tape or hard disk (HDD); optical storage devices, such as compact disk (CD-ROM); memory, such as random access memory (RAM) or flash memory; and / or wired / wireless communication links.

[0032] The embodiments of the present disclosure provide a data processing method for improving the intelligence level of large-model question and answer. The data processing method includes: determining the first input information of the large model. Based on the first input information, determining at least one target file, the target file represents the user-referenced object in the first input information. Based on the at least one target file, the first input information is supplemented to obtain the second input information. Based on the second input information, the output of the large model is determined. By automatically determining the target file related to the answer after the user inputs the question or before the user inputs the question, and then inputting the target file into the model and obtaining the answer, the intelligence level of the entire question and answer process is improved.

[0033] Figure 1 The application scenario diagram of the data processing method according to the embodiment of the present disclosure is schematically shown.

[0034] like Figure 1As shown, the application scenario 100 according to this embodiment may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for a communication link between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0035] Users can use terminal devices 101, 102, 103 to interact with server 105 through network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, 103, such as intelligent question-and-answer applications, security applications, shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only examples).

[0036] The terminal devices 101 , 102 , and 103 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, and desktop computers.

[0037] The server 105 may be a server that provides various services, such as a background management server (only an example) that provides support for websites browsed or applications logged in by users using the terminal devices 101, 102, and 103. The background management server may analyze and process the received data such as user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal device to intelligently answer questions raised by users.

[0038] It should be noted that the data processing method provided in the embodiment of the present disclosure can generally be executed by the server 105 or by the terminal device. Accordingly, the data processing device provided in the embodiment of the present disclosure can generally be set in the server 105. The data processing method provided in the embodiment of the present disclosure can also be executed by a server or server cluster that is different from the server 105 and can communicate with the terminal devices 101, 102, 103 and / or the server 105. Accordingly, the data processing device provided in the embodiment of the present disclosure can also be set in a server or server cluster that is different from the server 105 and can communicate with the terminal devices 101, 102, 103 and / or the server 105.

[0039] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0040] The following will be based on Figure 1 The scene described by Figure 2~Figure 10The data processing method of the disclosed embodiment is described in detail.

[0041] Figure 2 The flowchart of the data processing method according to the embodiment of the present disclosure is schematically shown.

[0042] According to the embodiments of the present disclosure, Figure 2 As shown, the data processing method of this embodiment includes, for example, operations S210 to S240.

[0043] In operation S210, first input information of a large model is determined.

[0044] For example, the first input information can be a question input by the user, such as instructions, prompt words, and directional text. The process of knowledge-based question answering (KB-QA) usually includes the user inputting a question, and then the system uses the pre-built knowledge base and question-answering model to parse the question and retrieve relevant information from the knowledge base to generate an answer. This process can be seen as an input-processing-output process, where the input is the user's question, the processing is completed by the question-answering model, and the output is the answer extracted and generated from the knowledge base.

[0045] Among them, the knowledge base is, for example, a database that stores a large amount of structured or semi-structured information. In the KB-QA system, the knowledge base usually contains information such as entities, attributes, and relationships, which are the basis for generating accurate answers. The knowledge base can be general (such as online encyclopedia data) or targeted at specific fields or applications (such as medical knowledge base).

[0046] The big model (question-answering model) is, for example, a component responsible for parsing user questions, retrieving relevant information from the knowledge base, and generating answers. It usually includes natural language processing (NLP) technologies such as word segmentation, part-of-speech tagging, named entity recognition, semantic understanding, etc., as well as information retrieval and machine learning algorithms. The question-answering model needs to be able to accurately understand the intent of the user's question and convert it into a query to the knowledge base.

[0047] Based on different system application scenarios, performance requirements, data resources and other factors, the question-answering model can be a variety of large models.

[0048] Exemplarily, the large model can be a question-answering model based on keyword matching. This model finds the corresponding answer by matching the keywords in the user's question with the keywords in the preset answer library. Its implementation is relatively simple and is usually suitable for handling some relatively direct and clear questions.

[0049] Exemplarily, the large model can also be a question-answering model based on natural language understanding (NLU). This type of model uses natural language processing technology to deeply analyze user questions, understand their semantics and context, and find more accurate answers. This usually involves natural language processing tasks such as part-of-speech tagging, named entity recognition, and syntactic parsing. Question-answering models based on natural language understanding usually require large-scale training data and complex algorithm support, but they can handle a wider range of more complex question types.

[0050] Exemplarily, the large model can also be a question-answering model based on deep learning. These models use neural networks to learn large amounts of text data, and can automatically extract features and generate answers. Common deep learning models include BERT, RoBERTa, etc. These models can capture rich semantic information and generate high-quality answers when processing natural language questions.

[0051] Exemplarily, the large model can also be a question-answering model based on a knowledge graph. A knowledge graph is a structured way of representing knowledge that organizes information such as entities, attributes, and relationships in the form of a graph, making it easier for computers to perform efficient queries and reasoning. A question-answering model based on a knowledge graph can use the information in the knowledge graph to parse user questions and find relevant answers through a graph query algorithm. This model has advantages in dealing with problems involving multiple entities and complex relationships.

[0052] In practical applications, hybrid models are often used to improve the performance and accuracy of question-answering systems. Hybrid models can combine the advantages of multiple question-answering models, such as combining keyword matching-based methods with deep learning-based methods, or combining natural language understanding-based methods with knowledge graph-based methods. By combining the advantages of multiple technologies, hybrid models can handle a wider range of more complex question types and provide higher-quality answers.

[0053] For another example, the first input information may also be keywords, paragraphs, and files automatically recognized by the system, such as keywords of a currently received email, a document currently being previewed, and text currently in the clipboard.

[0054] For example, the user enters a specific question about a weekly report summary, translation, email, or document reading through the interface, and the question is used as the first input information of the large model. For example, the user may enter a question like "How was the project progress last week?"

[0055] In operation S220, at least one target file is determined based on the first input information, where the target file represents an object referred to by a user in the first input information.

[0056] The target file is, for example, a file associated with the first input information, that is, a file that the user wants or needs to use when answering the current question. The target file can be a file stored in the knowledge base, or a file that is not currently stored in the knowledge base. The question-answering model works, for example, in the form of a question-answering program.

[0057] The user's reference object is related to the user's intention. Since the user can directly open the Q&A program (without performing other program operations) and enter the question, or can open the Q&A program and enter the question while using other applications, the target file can also be determined based on the clarity of the user's intention. When the user directly opens the Q&A program and enters the question, there is a clear intention, and the target file can be determined based on the clear intention. When the user opens the Q&A program while using other programs, the intention is not clear before the user enters the question, but the files that the user may need and are associated with it can be determined as potential target files to be selected based on the content identified from the current program.

[0058] For example, the system analyzes the first input information and identifies the user's reference object, such as "last week's project progress". Subsequently, the system searches for files related to "last week's project progress" in the pre-established knowledge base. These files may include last week's project report, meeting minutes or other related documents, which are determined as target files.

[0059] For example, the target file includes any one or more of documents, pictures, audio and video.

[0060] In operation S230, the first input information is supplemented based on at least one target file to obtain second input information.

[0061] You can add the target file directly to the input box, or you can extract information from the target file and then add the extracted information to the input box.

[0062] For example, the system reads the contents of the target file and extracts key information related to the first input information. This information is used to supplement the first input information to form a more complete and specific second input information. For example, the second input information may include the first input information plus specific project progress details extracted from the target file.

[0063] In operation S240, based on the second input information, an output of the large model is determined.

[0064] For example, the second input information is fed into the large model, and the large model generates a response based on the information. The response may be a brief summary or a detailed explanation or analysis.

[0065] The method of this embodiment improves the intelligence level of the entire question-and-answer process by automatically determining target documents related to the answer and inputting their content into the model. This method not only reduces the time for users to manually search for and provide relevant information, but also improves the accuracy and efficiency of question-and-answer.

[0066] Figure 3 Schematically shows a flowchart of a method for determining a target document according to an embodiment of the present disclosure.

[0067] According to an embodiment of the present disclosure, as Figure 3 shown, in addition to including the operations S210 - S240 described above with reference to Figure 2 description, the method of this embodiment also determines at least one target document based on the first input information, for example, through at least one of the operations S321 - S323. For the sake of brevity of description, the description of the operations S210 - S240 is omitted here, and the subsequent related method embodiments are the same by analogy and will not be elaborated further.

[0068] In operation S321, in response to the user inputting a first text to the large model, multiple knowledge bases are searched to obtain at least one target document, and the multiple knowledge bases are associated with the large model.

[0069] The user inputs a specific question or text, such as about weekly report summary, translation, email, or document viewing, through the interface, and this question or text serves as the first input information of the large model. Here, the first input information can be one of two forms: the first text or the determined first associated object.

[0070] When the first input content is the first text, in response to the user inputting the first text to the large model, the system searches multiple knowledge bases associated with the large model. These knowledge bases may include internal document libraries, project management systems, email servers, etc., which store various information related to the user's question.

[0071] The system matches and retrieves in the knowledge base according to the keywords, phrases, or context information in the first text, so as to obtain at least one target document related to the first text.

[0072] In operation S322, in response to the user's operation on the current application, the first associated object of the page displayed by the current application is determined.

[0073] In operation S323, multiple knowledge bases are searched based on the first associated object to obtain at least one target document.

[0074] When the first input content is a determined first associated object, in response to the user's operation on the current application (such as clicking, selecting, etc.), the system determines the first associated object of the page displayed by the current application. This associated object may be a link, a button, a selected text paragraph, etc. on the page, which represents the content that the user is currently paying attention to or may be interested in.

[0075] Based on this first associated object, the system searches multiple knowledge bases associated with the large model. This search may be more specific and targeted because the associated object has provided the user with a clear context or direction.

[0076] Through matching and searching, the system obtains at least one target file related to the first associated object.

[0077] The data processing method of this embodiment considers the first input content in two different situations: the first text and the determined first associated object. This method not only improves the accuracy and efficiency of the search, but also makes the entire question-and-answer process more intelligent and personalized. By combining the target file determination methods in these two situations, the system can more comprehensively meet the user's knowledge base question-and-answer needs in scenarios such as weekly report summaries, translations, emails, and document reading.

[0078] Figure 4 A flow chart of a method for determining a target file according to another embodiment of the present disclosure is schematically shown.

[0079] According to the embodiments of the present disclosure, Figure 4 As shown, for example, by performing at least one of operations S4211 to S4212 in response to a user inputting a first text into a large model, multiple knowledge bases are searched to obtain at least one target file.

[0080] In operation S4211, in response to the user inputting instruction information to the large model, multiple knowledge bases are searched to obtain at least one target file, and the instruction information represents the user's action instructions for the large model.

[0081] The user inputs a question, an instruction or a text to the large model through the interface, and this information is used as the first input information of the large model. In this embodiment, the first input information can be instruction information or reference information input by the user.

[0082] When the first input information is instruction information, the instruction information represents a specific action instruction of the user for the large model, such as "find the minutes of last week's meeting", "translate this text into English", etc.

[0083] In response to these instruction messages, the system searches multiple knowledge bases associated with the large model, which may include internal documents, email archives, translation memories, etc.

[0084] The system analyzes the instruction information, identifies the user's intention, and matches and retrieves relevant target files in the knowledge base. For example, for the instruction "Find the meeting minutes from last week", the system may search the project management system or the document library to find the documents related to the meeting last week.

[0085] For example, when the user clearly issues an instruction such as "Summarize", non-picture files in the knowledge base are automatically filtered. And when entering "Translate", non-system language files in the knowledge base are automatically filtered, etc.

[0086] In operation S4212, in response to the user inputting referential information to the large model, multiple knowledge bases are searched to obtain at least one target file, where the referential information represents the target referential object stored in multiple knowledge bases.

[0087] When the first input information is referential information, the referential information represents a specific target referential object stored in multiple knowledge bases or in the historical conversation, such as "The progress report of Project A", "The email about the budget sent by Zhang San", etc.

[0088] The user may directly input this referential information, or indirectly provide it by selecting links, buttons, etc. on the interface.

[0089] Based on this referential information, the system conducts an exact search in the knowledge base to find the target files that exactly match or are highly relevant to the referential object.

[0090] For example, when obtaining directive words in the input content such as "The PPT just added to the knowledge base", "The graph just received", "The weekly report", "The second one in the previous question", etc., after retrieving the relevant target files in the knowledge base, the corresponding PPT, graph, weekly report, and question are automatically added to the input box.

[0091] In both of these situations, the system may return multiple target files as search results, and these files are highly relevant to the instruction information or referential information input by the user.

[0092] Based on at least one target file, the first input information is supplemented to obtain the second input information (this step may involve extracting, integrating, and reorganizing the content of the target file to form a more complete and specific input information for the large model to process, but the specific implementation method depends on the characteristics and application scenarios of the large model).

[0093] Based on the second input information, the output of the large model is determined (the large model generates an answer or executes relevant tasks according to the second input information, and the output may include various forms such as text, image, audio, etc.).

[0094] The data processing method of this embodiment takes into account two different situations: instruction information and reference information input by the user. By parsing the instructions and reference objects input by the user, the system can perform accurate searches in multiple knowledge bases to find target files that are highly relevant to the user's intentions. This method not only improves the accuracy and efficiency of the search, but also makes the entire question-and-answer or task processing process more intelligent and adaptive. By combining the target file determination methods in these two situations, the system can more comprehensively meet the needs of users in various application scenarios, such as document retrieval, translation, email processing, etc.

[0095] Figure 5 A flowchart of a method for determining an associated object according to an embodiment of the present disclosure is schematically shown.

[0096] According to the embodiments of the present disclosure, Figure 5 As shown, for example, by performing at least one of operations S5221 to S5222 in response to a user's operation on the current application, a first associated object of the page displayed by the current application is determined.

[0097] In operation S5221, information is extracted from a display page of the current application to obtain a first associated object.

[0098] When using a smartphone or computer, users may switch between different applications, which may include social media, email clients, browsers, etc. Suppose a user is browsing an email about "the development trend of artificial intelligence", and then he opens a question-and-answer software to get more information about this topic. At this time, the question-and-answer software needs to intelligently identify the user's possible intentions and provide relevant question suggestions or answers accordingly.

[0099] The system can extract based on the interface information of the current application. When the user switches from the email client to the question-and-answer software, the question-and-answer software first captures the current display interface of the email client (such as through a screenshot or an API interface).

[0100] The screenshot is processed using OCR (optical character recognition) or image recognition technology to extract key information, such as the email subject "Development Trends of Artificial Intelligence" and keywords in the text.

[0101] The extracted information is used as the first associated object for subsequent knowledge base retrieval.

[0102] In operation S5222, a first associated object is obtained based on the association between the current application and the large model, wherein the current application and the large model are respectively associated with the same user account.

[0103] Assume that the user is logged into the same user account when using the email client and the question-answering software (eg, through a single sign-on system).

[0104] The question-and-answer software uses this account association to directly access or request the API interface provided by the email client to obtain the email information that the user is currently viewing or has recently viewed.

[0105] Due to account association, the Q&A software can directly obtain key information in the email, such as email subject, sender, recipient, email body content, etc., as the first associated object without taking screenshots or performing OCR processing.

[0106] The question-answering software combines the first associated objects obtained by the above two methods and performs a comprehensive analysis to more accurately determine the user's intention.

[0107] At the same time, the question-and-answer software can further optimize the search content and improve the user experience based on the user's historical behavior, preferences and other information.

[0108] Through the above embodiments, determining the first associated object of the page displayed by the current application can be achieved not only by extracting information from the display page of the current application, but also based on the association between the current application and the question-and-answer model (or question-and-answer software). This association may be caused by the user using multiple applications under the same account, or the data sharing between applications through an API interface. In actual applications, intelligent applications such as question-and-answer software can flexibly combine these two methods, as well as information such as the user's historical behavior and preferences, to more accurately identify user intentions and provide personalized services. This method not only improves the intelligence level of the application, but also greatly improves the user experience.

[0109] Figure 6 A flowchart of a method for determining a target file according to yet another embodiment of the present disclosure is schematically shown.

[0110] According to the embodiments of the present disclosure, Figure 6 As shown, at least one target file is determined based on the first input information through operations S621-S622, for example.

[0111] In operation S621 , in response to a user operation on a current application, a second associated object of a page displayed by the current application is determined.

[0112] In operation S622, it is determined that the second associated object is at least one target file.

[0113] In some embodiments, for example, when a user is previewing a document and opens the question-and-answer software, the user's intention may be to obtain relevant information or answers in the currently previewed document, and thus the previewed document may be directly used as the target file.

[0114] A user opens a document in an app for preview.

[0115] When the user stays on a document preview page, the page is identified as the currently focused page.

[0116] The system can further analyze the content of the current page of interest and determine one or more key elements (such as titles, paragraphs, charts, etc.). These key elements represent information points that the user may be interested in or need to know.

[0117] In this scenario, since the user is clearly previewing a document and the application is able to identify the key elements that the user is currently focusing on, the document (or a specific part of the document) can be directly determined as the target file.

[0118] There is no need to conduct additional knowledge base searches or complex analysis processes. The application can directly find relevant information or generate answers in the target file based on the questions or needs raised by the user.

[0119] The method of this embodiment can meet the immediate needs of users more directly and accurately.

[0120] Figure 7 A flow chart of a method for obtaining second input information according to an embodiment of the present disclosure is schematically shown.

[0121] According to an embodiment of the present disclosure, the first input information includes at least one of a first text, an associated object, and a second text, and the associated object is determined based on the user's operation on the current application. Figure 7 As shown, for example, by performing at least one of operations S731 to S732, the first input information is supplemented based on at least one target file to obtain the second input information.

[0122] In operation S731, a first target file is added to the first text to obtain second input information, and the first target file is determined based on the first text.

[0123] In the case where the first input information is the first text, it is assumed that the user is using a document editing software and is preparing to ask a question about "the application of artificial intelligence in the medical field". The user first enters the first text in the input box: "The application of artificial intelligence in the medical field is becoming more and more extensive, especially in diagnosis..." At this time, the software detects that the user has entered keywords related to "artificial intelligence" and "medical field".

[0124] The algorithm in the background of the software searches for documents, research reports or papers related to "artificial intelligence" and "medical field" in the user's local file library or cloud storage based on the first text entered by the user. For example, a PDF document titled "The Latest Advances in Artificial Intelligence in Medical Diagnosis" was found.

[0125] The software automatically adds the content or summary of this PDF document (as the first target file) to the article the user is editing as supplementary information to the first text entered by the user, forming the second input information. In this way, the user's question is enriched with supplementary materials without the need for manual search and insertion.

[0126] In operation S732, a second target file is added to the second text to obtain second input information, and the second target file is determined based on the associated object.

[0127] In the case where the first input information is an associated object and a second text, suppose the user has just opened a question-and-answer software and is ready to ask a question about "how to learn programming efficiently". At this time, the user has not yet entered any text in the input box (that is, the second text is empty), but the software has identified the user's intention (that is, the associated object) through the user's historical behavior (such as the programming learning website recently browsed, the programming books purchased, etc.).

[0128] Based on the identified user intent, the software searches the user's resource library for tutorials, video links, or study notes related to "Efficient Learning Programming." For example, it finds a Word document titled "Efficient Programming Learning Guide."

[0129] When a user enters "What are some efficient ways to learn programming?", the software can automatically add the second target file (a summary or link of an efficient programming learning guide) found previously to the user's question as additional information and send it to the Q&A community or search engine. This not only improves the quality of the question, but also increases the possibility of obtaining useful answers.

[0130] This example demonstrates how to automatically add relevant target files to an input box based on two different input information. In the first case, when the user has already entered some text, the software can retrieve and add relevant files based on the text; in the second case, even if the user has just started to enter (or has not yet entered), the software can prepare and add relevant files in advance by predicting the user's intention. Both methods greatly improve the user's work efficiency and the accuracy of information acquisition.

[0131] Figure 8 The flowchart schematically shows a data processing method according to another embodiment of the present disclosure.

[0132] According to the embodiments of the present disclosure, Figure 8 As shown, the data processing method of this embodiment also includes operations S810 to S830.

[0133] In operation S810, multiple knowledge bases are searched based on the associated objects to obtain at least one file to be selected.

[0134] In operation S820, at least one file to be selected is displayed.

[0135] In operation S830, in response to a user's selection of at least one file to be selected, a second target file is determined.

[0136] In some embodiments, it is assumed that a user is using a question-and-answer platform in a professional field, which integrates multiple professional knowledge bases to help users quickly find and solve professional problems. The user has just opened the platform and is about to enter a question about "machine learning algorithm optimization", but has not yet entered any text in the input box.

[0137] Although the user has not yet entered any text, the platform has identified the user's current intention to learn about "machine learning algorithm optimization" by analyzing the user's historical behavior (such as the pages they have just visited, the keywords they have searched for, the discussions they have participated in, etc.). This intention is used as an associated object.

[0138] The platform uses this associated object to search multiple integrated professional knowledge bases (such as academic paper libraries, professional forum posts, expert blogs, etc.) to find content related to "machine learning algorithm optimization."

[0139] After the search is completed, the platform selects several candidate files that best match the user's needs from the search results. These files may include abstracts of academic papers, forum discussion highlights, expert blog articles, etc.

[0140] The platform displays these selected files to users in the form of lists or cards. Each entry contains key information such as the file's title, summary, source, etc., so that users can quickly understand the file content.

[0141] After browsing the list of files to be selected, the user selects one or more files of interest according to his or her needs.

[0142] In response to the user's selection, the platform determines the file selected by the user as the second target file. These files may contain specific algorithm optimization techniques, case analysis or expert advice that the user wants to know.

[0143] For example, in other application scenarios, you can also obtain the keywords of the currently received email, and when you activate the Q&A software, you will be prompted to find relevant files in the knowledge base and ask whether to add the relevant files to the input box. You can also obtain the document currently being previewed and when you activate the Q&A software, you will be prompted to ask whether to add the preview document to the input box. You can also obtain the text currently in the clipboard and when you activate the Q&A software, you will be prompted to ask whether to add the text to the input box.

[0144] It should be noted that the display of the text in the preview document and the clipboard does not require searching the knowledge base, but can directly use it as the target file or target content.

[0145] The method of this embodiment makes full use of the user's historical behavior and current intentions, as well as the professional knowledge base resources integrated by the platform, to provide users with a personalized and efficient information acquisition experience. Even if the user has not yet entered specific query text, the platform can predict the user's needs through intelligent analysis and actively recommend relevant knowledge resources. This not only improves the user's work efficiency, but also enhances the platform's user experience and competitiveness.

[0146] Fig. 9 A flow chart of a method for determining first input information according to an embodiment of the present disclosure is schematically shown.

[0147] According to the embodiments of the present disclosure, Fig. 9 As shown, for example, the first input information of the large model is determined through operations S911 to S913.

[0148] In operation S911, copy information of a user is acquired.

[0149] In operation S912, the copy information is displayed.

[0150] In operation S913, in response to a user's selection of copy information, first input information is determined.

[0151] In some embodiments, it is assumed that a user is using an intelligent assistant application that integrates a large language model (LLM), which is designed to help the user quickly process text information, such as composing emails, editing documents, or generating creative content. The user copies a piece of text in an editing interface and hopes to use the LLM to analyze, expand, or rewrite the text.

[0152] A user selects and copies a piece of text in an editing interface (such as a text editor, email client, or web page). This text can be any type of content, such as a news report, an abstract of an academic paper, or a personal thought.

[0153] The smart assistant application automatically captures the text information copied by the user through the system clipboard monitoring function.

[0154] Once the copied information is captured, the smart assistant application will display the text on its interface. The display form can be the original text directly displayed or displayed in an editable text box so that the user can modify or supplement it at any time.

[0155] In addition, the application can also provide some additional information, such as the source of the copied information (if available), word count, keyword extraction, etc., to help users better understand the text.

[0156] After viewing the displayed copy information, the user may directly accept the text as input for the LLM, or may make some modifications or additions.

[0157] The intelligent assistant application responds to the user's selection or modification operation and transmits the finalized text content as the first input information to the LLM. This first input information may be completely based on the original text copied by the user, or may include the user's modifications and supplements.

[0158] This embodiment obtains the user's copied information, displays it in the smart assistant application, and possibly modifies it, and finally determines the first input information of the large language model. This method makes full use of the user's daily operation habits (copying text) and combines it with the powerful functions of the smart assistant application to provide users with a convenient and efficient text processing solution. Users can quickly use LLM to perform text analysis, expansion or rewriting without manually entering or pasting text, which greatly improves work efficiency.

[0159] Fig.10 A flow chart of a method for determining a large model output according to an embodiment of the present disclosure is schematically shown.

[0160] According to the embodiments of the present disclosure, Fig.10 As shown, for example, through operations S1041-S1042, the output of the large model is determined based on the second input information.

[0161] In operation S1041, information extraction is performed on the second input information to obtain index information.

[0162] In operation S1042, at least one target file is retrieved based on the index information, and the retrieval results are integrated to obtain an output of a large model.

[0163] In some embodiments, when determining the large model output, in-depth information extraction is performed on the second input information, and the previously determined target files are directly retrieved and integrated based on the extracted index information to generate a more accurate and comprehensive answer.

[0164] For example, the user enters a question through the interface, such as "What were the key technical challenges in last week's project?" This question serves as the first input information for the big model.

[0165] The system analyzes the first input information and identifies the keywords “last week’s project” and “key technical challenges”.

[0166] Search for documents related to these keywords in the pre-established knowledge base, such as last week's project report, technical discussion records, etc., and identify them as target documents.

[0167] The system reads the contents of the target file and extracts information related to “the key technical challenges in last week’s project.”

[0168] This information or the entire target file is supplemented to the first input information to form the second input information, which may include the problem plus a description of the specific technical challenge extracted from the target file.

[0169] The system performs further information extraction on the second input information and identifies more specific keywords or phrases, such as “key technology,” “challenge description,” “solution,” etc., which serve as index information.

[0170] The previously determined target files are retrieved using the index information to find parts that are more closely related to the index information. For example, the target files may need to be divided or extracted more finely to ensure that the retrieved content is highly matched to the user's query intent.

[0171] The system integrates the search results and brings together relevant information. For example, the integrated information may need further editing or formatting to ensure the clarity and readability of the output.

[0172] Ultimately, the system generates a large model output, which might be a detailed answer containing a description of the key technical challenges in last week’s project, possible solutions, and relevant data or graphs.

[0173] The method of this embodiment not only improves the accuracy and comprehensiveness of the question and answer, but also enhances the system's ability to understand the user's query intent. Through a more sophisticated information extraction and retrieval process, the system can generate more specific and targeted answers, thereby better meeting the needs of users. In addition, this method narrows the search scope and improves the efficiency of question and answer by retrieving pre-screened target files.

[0174] Based on the above method, the present disclosure also provides a data processing device. Fig.11 The data processing device is described in detail.

[0175] Fig.11 The structure block diagram of a data processing device according to an embodiment of the present disclosure is schematically shown.

[0176] like Fig.11 As shown, the data processing device 1100 of this embodiment includes, for example: a first determination module 1110, a second determination module 1120, a supplementation module 1130 and a third determination module 1140. The data processing device 1100 can execute the above reference Figure 2~Figure 10 Describe the method to achieve efficient response to the questions asked.

[0177] Specifically, the first determination module 1110 is used to determine the first input information of the large model. In one embodiment, the first determination module 1110 can be used to perform the operation S210 described above, which will not be described in detail here.

[0178] The second determination module 1120 is used to determine at least one target file based on the first input information, where the target file represents the user-referenced object in the first input information. In one embodiment, the second determination module 1120 can be used to perform the operation S220 described above, which will not be described in detail here.

[0179] The supplementing module 1130 is used to supplement the first input information based on at least one target file to obtain the second input information. In one embodiment, the supplementing module 1130 can be used to perform the operation S230 described above, which will not be described in detail here.

[0180] The third determination module 1140 is used to determine the output of the large model based on the second input information. In one embodiment, the third determination module 1140 can be used to perform the operation S240 described above, which will not be described in detail here.

[0181] It is understandable that the first determination module 1110, the second determination module 1120, the supplementary module 1130 and the third determination module 1140 can be implemented in one module, or any one of the modules can be split into multiple modules. Alternatively, at least part of the functions of one or more of these modules can be combined with at least part of the functions of other modules and implemented in one module. According to an embodiment of the present disclosure, at least one of the first determination module 1110, the second determination module 1120, the supplementary module 1130 and the third determination module 1140 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application specific integrated circuit (ASIC), or can be implemented in hardware or firmware in any other reasonable way of integrating or packaging the circuit, or in a suitable combination of software, hardware and firmware. Alternatively, at least one of the first determination module 1110, the second determination module 1120, the supplementation module 1130 and the third determination module 1140 may be at least partially implemented as a computer program module, and when the program is executed by a computer, the function of the corresponding module may be executed.

[0182] Fig.12 A block diagram of an electronic device suitable for implementing a data processing method according to an embodiment of the present disclosure is schematically shown.

[0183] like Fig.12As shown, the electronic device 1200 according to an embodiment of the present disclosure includes a processor 1201, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1202 or a program loaded from a storage part 1208 to a random access memory (RAM) 1203. The processor 1201 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or a related chipset and / or a dedicated microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 1201 may also include an onboard memory for caching purposes. The processor 1201 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present disclosure.

[0184] In RAM 1203, various programs and data required for the operation of electronic device 1200 are stored. Processor 1201, ROM 1202 and RAM 1203 are connected to each other via bus 1204. Processor 1201 performs various operations of the method flow according to the embodiment of the present disclosure by executing the program in ROM 1202 and / or RAM 1203. It should be noted that the program can also be stored in one or more memories other than ROM 1202 and RAM 1203. Processor 1201 can also perform various operations of the method flow according to the embodiment of the present disclosure by executing the program stored in the one or more memories.

[0185] According to an embodiment of the present disclosure, the electronic device 1200 may further include an input / output (I / O) interface 1205, which is also connected to the bus 1204. The electronic device 1200 may further include one or more of the following components connected to the I / O interface 1205: an input portion 1206 including a keyboard, a mouse, etc.; an output portion 1207 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage portion 1208 including a hard disk, etc.; and a communication portion 1209 including a network interface card such as a LAN card, a modem, etc. The communication portion 1209 performs communication processing via a network such as the Internet. A drive 1210 is also connected to the I / O interface 1205 as needed. A removable medium 1211, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 1210 as needed, so that a computer program read therefrom is installed into the storage portion 1208 as needed.

[0186] The input part 1206 is used to input the first input information and the second input information to the large model. The storage part 1208 is used to store at least one target file. The one or more processors 1201 are used to determine at least one target file based on the first input information, where the target file represents the user-referenced object in the first input information. Based on the at least one target file, the input content of the large model is updated to obtain the second input information. Based on the second input information, the output of the large model is determined.

[0187] The present disclosure also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or may exist independently without being assembled into the device / apparatus / system. The above computer-readable storage medium carries one or more programs, and when the above one or more programs are executed, the data processing method according to the embodiment of the present disclosure is implemented.

[0188] According to an embodiment of the present disclosure, a computer-readable storage medium may be a non-volatile computer-readable storage medium, for example, may include but is not limited to: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, an apparatus, or a device. For example, according to an embodiment of the present disclosure, a computer-readable storage medium may include the ROM 1202 and / or RAM 1203 described above and / or one or more memories other than ROM 1202 and RAM 1203.

[0189] The embodiment of the present disclosure also includes a computer program product, which includes a computer program, and the computer program contains program code for executing the method shown in the flowchart. When the computer program product is run in a computer system, the program code is used to enable the computer system to implement the data processing method provided by the embodiment of the present disclosure.

[0190] The above functions defined in the system / device of the embodiment of the present disclosure are performed when the computer program is executed by the processor 1201. According to the embodiment of the present disclosure, the system, device, module, unit, etc. described above can be implemented by a computer program module.

[0191] In one embodiment, the computer program may rely on tangible storage media such as optical storage devices, magnetic storage devices, etc. In another embodiment, the computer program may also be transmitted and distributed in the form of signals on a network medium, and downloaded and installed through the communication part 1209, and / or installed from the removable medium 1211. The program code contained in the computer program may be transmitted using any appropriate network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0192] In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 1209, and / or installed from the removable medium 1211. When the computer program is executed by the processor 1201, the above functions defined in the system of the embodiment of the present disclosure are performed. According to the embodiment of the present disclosure, the system, device, means, module, unit, etc. described above can be implemented by a computer program module.

[0193] According to an embodiment of the present disclosure, the program code for executing the computer program provided by the embodiment of the present disclosure can be written in any combination of one or more programming languages. Specifically, these computing programs can be implemented using high-level process and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, Java, C++, python, "C" language or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, partially on the remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (for example, using an Internet service provider to connect through the Internet).

[0194] The flow charts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0195] It will be appreciated by those skilled in the art that the features described in the various embodiments and / or claims of the present disclosure may be combined and / or combined in a variety of ways, even if such combinations and / or combinations are not explicitly described in the present disclosure. In particular, the features described in the various embodiments and / or claims of the present disclosure may be combined and / or combined in a variety of ways without departing from the spirit and teachings of the present disclosure. All of these combinations and / or combinations fall within the scope of the present disclosure.

[0196] The embodiments of the present disclosure are described above. However, these embodiments are only for illustrative purposes and are not intended to limit the scope of the present disclosure. Although the embodiments are described above separately, this does not mean that the measures in the various embodiments cannot be used in combination to advantage. The scope of the present disclosure is defined by the attached claims and their equivalents. Without departing from the scope of the present disclosure, those skilled in the art may make a variety of substitutions and modifications, which should all fall within the scope of the present disclosure.

Claims

1. A data processing method, comprising: Determine the first input information of the large model; Based on the first input information, determining at least one target file, the target file representing the user-referenced object in the first input information; Based on the at least one target file, supplement the first input information to obtain second input information; Based on the second input information, an output of the large model is determined.

2. The method according to claim 1, wherein: The determining at least one target file based on the first input information includes at least one of the following: In response to a user inputting a first text into the macro model, searching a plurality of knowledge bases to obtain the at least one target file, the plurality of knowledge bases being associated with the macro model; In response to a user operation on a current application, determining a first associated object of a page displayed by the current application; The multiple knowledge bases are searched based on the first associated object to obtain the at least one target file.

3. The method according to claim 2, wherein: In response to the user inputting the first text into the large model, searching multiple knowledge bases to obtain the at least one target file includes at least one of the following: In response to a user inputting instruction information into the large model, searching the multiple knowledge bases to obtain the at least one target file, the instruction information representing the user's action instruction for the large model; In response to a user inputting reference information into the large model, the multiple knowledge bases are searched to obtain the at least one target file, wherein the reference information represents a target reference object stored in the multiple knowledge bases.

4. The method according to claim 2, wherein: The determining, in response to the user's operation on the current application, the first associated object of the page displayed by the current application includes at least one of the following: Extracting information from the display page of the current application to obtain the first associated object; Based on the association between the current application and the large model, the first association object is obtained, wherein the current application and the large model are respectively associated with the same user account.

5. The method according to claim 1, wherein: The determining at least one target file based on the first input information further includes: In response to a user operation on a current application, determining a second associated object of a page displayed by the current application; The second associated object is determined to be the at least one target file.

6. The method according to claim 1, wherein: The first input information includes at least one of a first text, an associated object, and a second text, wherein the associated object is determined based on a user's operation on the current application; The step of supplementing the first input information based on the at least one target file to obtain the second input information includes at least one of the following: adding a first target file to the first text to obtain the second input information, wherein the first target file is determined based on the first text; A second target file is added to the second text to obtain the second input information, wherein the second target file is determined based on the associated object.

7. The method according to claim 6, wherein: The method further comprises: Searching multiple knowledge bases based on the associated object to obtain at least one file to be selected; Displaying the at least one file to be selected; In response to the user's selection of the at least one file to be selected, the second target file is determined.

8. The method according to claim 1, wherein: The first input information of determining the large model includes: Get the user's copy information; Displaying the copied information; In response to a user's selection of the copy information, the first input information is determined.

9. The method according to claim 1, wherein: The step of determining the output of the large model based on the second input information includes: performing information extraction on the second input information to obtain index information; Based on the index information, the at least one target file is searched, and the search results are integrated to obtain the output of the large model.

10. An electronic device comprising: An input part, used for inputting first input information and second input information into the large model; A storage part, used for storing at least one target file; One or more processors, configured to determine, based on the first input information, the at least one target file, the target file representing the user-referenced object in the first input information; Based on the at least one target file, updating the input content of the large model to obtain the second input information; Based on the second input information, an output of the large model is determined.

Images