Information retrieval method and device, electronic equipment and storage medium

Abstract

The invention provides an information retrieval method and device, electronic equipment and a storage medium, and relates to the technical field of data processing. The method comprises the steps of determining a target business scene corresponding to retrieval information in response to the obtained retrieval information of a user, and ensuring that a target knowledge fragment obtained according to the retrieval information belongs to the corresponding business scene; on the basis, matching a target knowledge fragment corresponding to the retrieval information in a target knowledge set according to semantic features corresponding to files in the target knowledge set in the target business scene; and according to the target knowledge fragment and the retrieval information, generating reply information corresponding to the retrieval information, and because the target knowledge set pre-stores the file corresponding to the business scene and the semantic feature of the file, the reply information can be replied according to the reply information. Therefore, the target knowledge fragments corresponding to the retrieval information can be further screened from the files of the target knowledge set according to the semantic features of the files, and the accuracy of the retrieval result is improved to a certain extent.

Description

Technical Field

[0001] This disclosure belongs to the field of data processing technology, and specifically relates to an information retrieval method, apparatus, electronic device and storage medium. Background Technology

[0002] Information retrieval involves searching a pre-defined database for results corresponding to the searched information. This database stores a large number of unprocessed documents, images, logs, and other files. Typically, after obtaining the search information, the database is used to match files with the searched information based on their attribute information, thus yielding the relevant search results.

[0003] Since the retrieval in the relevant technology can only be performed based on the attribute information of each file in the preset database, it cannot further identify the content of the file. Therefore, if there is a mismatch between the file content and the attribute information, the retrieval results will be inaccurate. Summary of the Invention

[0004] This disclosure presents an information retrieval method, apparatus, electronic device, and storage medium.

[0005] A first aspect of this disclosure provides an information retrieval method, the method comprising: In response to obtaining the search information, the target business scenario corresponding to the search information is determined; Based on the semantic features of the files in the target knowledge set corresponding to the target business scenario, target knowledge fragments corresponding to the search information are matched in the target knowledge set; wherein, the target knowledge set pre-stores files corresponding to the target business scenario and the semantic features of the files; Based on the target knowledge fragment and the search information, generate response information corresponding to the search information.

[0006] In this embodiment of the disclosure, determining the target business scenario corresponding to the retrieved information includes: Based on the information type of the retrieved information, a target preprocessing method corresponding to the retrieved information is determined; the information type represents the signal form of the retrieved information. Based on the target preprocessing method, key information in the retrieved information is identified; wherein, the key information represents the business type involved in the retrieved information; Based on the key information, the target business scenario corresponding to the search information is determined.

[0007] In this embodiment of the disclosure, the key information includes at least one keyword; determining the target business scenario corresponding to the search information based on the key information includes: For any given keyword, determine the initial business scenario corresponding to that keyword; Given that the initial business scenario corresponding to each keyword is determined, the target business scenario is determined based on the number of initial business scenarios corresponding to each keyword and the type of the initial business scenario.

[0008] In this embodiment of the disclosure, matching the target knowledge fragment corresponding to the retrieval information in the target knowledge set includes: Based on the mapping relationship between the target business scenario and the knowledge set, determine the target knowledge set corresponding to the target business scenario; Based on the key information in the retrieval information, target semantic features that match the key information are retrieved from the target knowledge set; The information corresponding to the target semantic features is taken as the target knowledge fragment.

[0009] In this embodiment of the disclosure, retrieving target semantic features matching the key information in the target knowledge set based on the key information in the retrieval information includes: Determine the similarity between the semantic features of the key information and the semantic features corresponding to each file in the target knowledge set, and obtain the similarity of the semantic features corresponding to each file; The semantic features whose similarity meets the preset similarity conditions are used as the target semantic features.

[0010] In this embodiment of the disclosure, generating response information corresponding to the search information based on the target knowledge fragment further includes: By combining the target knowledge fragment and the search information, a prompt message is generated; The preset large language model is invoked to generate response information corresponding to the search information based on the prompt information.

[0011] In this embodiment of the disclosure, the step of splicing the target knowledge fragment and the search information to generate a prompt message includes: If the target knowledge fragment includes multiple fragments, the target knowledge fragments are sequentially combined according to the order of the keywords in the key information in the search information to generate a knowledge fragment sequence. The knowledge fragment sequence and the retrieval information are combined to generate a prompt message.

[0012] In this embodiment of the disclosure, the method further includes: In response to obtaining an updated file, the system identifies the business scenario to which the updated file belongs; and invokes a preset model to convert the text of the updated file into a vector to obtain the semantic features of the updated file. The updated file and its semantic features are stored in the knowledge set corresponding to the business scenario to obtain the updated knowledge set.

[0013] A second aspect of this disclosure provides an information retrieval device, the device comprising: The determination module is used to determine the target business scenario corresponding to the user's search information in response to obtaining the user's search information; The retrieval module is used to match target knowledge fragments corresponding to the retrieval information in the target knowledge set based on the semantic features of files in the target knowledge set corresponding to the target business scenario; wherein, the target knowledge set pre-stores files corresponding to the target business scenario and the semantic features of the files; The generation module is used to generate response information corresponding to the search information based on the target knowledge fragment and the search information.

[0014] An embodiment of the third aspect of this disclosure provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method described in the first aspect or any optional embodiment of the first aspect.

[0015] An embodiment of the fourth aspect of this disclosure provides a computer-readable storage medium having a computer program stored thereon, the program being executed by a processor to implement the method described in the first aspect or any optional embodiment of the first aspect.

[0016] The technical solutions provided in this disclosure have at least the following technical effects or advantages: In response to obtaining user search information, this embodiment determines the target business scenario corresponding to the search information, ensuring that the target knowledge fragment obtained based on the search information belongs to the corresponding business scenario. Based on this, according to the semantic features of files in the target knowledge set corresponding to the target business scenario, the target knowledge fragment corresponding to the search information is matched in the target knowledge set. Based on the target knowledge fragment and the search information, response information corresponding to the search information is generated. Since the target knowledge set pre-stores files corresponding to the business scenario and the semantic features of those files, the target knowledge fragment corresponding to the search information can be further filtered from each file in the target knowledge set based on the semantic features of the files, thereby improving the accuracy of the search results to a certain extent.

[0017] Additional aspects and advantages of this disclosure will be set forth in part in the description which follows, and in part will be obvious from the description or may be learned by practice of this disclosure. Attached Figure Description

[0018] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this disclosure. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A flowchart of an information retrieval method provided in an embodiment of this disclosure is shown; Figure 2 A flowchart of an information retrieval method provided in an embodiment of this disclosure is shown; Figure 3 A flowchart of an information retrieval method provided in an embodiment of this disclosure is shown; Figure 4 A schematic diagram of the structure of an information retrieval device provided in another embodiment of this disclosure is shown; Figure 5 A schematic diagram of the structure of an electronic device provided in an embodiment of the present disclosure is shown; Figure 6 A schematic diagram of a storage medium provided according to an embodiment of the present disclosure is shown. Detailed Implementation

[0019] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0020] It should be noted that, unless otherwise stated, the technical or scientific terms used in this disclosure shall have the ordinary meaning as understood by one of ordinary skill in the art to which this disclosure pertains.

[0021] The following describes the implementation scenarios and related technologies involved in the embodiments of this disclosure.

[0022] Autonomous driving refers to the technology that enables vehicles to navigate and drive automatically without the active operation of a human driver. The research and development process for related functions in the field of autonomous driving generates a large number of documents, including requirement documents, design documents, software code, algorithm documents, and test documents. Related personnel need to perform coding and development operations based on the relevant materials. In related technologies, these documents from the research and development process are directly stored in a pre-defined database. When managing or retrieving these files, searches can only be performed based on file attribute information, such as filename, file type, and modification time. If the actual content of a file does not correspond to the attribute information, or if the attribute information cannot fully summarize / cover the actual content, inaccurate files may be retrieved. This necessitates manual retrieval, which is inefficient and makes it difficult to accurately locate the required content.

[0023] In view of the above, this disclosure provides an information retrieval method, apparatus, electronic device, and storage medium. The technical solutions of this disclosure are described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this disclosure will now be described with reference to the accompanying drawings.

[0024] like Figure 1 As shown in the embodiments of this disclosure, an information retrieval method is provided, the method comprising: In step S11, in response to obtaining the search information, the target business scenario corresponding to the search information is determined.

[0025] For example, the information retrieval method of this disclosure can be applied to information retrieval in the research and development field. For instance, researchers can retrieve information related to research and development processes and tasks, and can also retrieve corresponding solutions when difficulties arise during the research and development process. In this disclosure, the information retrieval method can also be applied to an intelligent question-and-answer system based on a preset database.

[0026] In this embodiment of the disclosure, information retrieval can be based on inquiries entered by relevant personnel or file information to be retrieved. In the application scenarios of this embodiment, the R&D process involves files corresponding to different business scenarios, such as programming-related issues, autonomous driving simulation-related issues, and testing-related issues. Therefore, after obtaining the retrieval information, the target business scenario can be determined based on the retrieval information, so as to retrieve information from the files corresponding to that target business scenario and improve the accuracy of the retrieval.

[0027] In some embodiments, determining the target business scenario corresponding to the retrieved information can also be achieved in the following ways: determining the target preprocessing method corresponding to the retrieved information based on the information type of the retrieved information; wherein, the information type represents the signal form of the retrieved information; identifying key information in the retrieved information based on the target preprocessing method; wherein, the key information represents the business type involved in the retrieved information; and determining the target business scenario corresponding to the retrieved information based on the key information.

[0028] For example, the retrieved information can be text, image, or audio, and the methods for identifying semantic features in the retrieved information differ depending on the type of retrieved information. Preprocessing can be a method for identifying semantic features in the retrieved information. For instance, if the retrieved information is text, semantic recognition models can be used directly to identify the semantic features to obtain the key information; if the retrieved information is an image, the content of the image needs to be identified to obtain the key information. The information type of the retrieved information can be determined based on its file extension. For example, if the retrieved information is an image, the corresponding file extension could be bmp, jpg, png, or tif; if the retrieved information is text, the corresponding file extension could be dox or docx.

[0029] A mapping relationship between preprocessing methods and information types can be pre-built. For example, a mapping table of information types and preprocessing methods can be constructed. For text, preprocessing methods could include word segmentation, stop word removal, and stemming; for speech, preprocessing methods could include noise reduction, automatic speech-to-text conversion, and text cleaning; and for images, preprocessing methods could include optical character recognition (OCR) for text recognition and object detection (such as recognizing codes in images). Selecting the appropriate preprocessing flow based on the signal form ensures the accuracy of subsequent key information extraction.

[0030] For text, Natural Language Processing (NLP) techniques can be used to extract keywords and parse grammatical structures. For speech-to-text data, speech quality assessment (such as signal-to-noise ratio) can be performed first, followed by the application of text feature extraction methods. For image data, if optical character recognition extracts text, it is processed as text; if it is a non-text image (such as a picture), features are extracted using a Convolutional Neural Network (CNN) classification model.

[0031] The extracted features are matched with predefined dictionaries corresponding to specific business scenarios. For example, the keyword "turn on lane keeping assist" corresponds to the assisted driving business scenario; the image identification of "simulated rainstorm" corresponds to the extreme weather testing business scenario. This enables searching within files corresponding to specific business scenarios based on retrieval information, improving the accuracy of search results.

[0032] Taking the retrieved information as text as an example, the key information includes keywords corresponding to the text, and these keywords correspond to business types. The target business scenario can then be determined based on these keywords. In some embodiments, where the key information includes at least one keyword, an initial business scenario corresponding to any keyword is determined. If the initial business scenarios corresponding to each keyword are determined, the target business scenario is determined based on the number and type of the initial business scenarios corresponding to each keyword.

[0033] For example, if a single keyword is identified in the search results, the business scenario corresponding to that keyword can be directly used as the target business scenario. If multiple keywords are identified in the search results, one or more initial business scenarios can be determined for each keyword. The target business scenario is then determined based on the type and number of these initial scenarios. For example, the initial scenarios may belong to different types, such as driving strategy, hardware control, and safety warning, requiring analysis of their priority and compatibility. The initial scenarios for all keywords in a single query are deduplicated and merged to obtain a "Scenario - Occurrence Count" statistics table. The target business scenario is then determined based on the occurrence count of each business scenario.

[0034] In some embodiments, when multiple keywords are identified in the search information, one or more initial business scenarios can be determined based on each keyword. After determining the initial business scenarios corresponding to all keywords, the target business scenario can be determined based on the number of keywords corresponding to each initial business scenario. If multiple initial business scenarios have the same number of keywords, one initial business scenario can be randomly selected as the target business scenario from among the multiple initial business scenarios with the same number of keywords. If there is a correlation among the multiple initial business scenarios with the same number of keywords, all of the multiple initial business scenarios with the same number of keywords can be used as the target business scenario.

[0035] In step S12, based on the semantic features of the files in the target knowledge set corresponding to the target business scenario, the target knowledge fragments corresponding to the retrieved information are matched in the target knowledge set; wherein, the target knowledge set pre-stores the files corresponding to the target business scenario and the semantic features of the files.

[0036] For example, after classifying relevant documents in R&D testing according to business scenarios, a dynamic business scenario map can be constructed, and graph neural networks can be used to infer the implicit relationships between scenarios in real time, adapting to the rapid iteration characteristics of scenarios in autonomous driving R&D.

[0037] In the process of graph construction, nodes can be business scenarios, knowledge fragments, and technical modules, etc.; edges can be the correlation strength between scenarios and knowledge fragments, or the similarity between business scenarios. The correlation strength can be determined based on historical query frequency, or based on the similarity and commonalities between business scenarios, such as the commonalities between overtaking in rainy weather and overtaking in foggy weather. When a new business scenario (such as overtaking in snowy weather) appears, its similarity with existing scenarios is analyzed, the associated knowledge fragments are migrated to the new scenario nodes, and the edge weights are adjusted.

[0038] After constructing the business graph, query conditions can be expanded based on the relationships between various business scenarios within the graph. For example, scenarios similar to overtaking in rainy weather can be searched within the graph to supplement candidate files, such as overtaking in foggy weather. Finally, relevant target knowledge fragments are identified from these candidate files.

[0039] In some embodiments, matching the target knowledge fragments corresponding to the retrieved information in the target knowledge set can be achieved in the following ways: determining the target knowledge set corresponding to the target business scenario based on the mapping relationship between the target business scenario and the knowledge set; retrieving target semantic features that match the key information in the retrieved information from the target knowledge set; and using the information corresponding to the target semantic features as the target knowledge fragments.

[0040] The process involves retrieving target semantic features that match the key information from the search information, based on the key information in the target knowledge set. This includes: determining the similarity between the semantic features of the key information and the semantic features corresponding to each file in the target knowledge set, and obtaining the similarity between the semantic features corresponding to each file; and using the semantic features whose similarity meets the preset similarity conditions as target semantic features.

[0041] For example, after classifying relevant documents in R&D testing according to business scenarios, embedding models such as BERT can be used to transform the document content in the knowledge base corresponding to the business scenario into high-dimensional vectors, while extracting explicit semantic tags (such as keywords, entities, and scenario types). For example, the semantic features of the document "Rainy Day Sensor Cleaning Specification" are represented by a vector of [0.12, -0.05, ..., 0.34], and the explicit tags are {Scenario: Rainy Day, Module: Sensor, Operation: Cleaning}.

[0042] The process transforms user-input natural language questions into matchable semantic features and associates them with target business scenarios. The retrieved information undergoes word segmentation, part-of-speech tagging, and named entity recognition to extract key information. For example, a user question like "What should I do if the sensor malfunctions when overtaking in the rain?" yields keywords such as "rainy day," "overtaking," and "sensor malfunction." The retrieved information contains entities like "scenario," "operation," and "component," where the scenario is "rainy day," the operation is "overtaking," and the component is "sensor." Based on the keywords and entities, a predefined business scenario library is matched. The matched scenario is "overtaking in the rain" (associated with the knowledge set: rainy day driving strategy, sensor fault handling). The most semantically relevant knowledge fragment to the retrieved information is retrieved from the target knowledge set and used as the target knowledge fragment. "Most relevant" can be the knowledge fragment with the highest matching degree or that meets corresponding conditions.

[0043] For example, the system retrieves the knowledge fragments most semantically relevant to the target knowledge set. It calculates the cosine similarity between the semantic vector of the user's question and the vectors of the files in the target knowledge set, recalling the knowledge fragments with the highest similarity.

[0044] For example, the vector of the retrieved information is [0.15, -0.03, ..., 0.30]. The knowledge fragment vector library in the knowledge base includes fragment 1 and fragment 2. Fragment 1 ("Rainy Day Sensor Cleaning Specification") has a vector of [0.12, -0.05, ..., 0.34], with a similarity of 0.92 to the retrieved information vector. Fragment 2 ("Overtaking Trajectory Planning Guide") has a vector of [0.08, 0.10, ..., 0.25], with a similarity of 0.75 to the retrieved information vector. Since fragment 1 has a higher similarity, the target knowledge fragment can be fragment 1.

[0045] In step S13, response information corresponding to the search information is generated based on the target knowledge fragment and the search information.

[0046] For example, after determining the target knowledge fragment, the corresponding response information can be determined from the target knowledge fragment. In some embodiments, if the target knowledge fragment has completely covered the search information, a preset standardized answer is returned directly, or relevant information is extracted from multiple knowledge fragments and spliced ​​or reorganized into a coherent response, such as merging paragraphs from multiple documents; if the knowledge fragment is incomplete, supplementary information is obtained by reasoning based on context or external rules; if the information is insufficient, follow-up questions can be generated to clarify the requirements.

[0047] Of course, when generating response information based on target knowledge fragments and search information, the content in the target knowledge fragments can be further verified. When the same content in different target knowledge fragments points to different content, the content in the file with the most recent modification time can be selected as the main content.

[0048] In some embodiments, generation can also be achieved by: splicing target knowledge fragments and retrieval information to generate prompt information; and calling a preset large language model to generate response information corresponding to the retrieval information based on the prompt information.

[0049] For example, the retrieved information and knowledge fragments are logically concatenated according to the R&D scenario to form the response information corresponding to the prompt information generated by the guidance model. The prompt information can be used to standardize the output results of the preset large language model. For instance, in some embodiments, concatenating target knowledge fragments and retrieved information to generate prompt information includes: if there are multiple target knowledge fragments, sequentially combining the target knowledge fragments according to the order of keywords in the retrieved information to generate a knowledge fragment sequence; and combining the knowledge fragment sequence with the retrieved information to generate prompt information. For example, in the preprocessing stage, the keywords in the retrieved information are numbered sequentially. The order of the keywords can be determined by their position in the retrieved information or by their position in the corresponding semantic features of the retrieved information. Each target knowledge fragment has recorded the keywords it can cover. Only the target knowledge fragments covering the keywords need to be sequentially concatenated according to their order to obtain the knowledge fragment sequence; the knowledge fragment sequence and the retrieved information are then concatenated to obtain the corresponding prompt information.

[0050] Prompt information can also be generated in the following ways: a three-element prompt template can be pre-built, which can include {instruction area, context area, and question area}. The instruction area limits the model: Please answer based on the following information. If the information is insufficient, please explain. The context area concatenates the target knowledge fragments into a paragraph of text using line breaks, or it can perform deduplication and other cleaning operations before concatenating the target knowledge fragments. The question area can directly put the user's original search information (question).

[0051] To avoid excessively long prompts, the total number of characters in the prompts can be calculated. If it exceeds the preset single-use limit of the large language model, the question area and instruction area are left unchanged, and the corresponding target knowledge fragments are cyclically deleted from the end of the context area until the remaining length meets the threshold. Simultaneously, a keyword blacklist can be enabled, replacing sensitive words with "*" to ensure input security. When deleting target knowledge fragments, deletion can be based on the similarity between them. One or more highly similar target knowledge fragments can be deleted, while only one can be retained.

[0052] The prompt message is sent to the deployed large language model. The call parameters are fixed at temperature=0.3 and top-p=0.85 to ensure factual consistency and low randomness in the output results; only the first generated result is returned to avoid confusion from multiple candidates.

[0053] The original text returned by the model is first "de-instructionized"—automatically removing clichés such as "based on the above information"; then it is aligned with the knowledge fragments and the mentioned facts are marked with "[knowledge {i}]" links; finally, the purified text is pushed to the user as "response information corresponding to the search information", completing the whole process.

[0054] like Figure 2 The diagram illustrates the process for generating response information in the above embodiment. After obtaining the user's query information, the query information is preprocessed, including analysis and entity recognition. A vector (semantic feature) of the preprocessed query information is generated. Based on this vector, similarity matching is performed in the knowledge base of the corresponding business scenario to obtain the target knowledge fragment. Based on the query information and the target knowledge fragment, a preset large language model is invoked to generate the corresponding response information, which is then returned to the user.

[0055] Based on the above embodiments, in order to ensure the real-time nature of the obtained response information, the knowledge set corresponding to each business scenario can be updated. Therefore, this embodiment of the disclosure further includes: in response to obtaining the updated file, identifying the business scenario to which the updated file belongs; and calling a preset model to convert the text of the updated file into a vector to obtain the semantic features of the updated file; and storing the updated file and the semantic features of the updated file into the knowledge set corresponding to the business scenario to obtain the updated knowledge set.

[0056] For example, the updated file can be an updated version of a file already stored in the knowledge base corresponding to the business scenario, a file of a certain type not stored in the knowledge base, or a file corresponding to a new business scenario. After obtaining the updated file, its semantic features can be obtained by identifying its file type. The file type of the updated file can be a document or an image, etc. Different semantic recognition methods are used for different file types. For details, please refer to the method of obtaining semantic features based on the preprocessing method of the retrieved information mentioned above, which will not be elaborated here.

[0057] like Figure 3 As shown, the updated file can be a new document or group chat record during the R&D process. The updated file is integrated with the corresponding files in the knowledge sets of the original business scenario. The integration process can involve preprocessing the updated file, such as cleaning, word segmentation, and entity processing. After preprocessing, the updated file is stored in the knowledge set of the corresponding business scenario. During storage, the preprocessed updated file is converted into a vector to obtain the corresponding semantic features. This semantic feature can then be used for matching after obtaining retrieval information. The semantic feature is then stored, and the file update is completed.

[0058] After determining the semantic features of the update file, if a historical file corresponding to the business scenario is already stored in the existing knowledge base, the update is performed based on the semantic features, or the previous historical file is deleted before the update file is stored, to avoid file redundancy and reduce space consumption. If the update file does not have a corresponding business scenario or a corresponding historical file in the knowledge set, the corresponding business scenario and knowledge set are created directly before the update file is stored.

[0059] The information retrieval method corresponding to the embodiments of this disclosure is more suitable for scenarios with high requirements for data security and confidentiality. By retrieving relevant content from internal documents, the information retrieval method corresponding to the embodiments of this disclosure not only improves the efficiency and real-time performance of information retrieval, but also ensures the security of internal data to a certain extent.

[0060] In summary, this embodiment of the present disclosure, in response to obtaining the user's search information, determines the target business scenario corresponding to the search information, ensuring that the target knowledge fragment obtained based on the search information belongs to the corresponding business scenario. Based on this, according to the semantic features of the files in the target knowledge set corresponding to the target business scenario, it matches the target knowledge fragment corresponding to the search information in the target knowledge set. Based on the target knowledge fragment and the search information, it generates response information corresponding to the search information. Since the target knowledge set pre-stores files corresponding to the business scenario and their semantic features, it can further filter the target knowledge fragments corresponding to the search information from each file in the target knowledge set based on the file's semantic features, thereby improving the accuracy of the search results to a certain extent.

[0061] correspond Figure 1 The illustrated information retrieval method, in this disclosure embodiment also provides an information retrieval device, such as... Figure 6 As shown, the device includes: The determination module 401 is used to determine the target business scenario corresponding to the search information in response to obtaining the user's search information; The retrieval module 402 is used to match the target knowledge fragment corresponding to the retrieval information in the target knowledge set according to the semantic features of the files in the target knowledge set corresponding to the target business scenario; wherein, the target knowledge set pre-stores the files corresponding to the target business scenario and the semantic features of the files; The generation module 403 is used to generate response information corresponding to the search information based on the target knowledge fragment and the search information.

[0062] In an optional implementation, the determining module is further configured to: determine a target preprocessing method corresponding to the search information based on the information type of the search information; the information type characterizes the signal form of the search information; identify key information in the search information based on the target preprocessing method; wherein the key information characterizes the business type involved in the search information; and determine the target business scenario corresponding to the search information based on the key information.

[0063] In one optional implementation, the key information includes at least one keyword; the determining module is further configured to: determine the initial business scenario corresponding to any keyword; and, given the initial business scenarios corresponding to each keyword, determine the target business scenario based on the number of initial business scenarios corresponding to each keyword and the type of the initial business scenario.

[0064] In an optional implementation, the retrieval module is further configured to: determine the target knowledge set corresponding to the target business scenario based on the mapping relationship between the target business scenario and the knowledge set; retrieve target semantic features matching the key information in the target knowledge set based on the key information in the retrieval information; and use the information corresponding to the target semantic features as the target knowledge fragment.

[0065] In an optional implementation, the retrieval module is further configured to: determine the similarity between the semantic features of the key information and the semantic features corresponding to each file in the target knowledge set, thereby obtaining the similarity of the semantic features corresponding to each file; and use the semantic features whose similarity satisfies a preset similarity condition as the target semantic features.

[0066] In an optional implementation, the generation module is further configured to: concatenate the target knowledge fragment and the search information to generate prompt information; and call a preset large language model to generate response information corresponding to the search information based on the prompt information.

[0067] In an optional implementation, the generation module is further configured to: if the target knowledge fragment includes multiple fragments, sequentially combine the target knowledge fragments according to the order of keywords in the key information in the search information to generate a knowledge fragment sequence; combine the knowledge fragment sequence and the search information to generate prompt information.

[0068] In an optional implementation, the apparatus is further configured to: in response to obtaining an updated file, identify the business scenario to which the updated file belongs; and invoke a preset model to convert the text of the updated file into a vector to obtain the semantic features of the updated file; and store the updated file and the semantic features of the updated file into a knowledge set corresponding to the business scenario to obtain an updated knowledge set.

[0069] The information retrieval device and the information retrieval method provided in the above embodiments of this disclosure are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the applications stored therein.

[0070] This disclosure also provides an electronic device for performing the information retrieval method described above. Please refer to... Figure 5 This illustrates a schematic diagram of an electronic device provided by some embodiments of the present disclosure. For example... Figure 5 As shown, the electronic device 5 includes: a processor 500, a memory 501, a bus 502, and a communication interface 503. The processor 500, the communication interface 503, and the memory 501 are connected via the bus 502. The memory 501 stores a computer program that can run on the processor 500. When the processor 500 runs the computer program, it executes the information retrieval method provided in any of the foregoing embodiments of this disclosure.

[0071] It should be noted that the memory 501 may also store the aforementioned monitoring model. When the electronic device is implemented as an access point device, the electronic device may also include a display.

[0072] The memory 501 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Communication between the virtual devices in the system is achieved through at least one communication interface 503 (which can be wired or wireless), such as the Internet, wide area network, local area network, or metropolitan area network.

[0073] Bus 502 can be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. The memory 501 is used to store programs. After receiving an execution instruction, the processor 500 executes the program. The information retrieval method disclosed in any of the foregoing embodiments of this disclosure can be applied to the processor 500, or implemented by the processor 500.

[0074] The processor 500 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of the processor 500 or by instructions in software form. The processor 500 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this disclosure. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this disclosure can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules may reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory 501. The processor 500 reads the contents of memory 501 and, in conjunction with its hardware, completes the steps of the above method.

[0075] The electronic devices and information retrieval methods provided in this disclosure are based on the same inventive concept and have the same beneficial effects as the methods they employ, operate, or implement.

[0076] This disclosure also provides a computer-readable storage medium corresponding to the information retrieval method provided in the foregoing embodiments. Please refer to... Figure 6 The computer-readable storage medium shown is an optical disc 30, on which a computer program (i.e., a program product) is stored. When the computer program is run by a processor, it executes the information retrieval method provided in any of the foregoing embodiments.

[0077] It should be noted that examples of the computer-readable storage medium may also 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 optical and magnetic storage media, which will not be elaborated here.

[0078] The computer-readable storage medium provided in the above embodiments of this disclosure and the information retrieval method provided in the embodiments of this disclosure are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the applications stored therein.

[0079] Although alternative embodiments of this disclosure have been described, those skilled in the art, upon learning the basic inventive concept, can make further changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this disclosure.

[0080] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of this disclosure. It should be understood that the above description is only a specific embodiment of this disclosure and is not intended to limit the scope of protection of this disclosure. Any modifications, equivalent substitutions, improvements, etc., made on the basis of the technical solution of this disclosure should be included within the scope of protection of this invention.

Claims

1. An information retrieval method, characterized in that, The method includes: In response to obtaining the search information, the target business scenario corresponding to the search information is determined; Based on the semantic features of the files in the target knowledge set corresponding to the target business scenario, target knowledge fragments corresponding to the search information are matched in the target knowledge set; wherein, the target knowledge set pre-stores files corresponding to the target business scenario and the semantic features of the files; Based on the target knowledge fragment and the search information, generate response information corresponding to the search information.

2. The method according to claim 1, characterized in that, Determining the target business scenario corresponding to the retrieved information includes: Based on the information type of the retrieved information, a target preprocessing method corresponding to the retrieved information is determined; the information type represents the signal form of the retrieved information. Based on the target preprocessing method, key information in the retrieved information is identified; wherein, the key information represents the business type involved in the retrieved information; Based on the key information, the target business scenario corresponding to the search information is determined.

3. The method according to claim 2, characterized in that, The key information includes at least one keyword; determining the target business scenario corresponding to the search information based on the key information includes: For any given keyword, determine the initial business scenario corresponding to that keyword; Given that the initial business scenario corresponding to each keyword is determined, the target business scenario is determined based on the number of initial business scenarios corresponding to each keyword and the type of the initial business scenario.

4. The method according to claim 3, characterized in that, The step of matching the target knowledge fragment corresponding to the search information in the target knowledge set includes: Based on the mapping relationship between the target business scenario and the knowledge set, determine the target knowledge set corresponding to the target business scenario; Based on the key information in the retrieval information, target semantic features that match the key information are retrieved from the target knowledge set; The information corresponding to the target semantic features is taken as the target knowledge fragment.

5. The method according to claim 4, characterized in that, The step of retrieving target semantic features that match the key information in the target knowledge set based on the key information in the retrieval information includes: Determine the similarity between the semantic features of the key information and the semantic features corresponding to each file in the target knowledge set, and obtain the similarity of the semantic features corresponding to each file; The semantic features whose similarity meets the preset similarity conditions are used as the target semantic features.

6. The method according to claim 3, characterized in that, The step of generating response information corresponding to the search information based on the target knowledge fragment further includes: By combining the target knowledge fragment and the search information, a prompt message is generated; The preset large language model is invoked to generate response information corresponding to the search information based on the prompt information.

7. The method according to claim 6, characterized in that, The step of splicing the target knowledge fragment and the search information to generate prompt information includes: If the target knowledge fragment includes multiple fragments, the target knowledge fragments are sequentially combined according to the order of the keywords in the key information in the search information to generate a knowledge fragment sequence. The knowledge fragment sequence and the retrieval information are combined to generate a prompt message.

8. The method according to claim 1, characterized in that, The method further includes: In response to obtaining an updated file, the system identifies the business scenario to which the updated file belongs; and invokes a preset model to convert the text of the updated file into a vector to obtain the semantic features of the updated file. The updated file and its semantic features are stored in the knowledge set corresponding to the business scenario to obtain the updated knowledge set.

9. An information retrieval device, characterized in that, The device includes: The determination module is used to determine the target business scenario corresponding to the user's search information in response to obtaining the user's search information; The retrieval module is used to match target knowledge fragments corresponding to the retrieval information in the target knowledge set based on the semantic features of files in the target knowledge set corresponding to the target business scenario; wherein, the target knowledge set pre-stores files corresponding to the target business scenario and the semantic features of the files; The generation module is used to generate response information corresponding to the search information based on the target knowledge fragment and the search information.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory, wherein the processor, when executing the computer program, implements the method of any one of claims 1 to 8.

11. A computer-readable storage medium storing a computer program that, when executed by a processor, implements the method of any one of claims 1 to 8.

Images