Retrieval method, device and system based on industry knowledge labels

Abstract

The invention relates to a retrieval method, device and system based on industry knowledge labels. The method comprises the following steps: in response to received question information of a user, identifying a question intention in the question information by adopting a pre-trained large language model, and generating an intention label for the question intention by adopting the large language model, a plurality of candidate labels matched with the intention label and candidate industry knowledge corresponding to the candidate labels are screened from a preset knowledge base, a candidate industry knowledge set is obtained, the problem and the candidate industry knowledge set are input into a pre-trained label recognition model, and a retrieval result generated by the label recognition model based on the candidate industry knowledge set is obtained. According to the scheme, the retrieval efficiency and the accuracy of the retrieval result can be improved.

Description

Technical Field

[0001] The present disclosure relates to the field of data retrieval technology, and in particular to a retrieval method, device and system based on industry knowledge tags. Background Art

[0002] In related technologies, the rapid growth of data volume makes traditional retrieval methods unable to cope with massive amounts of information, especially in specific industries, where users' information search needs are often highly specialized and complex. Existing vector retrieval technologies mostly rely on surface similarity for information matching, which causes users to be troubled by a large amount of irrelevant information during the search process. This information redundancy not only wastes users' time, but also weakens the effectiveness of information retrieval, thereby affecting the efficiency of decision-making. Summary of the invention

[0003] In order to overcome the problems existing in the related art, the present disclosure provides a retrieval method, device and system based on industry knowledge tags.

[0004] According to a first aspect of an embodiment of the present disclosure, a retrieval method based on industry knowledge tags is provided, comprising:

[0005] The retrieval method based on industry knowledge tags is characterized by comprising:

[0006] In response to receiving a question message from a user, using a pre-trained large language model to identify the question intention in the question message;

[0007] Using the large language model to generate an intent label for the question intent;

[0008] Screening multiple candidate tags matching the intention tag and candidate industry knowledge corresponding to the multiple candidate tags from a preset knowledge base to obtain a candidate industry knowledge set; the knowledge base is pre-constructed based on industry expert knowledge, and each data block in the knowledge base is marked with a corresponding tag; the candidate industry knowledge includes at least one data block;

[0009] The question and the candidate industry knowledge set are input into a pre-trained label recognition model to obtain a retrieval result generated by the label recognition model based on the candidate industry knowledge set.

[0010] In some embodiments of the present application, in response to receiving a question message from a user, using a pre-trained large language model to identify the question intention in the question message includes:

[0011] Inputting the question information into the large language model;

[0012] Using the large language model to filter the question information based on the industry expert knowledge, to obtain sub-information associated with the industry knowledge in the question information;

[0013] The large language model is used to identify the question intention in the sub-information.

[0014] In some embodiments of the present application, the data block is marked with multiple tags; the method further includes:

[0015] For each data block in the knowledge base, corresponding tags are generated according to a plurality of preset levels, and each high-level tag is associated with at least one low-level tag.

[0016] In some embodiments of the present application, the method further includes:

[0017] Generate a high-dimensional vector for each data block in the knowledge base;

[0018] Store the high-dimensional vector of each data block into the vector library;

[0019] The step of inputting the question information and the candidate industry knowledge set into a pre-trained label recognition model to obtain a retrieval result generated by the label recognition model based on the candidate industry knowledge set includes:

[0020] Inputting the question information, the candidate industry knowledge set, and the first high-dimensional vector corresponding to the candidate industry knowledge set into a pre-trained label recognition model;

[0021] Using the label recognition model to convert the question information into a second high-dimensional vector, matching the second high-dimensional vector with the first high-dimensional vector, and obtaining a target vector in the first high-dimensional vector that matches the second high-dimensional vector;

[0022] The search result is generated using the candidate industry knowledge corresponding to the target vector.

[0023] In some embodiments of the present application, the method further includes:

[0024] generating query information for the search results;

[0025] Sending the inquiry information to the terminal device of the user so that the terminal device displays the inquiry information;

[0026] In response to receiving reply information sent by the terminal device, adjusting the labels in the knowledge base based on the reply information; the reply information is reply information input by the user in response to the inquiry information.

[0027] According to a second aspect of an embodiment of the present disclosure, a retrieval device based on industry knowledge tags is provided, comprising:

[0028] A recognition unit, configured to, in response to receiving a question message from a user, use a pre-trained large language model to recognize the question intention in the question message;

[0029] A first generating unit, configured to generate an intention label for the question intention by using the large language model;

[0030] A matching unit, configured to screen multiple candidate tags matching the intended tag and candidate industry knowledge corresponding to the multiple candidate tags from a preset knowledge base to obtain a candidate industry knowledge set; the knowledge base is pre-constructed based on industry expert knowledge, and each data block in the knowledge base is annotated with a corresponding tag; the candidate industry knowledge includes at least one data block;

[0031] The second generating unit is used to input the question and the candidate industry knowledge set into a pre-trained label recognition model to obtain a retrieval result generated by the label recognition model based on the candidate industry knowledge set.

[0032] In some embodiments of the present application, the identification unit is specifically used for:

[0033] Inputting the question information into the large language model;

[0034] Using the large language model to filter the question information based on the industry expert knowledge, to obtain sub-information associated with the industry knowledge in the question information;

[0035] The large language model is used to identify the question intention in the sub-information.

[0036] According to a third aspect of an embodiment of the present disclosure, an electronic device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method described in any one of the first aspects is implemented.

[0037] According to a fourth aspect of an embodiment of the present disclosure, there is provided a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the method according to any one of the first aspects is implemented.

[0038] According to a fifth aspect of an embodiment of the present disclosure, a computer program product is provided, comprising a computer program, wherein the computer program implements the method as described in any one of the first aspects when executed by a processor.

[0039] The technical solution provided by the embodiments of the present disclosure may include the following beneficial effects: by responding to the user's question information, using a pre-trained large language model to identify the question intention in the question information, using the large language model to generate an intention label for the question intention, screening multiple candidate labels matching the intention label and candidate industry knowledge corresponding to the multiple candidate labels from a preset knowledge base, and obtaining a candidate industry knowledge set, the knowledge base is pre-constructed based on the knowledge of industry experts, each data block in the knowledge base is annotated with a corresponding label, the candidate industry knowledge includes at least one data block, the question and the candidate industry knowledge set are input into the pre-trained label recognition model, and the retrieval result generated by the label recognition model based on the candidate industry knowledge set is obtained. By identifying the user's intention, the candidate industry knowledge related to the question intention is screened out from the knowledge base using the intention label that can reflect the question intention, which not only improves the efficiency of the retrieval, but also enables the retrieval results selected from the candidate industry knowledge to be more in line with the user's needs, thereby improving the accuracy of the retrieval results.

[0040] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0042] Figure 1 The figure is a flow chart of a retrieval method based on industry knowledge tags according to an exemplary embodiment.

[0043] Figure 2 The invention is a block diagram of a retrieval device based on industry knowledge tags according to an exemplary embodiment.

[0044] Figure 3 It is a block diagram of a device for a retrieval method based on industry knowledge tags according to an exemplary embodiment. DETAILED DESCRIPTION

[0045] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0046] The terms used in the disclosed embodiments are only for the purpose of describing specific embodiments, and are not intended to limit the disclosed embodiments. The singular forms "a", "an" and "the" used in the disclosed embodiments and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings.

[0047] It should be understood that although the terms first, second, third, etc. may be used to describe various information in the disclosed embodiments, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the disclosed embodiments, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the words "if" and "if" as used herein may be interpreted as "at" or "when" or "in response to determination".

[0048] In addition, various forms of processes shown in the embodiments of the present disclosure may be used to reorder, add or delete steps. For example, the steps described in the present application may be performed in parallel, sequentially or in different orders, as long as the desired results of the technical solution disclosed in the present application can be achieved, and this document does not limit this.

[0049] In related technologies, the rapid growth of data volume makes traditional retrieval methods unable to cope with massive amounts of information, especially in specific industries, where users' information search needs are often highly specialized and complex. Existing vector retrieval technologies mostly rely on surface similarity for information matching, causing users to be troubled by a large amount of irrelevant information during the search process. This information redundancy not only wastes users' time, but also weakens the effectiveness of information retrieval, thereby affecting the efficiency of decision-making. Traditional search engines rely on keyword matching and often cannot deeply understand users' intentions, especially when involving multi-level contexts and industry-specific terms. This limitation makes it difficult for users to find information that truly meets their needs.

[0050] In order to solve the above problems, the present disclosure provides a retrieval method, device and system based on industry knowledge labels, which responds to the user's question information, uses a pre-trained large language model to identify the question intention in the question information, uses the large language model to generate an intention label for the question intention, and screens multiple candidate labels matching the intention label and candidate industry knowledge corresponding to the multiple candidate labels from a preset knowledge base to obtain a candidate industry knowledge set. The knowledge base is pre-constructed based on the knowledge of industry experts, and each data block in the knowledge base is annotated with a corresponding label. The candidate industry knowledge includes at least one data block. The question and the candidate industry knowledge set are input into the pre-trained label recognition model to obtain the retrieval result generated by the label recognition model based on the candidate industry knowledge set. By identifying the user's intention and using the intention label that can reflect the question intention to screen out the candidate industry knowledge related to the question intention from the knowledge base, not only the efficiency of retrieving industry knowledge is improved, but also the retrieval results selected from the candidate industry knowledge can be made more in line with the user's needs, thereby improving the accuracy of the retrieval results.

[0051] Figure 1 is a flowchart of a retrieval method based on industry knowledge tags according to an exemplary embodiment. Figure 1 As shown, it should be noted that the retrieval method based on industry knowledge tags in the embodiment of the present application is applied to a retrieval device based on industry knowledge tags. Figure 1 As shown, the method may include the following steps:

[0052] Step 101, in response to receiving a question message from a user, using a pre-trained large language model to identify the question intention in the question message.

[0053] It can be understood that the above question information may be information input by the user through the terminal device in order to retrieve target industry knowledge.

[0054] In one example, the large language model may be pre-trained based on industry knowledge data.

[0055] In one embodiment, after receiving the question information sent by the terminal device, the question information is input into the large language model, and the large language model recognizes the question information to obtain the user's question intention in the question information.

[0056] In some embodiments of the present application, step 101 may specifically include the following steps:

[0057] Input question information into the large language model;

[0058] Use a large language model to filter and process question information based on industry expert knowledge to obtain sub-information related to industry knowledge in the question information;

[0059] Using large language models to identify question intent in sub-messages.

[0060] It should be noted that in some cases, the amount of information in the question information entered by the user is large, but there is a lot of information in the question information that is irrelevant to the question intention. Therefore, the large language model can be used to filter out sub-information related to industry knowledge from the question information, so as to more accurately identify the user's question intention and provide more accurate retrieval results.

[0061] In addition, the introduction of industry expert knowledge ensures that the system can accurately reflect industry-specific terms, concepts and information structures, thereby providing more targeted and in-depth retrieval results when processing queries in specific fields, avoiding the information bias that may be caused by general tags.

[0062] In one embodiment, the large language model can be pre-trained for question intent recognition based on industry expert knowledge, and the trained large language model can be used to filter and process question information to obtain sub-information related to industry knowledge in the question information, and the large language model can be used to identify the question intent in the sub-information.

[0063] Step 102: Use a large language model to generate an intent label for the question intent.

[0064] It should be noted that the use of a large language model to generate intent labels for question intent can be implemented using existing technologies, which will not be elaborated here.

[0065] In one embodiment, an intent tag is generated for the question intention, so as to preliminarily filter the industry knowledge related to the user's question intention from the knowledge base, thereby narrowing the search scope.

[0066] Step 103, screening multiple candidate tags matching the intent tag and candidate industry knowledge corresponding to the multiple candidate tags from a preset knowledge base to obtain a candidate industry knowledge set.

[0067] The knowledge base is pre-built based on the knowledge of industry experts, and each data block in the knowledge base is marked with a corresponding label; the candidate industry knowledge includes at least one data block.

[0068] It should be noted that the intent label can be one or more.

[0069] In one embodiment, for each intent tag, multiple candidate tags matching the intent tag and candidate industry knowledge corresponding to the multiple candidate tags are screened from a preset knowledge base to obtain a candidate industry knowledge set.

[0070] As an example, candidate tags can be matched to intent tags by exact matching, that is, the matched candidate tags can be consistent with the intent tags. Candidate tags can also be matched to intent tags by similarity matching, that is, tags whose similarity to the intent tags is greater than a preset threshold can be used as candidate tags.

[0071] It is understandable that by filtering out candidate industry knowledge through intent tags, relevant data can be quickly located, which can significantly improve the speed and efficiency of information retrieval, reduce information redundancy, and optimize the user's information acquisition experience.

[0072] In some embodiments of the present application, the data block is marked with multiple tags, and the method further includes:

[0073] For each data block in the knowledge base, corresponding tags are generated according to multiple preset levels, and each high-level tag is associated with at least one low-level tag.

[0074] It should be noted that if the labels in the database are not perfect, the labels in the database may fail to match the intended labels. In order to improve the success rate of label matching, multi-level labels can be generated for the data blocks in the database.

[0075] In addition, by generating multi-level tags, we can better understand the user's search intent, especially in industries such as medicine or law, where users may use different terms to describe the same concept. The construction of multi-level tags helps the model identify the correlation between terms and greatly improves retrieval accuracy.

[0076] In one example, the core concepts of a data block may be placed at the top level (ie, configured as high-level tags), and specific sub-concepts and related knowledge may be organized at the bottom level (ie, configured as low-level tags), thereby supporting hierarchical matching of complex queries.

[0077] In some embodiments of the present application, batch processing and asynchronous computing methods can be used to mark data blocks, thereby increasing processing speed.

[0078] Step 104, input the question and the candidate industry knowledge set into the pre-trained label recognition model to obtain the retrieval result generated by the label recognition model based on the candidate industry knowledge set.

[0079] As an example, the tag recognition model uses technologies such as convolutional neural networks (CNN) and recurrent neural networks (RNN) combined with word embedding technology in natural language processing (NLP) to quickly identify key tags in questions. In addition, the model uses transfer learning from pre-trained language models (such as BERT or GPT) to effectively capture contextual information in user queries, thereby identifying key tags corresponding to the industry knowledge tag system.

[0080] In one embodiment, the question and the candidate industry knowledge set are input into a pre-trained label recognition model. The label recognition model generates a retrieval label based on the question information, matches the retrieval label with the label in the candidate industry knowledge set to obtain a target label, and matches the candidate industry knowledge (i.e., candidate data block) corresponding to the target label to obtain a retrieval result.

[0081] In some embodiments of the present application, the method may further include the following steps:

[0082] Generate a high-dimensional vector for each data block in the knowledge base;

[0083] Store the high-dimensional vector of each data block into the vector library;

[0084] Step 104 may specifically include the following steps:

[0085] Inputting the question information, the candidate industry knowledge set, and the first high-dimensional vector corresponding to the candidate industry knowledge set into the pre-trained label recognition model;

[0086] The label recognition model is used to convert the question information into a second high-dimensional vector, and the second high-dimensional vector is matched with the first high-dimensional vector to obtain a target vector in the first high-dimensional vector that matches the second high-dimensional vector;

[0087] Generate retrieval results using the candidate industry knowledge corresponding to the target vector.

[0088] In one embodiment, a high-dimensional vector can be generated for each database in the knowledge base and stored in the vector library. The high-dimensional vector reflects the position of the data block in the multidimensional space, so that similar data blocks can be more closely clustered in the vector space, so that similarity retrieval can be quickly performed in the high-dimensional space through vector calculation technology (such as cosine similarity or Euclidean distance), so as to find the data most relevant to the user's query intention. Not only does it ensure efficient data management and convenience of subsequent retrieval, but it also captures the deep semantic features of the data block by creating a high-dimensional embedding vector for each data block, thereby improving the accuracy of retrieval.

[0089] In some embodiments of the present application, after step 104, the method may further include the following steps:

[0090] Generate query information for search results;

[0091] Sending the inquiry information to the user's terminal device so that the terminal device displays the inquiry information;

[0092] In response to receiving reply information sent by the terminal device, the tags in the knowledge base are adjusted based on the reply information; the reply information is reply information input by the user in response to the inquiry information.

[0093] In other embodiments of the present application, new professional terms and concepts can be regularly introduced into the knowledge base to ensure that it is always consistent with the industry frontier. The integration of industry expert knowledge enables the system to provide multi-dimensional information presentation to meet the needs of users in different scenarios, allowing users to obtain more comprehensive information such as the latest research, market analysis, and case studies, thereby improving the overall search depth and efficiency.

[0094] According to the retrieval method based on industry knowledge labels proposed in the embodiment of the present application, in response to receiving the user's question information, a pre-trained large language model is used to identify the question intention in the question information, and an intention label for the question intention is generated by the large language model. Multiple candidate labels matching the intention label and candidate industry knowledge corresponding to the multiple candidate labels are screened from the preset knowledge base to obtain a candidate industry knowledge set. The knowledge base is pre-constructed based on the knowledge of industry experts, and each data block in the knowledge base is annotated with a corresponding label. The candidate industry knowledge includes at least one data block. The question and the candidate industry knowledge set are input into the pre-trained label recognition model to obtain the retrieval result generated by the label recognition model based on the candidate industry knowledge set. By identifying the user's intention and using the intention label that can reflect the question intention to screen out the candidate industry knowledge related to the question intention from the knowledge base, not only the efficiency of the retrieval is improved, but also the retrieval results selected from the candidate industry knowledge can be made more in line with the user's needs, thereby improving the accuracy of the retrieval results.

[0095] Figure 2 is a block diagram of a retrieval device based on industry knowledge tags according to an exemplary embodiment. Figure 2 The device includes a recognition unit 201, a first generation unit 202, a matching unit 203 and a second generation unit 204.

[0096] The recognition unit 201 is used to recognize the question intention in the question information by using a pre-trained large language model in response to receiving the question information from the user;

[0097] A first generating unit 202, configured to generate an intent label for the question intent using a large language model;

[0098] The matching unit 203 is used to screen multiple candidate tags matching the intended tag and candidate industry knowledge corresponding to the multiple candidate tags from a preset knowledge base to obtain a candidate industry knowledge set; the knowledge base is pre-constructed based on industry expert knowledge, and each data block in the knowledge base is marked with a corresponding tag; the candidate industry knowledge includes at least one data block;

[0099] The second generating unit 204 is used to input the question and the candidate industry knowledge set into the pre-trained label recognition model to obtain the retrieval result generated by the label recognition model based on the candidate industry knowledge set.

[0100] In some embodiments of the present application, the identification unit 201 is specifically used for:

[0101] Input question information into the large language model;

[0102] Use a large language model to filter and process question information based on industry expert knowledge to obtain sub-information related to industry knowledge in the question information;

[0103] Using large language models to identify question intent in sub-messages.

[0104] In some embodiments of the present application, the data block is marked with multiple tags, and the device may further include:

[0105] The third generating unit is used to generate corresponding labels for each data block in the knowledge base according to a plurality of preset levels, wherein each high-level label is associated with at least one low-level label.

[0106] In some embodiments of the present application, the device may further include:

[0107] A fourth generating unit, used to generate a high-dimensional vector for each data block in the knowledge base;

[0108] A storage unit, used for storing the high-dimensional vector of each data block into a vector library;

[0109] The second generating unit 204 can be specifically used for:

[0110] Inputting the question information, the candidate industry knowledge set, and the first high-dimensional vector corresponding to the candidate industry knowledge set into the pre-trained label recognition model;

[0111] The label recognition model is used to convert the question information into a second high-dimensional vector, and the second high-dimensional vector is matched with the first high-dimensional vector to obtain a target vector in the first high-dimensional vector that matches the second high-dimensional vector;

[0112] Generate retrieval results using the candidate industry knowledge corresponding to the target vector.

[0113] In some embodiments of the present application, the device may further include:

[0114] A fifth generating unit, used to generate query information for the search result;

[0115] A sending unit, used to send the inquiry information to the user's terminal device so that the terminal device displays the inquiry information;

[0116] The adjustment unit is used to adjust the labels in the knowledge base based on the reply information received from the terminal device; the reply information is the reply information input by the user in response to the inquiry information.

[0117] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0118] According to a retrieval device based on industry knowledge labels proposed in an embodiment of the present application, in response to receiving a user's question information, a pre-trained large language model is used to identify the question intent in the question information, and an intent label for the question intent is generated using the large language model. Multiple candidate labels matching the intent label and candidate industry knowledge corresponding to the multiple candidate labels are screened from a preset knowledge base to obtain a candidate industry knowledge set. The knowledge base is pre-constructed based on industry expert knowledge, and each data block in the knowledge base is annotated with a corresponding label. The candidate industry knowledge includes at least one data block. The question and the candidate industry knowledge set are input into a pre-trained label recognition model to obtain a retrieval result generated by the label recognition model based on the candidate industry knowledge set. By identifying the user's intent and using the intent label that can reflect the question intent to screen out the candidate industry knowledge related to the question intent from the knowledge base, not only the efficiency of the retrieval is improved, but also the retrieval results selected from the candidate industry knowledge can be made more in line with the user's needs, thereby improving the accuracy of the retrieval results.

[0119] Figure 3 3 is a block diagram of a device for retrieval based on industry knowledge tags according to an exemplary embodiment. For example, the device 300 may be an electronic device, such as a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, etc.

[0120] Reference Figure 3 , the device 300 may include one or more of the following components: a processing component 302 , a memory 304 , a power component 306 , a multimedia component 308 , an audio component 310 , an input / output (I / O) interface 312 , a sensor component 314 , and a communication component 316 .

[0121] The processing component 302 generally controls the overall operation of the device 300, such as operations associated with display, phone calls, data communications, camera operations, and recording operations. The processing component 302 may include one or more processors 320 to execute instructions to complete all or part of the steps of the above-mentioned method. In addition, the processing component 302 may include one or more modules to facilitate the interaction between the processing component 302 and other components. For example, the processing component 302 may include a multimedia module to facilitate the interaction between the multimedia component 308 and the processing component 302.

[0122] The memory 304 is configured to store various types of data to support operations on the device 300. Examples of such data include instructions for any application or method operating on the device 300, contact data, phone book data, messages, pictures, videos, etc. The memory 304 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0123] The power component 306 provides power to the various components of the device 300. The power component 306 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power for the device 300.

[0124] The multimedia component 308 includes a screen that provides an output interface between the device 300 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense the boundaries of the touch or slide action, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 308 includes a front camera and / or a rear camera. When the device 300 is in an operating mode, such as a shooting mode or a video mode, the front camera and / or the rear camera may receive external multimedia data. Each front camera and rear camera may be a fixed optical lens system or have a focal length and optical zoom capability.

[0125] The audio component 310 is configured to output and / or input audio signals. For example, the audio component 310 includes a microphone (MIC), and when the device 300 is in an operating mode, such as a call mode, a recording mode, and a speech recognition mode, the microphone is configured to receive an external audio signal. The received audio signal can be further stored in the memory 304 or sent via the communication component 316. In some embodiments, the audio component 310 also includes a speaker for outputting audio signals.

[0126] I / O interface 312 provides an interface between processing component 302 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include but are not limited to: a home button, a volume button, a start button, and a lock button.

[0127] The sensor assembly 314 includes one or more sensors for providing various aspects of status assessment for the device 300. For example, the sensor assembly 314 can detect the open / closed state of the device 300, the relative positioning of components, such as the display and keypad of the device 300, the sensor assembly 314 can also detect the position change of the device 300 or a component of the device 300, the presence or absence of user contact with the device 300, the orientation or acceleration / deceleration of the device 300, and the temperature change of the device 300. The sensor assembly 314 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. The sensor assembly 314 may also include an optical sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 314 may also include an accelerometer, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

[0128] The communication component 316 is configured to facilitate wired or wireless communication between the device 300 and other devices. The device 300 can access a wireless network based on a communication standard, such as WiFi, 2G or 3G, or a combination thereof. In an exemplary embodiment, the communication component 316 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 316 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

[0129] In an exemplary embodiment, the apparatus 300 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components to perform the above method.

[0130] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory 304 including instructions, and the instructions can be executed by the processor 320 of the device 300 to perform the above method. For example, the non-transitory computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0131] In an exemplary embodiment, a computer program product is also provided, comprising a computer program, which implements the above method when executed by the processor 320 of the device 300 .

[0132] Those skilled in the art will readily appreciate other embodiments of the present invention after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary techniques in the art that are not disclosed in this disclosure. The specification and examples are to be considered exemplary only, and the true scope and spirit of the present invention are indicated by the following claims.

[0133] It should be understood that the present invention is not limited to the exact construction that has been described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

Claims

1. A retrieval method based on industry knowledge tags, characterized in that: include: In response to receiving a question message from a user, using a pre-trained large language model to identify the question intention in the question message; Using the large language model to generate an intent label for the question intent; Screening multiple candidate tags matching the intention tag and candidate industry knowledge corresponding to the multiple candidate tags from a preset knowledge base to obtain a candidate industry knowledge set; the knowledge base is pre-constructed based on industry expert knowledge, and each data block in the knowledge base is marked with a corresponding tag; the candidate industry knowledge includes at least one data block; The question and the candidate industry knowledge set are input into a pre-trained label recognition model to obtain a retrieval result generated by the label recognition model based on the candidate industry knowledge set.

2. The retrieval method based on industry knowledge tags according to claim 1 is characterized in that: In response to receiving the user's question information, using the pre-trained large language model to identify the question intention in the question information includes: Inputting the question information into the large language model; Using the large language model to filter the question information based on the industry expert knowledge, to obtain sub-information associated with the industry knowledge in the question information; The large language model is used to identify the question intention in the sub-information.

3. The retrieval method based on industry knowledge tags according to claim 1 is characterized in that: The data block has multiple tags; the method also includes: For each data block in the knowledge base, corresponding tags are generated according to a plurality of preset levels, and each high-level tag is associated with at least one low-level tag.

4. The retrieval method based on industry knowledge tags according to claim 1 is characterized in that: Also includes: Generate a high-dimensional vector for each data block in the knowledge base; Store the high-dimensional vector of each data block into the vector library; The step of inputting the question information and the candidate industry knowledge set into a pre-trained label recognition model to obtain a retrieval result generated by the label recognition model based on the candidate industry knowledge set includes: Inputting the question information, the candidate industry knowledge set, and the first high-dimensional vector corresponding to the candidate industry knowledge set into a pre-trained label recognition model; Using the label recognition model to convert the question information into a second high-dimensional vector, matching the second high-dimensional vector with the first high-dimensional vector, and obtaining a target vector in the first high-dimensional vector that matches the second high-dimensional vector; The search result is generated using the candidate industry knowledge corresponding to the target vector.

5. The retrieval method based on industry knowledge tags according to claim 1 is characterized in that: Also includes: generating query information for the search results; Sending the inquiry information to the terminal device of the user so that the terminal device displays the inquiry information; In response to receiving reply information sent by the terminal device, adjusting the labels in the knowledge base based on the reply information; The reply information is reply information input by the user in response to the inquiry information.

6. A retrieval device based on industry knowledge tags, characterized in that: include: A recognition unit, configured to, in response to receiving a question message from a user, use a pre-trained large language model to recognize the question intention in the question message; A first generating unit, configured to generate an intention label for the question intention by using the large language model; A matching unit, configured to screen multiple candidate tags matching the intended tag and candidate industry knowledge corresponding to the multiple candidate tags from a preset knowledge base to obtain a candidate industry knowledge set; the knowledge base is pre-constructed based on industry expert knowledge, and each data block in the knowledge base is annotated with a corresponding tag; the candidate industry knowledge includes at least one data block; The second generating unit is used to input the question and the candidate industry knowledge set into a pre-trained label recognition model to obtain a retrieval result generated by the label recognition model based on the candidate industry knowledge set.

7. The retrieval device based on industry knowledge tags according to claim 6 is characterized in that: The identification unit is specifically used for: Inputting the question information into the large language model; Using the large language model to filter the question information based on the industry expert knowledge, to obtain sub-information associated with the industry knowledge in the question information; The large language model is used to identify the question intention in the sub-information.

8. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method according to any one of claims 1 to 5 is implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

10. A computer program product, comprising a computer program, characterized in that The computer program implements the method according to any one of claims 1 to 5 when executed by a processor.

Images