Dialogue information extraction method and device based on large model, equipment and storage medium

Abstract

The invention discloses a dialogue information extraction method and device based on a large model, equipment and a storage medium, and the method comprises the steps: determining an information extraction scene, and determining a to-be-extracted slot position based on the information extraction scene; according to the to-be-extracted slot position, slot position information extraction is carried out on the to-be-processed problem, and extracted information and an extraction progress are obtained; and when the extraction progress is that the extraction is not completed, generating a reverse question, and when the extraction progress is that the extraction is completed, generating an answer result of the to-be-processed question based on the extracted information. According to the technical scheme provided by the embodiment of the invention, scene confirmation is carried out to determine slot position information needing to be extracted, then information extraction and self-reflection are carried out by utilizing the complex semantic understanding capability of the large model to ensure the integrity of the extracted content, and then flexible anti-questioning is carried out automatically according to the vacant slot position through self-questioning-self-answering of the large model, so that the efficiency of extracting the slot position information is improved. And finally, the whole extraction process is completed, and returning is carried out in a structured form.

Description

Technical Field

[0001] The present invention relates to the technical field of dialogue information extraction, and in particular to a dialogue information extraction method, device, equipment and storage medium based on a large model. Background Art

[0002] Information extraction models play a very important role in dialogue systems, mainly by converting unstructured information in dialogues into structured information. With the development of big model technology, many information extraction scenarios currently use big models to extract key information. In the case of zero or few samples, big models have also achieved good results.

[0003] In the past, the information extraction method for multi-round dialogues was often used to extract information using the method of intention judgment plus slot extraction. By designing the dialogue process of the entire system and the slot collection rhetorical question method, the information collection of the entire dialogue process is completed based on small model deep learning algorithms such as LSTM and BERT.

[0004] However, these technologies also have their limitations: 1) Methods based on small models, although they perform well in training, are not ideal for zero or few samples, and also have defects in model scenario migration; 2) Pre-designed dialogue processes require the intervention of developers or operators, which adds a lot of extra costs; 3) Defined processes and rhetorical questions are not user-friendly enough and are not flexible enough. Summary of the invention

[0005] The present invention provides a conversation information extraction method, device, equipment and storage medium based on a large model, so as to improve the accuracy, flexibility and friendliness of information extraction in multi-round conversations.

[0006] According to one aspect of the present invention, a method for extracting multi-round dialogue information based on a large model is provided, comprising:

[0007] Determine an information extraction scenario, and determine a slot to be extracted based on the information extraction scenario;

[0008] According to the slot to be extracted, extract the slot information of the problem to be processed, and obtain the extracted information and the extraction progress;

[0009] When the extraction progress is that the extraction is incomplete, a counter-question is generated. When the extraction progress is that the extraction is completed, an answer result of the pending question is generated based on the extracted information.

[0010] According to another aspect of the present invention, a multi-round dialogue information extraction device based on a large model is provided, comprising:

[0011] A slot determination module, used to determine an information extraction scenario, and determine a slot to be extracted based on the information extraction scenario;

[0012] An extraction reflection module is used to extract slot information of the problem to be processed according to the slot to be extracted, and obtain the extracted information and extraction progress;

[0013] The self-questioning and self-answering module is used to generate a counter-question when the extraction progress is that the extraction is incomplete, and to generate an answer result of the pending question based on the extracted information when the extraction progress is that the extraction is completed.

[0014] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0015] at least one processor;

[0016] and a memory communicatively connected to the at least one processor; wherein,

[0017] The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the large model-based dialogue information extraction method described in any embodiment of the present invention.

[0018] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the large model-based dialogue information extraction method described in any embodiment of the present invention when executed.

[0019] The technical solution of the embodiment of the present invention includes: determining the information extraction scenario, determining the slot to be extracted based on the information extraction scenario; extracting the slot information of the problem to be processed according to the slot to be extracted, and obtaining the extracted information and the extraction progress; generating a counter-question when the extraction progress is that the extraction is not completed, and generating the answer result of the problem to be processed based on the extracted information when the extraction progress is that the extraction is completed. The technical solution of the embodiment of the present invention solves the problem that the method based on small models in the existing information extraction technology is not ideal for zero samples or few samples, and there are defects in the migration of model scenarios. It is based on the pre-designed dialogue process, which requires the intervention of development or operation personnel, adding a lot of extra costs to obtain problems, and the use of defined processes and counter-questioning methods is not user-friendly enough and not flexible enough. First, the scenario is confirmed to determine the slot information that needs to be extracted. Secondly, the complex semantic understanding ability of the big model is used to extract information and self-reflect to ensure the completeness of the extracted content. Then, the big model's self-questioning and self-answering function is used to automatically ask flexible questions based on the vacant slots to ensure the flexibility and friendliness of the conversation. Finally, the entire extraction process is completed and returned in a structured form.

[0020] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0022] Figure 1 A flowchart of a method for extracting conversation information based on a large model provided in the first embodiment of the present invention;

[0023] Figure 2 A flowchart of a method for extracting conversation information based on a large model provided in Embodiment 2 of the present invention;

[0024] Figure 3 A schematic diagram of the structure of a conversation information extraction device based on a large model provided in Embodiment 3 of the present invention;

[0025] Figure 4 This is a schematic diagram of the structure of an electronic device provided in Embodiment 4 of the present invention. DETAILED DESCRIPTION

[0026] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0027] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0028] Embodiment 1

[0029] Figure 1 This is a flowchart of a method for extracting conversation information based on a large model provided in the first embodiment of the present invention. This embodiment is applicable to the case of extracting information from a problem to be processed. The method can be executed by a conversation information extraction device based on a large model. The device can be implemented in the form of hardware and / or software, and the device can be configured in a computer device. Figure 1 As shown, the method includes:

[0030] S110: Determine an information extraction scenario, and determine a slot to be extracted based on the information extraction scenario.

[0031] Among them, the information extraction scenario can be understood as the specific scenario in which the information extraction is applied. In different application scenarios, the information to be extracted is different, and the slot to be extracted can be understood as the type or category of information to be extracted. For example, in the fault repair scenario, the slot to be extracted can be the repair equipment, fault description, department information, etc.; in the e-commerce field, the information slot to be extracted can be the product name, price, etc.

[0032] It is understandable that when extracting conversation text information, the scope of information extraction, that is, the slot to be extracted, can be determined according to the information scenario. In actual applications, the information extraction scenario and slot can be updated based on the operation of business personnel to ensure that the extracted information matches the actual application requirements.

[0033] In an optional implementation, determining an information extraction scenario and determining a slot to be extracted based on the information extraction scenario may include: determining an information extraction scenario and determining target extraction slot description information that matches the information extraction scenario based on a mapping relationship between the information extraction scenario and the extraction slot description information; and determining a slot to be extracted based on the target extraction slot description information.

[0034] Among them, the mapping relationship between the information extraction scenario and the extraction slot description information can be a mapping relationship established in advance by business personnel, and the extraction slot description information refers to the information used to describe the slot to be extracted corresponding to the information extraction scenario; the target extraction slot description information can be the extraction slot description information corresponding to the current information extraction scenario.

[0035] Specifically, after clarifying the information extraction scenario, the target extraction slot description information matching the current information extraction scenario can be determined by querying or matching the mapping relationship. The target extraction slot description information usually describes the slot to be extracted in the current information extraction scenario.

[0036] S120. Extract slot information of the problem to be processed according to the slot to be extracted, and obtain extracted information and extraction progress.

[0037] Among them, the problem to be processed can be a user's question or conversation; the extracted information can be understood as the information extracted from the slot to be extracted in the text corresponding to the problem to be processed; the extraction progress is used to determine whether the information extraction of all slots to be extracted is completed.

[0038] In an embodiment of the present invention, the process of extracting slot information of a problem to be processed based on the slot to be extracted to obtain the extracted information includes: generating first guidance prompt information based on the slot to be extracted and the problem to be processed, guiding the large model to extract information corresponding to the slot to be extracted in the problem to be processed in a self-reflective mode based on the first guidance prompt information to obtain the extracted information, and determining whether the information corresponding to the slot to be extracted is completely extracted.

[0039] Among them, the first guidance prompt information can be information used to guide the large model to execute the extraction process. The extracted information refers to the information extracted by the large model after the extraction process. The extracted information can be information obtained after the extraction of all slots to be extracted, or it can be information obtained after the extraction of some slots to be extracted.

[0040] Specifically, the first guidance information can be generated according to the text of the slot to be extracted and the problem to be processed, guiding the large model to extract the information corresponding to the slot to be extracted in the problem to be processed in a self-reflective mode, obtain the extracted information, and determine whether the information corresponding to the slot to be extracted is complete. In the process of self-reflection, the large model will accurately extract the information corresponding to the slot to be extracted in the problem to be processed according to the instructions of the first guidance prompt information. At the same time, the integrity of the extracted information will be judged to ensure that the extracted information is comprehensive and accurate.

[0041] In an embodiment of the present invention, the first guidance prompt information is generated based on the slot to be extracted and the problem to be processed, including: the slot to be extracted is spliced ​​with the user description content as a system prompt word, and is input as the first guidance prompt information to the large model, so that the large model obtains the self-reflection result through decoding.

[0042] The user description content may refer to a pending problem input by the user.

[0043] Specifically, the slot to be extracted can be spliced ​​with the user description content and input into the large model as the first guidance prompt information, so that the large model can obtain the self-reflection result through decoding.

[0044] In the embodiment of the present invention, the large model obtains the self-reflection result through decoding, including: the large model determines the self-reflection result through decoding.

[0045] Among them, the self-reflection result is: determining to continue slot extraction and generating rhetorical questions, or determining that information collection is completed.

[0046] Specifically, the large model performs a self-reflection process through decoding, that is, determines whether the information extraction of all slots to be extracted has been completed. If the result of self-reflection is to determine to continue slot extraction and generate rhetorical questions, it means that the information extraction of all slots to be collected has not been completed, and it is necessary to further generate rhetorical questions for users to answer, so as to extract the information of the unextracted slots based on the user's answers. If the result of self-reflection is to determine that information collection is completed, it means that the extraction of information of all slots to be extracted has been completed. In an optional implementation method, according to the slots to be extracted, the slot information of the problem to be processed is extracted to obtain the extraction progress, including: determining the extracted slots according to the extracted information, and judging the extraction progress based on the extracted slots and the slots to be extracted.

[0047] The extracted slot refers to the slot from which information extraction has been completed.

[0048] Specifically, the extracted slots can be matched with the slots to be extracted to determine which slots have been successfully extracted and which slots have not been extracted. Then, the extraction progress is calculated based on the number of extracted slots and the total number of slots to be extracted. For example, if 3 slots have been extracted and there are a total of 5 slots to be extracted, then the extraction progress is 60%. If the progress reaches 100%, it means that the information of all slots to be extracted has been successfully extracted, that is, the extraction has been completed; if the progress is less than 100%, it means that there is still slot information that has not been extracted, and the information extraction work needs to be continued, that is, the information extraction is not completed.

[0049] S130. When the extraction progress is that the extraction is incomplete, generate a counter-question; when the extraction progress is that the extraction is completed, generate an answer result of the pending question based on the extracted information.

[0050] Specifically, when the extraction progress is complete, a counter-question can be generated, the user can answer the question, and then the unextracted slots are extracted again according to the user's answer. The process is repeated until the extraction progress is completed, and the answer results of the pending questions are generated based on the extracted information.

[0051] In an embodiment of the present invention, when the extraction progress is incomplete, generating a rhetorical question includes: determining unextracted slots based on the extracted slots and the slots to be extracted, generating second guidance prompt information based on the unextracted slots, the extracted information and the first guidance prompt information; and prompting the large model to generate a rhetorical question based on the second guidance prompt information.

[0052] Specifically, by comparing the extracted slots with the slots to be extracted, it is possible to determine which slots to be extracted have not been successfully extracted, and these unextracted slots are regarded as unextracted slots. Then, the first guidance prompt information is further supplemented and adjusted according to the unextracted slots and the extracted information to obtain the second guidance prompt information.

[0053] The second guiding prompt information can guide the large model to generate rhetorical questions. The rhetorical questions can be questions asked to the user about the information of the unextracted slots. The information of the unextracted slots can be obtained through interaction with the user through rhetorical questions.

[0054] In a preferred implementation, undrawn slots are determined based on the drawn slots and the slots to be drawn, and second guidance prompt information is generated based on the undrawn slots, the extracted information and the first guidance prompt information; based on the second guidance prompt information, the big model is prompted to generate rhetorical questions, specifically including: inputting scene information, user description and self-reflection results into the big model; the big model gradually generates questions for self-questioning, and the big model answers itself according to the self-questioned questions to generate rhetorical questions.

[0055] The rhetorical question is used to prompt the user to supplement the information; if the information collection is completed, the rhetorical question is used to prompt the user to confirm whether the extracted information is accurate.

[0056] Specifically, the scenario information, user description and self-reflection results are input into the large model. The large model can determine which slot information needs to be further extracted through reasoning. If the unextracted slot information includes "department information" and "contact name", the self-questioning question generated by the large model can be: Please ask the user questions based on the "department information" and "contact name" slots to be extracted. Correspondingly, the large model further generates a counter-question based on the self-questioning question and returns the counter-question to the user. For example, the generated counter-question is: What is your department and what is your contact name?

[0057] In an embodiment of the present invention, after prompting the large model to generate a rhetorical question based on the second guidance prompt information, it also includes: receiving the user's answer to the rhetorical question, and extracting the unextracted slot based on the answer.

[0058] Specifically, after generating the rhetorical question, the user's answer to the rhetorical question can be received, and the unextracted slots can be extracted from the answer content, and the above process is repeated until all the slots to be extracted are extracted.

[0059] In an embodiment of the present invention, when the extraction progress is that the extraction is completed, an answer result of the pending question is generated based on the extracted information, including: when the extraction progress is that the extraction is completed, generating an extraction information confirmation question based on the extracted information; receiving the user's answer result to the extraction information confirmation question and the extracted information, and generating the answer result of the pending question.

[0060] Specifically, when the extraction progress is completed, the extracted information is combined through the large model to generate an extracted information confirmation question, and the extracted information confirmation question is fed back to the user. The extracted information confirmation question is used to allow the user to confirm whether the extracted information is accurate. If the user's answer is confirmed to be correct, the answer to the pending question can be generated based on the extracted information.

[0061] The technical solution of the embodiment of the present invention includes: determining the information extraction scenario, determining the slot to be extracted based on the information extraction scenario; extracting the slot information of the problem to be processed according to the slot to be extracted, and obtaining the extracted information and the extraction progress; generating a counter-question when the extraction progress is that the extraction is not completed, and generating the answer result of the problem to be processed based on the extracted information when the extraction progress is that the extraction is completed. The technical solution of the embodiment of the present invention solves the problem that the method based on small models in the existing information extraction technology is not ideal for zero samples or few samples, and there are defects in the migration of model scenarios. It is based on the pre-designed dialogue process, which requires the intervention of development or operation personnel, adding a lot of extra costs to obtain problems, and the use of defined processes and counter-questioning methods is not user-friendly enough and not flexible enough. First, the scenario is confirmed to determine the slot information that needs to be extracted. Secondly, the complex semantic understanding ability of the big model is used to extract information and self-reflect to ensure the completeness of the extracted content. Then, the big model's self-questioning and self-answering function is used to automatically ask flexible questions based on the vacant slots to ensure the flexibility and friendliness of the conversation. Finally, the entire extraction process is completed and returned in a structured form.

[0062] Embodiment 2

[0063] Figure 2 This is a flowchart of a method for extracting conversation information based on a large model provided in Example 2 of the present invention. This example is a preferred example of the above example. For its specific implementation, please refer to the technical solution of this example. The technical terms that are the same or corresponding to the above example are not repeated here. Figure 2 As shown, the method includes:

[0064] S210: Confirm the information extraction scenario.

[0065] Specifically, before the real-time extraction of conversation information, the specific scenario needs to be confirmed to ensure that the slot to be extracted and the slot description are clear. Taking the repair scenario as an example, during the conversation process, it is necessary to collect "repair equipment", "fault description", "department information", "contact name", "contact number", and "address information".

[0066] S220. The large model extracts slots from user questions and performs integrity verification through self-reflection.

[0067] At the beginning of the conversation, the large model needs to extract slots from the user description based on the scenario information and confirm whether the slot information is extracted completely through self-reflection.

[0068] The large model is represented as LLM. The scene information will be concatenated with the system prompt word System Prompt and the user description content User as input to the LLM. The large model obtains the self-reflection Self-Though result through decoding.

[0069] Among them, the splicing input is: <system>Scene Information <user>User Description <self-though>

[0070] Taking the repair scenario as an example, the self-reflection results of the simulated large model are as follows:

[0071]

[0072] S230. According to the self-reflection results of the large model, the large model conducts self-questioning and self-answering to confirm whether the information collection is completed. If not, repeat S220.

[0073] Specifically, according to the self-reflection results of the large model, the large model conducts self-questioning and self-answering to determine whether to continue to extract slots to generate rhetorical questions or to complete information collection.

[0074] Among them, the input given to the large model is: <system>Scene Information <user>User Description <self-though>Reflection content <self-query>. Gradually generate questions for yourself and then answer them accordingly.

[0075] Taking the repair scenario as an example, the results of simulating a large model’s self-questioning and self-answering are as follows:

[0076]

[0077] If it is necessary to continue collecting information, the answer content is a counter-question to the user to guide the user to supplement the information, and then repeat step S220 to extract the information;

[0078] If the information collection is completed, the answer is to confirm with the user whether the extracted content is accurate, that is, step S240 is performed to confirm the information with the user.

[0079] S240: proceed according to the final confirmation result of the user, further reflect, if the confirmation is correct, end the dialogue and return the result in a structured manner; if there is any incorrect content, repeat step S230.

[0080] Taking the repair scenario as an example, the results of simulating a large model’s self-questioning and self-answering are as follows:

[0081]

[0082] The present invention aims to improve the accuracy, flexibility and friendliness of information extraction in multi-round dialogues. By designing a thinking chain of reflection-self-questioning-self-answering, the large model can self-confirm the accuracy of the extracted content and the integrity of the extracted content during the information extraction process, and automatically design flexible counter-questioning methods and information flows. The complex semantic understanding ability of the large model is utilized to avoid the configuration of fixed processes and fixed slots for collecting speech in traditional multi-round dialogues, thereby improving the fluency and intelligence of the overall dialogue. By adding the thinking chain logic of reflection-self-questioning-self-answering of the large model, the accuracy of information extraction in multi-round dialogues of the large model is improved.

[0083] The technical solution of the embodiment of the present invention includes: determining an information extraction scenario, determining a slot to be extracted based on the information extraction scenario; extracting slot information for a problem to be processed according to the slot to be extracted, and obtaining extracted information and extraction progress; generating a counter-question when the extraction progress is that the extraction is not completed, and generating an answer result of the problem to be processed based on the extracted information when the extraction progress is that the extraction is completed. The technical solution of the embodiment of the present invention solves the problem that the method based on a small model in the existing information extraction technology is not ideal for zero or few samples, and there are also defects in the migration of model scenarios, the problem that the development or operation personnel need to intervene based on the pre-designed dialogue process, which increases a lot of additional costs, and the problem that the defined process and counter-question method are not user-friendly and not flexible enough. First, the scenario is confirmed to determine the slot information that needs to be extracted. Secondly, the complex semantic understanding ability of the big model is used to extract information and self-reflect to ensure the completeness of the extracted content. Then, the big model's self-questioning and self-answering function is used to automatically ask flexible questions based on the vacant slots to ensure the flexibility and friendliness of the conversation. Finally, the entire extraction process is completed and returned in a structured form.

[0084] Embodiment 3

[0085] Figure 3 This is a schematic diagram of the structure of a conversation information extraction device based on a large model provided in the third embodiment of the present invention. Figure 3 As shown, the device comprises:

[0086] A slot determination module 310, used to determine an information extraction scenario, and determine a slot to be extracted based on the information extraction scenario;

[0087] An extraction reflection module 320 is used to extract slot information for the problem to be processed according to the slot to be extracted, and obtain the extracted information and extraction progress;

[0088] The self-questioning and self-answering module 330 is used to generate a counter-question when the extraction progress is that the extraction is incomplete, and to generate an answer result of the pending question based on the extracted information when the extraction progress is that the extraction is completed.

[0089] The technical solution of the embodiment of the present invention includes: determining the information extraction scenario, determining the slot to be extracted based on the information extraction scenario; extracting the slot information of the problem to be processed according to the slot to be extracted, and obtaining the extracted information and the extraction progress; generating a counter-question when the extraction progress is that the extraction is not completed, and generating the answer result of the problem to be processed based on the extracted information when the extraction progress is that the extraction is completed. The technical solution of the embodiment of the present invention solves the problem that the method based on small models in the existing information extraction technology is not ideal for zero samples or few samples, and there are defects in the migration of model scenarios. It is based on the pre-designed dialogue process, which requires the intervention of development or operation personnel, adding a lot of extra costs to obtain problems, and the use of defined processes and counter-questioning methods is not user-friendly enough and not flexible enough. First, the scenario is confirmed to determine the slot information that needs to be extracted. Secondly, the complex semantic understanding ability of the big model is used to extract information and self-reflect to ensure the completeness of the extracted content. Then, the big model's self-questioning and self-answering function is used to automatically ask flexible questions based on the vacant slots to ensure the flexibility and friendliness of the conversation. Finally, the entire extraction process is completed and returned in a structured form.

[0090] Optionally, the slot determination module 310 is specifically configured to:

[0091] Determine an information extraction scenario, and determine target extraction slot description information matching the information extraction scenario based on a mapping relationship between the information extraction scenario and the extraction slot description information;

[0092] The slot to be extracted is determined based on the target extraction slot description information.

[0093] Optionally, the extraction and reflection module 320 is specifically used for:

[0094] Generate first guidance prompt information based on the slot to be extracted and the problem to be processed, and guide the large model to extract the information corresponding to the slot to be extracted in the problem to be processed in a self-reflective mode based on the first guidance prompt information to obtain the extracted information, and determine whether the information corresponding to the slot to be extracted is completely extracted. Optionally, the extraction reflection module 320 includes:

[0095] The reflection result acquisition submodule is used to splice the slot to be extracted as a system prompt word with the user description content, and input it as the first guiding prompt information to the large model, so that the large model can obtain the self-reflection result through decoding.

[0096] Optionally, the reflection result acquisition submodule is specifically used for: the large model determines the self-reflection result through decoding, wherein the self-reflection result is: determining to continue slot extraction and generate rhetorical questions, or determining that information collection is completed.

[0097] Optionally, the self-questioning and self-answering module 330 includes:

[0098] a rhetorical question generation submodule, configured to determine an undrawn slot based on the drawn slot and the slot to be drawn, and generate a second guidance prompt information based on the undrawn slot, the drawn information and the first guidance prompt information;

[0099] Prompt the large model to generate a rhetorical question based on the second guidance prompt information.

[0100] Optionally, the rhetorical question generation submodule is specifically used to: input scenario information, user description and self-reflection results into the large model;

[0101] The large model gradually generates questions for self-questioning, and the large model answers itself according to the self-questioned questions to generate counter-questions;

[0102] Wherein, the rhetorical question is used to prompt the user to supplement the information;

[0103] If the information collection is complete, rhetorical questions are used to prompt the user to confirm whether the extracted information is accurate.

[0104] Optionally, the self-questioning and self-answering module 330 is specifically used for:

[0105] When the extraction progress is that the extraction is completed, generating extraction information confirmation questions based on the extracted information;

[0106] Receive the user's answer to the question to confirm the extracted information and the extracted information, and generate the answer to the question to be processed.

[0107] The big model-based conversation information extraction device provided in the embodiment of the present invention can execute the big model-based conversation information extraction method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0108] Embodiment 4

[0109] Figure 4 A schematic diagram of the structure of an electronic device provided for Embodiment 4 of the present invention. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0110] like Figure 4 As shown, the electronic device 10 includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0111] A number of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0112] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as the dialogue information extraction method based on a large model.

[0113] In some embodiments, the conversation information extraction method based on the large model can be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the conversation information extraction method based on the large model described above can be performed. Alternatively, in other embodiments, the processor 11 can be configured to execute the conversation information extraction method based on the large model in any other appropriate manner (for example, by means of firmware).

[0114] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0115] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0116] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0117] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0118] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0119] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0120] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0121] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention. < / user> < / system> < / user> < / system>

Claims

1. A method for extracting conversation information based on a large model, characterized in that: include: Determine an information extraction scenario, and determine a slot to be extracted based on the information extraction scenario; According to the slot to be extracted, extract the slot information of the problem to be processed to obtain the extracted information and the extraction progress; When the extraction progress is that the extraction is incomplete, a counter-question is generated. When the extraction progress is that the extraction is completed, an answer result of the pending question is generated based on the extracted information.

2. The method according to claim 1, characterized in that The determining of the information extraction scenario and determining the slot to be extracted based on the information extraction scenario includes: Determine an information extraction scenario, and determine target extraction slot description information matching the information extraction scenario based on a mapping relationship between the information extraction scenario and the extraction slot description information; The slot to be extracted is determined based on the target extraction slot description information.

3. The method according to claim 1, characterized in that The extracting slot information of the problem to be processed according to the slot to be extracted to obtain the extracted information includes: A first guidance prompt information is generated based on the slot to be extracted and the problem to be processed, and based on the first guidance prompt information, the large model is guided to extract the information corresponding to the slot to be extracted in the problem to be processed in a self-reflective mode to obtain the extracted information, and determine whether the information corresponding to the slot to be extracted is completely extracted.

4. The method according to claim 3, characterized in that The generating first guidance prompt information based on the slot to be extracted and the problem to be processed includes: The slot to be extracted is spliced ​​with the user description content as a system prompt word and input into the large model as the first guiding prompt information, so that the large model obtains the self-reflection result through decoding.

5. The method according to claim 4, characterized in that The large model obtains self-reflection results through decoding, including: The large model determines the self-reflection result through decoding, wherein the self-reflection result is: determining to continue slot extraction and generate rhetorical questions, or determining that information collection is completed.

6. The method according to claim 5, characterized in that When the extraction progress is that the extraction is not completed, generating rhetorical questions includes: Determine an undrawn slot based on the drawn slot and the slot to be drawn, and generate second guidance prompt information based on the undrawn slot, the drawn information and the first guidance prompt information; Based on the second guidance prompt information, the large model is prompted to generate a rhetorical question, Specifically include: Inputting scenario information, user descriptions and self-reflection results into the large model; The large model gradually generates questions for self-questioning, and the large model answers itself according to the self-questioned questions to generate counter-questions; Wherein, the rhetorical question is used to prompt the user to supplement the information; If the information collection is complete, rhetorical questions are used to prompt the user to confirm whether the extracted information is accurate.

7. The method according to claim 1, characterized in that When the extraction progress is complete, generating an answer result of the pending question based on the extracted information includes: When the extraction progress is that the extraction is completed, generating extraction information confirmation questions based on the extracted information; Receive the user's answer result to the extraction information confirmation question and the extracted information, and generate the answer result of the pending question.

8. A conversation information extraction device based on a large model, characterized in that: include: A slot determination module, used to determine an information extraction scenario, and determine a slot to be extracted based on the information extraction scenario; An extraction reflection module is used to extract slot information of the problem to be processed according to the slot to be extracted, and obtain the extracted information and extraction progress; The self-questioning and self-answering module is used to generate a counter-question when the extraction progress is that the extraction is incomplete, and to generate an answer result of the pending question based on the extracted information when the extraction progress is that the extraction is completed.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the large model-based dialogue information extraction method described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the large model-based dialogue information extraction method described in any one of claims 1 to 7 when executed.

Images