Information service method and system based on agent causal relationship reasoning

Abstract

The invention discloses an information service method and system based on agent causal relationship reasoning, and belongs to the technical field of artificial intelligence, and the method comprises the following steps: constructing a relationship extraction agent, carrying out entity extraction on an information file in a field, extracting entity attributes, and constructing an entity library; according to the entity library information, basic relationship extraction is carried out, and a basic relationship library is established; a causal reasoning agent is constructed, reasoning mining is carried out according to entity attributes and relation information in the basic relation library, and a causal relation library is established; carrying out acceleration processing on the agent base model by utilizing a model training reasoning acceleration technology based on matrix low-rank decomposition; a periodic business inspection service is provided; and a precise question and answer type feedback service is provided. According to the invention, causal reasoning and information mining are completed by using an intelligent means, and personalized services are provided for different business scene demands.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an information service method and system based on intelligent agent causal relationship reasoning. Background Art

[0002] With the deepening digital transformation of business systems, various business systems have shifted from a paper-based, heavily manual office model to a digital model based on computers and the internet. Information carriers have evolved from paper documents to a variety of data. Extracting useful information from this massive amount of data for operational purposes is a long-term goal for all business systems. In recent years, with the explosive growth of artificial intelligence technology, utilizing intelligent methods to mine data and improve the efficiency of business systems has become a key focus of technical research in the information services sector. Summary of the Invention

[0003] The technical task of the present invention is to address the above shortcomings and provide an information service method and system based on intelligent agent causal reasoning, so as to realize causal reasoning and information mining by intelligent means, and provide personalized services for different business scenarios.

[0004] The technical solution adopted by the present invention to solve its technical problem is: An information service method based on agent causal reasoning, the implementation of which includes the following steps: Step 1: Build a relationship extraction agent to extract entities from information files within the domain, extract entity attributes, and build an entity library; Step 2: Extract basic relationships based on the entity library information and establish a basic relationship library; Step 3: Build a causal reasoning agent, perform reasoning and mining based on entity attributes and relationship information in the basic relationship library, and establish a causal relationship library; Step 4: Use model training and inference acceleration technology based on matrix low-rank decomposition to accelerate the processing of the intelligent agent base model to reduce the dependence on computing resources during the model inference process; Step 5: Provide periodic business inspection services. Set inspection subjects and indicators based on routine inspection requirements in the business field, build an inspection agent, and conduct periodic inspections on the causal relationship database. Step 6: Provide accurate question-answering feedback services, build a question-answering agent, receive targeted needs input by users, query the causal relationship library, and perform reasoning to generate conclusions.

[0005] This method constructs an intelligent agent with customized functions to extract basic relationships and causal relationship reasoning from information files in the domain, builds an information database that meets the causal relationships in the business domain, and provides periodic business inspection services and precise question-and-answer feedback services. It realizes the use of intelligent means to complete causal reasoning and information mining, and provides personalized services for different business scenarios by building an intelligent agent with business-specific service capabilities.

[0006] Furthermore, in step 1, before entity extraction, entity recognition rules in the field are input to the agent by inputting prompt words, and the scope of entity attribute extraction is standardized to ensure the accuracy of entity recognition and business matching.

[0007] Furthermore, in step 2, during the basic relationship extraction process, full-scale relationship extraction is performed based on the entities and entity attributes in the entity library, or relationship extraction is performed in a targeted manner according to business needs.

[0008] Furthermore, in step 3, the causal relationship types include explicit causal relationships and implicit causal relationships. Explicit causal relationships represent associations that can be inferred from explicit relationships in an existing basic relationship library. Explicit causal relationships are mined through multi-layer direct association mining or through statistical mining of quantitative indicators. This capability can be achieved by simply building a training corpus from a certain number of business cases and fine-tuning the base model that drives the causal reasoning agent. Implicit causal relationships refer to further exploring overlapping relationships involving time and space, in addition to considering multi-layer direct correlations. To establish such capabilities, in addition to fine-tuning the base model that drives the causal reasoning agent through case studies, it is also necessary to develop workflows and standardize the direction and focus of the agent's information mining to meet business needs.

[0009] Furthermore, in step 4, the agent base model is accelerated using the model training and inference acceleration technology based on matrix low-rank decomposition. The specific steps are as follows: Step 4.1: Large Dimensional Matrix from Base Model Evenly selected Columns form a matrix ; Step 4.2: From Evenly selected Rows form a matrix ; Step 4.3: Perform singular value decomposition and determine the truncation rank of the matrix based on the truncation error ,get ,if ,Will Increase to , repeat steps 4.1 and 4.2; Step 4.4: From Evenly selected Rows form a matrix ; Step 4.5: The formed matrix and ,satisfy ,in , .

[0010] Furthermore, in step 5, the periodic service inspection service is specifically: The inspection subjects and inspection indicators are input into the inspection agent in the form of prompt words. The inspection agent performs periodic inspections on the causal relationship library according to the inspection rules. If target information appears, an alarm prompt and visual display will be issued. The inspection process and the causal relationship library construction process are in parallel. During the inspection process, the content of the causal relationship library can be updated in real time according to the changes in the information files.

[0011] Furthermore, in step 6, the precise question-and-answer feedback service is specifically as follows: The question-answering agent queries the causal relationship library based on the targeted needs of the user input and generates conclusions and visual displays through reasoning.

[0012] The present invention further claims protection for an information service system based on agent causal reasoning, comprising: The entity library is used to complete the entity extraction function and collect and store the extracted entities and entity attributes; The basic relationship library is used to extract basic entity relationships and store the extracted basic relationships; The causal relationship library is used to implement reasoning and mining based on entity attributes and relationship information in the basic relationship library to obtain causal relationships; The agent suite is used to build customized agents based on system requirements, including relationship extraction agents, causal reasoning agents, inspection agents, and question-answering agents. The model acceleration module uses model training and inference acceleration technology based on matrix low-rank decomposition to accelerate the processing of the intelligent agent base model, achieving compressed decomposition of large-dimensional matrices in the model to reduce the dependence on computing power resources during the model inference process; The application service module is used to provide periodic business inspection services and precise question-and-answer feedback services based on set requirements; The system realizes causal reasoning and information services through the above methods.

[0013] The present invention also claims protection for an information service device based on agent causal reasoning, characterized in that it comprises: at least one memory and at least one processor; The at least one memory is configured to store a machine-readable program; The at least one processor is configured to call the machine-readable program to implement the above method.

[0014] The present invention also claims protection for a computer-readable medium, characterized in that the computer-readable medium stores computer instructions, and when the computer instructions are executed by a processor, the above method can be implemented.

[0015] Compared with the prior art, the information service method and system based on agent causal reasoning of the present invention has the following beneficial effects: The present invention constructs customized agents with business attributes, such as relationship extraction agents and causal reasoning agents, to extract basic relationships and conduct causal reasoning on information files within the domain. It then sequentially constructs a basic relationship library and a causal library to form an information library that satisfies the causal logical relationships within the business domain. Inspection agents and question-and-answer agents are also constructed to provide periodic business inspection services and precise question-and-answer feedback services, automating the output of results and visualizing them. This creates an information service system based on agent-based causal reasoning, enabling the use of intelligent means to complete causal reasoning and information mining, and providing personalized services tailored to the needs of different business scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a flowchart of an information service method based on agent causal reasoning provided by one embodiment of the present invention; Figure 2 This is a schematic diagram of a periodic service inspection service provided by an embodiment of the present invention; Figure 3 This is a schematic diagram of a precise question-and-answer feedback service provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0017] The present invention will be further described below with reference to specific embodiments.

[0018] The present invention provides an information service method based on agent-based causal reasoning. By constructing customized agents with business attributes, such as a relationship extraction agent and a causal reasoning agent, the method extracts basic relationships and performs causal reasoning on information files within a domain. A basic relationship library and a causal library are then constructed, forming an information library that satisfies the causal relationships within the business domain. Furthermore, an inspection agent and a question-and-answer agent are constructed to provide periodic business inspection services and precise question-and-answer feedback services. Results are automatically output and visualized, enabling intelligent causal reasoning and information mining. Furthermore, personalized services are provided for different business scenarios.

[0019] like Figure 1 As shown, first, a relationship extraction agent is constructed, an entity library is built, and basic relationship extraction is performed to establish a basic relationship library; then a causal reasoning agent is constructed to establish a causal relationship library; then, inspection subjects and indicators are set according to the routine inspection requirements of the business field, and an inspection agent is constructed to perform periodic inspections on the causal relationship library and provide periodic business inspection services; a question-and-answer agent is constructed to receive targeted needs input by users, query the causal relationship library and perform reasoning to generate conclusions, and provide accurate question-and-answer feedback services.

[0020] The specific steps to implement this method are as follows: Step 1: Build a relationship extraction agent to extract entities from information files within the domain, extract entity attributes, and build an entity library. Before entity extraction, you need to input the entity identification rules within the domain into the agent by entering prompt words, and standardize the scope of entity attribute extraction to ensure the accuracy of entity identification and business matching. For example, enter prompt words into the relationship extraction agent: Please extract entities with "person (name)", "account (card number)", "property (address)", "vehicle (license plate number)", "telephone (number)", and "unit (name)" as the main objects, extract the corresponding entity attributes, and build an entity library.

[0021] Step 2: The relationship extraction agent extracts basic relationships based on the entity library information and establishes a basic relationship library. During the relationship extraction process, full-scale relationship extraction can be performed based on the entities and entity attributes in the entity library, or relationship extraction can be performed in a targeted manner according to business needs. For example, by inputting the prompt "relationship extraction centered on the entity 'person'" into the relationship extraction agent, it can extract relationships such as "person-to-person," "person-to-account," "person-to-property," "person-to-car," "person-to-telephone," and "person-to-organization" from the basic relationship library to establish a basic relationship library.

[0022] Step 3: Build a causal reasoning agent to perform reasoning and mining based on entity attributes and relationship information in the basic relationship library to establish a causal relationship library. When extracting causal relationships, it is necessary to distinguish between explicit causal relationships and implicit causal relationships.

[0023] Explicit causal relationships represent associations that can be inferred based on explicit relationships in the existing basic relationship library. This relationship can be mined through multi-layer direct association mining or through statistical mining of quantitative indicators. This capability can be achieved by establishing a certain number of business cases to build a training corpus and fine-tuning the base model that drives the causal reasoning agent.

[0024] Implicit causal relationships require not only considering multiple layers of direct associations but also overlapping relationships in time and space. Building this capability requires not only case-based fine-tuning of the base model driving the causal reasoning agent, but also the development of workflows to standardize the direction and focus of information mining to meet business needs. For example, to determine whether two people have met at a specific location, a workflow is established: "Determine whether the two people appeared in or arrived in a city during the same time period - query their accommodation and consumption records in that city to determine whether they appeared on the same street at the same time - query whether the two people had any phone calls or other connections during this period - analyze the mined information and output conclusions, while visualizing all mined relationships through graphs."

[0025] Step 4: Use the model training and inference acceleration technology based on matrix low-rank decomposition to accelerate the processing of the intelligent agent base model, and realize the compression decomposition of large-dimensional matrices in the model to reduce the dependence on computing power resources during the model inference process. The specific steps are as follows: Step 4.1: Large Dimensional Matrix from Base Model Evenly selected Columns form a matrix ; Step 4.2: From Evenly selected Rows form a matrix ; Step 4.3: Perform singular value decomposition and determine the truncation rank of the matrix based on the truncation error ,get ,if ,Will Increase to , repeat steps 4.1 and 4.2; Step 4.4: From Evenly selected Rows form a matrix ; Step 4.5: The formed matrix and ,satisfy ,in , .

[0026] Step 5: Provide periodic business inspection services, set inspection subjects and indicators according to the routine inspection requirements of the business field, build an inspection intelligent body, and conduct periodic inspections on the causal relationship library. Figure 2 As shown, inspection subjects and indicators are input into the inspection agent via prompts. The inspection agent then periodically inspects the causal relationship library according to the inspection rules. If target information appears, an alarm is generated and a visual display is provided. The inspection process and the causal relationship library construction process run in parallel. During the inspection process, the causal relationship library content can be updated in real time based on changes in the information file.

[0027] Step 6: Provide accurate question-answering feedback services, build a question-answering agent, receive targeted needs input by users, query the causal relationship library and perform reasoning to generate conclusions. Figure 3 As shown in the figure, the question-answering agent queries the causal relationship library based on the targeted needs of the user input and generates conclusions and visual displays through reasoning.

[0028] An embodiment of the present invention further provides an information service system based on agent causal reasoning, comprising: The entity library is used to complete the entity extraction function and collect and store the extracted entities and entity attributes; The basic relationship library is used to extract basic entity relationships and store the extracted basic relationships; The causal relationship library is used to implement reasoning and mining based on entity attributes and relationship information in the basic relationship library to obtain causal relationships; The agent suite is used to build customized agents based on system requirements, including relationship extraction agents, causal reasoning agents, inspection agents, and question-answering agents. The model acceleration module uses model training and inference acceleration technology based on matrix low-rank decomposition to accelerate the processing of the intelligent agent base model, achieving compressed decomposition of large-dimensional matrices in the model to reduce the dependence on computing power resources during the model inference process; The application service module is used to provide periodic business inspection services and precise question-and-answer feedback services based on set requirements; The system implements causal relationship reasoning and information service through the information service method based on agent causal relationship reasoning described in the above embodiment.

[0029] First, a relationship extraction agent is constructed to extract entities from information files within the domain, extract entity attributes, and construct an entity library. Then, based on the entity library information, basic relationships are extracted to establish a basic relationship library. Then, a causal reasoning agent is constructed to perform reasoning and mining based on entity attributes and relationship information in the basic relationship library, extract causal relationships, integrate basic relationship information, and establish a causal relationship library. Information service methods are divided into two types: periodic business inspection service and precise question-and-answer feedback service. The periodic business inspection service is used by the business system to set inspection subjects and indicators based on routine inspection requirements in the business field. The inspection agent drives periodic inspections of the causal relationship library. If target information appears, an alarm prompt and visual display are issued. The precise question-and-answer feedback service is used in business scenarios. When receiving targeted demand input, the question-and-answer agent queries the causal relationship library and generates conclusions and visual displays through reasoning.

[0030] The specific steps for the system to implement information services are as follows: Step 1: Build a relationship extraction agent to extract entities from domain information documents, extract entity attributes, and build an entity library. Before entity extraction, the agent must input entity identification rules within the domain by entering prompt words and standardize the scope of entity attribute extraction to ensure the accuracy of entity identification and business matching.

[0031] Step 2: The relationship extraction agent extracts basic relationships based on the entity library information and establishes a basic relationship library. During the relationship extraction process, full-scale relationship extraction can be performed based on the entities and entity attributes in the entity library, or relationship extraction can be performed in a targeted manner according to business needs. For example, by inputting the prompt "relationship extraction centered on the entity 'person'" into the relationship extraction agent, it can extract relationships such as "person-to-person," "person-to-account," "person-to-property," "person-to-car," "person-to-telephone," and "person-to-organization" from the basic relationship library to establish a basic relationship library.

[0032] Step 3: Build a causal reasoning agent to perform reasoning and mining based on entity attributes and relationship information in the basic relationship library to establish a causal relationship library. When extracting causal relationships, it is necessary to distinguish between explicit causal relationships and implicit causal relationships.

[0033] Explicit causal relationships represent associations that can be inferred based on explicit relationships in the existing basic relationship library. This relationship can be mined through multi-layer direct association mining or through statistical mining of quantitative indicators. This capability can be achieved by establishing a certain number of business cases to build a training corpus and fine-tuning the base model that drives the causal reasoning agent.

[0034] Implicit causal relationships mean that in addition to considering multi-layer direct correlations, the overlapping relationships in the time and space dimensions must also be considered. To establish such capabilities, in addition to case-based fine-tuning training of the base model that drives the causal reasoning agent, it is also necessary to develop workflows and standardize the direction and focus of the agent's information mining to meet business needs.

[0035] Step 4: Use the model training and inference acceleration technology based on matrix low-rank decomposition to accelerate the processing of the intelligent agent base model, and realize the compression decomposition of large-dimensional matrices in the model to reduce the dependence on computing power resources during the model inference process. The specific steps are as follows: Step 4.1: Large Dimensional Matrix from Base Model Evenly selected Columns form a matrix ; Step 4.2: From Evenly selected Rows form a matrix ; Step 4.3: Perform singular value decomposition and determine the truncation rank of the matrix based on the truncation error ,get ,if ,Will Increase to , repeat steps 4.1 and 4.2; Step 4.4: From Evenly selected Rows form a matrix ; Step 4.5: The formed matrix and ,satisfy ,in , .

[0036] Step 5: Provide periodic business inspection services, set inspection subjects and indicators according to the routine inspection requirements of the business field, build an inspection intelligent body, and conduct periodic inspections on the causal relationship library. Figure 2 As shown, inspection subjects and indicators are input into the inspection agent via prompts. The inspection agent then periodically inspects the causal relationship library according to the inspection rules. If target information appears, an alarm is generated and a visual display is provided. The inspection process and the causal relationship library construction process run in parallel. During the inspection process, the causal relationship library content can be updated in real time based on changes in the information file.

[0037] Step 6: Provide accurate question-answering feedback services, build a question-answering agent, receive targeted needs input by users, query the causal relationship library and perform reasoning to generate conclusions. Figure 3As shown in the figure, the question-answering agent queries the causal relationship library based on the targeted needs of the user input and generates conclusions and visual displays through reasoning.

[0038] An embodiment of the present invention further provides an information service device based on agent causal reasoning, characterized in that it includes: at least one memory and at least one processor; The at least one memory is configured to store a machine-readable program; The at least one processor is used to call the machine-readable program to implement the information service method based on agent causal reasoning described in the above embodiment.

[0039] Embodiments of the present invention further provide a computer-readable medium storing computer instructions that, when executed by a processor, implement the information service method based on agent-based causal reasoning described in the above embodiments. Specifically, a system or device equipped with a storage medium can be provided. The storage medium stores software program code that implements the functions of any of the above embodiments, and the computer (or CPU or MPU) of the system or device can read and execute the program code stored in the storage medium.

[0040] In this case, the program code itself read from the storage medium can realize the function of any one of the above-mentioned embodiments, and thus the program code and the storage medium storing the program code constitute part of the present invention.

[0041] Examples of storage media for providing program code include floppy disks, hard disks, magneto-optical disks, optical disks (e.g., CD-ROMs, CD-Rs, CD-RWs, DVD-ROMs, DVD-RAMs, DVD-RWs, and DVD+RWs), magnetic tapes, non-volatile memory cards, and ROMs. Alternatively, the program code may be downloaded from a server computer via a communications network.

[0042] In addition, it should be clear that the functions of any of the above embodiments can be achieved not only by executing the program code read by the computer, but also by enabling the operating system operating on the computer to complete part or all of the actual operations based on the instructions of the program code.

[0043] In addition, it can be understood that the program code read from the storage medium is written into the memory provided in the expansion board inserted into the computer or into the memory provided in the expansion unit connected to the computer, and then based on the instructions of the program code, the CPU installed on the expansion board or expansion unit is enabled to perform part or all of the actual operations, thereby realizing the functions of any of the above embodiments.

[0044] The present invention has been shown and described in detail above through the accompanying drawings and preferred embodiments. However, the present invention is not limited to these disclosed embodiments. Based on the above multiple embodiments, those skilled in the art can know that the code review methods in the above different embodiments can be combined to obtain more embodiments of the present invention, and these embodiments are also within the scope of protection of the present invention.

Claims

1. An information service method based on agent causal reasoning, characterized in that: The implementation of this method includes the following steps: Step 1: Build a relationship extraction agent to extract entities from information files within the domain, extract entity attributes, and build an entity library; Step 2: Extract basic relationships based on the entity library information and establish a basic relationship library; Step 3: Construct a causal reasoning agent to perform reasoning and mining based on entity attributes and relationship information in the basic relationship library to establish a causal relationship library; Step 4: Use model training and inference acceleration technology based on matrix low-rank decomposition to accelerate the processing of the intelligent agent base model to reduce the dependence on computing resources during the model inference process; Step 5: Provide periodic business inspection services. Set inspection subjects and indicators based on routine inspection requirements in the business field, build an inspection agent, and conduct periodic inspections on the causal relationship database. Step 6: Provide accurate question-answering feedback services, build a question-answering agent, receive targeted needs input by users, query the causal relationship library, and perform reasoning to generate conclusions.

2. The information service method based on agent causal reasoning according to claim 1, characterized in that: In step 1, before entity extraction, entity recognition rules in the field are input to the agent by inputting prompt words, and the scope of entity attribute extraction is standardized to ensure the accuracy of entity recognition and business matching.

3. The information service method based on agent causal reasoning according to claim 1, characterized in that: In step 2, during the basic relationship extraction process, full-scale relationship extraction is performed based on the entities and entity attributes in the entity library, or relationship extraction is performed in a targeted manner according to business needs.

4. The information service method based on agent causal reasoning according to claim 1, characterized in that: In step 3, the causal relationship types include explicit causal relationships and implicit causal relationships. Explicit causal relationships refer to the association relationships that can be inferred based on the explicit relationships in the existing basic relationship database. The mining of explicit causal relationships is obtained through multi-layer direct association mining or through statistical mining of quantitative indicators. Implicit causal relationships refer to overlapping relationships that are further explored, including time and space dimensions, in addition to multiple layers of direct correlations.

5. The information service method based on agent causal reasoning according to claim 1, characterized in that: In step 4, the agent base model is accelerated using the model training and inference acceleration technology based on matrix low-rank decomposition. The specific steps are as follows: Step 4.1: Large Dimensional Matrix from Base Model Evenly selected Columns form a matrix ; Step 4.2: From Evenly selected Rows form a matrix ; Step 4.3: Perform singular value decomposition and determine the truncation rank of the matrix based on the truncation error ,get ,if ,Will Increase to , repeat steps 4.1 and 4.2; Step 4.4: From Evenly selected Rows form a matrix ; Step 4.5: The formed matrix and ,satisfy ,in , .

6. The information service method based on agent causal reasoning according to claim 1, characterized in that: In step 5, the periodic service inspection service is specifically as follows: The inspection subjects and inspection indicators are input into the inspection agent in the form of prompt words. The inspection agent performs periodic inspections on the causal relationship library according to the inspection rules. If target information appears, an alarm prompt and visual display will be issued. The inspection process and the causal relationship library construction process are in parallel. During the inspection process, the content of the causal relationship library is updated in real time according to the changes in the information files.

7. The information service method based on agent causal reasoning according to claim 1, characterized in that: In step 6, the precise question-and-answer feedback service is specifically as follows: The question-answering agent queries the causal relationship library based on the targeted needs of the user input and generates conclusions and visual displays through reasoning.

8. An information service system based on agent causal reasoning, characterized in that: include: The entity library is used to complete the entity extraction function and collect and store the extracted entities and entity attributes; The basic relationship library is used to extract basic entity relationships and store the extracted basic relationships; The causal relationship library is used to implement reasoning and mining based on entity attributes and relationship information in the basic relationship library to obtain causal relationships; The agent suite is used to build customized agents based on system requirements, including relationship extraction agents, causal reasoning agents, inspection agents, and question-answering agents. The model acceleration module uses model training and inference acceleration technology based on matrix low-rank decomposition to accelerate the processing of the intelligent agent base model, achieving compressed decomposition of large-dimensional matrices in the model to reduce the dependence on computing power resources during the model inference process; The application service module is used to provide periodic business inspection services and precise question-and-answer feedback services based on set requirements; The system realizes causal reasoning and information services through the method described in any one of claims 1 to 7.

9. An information service device based on agent causal reasoning, characterized in that: include: at least one memory and at least one processor; The at least one memory is configured to store a machine-readable program; The at least one processor is configured to call the machine-readable program to implement the method according to any one of claims 1 to 7.

10. A computer-readable medium, characterized in that The computer readable medium stores computer instructions, which, when executed by a processor, can implement the method according to any one of claims 1 to 7.

Images