Intelligent discovery system and method for providing technical problem joint attack path

Abstract

The invention discloses an intelligent discovery system for providing a joint technical problem attack path, and the system comprises a data collection subsystem which collects scientific and technological achievement data to form a database; the data governance subsystem is used for governing the data to form standardized available data and storing the standardized available data in different modes to construct a data base, and the data governance subsystem comprises a model fine adjustment module which is used for performing fine adjustment on a large language model based on the collected data so as to enable the large language model to recognize technical problems from questions of users and generate corresponding technical overview; and an application subsystem configured to process user questions using the first trimmed large language model to identify technical questions; mining a technical text from a scientific and technological achievement database based on the identified technical problem; identifying talents and institutions associated with the mined technical text; portraying talents and organizations; determining Chinese and foreign technical differences based on the technical text and the portrait; and generating a joint attack path based on the difference and the portrait. A corresponding method is also disclosed.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence applications, and more specifically, to an intelligent discovery system and method for providing a joint solution to technical problems. Background Art

[0003] First of all, it is not easy to clarify the underlying technical problems that need to be solved in a technical difficulty, and it often takes a lot of time to conduct research.

[0004] Secondly, even if the technical problem is found, it will take a lot of time to find talents and institutions capable of solving the technical problem.

[0005] Third, even if we know the relevant talents and institutions, it is still very difficult to organize them effectively to form the most appropriate joint research path.

[0006] Therefore, there is an urgent need in the art for an intelligent discovery system and method that provides a joint solution path for technical problems. The system and method can solve the above problems existing in the prior art, help relevant innovation entities or science and technology management agencies to easily and intuitively analyze the technical problems behind technical difficulties, and efficiently find a joint solution path for the technical problems to be solved. Summary of the invention

[0007] A brief summary of one or more aspects is given below to provide a basic understanding of these aspects. This summary is not an exhaustive overview of all conceived aspects, and is neither intended to identify the key or critical elements of all aspects nor to define the scope of any or all aspects. Its only purpose is to give some concepts of one or more aspects in a simplified form as a prelude to a more detailed description that will be given later.

[0008] According to a first aspect of the present application, there is provided an intelligent discovery system for providing a joint solution path for technical problems, comprising:

[0009] A data collection subsystem is used to collect scientific and technological achievement data from data sources to form a scientific and technological achievement database;

[0010] The data governance subsystem is used to manage the data in the scientific and technological achievements database to form standardized usable data, and store it in different ways to build the data base of the intelligent discovery system, wherein the data governance subsystem includes:

[0011] A model fine-tuning module, used to fine-tune the large language model based on the collected data, so that the large language model can identify technical problems from user questions and generate corresponding technical summaries; and

[0012] The application subsystem is configured as follows:

[0013] Processing the user question using the first fine-tuned large language model to identify a technical problem;

[0014] Based on the identified technical problems, mining technical texts from the scientific and technological achievements database;

[0015] Identify the people and institutions associated with the mined technical texts;

[0016] Profiling the identified talents and institutions;

[0017] Determine the differences between Chinese and foreign technologies based on the mined technical texts and the portrait; and

[0018] The joint research path is generated based on the determined technological differences between China and foreign countries and the portrait.

[0019] According to a preferred embodiment of the present application, the data base comprises:

[0020] A search engine for providing a search function for at least one of the following data: collected data, talent evaluation and profile data, and institution evaluation and profile data;

[0021] A graph database, for storing at least one of the following relationship graph data: a technical subject relationship graph, a talent relationship graph, an institution relationship graph, a talent-institution relationship graph, and a scientific research achievement relationship graph;

[0022] A vector database, used to store the result data after large-model vector encoding of the collected data; and

[0023] A relational database used to store common configuration data.

[0024] According to a preferred embodiment of the present application, the data governance subsystem further includes:

[0025] Name disambiguation module, which disambiguates names based on the natural language pre-training model and the external features of the collected data;

[0026] An organization tree construction module is used to extract organizations from the collected data and construct an organization tree;

[0027] Talent research capability evaluation module, used to evaluate the research capabilities of talents in related industries; and

[0028] The module for evaluating the innovation capability of an institution is used to evaluate the innovation capability of the institution based on the collected data.

[0029] According to a preferred embodiment of the present application, evaluating the scientific research capabilities of talents in related industries includes:

[0030] Construct a talent research capability evaluation model based on the data iteration generation process;

[0031] Identify and eliminate invalid citation relationships between talents;

[0032] Assign corresponding citation relationship weights to the cited persons based on the innovation ability of the citer; and

[0033] Evaluate the scientific research ability of the talent based on the citation relationship weight.

[0034] According to a preferred embodiment of the present application, the talent scientific research ability evaluation module also compares and fine-tunes the output results of the talent scientific research ability evaluation model with the expert review results to further calibrate the talent scientific research ability evaluation model.

[0035] According to a preferred embodiment of the present application, the innovative capabilities of the organization evaluating the collected data include:

[0036] The innovation capability of an organization is evaluated based on a multi-dimensional institutional innovation evaluation model, which includes one or more of the following dimensions: innovation hard power, market connection capability, innovation resource integration capability, and innovation team growth capability.

[0037] According to a second aspect of the present application, a method for providing a joint solution path for technical problems using the above-mentioned intelligent discovery system is provided, and the method comprises the following steps:

[0038] Receive user questions;

[0039] Processing the user question using the first fine-tuned large language model to identify a technical problem;

[0040] Based on the identified technical problems, mining technical texts from the scientific and technological achievements database;

[0041] Identify the people and institutions associated with the mined technical texts;

[0042] Profiling the identified talents and institutions;

[0043] Determine the differences between Chinese and foreign technologies based on the identified technical texts and the portrait; and

[0044] Based on the identified technological differences between China and foreign countries and this portrait, a joint research path is generated.

[0045] According to a preferred embodiment of the present application, based on the identified technical problem, mining the technical text further includes:

[0046] Based on the identified technical problems, screening relevant scientific and technological achievement data from the scientific and technological achievement database to form a solution data set; and

[0047] The data in the solution dataset is read using a second fine-tuned large language model to generate an overview of the current status of solving the technical problem and identify subdivided technical areas of the technical problem.

[0048] According to a preferred embodiment of the present application, determining the difference between Chinese and foreign technologies based on the identified technical text and the portrait further includes:

[0049] The corresponding visualization chart is automatically generated based on the recognized technical text and the portrait.

[0050] According to a preferred embodiment of the present application, generating the joint research path based on the determined Chinese and foreign technology differences and the portrait further includes:

[0051] Determine the technical problems that need to be overcome based on the identified technical differences between China and foreign countries;

[0052] Combine talents and institutions associated with the identified technical problems to be solved to form multiple candidate joint research paths; and

[0053] Based on the portrait, a comparative analysis is performed on each of the multiple candidate joint research paths to determine the recommended research path.

[0054] According to a preferred embodiment of the present application, the application subsystem is further configured to:

[0055] A tackling plan is generated based on the recommended tackling path.

[0056] According to a preferred embodiment of the present application, generating a tackling solution based on the recommended tackling path further includes:

[0057] According to the talent and institution information at each node on the recommended research path, extract the corresponding scientific research results data;

[0058] Conduct statistical analysis on various scientific research results based on the extracted scientific research results data;

[0059] Ask the second fine-tuned large language model questions using a predetermined set of questions for building a solution;

[0060] Using the second fine-tuned large language model to read the scientific research results text in the extracted scientific research results data, and forming an answer set according to the question set; and

[0061] Based on the answer set and in combination with the result of the statistical analysis, the solution is generated.

[0062] To achieve the foregoing and related ends, the one or more aspects include the features fully described below and particularly pointed out in the appended claims. The following description and drawings set forth in detail certain illustrative features of the one or more aspects. However, these features are merely indicative of several of the various ways in which the principles of the various aspects may be employed, and this description is intended to cover all such aspects and their equivalents. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to understand in detail the manner in which the above-stated features of the present application are used, a more specific description of the above briefly summarized contents may be made with reference to various aspects, some of which are illustrated in the accompanying drawings. However, it should be noted that the accompanying drawings illustrate only certain typical aspects of the present application and should not be considered to limit its scope, as the present description may allow for other equally effective aspects.

[0064] In the attached picture:

[0065] Figure 1 is a schematic flow chart illustrating an intelligent discovery method 10 for providing a joint solution path for a technical problem as described herein;

[0066] Figure 2 is a schematic structural block diagram illustrating an intelligent discovery system 20 for providing a joint solution path for technical problems as described herein;

[0067] Figure 3 is a schematic flow chart illustrating data collection by a data collection subsystem 201 of an intelligent discovery system 20 for providing a joint solution path for technical problems as described herein;

[0068] Figure 4 is a schematic flow chart illustrating adapter fine-tuning according to the data governance subsystem 202 of the intelligent discovery system 20 for providing a joint solution path for technical problems described herein;

[0069] Figure 5 Is the explanation Figure 4 A schematic flow chart of name disambiguation performed by the name disambiguation module of the data governance subsystem 202 shown in FIG.

[0070] Figure 6 Is the explanation Figure 4 A schematic flow chart of the organization tree construction module of the data governance subsystem 202 performing organization tree construction as shown in FIG.

[0071] Figure 7a Is the explanation Figure 4 A schematic flow chart of the data governance subsystem 202 performing talent research capability evaluation as shown in FIG.

[0072] Figure 7b Is the explanation Figure 4 A simplified talent reference relationship diagram of the process of evaluating the scientific research ability of talents by the data governance subsystem 202 shown in FIG.

[0073] Figure 8 Is the explanation Figure 4 A schematic diagram of the data governance subsystem 202 for evaluating the innovation capability of an institution as shown in FIG.

[0074] Fig. 9 is a schematic diagram illustrating the structure of a data base for an intelligent discovery system and method for providing a joint solution path for technical problems according to the description of this article;

[0075] Fig.10 is a schematic diagram for explaining technology screening by the application subsystem 203 of the intelligent discovery system 20 for providing a joint solution path for technical problems as described herein;

[0076] Fig.11 is a schematic flow chart illustrating the formation of a technical overview according to the application subsystem 203 of the intelligent discovery system 20 for providing a joint solution path for technical problems described herein;

[0077] Fig.12 is a schematic diagram for explaining the technology comparison to determine the difference between Chinese and foreign technologies according to the application subsystem 203 of the intelligent discovery system 20 for providing a joint solution path for technical problems described herein;

[0078] Fig.13 is a schematic diagram for explaining the acquisition of a technical path by the application subsystem 203 of the intelligent discovery system 20 for providing a joint solution path for technical problems as described herein; and

[0079] Fig.14 It is a schematic diagram for explaining the generation of a solution by the application subsystem 203 of the intelligent discovery system 20 for providing a joint solution path for technical problems as described in this article. DETAILED DESCRIPTION

[0080] The detailed description set forth below in conjunction with the accompanying drawings is intended as a description of various configurations and is not intended to represent the only configurations in which the concepts described herein may be practiced. This detailed description includes specific details to provide a thorough understanding of the various concepts. However, it will be apparent to those skilled in the art that these concepts may be practiced without these specific details. In some instances, well-known components are shown in block diagram form in order to avoid diluting such concepts.

[0081] It is to be understood that other embodiments would be apparent based on the present disclosure, and that system, structural, process, or mechanical changes may be made without departing from the scope of the present disclosure.

[0082] As mentioned above, there are many obstacles in the prior art in finding a solution path based on technical problems. In order to overcome these problems and help technicians and related institutions better find a solution path, the present application provides an intelligent discovery system and method for providing a joint solution path for technical problems.

[0083] See also Figure 1 and Figure 2 , aspects are described with reference to one or more components and one or more methods that can perform the actions or functions described herein. In one aspect, the term "component" as used herein can be one of the parts that make up a system, can be hardware or software or some combination thereof, and can be divided into other components. Although the following Figure 2 The operations described in the present invention are presented in a particular order and / or as being performed by example components, but it should be understood that the order of these actions and the components performing the actions may vary depending on the implementation. In addition, it should be understood that the following actions or functions may be performed by a specially programmed processor, a processor executing specially programmed software or computer-readable media, or by any other combination of hardware components and / or software components capable of performing the described actions or functions.

[0084] Figure 1 1 is a schematic flow chart illustrating an intelligent discovery method 10 for providing a joint solution path for technical problems according to the description of this article. The method 10 can be executed by an intelligent discovery system running on a server. The specific composition and architecture of the intelligent discovery system will be described in detail below. Figure 2 Describe in more detail.

[0085] First, at step 101, a user question is received. The user can ask a user question to the intelligent discovery system. For example, the user can input the information he wants to know about the technical problem he wants to solve into the intelligent discovery system in the form of a question in natural language. For example, what is the core key technology of self-lubricating bearings?

[0086] Then, at step 102, the user's question is processed using the first fine-tuned large language model to identify technical problems. After the intelligent discovery system receives the user's question, the corresponding technical problem is identified from the user's question. The user's natural language question may mention technical difficulties, but it may not be able to clearly describe the technical problem corresponding to the technical difficulty, or the real technical problem at the bottom of the technical difficulty, so it is necessary to identify the real technical problem from the user's natural language. As a non-limiting example, the identification of technical problems can be based on a pre-constructed scientific and technological achievement database (e.g., patents, papers, etc.), and is carried out through a fine-tuned open source large model. For example, by fine-tuning an open source large model (e.g., Google's BERT) to form a fine-tuned model A, and then using the fine-tuned model A to semantically understand the user's question, and convert it into a vector, and compare the similarity with the scientific and technological achievement vector in the scientific and technological achievement database, so as to screen out a matching technical problem solution data set.

[0087] By adjusting the pre-trained model, that is, further improving the model that has learned the general pattern from the large-scale data set in terms of special tasks or data sets, the general large model can be more adapted to the special needs of the field and the performance of the model can be better improved. Then proceed to step 103.

[0088] At step 103, based on the identified technical problem, technical text is mined from the scientific and technological achievements database. As a non-limiting example, this step can also be based on the aforementioned pre-built scientific and technological achievements database, and is performed by another fine-tuned open source large model. For example, by fine-tuning another open source large model (such as Facebook's LLaMA) to form another fine-tuned model B, and then using the fine-tuned model B to read the text in the solution data set screened in step 102, generate subdivided technical fields, and form a current overview of the technical problem. It can be understood that different large language models are used here for fine-tuning in order to adapt to different tasks. Different models have their own types of tasks that they are good at handling. For example, some models are better at semantic understanding and generating dialogues, while other models are better at reading text and summarizing.

[0089] The scientific and technological achievements database includes numerous domestic and foreign technical documents, patents, journals, etc. The number of similar technical texts that have been screened can be adjusted in real time based on user needs, such as 500, 1,000, etc.

[0090] Fine-tuning the open source large model can be performed by using the adapter fine-tuning method and combining it with the transformer mechanism to fine-tune the large language model.

[0091] Next, at step 104 , talents and organizations associated with the mined technical text are identified.

[0092] Talent and institution identification is mainly based on the identified subdivided technical fields, sorting out talents and institutions mainly engaged in the industry at home and abroad. As an example, identification can be achieved by extracting relevant information from the scientific and technological achievements text data in the previously screened solution data set that matches the technical problem, such as the author of the paper, the inventor and applicant of the patent, etc. In addition, as needed, search engines can be used to search external data sources based on this information to further identify the talents and institutions that have made contributions behind the relevant scientific and technological achievements. Optionally, the corresponding talents and institutions can also be disambiguated during the process of talent and institution identification. The disambiguation of talents and institutions will be discussed later in conjunction with Figure 5 and Figure 6 Describe in more detail.

[0093] Next, in step 105, the identified talents and institutions are profiled. The profile includes evaluating the talents and institutions from different dimensions. The specific process of talents and evaluation will be described in detail below. Figure 7a , Figure 7b and Figure 8 The method 10 then proceeds to step 106 .

[0094] At step 106, the difference between Chinese and foreign technologies is determined based on the identified technical text and the portrait. The specific process of determining the difference between Chinese and foreign technologies will be described in detail below. Fig.12 Describe in more detail.

[0095] Optionally, determining the difference between Chinese and foreign technologies based on the identified technical text and the portrait may further include automatically generating corresponding visualization charts based on the identified technical text and the portrait to present the results to the user more intuitively.

[0096] At step 107, a joint research path is generated based on the determined technical differences between China and foreign countries and the portrait. Based on the technical differences between China and foreign countries determined in step 106, a joint research path can be generated for each sub-field that lags behind foreign countries (i.e., a corresponding technical problem that needs to be technically tackled).

[0097] The joint research path is a collection of multiple research paths, each of which is a combination of identified relevant talents and institutions. Optionally, the institutions here can be further subdivided into enterprises and scientific research institutions through identification, so as to realize the joint research path composed of the three parties of industry, academia and research, thereby providing a reference for users (such as any party among industry, academia and research or other entities responsible for leading technical research) to form and implement joint research plans.

[0098] As a non-limiting example, generating a joint research path based on the determined Chinese and foreign technical differences and portraits may further include: determining the technical problems that need to be overcome based on the determined Chinese and foreign technical differences; combining the talents and institutions associated with the determined technical problems that need to be overcome to form multiple candidate joint research paths; and comparing and analyzing each path in the multiple candidate joint research paths based on the portraits to determine the recommended research path. As mentioned earlier, the talents and institutions involved in each path can be evaluated through portraits, such as the scientific research capabilities of computing talents, the innovation of institutions, etc., so that each path can be comprehensively evaluated based on the evaluation results to obtain the respective evaluation results for different paths. This comprehensive evaluation can similarly be obtained by setting different dimensions and different weights for talents and institutions as needed, and performing weighted fusion. On this basis, by comparing the evaluation results between different paths, the recommended research path can be obtained.

[0099] Optionally, method 10 may further include generating a recommended joint research plan based on the determined recommended research path. Generating a joint research plan refers to providing the determined recommended path to the user in the form of text, charts, etc. as a specific description of the recommended research path. For example, the system can once again utilize the advantages of the pre-trained large language model in text generation to provide the talents and institutions involved in the research path, and introduce their respective backgrounds, achievements, advantages, etc. Optionally, the joint research plan may also include cooperation models, cooperation plans, etc. of various relevant parties such as talents and institutions for reference. The specific process of generating a joint research plan will be combined below. Fig.14 Describe in more detail.

[0100] Figure 2 A schematic structural block diagram of an intelligent discovery system 20 for providing a joint solution path for technical problems described herein is illustrated in FIG. Figure 2 As shown in , the intelligent discovery system 20 includes: a data collection subsystem 201, a data governance subsystem 202, and an application subsystem 203.

[0101] The following is combined with Figure 3-14 The functions of each subsystem are introduced in detail.

[0102] Figure 3 2 illustrates a schematic flow chart of data collection by the data collection subsystem 201 of the intelligent discovery system 20 for providing a joint solution path for technical problems described herein. Figure 3As shown in , the data collection subsystem 201 is used to collect scientific and technological achievement data from the data source to form a scientific and technological achievement database. Data collection can be performed based on various methods in the prior art, including but not limited to: collecting data through database connection, API interface, file import, real-time data stream, web crawler, manual entry, radio frequency technology, sensors and gateways, etc. Figure 3 middle,

[0103] Taking web crawlers as an example, data collection is based on a self-learning mechanism. Given a seed URL, the crawler compares two similar pages according to the DOM structure tree, extracts the same DOM structure as a template, intelligently extracts meaningful content from the page, and performs adaptive collection of the target site.

[0104] Figure 4 hereinafter illustrates a schematic flow chart of adapter fine-tuning according to the data governance subsystem 202 of the intelligent discovery system 20 for providing a joint solution path for technical problems as described herein.

[0105] like Figure 4 As shown in , the data governance subsystem 202 is used to govern the data in the scientific and technological achievement database to form standardized usable data, and store it in different ways to build the data base of the intelligent discovery system 20. The data governance subsystem 202 includes a model fine-tuning module 2021. The model fine-tuning module 2021 is used to fine-tune the large language model based on the collected scientific and technological achievement data, so that the large language model has the ability to identify the technical field from the user's question and generate a corresponding technical review. For details, please refer to the above description of steps 102-104 of method 10.

[0106] As a non-limiting example, based on the open source big model, collected text information such as papers (for example, 400 million) and patents (for example, 100 million) are imported from the outside, and then the adapter fine-tuning (ADAPTER-FINE-TUNE) method is used in combination with the TRANSFORMER mechanism to fine-tune the neural network model. The goal is to allow the big model to learn and understand various types of knowledge in scientific and technological literature, and serve as a baseline model for subsequent natural language processing tasks, that is, to form a data base.

[0107] Figure 5 It explains how Figure 4 Schematic flow chart of name disambiguation performed by the name disambiguation module of the data governance subsystem 202 shown in FIG. As a non-limiting example, Figure 5 As shown in , the name disambiguation module can improve the accuracy of name disambiguation by adding external features such as institutions, collaborators, and publication time based on the similarity comparison of content features such as titles, keywords, and abstracts of papers / patents by the natural language pre-training model.

[0108] Specifically, the name disambiguation module cleans the institution names by splitting parent-child institutions, merging institutions with the same name, and building an institution tree to form an institution database; at the same time, it extracts information on talents and their institutions, extracts names, and preliminarily merges institutions with the same name to form person-institution associations and author-paper associations.

[0109] Next, we extract the features of talents with the same name to find the similarity features of the collaborators, and then perform name disambiguation. By eliminating ambiguities according to rules and building a talent history organization, the accuracy of name disambiguation can be significantly improved.

[0110] Figure 6 It explains how Figure 4 Schematic flow chart of the organization tree construction module of the data governance subsystem 20 shown in FIG. As a non-limiting example, the organization tree construction module can extract organizations by analyzing the foreign investment situation in the industrial and commercial database, and merge parent and child organizations for technology-based enterprises, thereby establishing a Chinese-English mapping relationship between enterprise names, and further constructing an organization tree.

[0111] Figure 7a It explains how Figure 4 A schematic flow chart of the talent scientific research capability evaluation process of the data governance subsystem 202 shown in FIG.

[0112] The talent research capability model uses graph neural networks to build a scientific and technological talent innovation capability evaluation model based on the data iteration generation process. After identifying and eliminating invalid citation relationships between scientific and technological talents, different citation relationship weights are given to the cited persons according to the different levels of innovation capabilities of the citers, which more accurately reflects the innovation capabilities of the cited persons in the field, thereby eliminating the problems caused by the simple evaluation of the number of citations. In addition to using the characteristics of the data itself for evaluation, the algorithm will also apply talent knowledge to form interactive optimization. The results of each model output will be compared and fine-tuned with the results of expert reviews to achieve optimal calibration of the model.

[0113] The process mainly includes the following steps.

[0114] First, select the talents whose scientific research ability needs to be evaluated from the database. Then, extract the features of the selected talents. The features are mainly based on the citation relationship of the talents. For example, Figure 7bIn the simplified citation relationship network diagram shown in , nodes A, B, C, and D represent authors, directed edges between nodes represent citation relationships between authors, and the weights on the edges represent the probability of random jumps between authors. Suppose there is a browser randomly walking on the citation relationship network. If the browser is at author A, the next step is transferred to authors B, C, and D with a probability of 1 / 3. If the browser is at author B, the next step is transferred to authors A and D with a probability of 1 / 2. If the browser is at author C, the next step is transferred to author A with a probability of 1. If the browser is at author D, the next step is transferred to authors B and C with a probability of 1 / 2.

[0115] Intuitively, if an author has more links pointing to him, the probability of randomly jumping to him is higher, the CA+ value of the author is higher, and the author is more important. If an author has a higher CA+ value pointing to him, the probability of randomly jumping to him is higher, the CA+ value of the author is higher, and the author is more important. The CA+ value depends on the topology of the network. Once the topology (connection relationship) of the network is determined, the CA+ value is determined.

[0116] Furthermore, the scientific research ability of talents can be evaluated. The calculation of CA+ value can be performed on the directed graph of the citation network, which is usually an iterative process. First, an initial distribution is assumed, and the CA+ values ​​of all authors are calculated through iteration until convergence.

[0117] This calculation method is known in the art, and therefore will not be described in detail here so as not to affect the description of the main technical solution.

[0118] In order to improve the accuracy of the evaluation results of the scientific research ability of talents, the evaluation results can also be compared with the expert evaluation results. The expert evaluation results come from the evaluation of the talent by peer experts in the industry. If the two are consistent, the construction of the evaluation model is completed. If they are not consistent, you can return to the feature extraction step to re-evaluate.

[0119] Figure 8 It explains how Figure 4 A schematic flowchart of the data governance subsystem for evaluating institutional innovation is shown in the figure.

[0120] The innovation capability of an enterprise is not only reflected in the number of patents, R&D investment and other result data. Through the construction of an institutional innovation model, we can use process data to identify the true innovation capability of an enterprise from the dimensions of innovation hard power, market connection, innovation resource integration, and innovation team growth, and build an innovation portrait of the enterprise, which will help to accurately screen the research targets in the future.

[0121] The hard power of innovation includes but is not limited to patent strength, innovation radiation, and technological cutting-edge. Market connection power includes but is not limited to market layout, cross-border integration, technological breadth, and industry-research integration. Innovation resource integration power includes but is not limited to innovation chain integration and capital chain integration. Innovation team growth power includes but is not limited to core team innovation, R&D team innovation, and R&D team stability.

[0122] As described above, the data governance subsystem governs all data in the scientific and technological achievement database to form standardized usable data, and stores it in different ways to build the data base of the intelligent discovery system. Fig. 9 , a schematic diagram of the structure of the data base 30 according to the intelligent discovery system and method for providing a joint solution path for technical problems described in this article is illustrated.

[0123] like Fig. 9 As shown in FIG. 3 , the data base 30 includes a search engine 301 , a graph database 302 , a vector database 303 and a relational database 304 .

[0124] The search engine 301 is used to provide a search function for at least one of the following data: collected data, talent evaluation and profile data, and institution evaluation and profile data. The graph database 302 is used to store at least one of the following relationship graph data: technical subject relationship graph, talent relationship graph, institution relationship graph, talent-institution relationship graph, and scientific research results relationship graph. The vector database 303 is used to store the result data after large-model vectorization encoding of the collected data. The relationship database 304 is used to store public configuration data.

[0125] The following combination Figure 10 to Figure 14 To further describe the application subsystem 203 of the intelligent discovery system 20 in detail.

[0126] Fig.10 The schematic diagram of the application subsystem 203 for technology screening is explained in FIG. As mentioned above, the main goal of technology screening is to encode the questions asked by users in natural language using the first fine-tuned open source large model, convert them into vectors, and compare the similarity with all scientific research results vectors in the vector database. After manually confirming the similarity threshold, the technical problems are identified and the solution data sets that match the technical problems are screened out. Then, the data in the solution data set is read using the second fine-tuned large language model to generate a summary of the current status of solving the technical problems and identify the subdivided technical fields of the technical problems.

[0127] Fig.112 illustrates a schematic flow chart of forming a technical overview according to the application subsystem 203 of the intelligent discovery system 20 described in this article for providing a joint solution path for technical problems.

[0128] The main goal of a technology review is to be able to identify technical areas from all solutions associated with technical problems and to generate a corresponding technology review.

[0129] First, read the technical text through the big model to let the big model understand the knowledge points of the technical text.

[0130] Second, organize domain experts to construct questions that help with domain classification, ask questions to the big model, and assist the big model in generating subdivided technical fields, that is, manually construct a set of classification questions.

[0131] Third, classify the results data according to subdivided technical fields.

[0132] Fourth, organize experts in the field to sort out the key points needed for the technical review and construct a corresponding set of questions. For example, "What are the current technological advances in the ** field? Who proposed them?" "What are the ways to solve ** problems? What are the specific ways?" "For a certain technical problem, which projects have been implemented and how were they implemented? Who is the leader?" and so on.

[0133] Fifth, around different sub-fields, machines generate answers based on questions and eventually organize them into technical reviews.

[0134] Fig.12 A schematic diagram of a technology comparison to determine the difference between Chinese and foreign technologies is illustrated in FIG. 2 , according to the application subsystem 203 of the intelligent discovery system 20 described in this article for providing a joint solution path for technical problems.

[0135] The main goal of technology comparison is to sort out the talents and institutions mainly engaged in the industry at home and abroad within a certain subdivided technology field, and compare them from different dimensions to find the gap between the two sides.

[0136] First, extract the related talents and institutions from the achievement data related to the subdivided fields, and associate the corresponding achievement data.

[0137] Secondly, multi-dimensional comparisons can be conducted, including comparison of talent concentration, talent level, technology layout, and technological advancement.

[0138] Talent concentration comparison means that through the institutions to which the authors in papers and patents belong, we can know that they belong to different institutions, and then count the number of talents in the institutions.

[0139] Talent level comparison refers to obtaining the ranking of talents in different fields and thus obtaining the level of talents by evaluating their scientific research capabilities.

[0140] Technology layout comparison refers to the use of natural language processing technology to extract topics from papers and patents, and further count the word frequency of the topics to see the technology layout over a period of time.

[0141] The comparison of technological advancement refers to the construction of a disruptive index. The disruptive index of all achievements can be calculated according to different countries. Through simple statistics, the differences in technological advancement between different institutions and countries can be known. Before this, a paper in Nature proposed the concept of "disruptive index", which will not be elaborated here.

[0142] Furthermore, based on the comparison results, corresponding visual charts are automatically generated to identify the differences between Chinese and foreign technologies.

[0143] Fig.13 A schematic diagram of obtaining a technical path by the application subsystem 203 of the intelligent discovery system 20 for providing a joint solution path for technical problems described in this article is explained in FIG.

[0144] The main goal of acquiring technology paths is to find talents and institutions that can solve technical problems in the country, in order to seek a combination of talents and institutions to meet the needs of joint research in certain subdivided technology fields.

[0145] First, extract relevant domestic talents and institutions from the achievement data related to the subdivided fields, and associate the corresponding achievement data.

[0146] Second, we use talent and institution portraits to identify the leading enterprises, leading scientific research institutions and high-end talents in various sub-sectors in China. We can use labels to distinguish between enterprises and scientific research institutions.

[0147] Third, by extracting and combining talents, enterprises and scientific research institutions respectively, several joint research paths are generated.

[0148] Fourth, build an evaluation index system for research paths, and obtain the final score for each path by evaluating the talent level, technology maturity, technological advancement, scientific research honors, and technological influence at each node.

[0149] Fifth, generate recommended joint research paths based on the scoring results.

[0150] Fig.14 2 illustrates a schematic diagram of generating a solution according to the application subsystem 203 of the intelligent discovery system 20 described in this article for providing a joint solution path for technical problems.

[0151] As mentioned above, based on the scoring results of the system, a recommended joint attack path can be generated. Based on the recommended joint path, the following steps can be performed.

[0152] First, based on the information of institutions (enterprises + scientific research institutions) and talents at each node on the research path, the corresponding scientific research results data are extracted.

[0153] Second, based on the extracted scientific research results data, statistical analysis of various scientific research results is carried out, including statistics on the technical achievements of various institutions on various research paths.

[0154] Third, experts in various fields will construct a set of general questions for solving problems and use them to ask questions to the big model.

[0155] Fourth, use a large model to read the extracted scientific research results data, form a set of answers based on the set of questions constructed by experts, and then combine the results of statistical analysis to form a research path and generate a joint research plan.

[0156] According to various aspects, elements, or any part of elements, or any combination of elements of the present disclosure, a "processing system" including one or more processors can be implemented. Examples of processors include: microprocessors, microcontrollers, digital signal processors (DSPs), field programmable gate arrays (FPGAs), programmable logic devices (PLDs), state machines, gated logic, discrete hardware circuits, and other suitable hardware configured to perform various functionalities described throughout the present disclosure. One or more processors in a processing system can execute software. Software should be broadly interpreted as meaning instructions, instruction sets, codes, code segments, program codes, programs, subroutines, software modules, applications, software applications, software packages, routines, subroutines, objects, executables, threads of execution, procedures, functions, etc., whether it is described in software, firmware, middleware, microcode, hardware description language, or other terms. Software may reside on a computer-readable medium. The computer-readable medium may be a non-transient computer-readable medium. As examples, non-transitory computer-readable media include: magnetic storage devices (e.g., hard disks, floppy disks, magnetic strips), optical disks (e.g., compact disks (CDs), digital versatile disks (DVDs)), smart cards, flash memory devices (e.g., memory cards, memory sticks, key drives), random access memory (RAM), read-only memory (ROM), programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), registers, removable disks, and any other suitable medium for storing software and / or instructions that can be accessed and read by a computer. As examples, computer-readable media may also include carrier waves, transmission lines, and any other suitable medium for transmitting software and / or instructions that can be accessed and read by a computer. The computer-readable medium may reside in the processing system, be external to the processing system, or be distributed across multiple entities including the processing system. The computer-readable medium may be implemented in a computer program product. As examples, a computer program product may include a computer-readable medium in packaging materials. Those skilled in the art will recognize how to best implement the described functionality presented throughout this disclosure depending on the specific application and the overall design constraints imposed on the overall system.

[0157] It should be understood that the specific order or hierarchy of steps in the disclosed methods is an illustration of an exemplary process. Based on design preferences, it should be understood that the specific order or hierarchy of steps in the methods or methodologies described herein may be rearranged. The attached method claims present elements of the various steps in a sample order and are not meant to be limited to the specific order or hierarchy presented unless specifically stated herein.

[0158] The previous description is provided to enable any person skilled in the art to practice the various aspects described herein. Various modifications to these aspects will be readily apparent to those skilled in the art, and the universal principles defined herein may be applied to other aspects. Therefore, the claims are not intended to be limited to the various aspects shown herein, but should be granted the full scope consistent with the language of the claims, wherein the singular reference to the element is not intended to mean "there is and only one" (unless specifically stated) but "one or more". Unless specifically stated otherwise, the term "some" refers to one or more. The phrase "at least one" quoting a list of items refers to any combination of these items, including a single member. As an example, "at least one of a, b or c" is intended to cover: at least one a; at least one b; at least one c; at least one a and at least one b; at least one a and at least one c; at least one b and at least one c; and at least one a, at least one b and at least one c. The elements of the various aspects described throughout this disclosure are all structurally and functionally equivalent schemes currently or hereafter known to those of ordinary skill in the art, and are expressly incorporated herein by reference, and are intended to be covered by the claims. Moreover, nothing disclosed herein is intended to be dedicated to the public regardless of whether such disclosure is explicitly recited in the claims.

Claims

1. An intelligent discovery system for providing joint solution paths for technical problems, include: A data collection subsystem is used to collect scientific and technological achievement data from data sources to form a scientific and technological achievement database; A data governance subsystem is used to manage the data in the scientific and technological achievements database to form standardized usable data, and store it in different ways to build the data base of the intelligent discovery system, wherein the data governance subsystem includes: A model fine-tuning module, used to fine-tune the large language model based on the collected data, so that the large language model can identify technical issues from user questions and generate corresponding technical summaries; and An application subsystem, wherein the application subsystem is configured to: Processing the user question using the first fine-tuned large language model to identify a technical problem; Based on the identified technical problems, mining technical texts from the scientific and technological achievements database; Identify the people and institutions associated with the mined technical texts; Profiling the identified talents and institutions; Determine the differences between Chinese and foreign technologies based on the mined technical texts and the portraits; and The joint research path is generated based on the determined technological differences between China and foreign countries and the portrait.

2. The intelligent discovery system according to claim 1, It is characterized in that The data base comprises: A search engine for providing a search function for at least one of the following data: collected data, talent evaluation and profile data, and institution evaluation and profile data; A graph database, for storing at least one of the following relationship graph data: a technical subject relationship graph, a talent relationship graph, an institution relationship graph, a talent-institution relationship graph, and a scientific research achievement relationship graph; A vector database, used to store the result data after large-model vector encoding of the collected data; and A relational database used to store common configuration data.

3. The intelligent discovery system according to claim 1, It is characterized in that The data governance subsystem further includes: A name disambiguation module, used to disambiguate names based on a natural language pre-trained model and in combination with external features of the collected data; An organization tree construction module is used to extract organizations from the collected data and construct an organization tree; Talent research capability evaluation module, used to evaluate the research capabilities of talents in related industries; and The module for evaluating the innovation capability of an institution is used to evaluate the innovation capability of the institution based on the collected data.

4. The intelligent discovery system according to claim 3, It is characterized in that Evaluation of the scientific research capabilities of talents in related industries includes: Construct a talent research capability evaluation model based on the data iteration generation process; Identify and eliminate invalid citation relationships between talents; Assign corresponding citation relationship weights to the cited persons based on the innovation ability of the citer; and The scientific research ability of the talents is evaluated based on the citation relationship weight.

5. The intelligent discovery system according to claim 4, It is characterized in that The talent scientific research ability evaluation module also compares and fine-tunes the output results of the talent scientific research ability evaluation model with the expert review results to further calibrate the talent scientific research ability evaluation model.

6. The intelligent discovery system according to claim 3, It is characterized in that The data collected to evaluate the innovation capabilities of the organization include: The innovation capability of an organization is evaluated based on a multi-dimensional organization innovation evaluation model, wherein the multiple dimensions include one or more of the following dimensions: innovation hard power, market connection capability, innovation resource integration capability, and innovation team growth capability.

7. A method for providing a joint solution path for technical problems using the intelligent discovery system according to any one of claims 1 to 6, the method comprising the following steps: Receive user questions; Processing the user question using the first fine-tuned large language model to identify a technical problem; Based on the identified technical problems, mining technical texts from the scientific and technological achievements database; Identify the people and institutions associated with the mined technical texts; Profiling the identified talents and institutions; Determine the differences between Chinese and foreign technologies based on the identified technical texts and the portraits; as well as Based on the determined technological differences between China and foreign countries and the portrait, a joint research path is generated.

8. The method according to claim 7, It is characterized in that Based on the identified technical problem, mining the technical text further includes: Based on the identified technical problem, screening relevant scientific and technological achievement data from the scientific and technological achievement database to form a solution data set; and The data in the solution dataset is read using a second fine-tuned large language model to generate a current overview of solutions to the technical problem and identify subdivided technical fields of the technical problem.

9. The method according to claim 7, It is characterized in that Determining the differences between Chinese and foreign technologies based on the identified technical text and the portrait further includes: A corresponding visualization chart is automatically generated based on the identified technical text and the portrait.

10. The method according to claim 7, It is characterized in that Generating the joint research path based on the determined Chinese and foreign technology differences and the portrait further includes: Determine the technical problems that need to be overcome based on the identified technical differences between China and foreign countries; Combine talents and institutions associated with the identified technical problems to be solved to form multiple candidate joint research paths; and Based on the portrait, a comparative analysis is performed on each of the multiple candidate joint tackling paths to determine a recommended tackling path.

11. The method according to claim 10, It is characterized in that The application subsystem is further configured to: A tackling plan is generated based on the recommended tackling path.

12. The method according to claim 11, It is characterized in that Generating a solution based on the recommended solution path further includes: Extract corresponding scientific research achievement data based on the talent and institution information at each node on the recommended research path; Conduct statistical analysis on various scientific research results based on the extracted scientific research results data; Ask the second fine-tuned large language model questions using a predetermined set of questions for building a solution; Using the second fine-tuned large language model to read the scientific research achievement text in the extracted scientific research achievement data, and forming an answer set according to the question set; and The solution is generated based on the answer set and in combination with the result of the statistical analysis.

Images