Method and device for evaluating understanding and innovation inspiration ability based on multi-person discussion text

Abstract

This application provides a method and apparatus for evaluating comprehension and innovation inspiration capabilities based on multi-person discussion texts, relating to the field of comprehension and innovation inspiration capability evaluation technology. The method includes: acquiring multiple dialogue texts obtained from multi-person discussions, and constructing an influence graph representing the problem domain and solution domain based on the multiple dialogue texts; one dialogue text per participant; employing a heuristic method, utilizing a natural language model to process the multiple dialogue texts, evaluating the comprehension ability of each participant, and evaluating each participant's problem-solving ability and creative thinking based on the influence graph. The method and apparatus provided in this application for evaluating comprehension and innovation inspiration capabilities based on multi-person discussion texts are used to comprehensively and accurately evaluate the comprehension and innovation inspiration capabilities of each participant in a multi-person discussion scenario.

Description

Technical Field

[0001] This application relates to the field of comprehension and innovation inspiration ability assessment technology, and in particular to a method and apparatus for comprehension and innovation inspiration ability assessment based on multi-person discussion text. Background Technology

[0002] In multi-person discussion scenarios, participants can develop an understanding of the problem and generate solutions to open tasks, demonstrating each participant's ability to explain and inspire innovation.

[0003] In related technologies, assessment systems for understanding and innovation capabilities are often based on individual decision-making scenarios, heavily reliant on the subjective experience and judgment of experts, and lack objective system assessments with theoretical support.

[0004] Therefore, there is an urgent need for an evaluation method that can comprehensively and accurately assess the understanding and innovative inspiration capabilities of each participant in a multi-person discussion scenario. Summary of the Invention

[0005] The purpose of this application is to provide a method and apparatus for evaluating the comprehension and creative inspiration capabilities of participants in multi-person discussion texts, which can be used to comprehensively and accurately evaluate the comprehension and creative inspiration capabilities of each participant in a multi-person discussion scenario.

[0006] This application provides a method for assessing comprehension and creative inspiration abilities based on multi-person discussion texts, including:

[0007] Multiple dialogue texts obtained from multi-person discussions are acquired, and an influence graph representing the problem domain and solution domain is constructed based on the multiple dialogue texts; one dialogue text corresponds to one participant; a heuristic method is used to process the multiple dialogue texts using a natural language model to evaluate each participant's comprehension ability, and the problem-solving ability and creative thinking of each participant are evaluated based on the influence graph.

[0008] Optionally, the step of constructing an influence graph representing the problem domain and the solution domain based on the plurality of dialogue texts includes: constructing a corpus based on a domain knowledge base related to the dialogue topic and the plurality of dialogue texts, and constructing an influence graph representing the problem domain and the solution domain based on the corpus; wherein, the nodes of the influence graph include: opportunity nodes, decision nodes, and utility nodes; the nodes of the influence graph are used to represent the information involved in the dialogue texts; the directed edges of the influence graph are used to represent the correlation between information.

[0009] Optionally, the step of constructing an influence graph representing the problem domain and the solution domain based on the corpus includes: generating multiple candidate nodes by using a preset word processing method and combining a stop word dictionary with domain knowledge related to the dialogue topic; extracting noun phrases from the corpus using a natural language model as the filling content for each candidate node among the multiple candidate nodes, and merging semantically similar nodes to obtain multiple nodes to be classified; performing a classification operation on the multiple nodes to be classified using a natural language model to obtain a classification result; wherein the classification operation is used to classify the nodes of the influence graph into: opportunity nodes, decision nodes, and utility nodes; the preset word processing method includes at least one of the following: word segmentation and syntactic analysis.

[0010] Optionally, the step of constructing an influence graph representing the problem domain and the solution domain based on the corpus includes: extracting explicit directed relations from the multiple dialogue texts based on causal templates and statement templates; extracting related relations from multiple nodes of the influence graph using a natural language model to obtain multiple directed edges of the influence graph; wherein, the directed edge between any two nodes is obtained based on the causal relationship and / or conditional relationship between the text information corresponding to the two nodes.

[0011] Optionally, the problem-solving ability includes at least one of the following: observation ability, comprehension ability, and reasoning ability; the evaluation of each participant's problem-solving ability based on the influence graph includes: evaluating each participant's observation ability based on the number of opportunity nodes hit in the influence graph; and / or, evaluating each participant's comprehension ability based on the number of utility nodes hit in the influence graph; and / or, evaluating each participant's comprehension ability based on the frequency of each type of reasoning pattern appearing in each participant's multi-type reasoning patterns; wherein, the multi-type reasoning patterns include: syllogism, relational reasoning, inductive reasoning... The system includes abductive reasoning and causal reasoning; the syllogism in the influence graph is characterized by abstract nodes pointing to fact nodes; the relational reasoning in the influence graph is characterized by a predetermined number of directed edges connecting fact nodes sequentially; the inductive reasoning in the influence graph is characterized by abstract nodes pointing to abstract nodes, or fact nodes pointing to abstract nodes; the abductive reasoning in the influence graph is characterized by abstract nodes pointing to fact nodes, or fact nodes pointing to abstract nodes, and there is a causal relationship; the fact nodes are nodes in the influence graph that contain factual information; the abstract nodes are nodes in the influence graph that contain abstract information.

[0012] Optionally, the creative thinking includes at least one of the following: inspiration and transferability, innovation and imagination, and thinking characteristics for generating innovative solutions; the evaluation of each participant's creative thinking based on the influence map includes: determining the mention order and hit rate of each participant's opportunity nodes, decision nodes, and utility nodes in the influence map based on each participant's dialogue text, and evaluating each participant's inspiration and transferability based on the mention order and hit rate; and / or, evaluating each participant's innovation and imagination based on the semantic features and mention frequency of the decision nodes corresponding to each participant's dialogue text in the influence map; and / or, determining the target number of times each participant inspires and transfers ideas between value-based decisions and solutions based on each participant's dialogue text, and evaluating each participant's thinking characteristics for generating innovative solutions based on the target number of times.

[0013] Optionally, the heuristic approach, utilizing a natural language model to process the multiple dialogue texts and assess each participant's comprehension ability, includes: preprocessing each participant's dialogue text and inputting the processed dialogue text into the natural language model; using guide words corresponding to the dialogue topic to control the natural language model to assess each participant's three-level comprehension ability; wherein, the assessment of the three-level comprehension ability includes: assessing the low-level comprehension ability based on the number of key points appearing in the dialogue text, assessing the intermediate-level comprehension ability based on the number of conceptual statements appearing in the dialogue text, and assessing the high-level comprehension ability based on the number of specific examples combining historical and current realities appearing in the dialogue text.

[0014] Optionally, the multiple dialogue texts are dialogue texts of participants from multiple groups; after employing a heuristic method to process the multiple dialogue texts using a natural language model to evaluate each participant's comprehension ability, and evaluating each participant's problem-solving ability and creative thinking based on the influence graph, the method further includes: performing a comprehensive evaluation of each group based on the product of the node relevance, number of hit nodes, and average degree of high-connection nodes of the dialogue texts of each participant in each group, to obtain a comprehensive evaluation result for each group; and performing component calibration based on the comprehensive evaluation results of each group to evaluate the performance of participants in groups with different dialogue topics.

[0015] This application also provides a device for assessing comprehension and creative inspiration abilities based on multi-person discussion texts, including:

[0016] The module is used to acquire multiple dialogue texts obtained from multi-person discussions; the module is used to construct an influence map representing the problem domain and solution domain based on the multiple dialogue texts; one dialogue text corresponds to one participant; the module is used to process the multiple dialogue texts using heuristic methods and natural language models to evaluate the comprehension ability of each participant, and to evaluate the problem-solving ability and creative thinking of each participant based on the influence map.

[0017] Optionally, the construction module is specifically used to construct a corpus based on a domain knowledge base related to the dialogue topic and the multiple dialogue texts, and to construct an influence graph representing the problem domain and the solution domain based on the corpus; wherein, the nodes of the influence graph include: opportunity nodes, decision nodes and utility nodes; the nodes of the influence graph are used to represent the information involved in the dialogue texts; the directed edges of the influence graph are used to represent the correlation between information.

[0018] Optionally, the construction module is specifically used to generate multiple candidate nodes by employing a preset word processing method and combining a stop word dictionary with domain knowledge related to the dialogue topic; the construction module is further used to extract noun phrases from the corpus using a natural language model as the filling content for each candidate node among the multiple candidate nodes, and to merge semantically similar nodes to obtain multiple nodes to be classified; the construction module is further used to perform a classification operation on the multiple nodes to be classified using a natural language model to obtain a classification result; wherein, the classification operation is used to divide the nodes of the influence graph into: opportunity nodes, decision nodes, and utility nodes; the preset word processing method includes at least one of the following: word segmentation method and syntactic analysis method.

[0019] Optionally, the construction module is specifically used to extract explicit directed relations from the multiple dialogue texts based on causal templates and statement templates; the construction module is also specifically used to extract related relations from multiple nodes of the influence graph using a natural language model to obtain multiple directed edges of the influence graph; wherein, the directed edge between any two nodes is obtained based on the causal relationship and / or conditional relationship between the text information corresponding to the two nodes.

[0020] Optionally, the problem-solving ability includes at least one of the following: observation ability, comprehension ability, and reasoning ability; the evaluation module is specifically used to evaluate each participant's observation ability based on the number of opportunity nodes hit in the influence graph; the evaluation module is further used to evaluate each participant's comprehension ability based on the number of utility nodes hit in the influence graph; the evaluation module is further used to evaluate each participant's comprehension ability based on the frequency of occurrence of each type of reasoning pattern in each participant's multi-type reasoning patterns; wherein, the multi-type reasoning patterns include: syllogism, relational reasoning, and inductive reasoning. And abductive reasoning; the syllogism in the influence graph is in the form of abstract nodes pointing to fact nodes; the relational reasoning in the influence graph is in the form of directed edges connecting a predetermined number of fact nodes in sequence; the inductive reasoning in the influence graph is in the form of abstract nodes pointing to abstract nodes, or fact nodes pointing to abstract nodes; the abductive reasoning in the influence graph is in the form of abstract nodes pointing to fact nodes, or fact nodes pointing to abstract nodes, and there is a causal relationship; the fact node is: a node in the influence graph that contains factual information; the abstract node is: a node in the influence graph that contains abstract information.

[0021] Optionally, the creative thinking includes at least one of the following: inspiration and transferability, innovation and imagination, and thinking characteristics for generating innovative solutions; the evaluation module is specifically used to determine the mention order and hit rate of each participant's opportunity nodes, decision nodes, and utility nodes in the influence diagram based on each participant's dialogue text, and to evaluate each participant's inspiration and transferability based on the mention order and hit rate; the evaluation module is further used to evaluate each participant's innovation and imagination based on the semantic features and mention frequency of the decision nodes corresponding to each participant's dialogue text in the influence diagram; the evaluation module is further used to determine the target number of times each participant inspires and transfers ideas between value-based decisions and solutions based on each participant's dialogue text, and to evaluate each participant's thinking characteristics for generating innovative solutions based on the target number of times.

[0022] Optionally, the evaluation module is specifically used to preprocess the dialogue text of each participant and input the processed dialogue text into a natural language model. It uses guide words corresponding to the dialogue topic to control the natural language model to evaluate each participant's three-level comprehension ability. The evaluation of the three-level comprehension ability includes: evaluating the low-level comprehension ability based on the number of key points appearing in the dialogue text; evaluating the intermediate-level comprehension ability based on the number of conceptual statements appearing in the dialogue text; and evaluating the high-level comprehension ability based on the number of specific examples combining historical and current realities appearing in the dialogue text.

[0023] Optionally, the multiple dialogue texts are dialogue texts of participants from multiple groups; the evaluation module is further configured to perform a comprehensive evaluation of each group based on the product of the node relevance, the number of hit nodes, and the average degree of high-connection nodes of the dialogue texts of each participant in each group, to obtain a comprehensive evaluation result for each group; the evaluation module is further configured to perform component calibration based on the comprehensive evaluation result of each group, so as to evaluate the performance of participants in groups with different dialogue topics.

[0024] This application also provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the method for assessing the understanding and innovative inspiration capabilities based on multi-person discussion text as described above.

[0025] This application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of any of the above-described methods for assessing the understanding and innovative inspiration capabilities based on multi-person discussion text.

[0026] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method for assessing the understanding and creative inspiration capabilities based on multi-person discussion text.

[0027] The method and apparatus for evaluating comprehension and innovative inspiration capabilities based on multi-person discussion texts provided in this application first acquire multiple dialogue texts obtained from multi-person discussions, and construct an influence graph representing the problem domain and solution domain based on the multiple dialogue texts; one dialogue text corresponds to one participant; then, a heuristic method is used to process the multiple dialogue texts using a natural language model to evaluate the comprehension ability of each participant, and to evaluate each participant's problem-solving ability and creative thinking based on the influence graph. In this way, a comprehensive and accurate evaluation of the comprehension and innovative inspiration capabilities of each participant in a multi-person discussion scenario can be achieved. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 This is a flowchart illustrating the assessment method for understanding and inspiring creativity based on multi-person discussion texts provided in this application;

[0030] Figure 2 This is a schematic diagram of the structure of the assessment device for understanding and creative inspiration based on multi-person discussion text provided in this application;

[0031] Figure 3 This is a schematic diagram of the structure of the electronic device provided in this application. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0033] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0034] In related technologies, the assessment of the understanding and innovative inspiration capabilities of each participant in a multi-person discussion scenario is often based solely on simplified single-person decision-making tasks. Therefore, it fails to encompass the understanding and inspiration capabilities demonstrated during the clash and debate of viewpoints, brainstorming, etc., and cannot capture interpersonal characteristics such as teamwork, leadership, and interpersonal skills, or group behavioral characteristics such as mutual inspiration and collective negotiation. Such systems often lack the support of decision analysis theory, and their evaluation dimensions are relatively unsystematic and fragmented, or only applicable to single tasks and unable to be transferred to general problem understanding and innovative inspiration scenarios. These systems often provide overall evaluations through expert opinions, which have a certain degree of subjectivity and lack quantitative evaluation of details. In particular, past assessment systems are difficult to apply to decision-making tasks with clearly defined boundaries (information, value, and options) and well-defined objectives.

[0035] To address the aforementioned technical problems in related technologies, this application provides a method for assessing comprehension and innovation-inspiring abilities based on multi-person discussion texts. This method consists of four modules: problem comprehension ability assessment; problem-solving ability assessment; creative thinking assessment; and inter-group calibration. The method uses natural language models and influence graphs for text processing and analysis. It semi-quantitatively measures participants' comprehension abilities using the summarizing capabilities of natural language models. An influence graph represents the problem domain and solution domain when participants complete innovative tasks. Three types of nodes (opportunity nodes, decision nodes, and utility nodes) in the influence graph represent the information involved in the text. Directed edges in the influence graph represent the directed relationships between information. The method measures participants' innovation-inspiring abilities through node coverage, semantic features of nodes, and the connection order of directed edges.

[0036] In the problem comprehension assessment module, this method adjusts the output of the natural language model through prompting engineering, enabling it to automatically summarize and extract content from participants' discussion texts, classify the texts, and provide a semi-quantitative assessment of participants' comprehension abilities based on a three-level comprehension framework. In the problem-solving ability assessment module, this method, based on syntactic analysis methods (word segmentation, association analysis) and the natural language model, combined with manual judgment and verification, extracts nodes and directed edges from participants' discussion texts, constructs an influence graph representing the problem domain and solution domain, and assesses the comprehensiveness of participants' information considerations, value considerations, and the richness of their reasoning methods using decision analysis theory. In the creative thinking assessment module, this method uses the mention order and proportion of three types of nodes in the influence graph to assess participants' inspiration and transfer abilities, uses the semantic features and mention frequency of decision nodes to assess participants' innovativeness and imagination, and uses the occurrence of two relatively ideal thinking modes (value-based decision-making and inter-solution association) to assess the characteristics of participants' thinking in generating innovative solutions. In the inter-group calibration module, this method performs inter-group calibration on the performance of each participant based on the node relevance, number of hit nodes, and average degree of high-connection nodes in each group's discussion text.

[0037] The following is a detailed description of the technical terms involved in the method for assessing the understanding and innovative inspiration capabilities based on multi-person discussion texts provided in the embodiments of this application:

[0038] 1. Problem domain and solution domain

[0039] In domain modeling, the problem space refers to all the information that defines the problem and constrains the solution, including objectives, problem background, basic functions of the solution, or other rules; the solution space defines the abstract environment for developing the solution, encompassing the design and functionality of the solution itself, the process of deriving the solution, etc. When completing the innovation inspiration task, participants need to observe, understand, and reason about the problem space (in-depth analysis and exploration), and complete the inspiration and transfer from the problem space to the solution space.

[0040] 2. Level 3 comprehension ability

[0041] Human beings' ability to understand abstract things can be divided into three levels: the low level refers to the ability to identify the thing or distinguish the difference between two things; the intermediate level refers to the ability to understand the essence and internal connection of the thing, that is, to understand its connotation; and the high level refers to the ability to concretize (from abstract to concrete) or systematize (from part to whole) the thing, or to transfer it between different scenarios.

[0042] Taking the understanding of a certain topic as an example, a low-level understanding is being able to answer questions about the content of the topic, an intermediate-level understanding is being able to explain the origin, characteristics, core concepts, etc. of the topic, and an advanced-level understanding is being able to elaborate on specific examples that embody the topic by combining historical and current realities.

[0043] 3. Project Notification

[0044] Cueing engineering is a technique for natural language models that guides them to generate outputs that meet specific requirements or achieve specific tasks by providing explicit, clear, and well-structured inputs. Common elements of cueing engineering include examples (which enable the language model to learn and imitate), descriptions and context (which enable the language model to generate more relevant outputs), roles (which assign an "identity" to the language model), and parameters (including temperature and reliability, maximum length, stopping sequences, frequency, and presence penalties, etc.).

[0045] This application uses a natural language model to complete natural language processing tasks such as content (nodes and directed relations) extraction, keyword matching, and text allocation, and provides prompts such as examples and parameters based on task types.

[0046] 4. Four types of reasoning patterns

[0047] This application's embodiments divide the information in the text into two categories: factual and abstract. It examines the semantic features (the order in which the two types of information are mentioned in the corpus) and the patterns in the influence graph (the directed relationship between the two types of information) of four commonly used reasoning patterns in problem-solving (syllogism, relational reasoning, inductive reasoning, and abductive reasoning).

[0048] Syllogism refers to the conclusion drawn based on a major premise (general principle) and a minor premise (specific case) (a judgment made on a specific case based on a general principle). In the corpus, this is manifested as abstract information appearing first, followed by factual information. In the influence graph, the pattern is that abstract nodes point to factual nodes.

[0049] Relational reasoning refers to reasoning in which at least one of the premises is a relational proposition (between facts). In the corpus, this is manifested as the appearance of factual information three times, and in the influence graph, it is manifested as the existence of directed edges connecting three fact nodes in sequence.

[0050] Inductive reasoning refers to inferring the attributes (abstract) of all objects based on the observable attributes (which may be abstract or concrete) of some objects in a certain category. In the corpus, this is manifested as the first appearance of abstract or factual information, followed by the appearance of abstract information. In the influence graph, the pattern is that abstract or factual nodes point to abstract nodes.

[0051] Abductive reasoning refers to reasoning from a set of facts to the best explanation. In the corpus, it is manifested as factual information appearing first, followed by abstract or factual information. In the influence graph, the pattern is that abstract or factual nodes point to factual nodes, and the directed relationship is a causal relationship (verified by randomized experiments).

[0052] In this application embodiment, the corresponding reasoning pattern is identified based on the combination of the above four semantic features and influence graph patterns, using paragraphs as the unit.

[0053] 5. Semantic Network

[0054] Semantic networks refer to the visualization of information in text in a network format based on the relationships between concepts. This application's embodiments use pre-trained word vectors (Wikipedia Chinese word vectors) to embed text information in a high dimension. Principal Component Analysis (PCA) is then used to transform the high-dimensional embedding vectors into two-dimensional vectors, which represent the two-dimensional coordinates of nodes in the semantic network. The embedding position and distance of nodes in the semantic network can intuitively represent their similarity and association, and are used to evaluate the innovativeness of the solution.

[0055] 6. Value-based decision making

[0056] This application's embodiments map information in text to three types of nodes (opportunity nodes, decision nodes, and utility nodes) in the influence graph, examining the semantic features of value-based decision-making and patterns in the influence graph. Value-based decision-making is a relatively ideal thinking pattern, referring to first identifying important goals of concern (utility nodes), and then using these important goals to guide the decision-making process (decision nodes), reflecting the participants' ability to grasp key values. Within a given time window (paragraph), value-based decision-making in the corpus is manifested by the simultaneous appearance of utility nodes and decision nodes (the order is not required), and the pattern in the influence graph is utility nodes pointing to decision nodes or decision nodes pointing to utility nodes.

[0057] 7. Inter-scheme association

[0058] This application's embodiments map information in the text to three types of nodes in the influence graph, examining the semantic features of associations between solutions. Associations between solutions represent another relatively ideal thinking pattern, referring to the simultaneous proposal of multiple related solutions (decision nodes), reflecting the participants' active and agile thinking. Within a given time window (paragraph), associations between solutions in the corpus are manifested as the simultaneous appearance of two or more related decision nodes (semantic similarity greater than a similarity threshold).

[0059] 8. Association Analysis

[0060] Association analysis refers to identifying common phrases and fixed collocations in text based on the occurrence and co-occurrence frequency of word segmentation results. This application employs the Apriori (association rule mining) algorithm to perform association mining after removing stop words from the word segmentation results. By adjusting parameters (minimum support, maximum length, threshold), candidate common phrases and fixed collocations are obtained. After manual judgment and verification, these are used to supplement the phrases extracted by the natural language model.

[0061] The following description, in conjunction with the accompanying drawings, details the method for evaluating the understanding and innovative inspiration capabilities of multi-person discussion texts provided in this application, through specific embodiments and application scenarios.

[0062] like Figure 1 As shown in the embodiment of this application, a method for assessing the understanding and innovative inspiration capabilities of multi-person discussion texts is provided. This method may include the following steps 101 and 102:

[0063] Step 101: Obtain multiple dialogue texts based on multi-person discussions, and construct an influence diagram representing the problem domain and solution domain based on the multiple dialogue texts.

[0064] In this system, each participant corresponds to one dialogue text.

[0065] For example, in this embodiment of the application, on-site shorthand and speech transcription can be used to convert the audio of a multi-person discussion into the original dialogue text. The original dialogue text is then preprocessed, including proofreading and removing interjections and pauses. For the original dialogue text with improper punctuation and segmentation, a natural language model is used to re-segment it by incorporating semantic information. In this way, the dialogue texts of each participant in the multi-person discussion can be obtained, i.e., the aforementioned multiple dialogue texts.

[0066] For example, in the embodiments of this application, the prompt word for the natural language model can also be fine-tuned based on prompt engineering, so that it can stably and repeatably classify the input text according to the given topic (i.e. select the topic that is closest to the general meaning of the input text) and report the confidence level (for example, the prompt can be "select the topic that is closest to the text from the above topics and score the degree of closeness").

[0067] For example, in this embodiment of the application, the guiding words for the natural language model can also be fine-tuned based on prompt engineering, so that it can stably and repeatedly output the content extraction and summary of the input text. Content extraction is to directly extract the target words or sentences from the input text (for example, the prompt can be "extract the content that is precisely mentioned in the original text"), and content summary is to search for relevant paragraphs in the input text according to the prompt, and summarize and refine them (for example, the prompt can be "if the text mentions this content, list the main points of the original text that mention this content").

[0068] For example, in the embodiments of this application, an influence map representing the problem domain and the solution domain can be constructed using a natural language model, and the comprehensiveness of the participants' information considerations, the comprehensiveness of their value considerations, and the richness of their reasoning methods can be evaluated based on the influence map.

[0069] Specifically, the step 101 above, which involves constructing an influence graph representing the problem domain and the solution domain based on the multiple dialogue texts, may further include the following step 101a:

[0070] Step 101a: Construct a corpus based on a domain knowledge base related to the dialogue topic and the multiple dialogue texts, and construct an influence graph representing the problem domain and the solution domain based on the corpus.

[0071] The nodes of the influence graph include: opportunity nodes, decision nodes, and utility nodes; the nodes of the influence graph are used to represent the information involved in the dialogue text; the directed edges of the influence graph are used to represent the correlation between information.

[0072] For example, in constructing an influence graph, guidelines and introductory texts can be selected as the domain knowledge base, and a corpus can be jointly constructed with texts from multiple discussions. Based on the corpus, an influence graph representing the problem domain and the solution domain is constructed. Nodes in the influence graph are generated using three methods (syntactic analysis, natural language modeling, and manual verification and supplementation), and directed edges in the influence graph are generated using three methods (template-based extraction, natural language modeling, and manual verification and supplementation).

[0073] Specifically, the step of constructing the nodes of the influence graph in step 101a above may also include the following steps 101a1 to 101a3:

[0074] Step 101a1: Using a preset word processing method, combined with a stop word dictionary that has domain knowledge related to the dialogue topic, multiple candidate nodes are generated.

[0075] Step 101a2: Use a natural language model to extract noun phrases from the corpus as the filling content for each candidate node among the multiple candidate nodes, and merge semantically similar nodes to obtain multiple nodes to be classified.

[0076] Step 101a3: Perform a classification operation on the multiple nodes to be classified using a natural language model to obtain the classification result.

[0077] The classification operation is used to divide the nodes of the influence graph into: opportunity nodes, decision nodes, and utility nodes; the preset word processing method includes at least one of the following: word segmentation method and syntactic analysis method.

[0078] For example, in the node generation stage, word segmentation (jieba) and syntactic analysis (spacy) methods are used, combined with a stop dictionary with domain knowledge, to generate a candidate node list, and common phrases and fixed collocations are supplemented by association analysis; nodes are directly extracted from the corpus using a natural language model (for example, the prompt could be "Extract all nodes mentioned in the original text, the form of the nodes is a noun phrase; this is a named entity recognition task, each node to be identified is a combination of an entity and a certain attribute"), and nodes with similar semantics are merged (for example, the prompt could be "For a group of nodes with very similar semantics, please keep only the first node"); the nodes obtained by the first two methods are confirmed and classified (opportunity nodes, decision nodes, utility nodes), and supplemented based on human knowledge.

[0079] Specifically, in step 101a above, the step of constructing the directed edges of the influence graph may further include the following steps 101a4 and 101a5:

[0080] Step 101a4: Extract explicit directed relations from the multiple dialogue texts based on causal templates and statement templates.

[0081] Step 101a5: Use a natural language model to extract correlations from multiple nodes of the influence graph to obtain multiple directed edges of the influence graph.

[0082] The directed edge between any two nodes is obtained based on the causal relationship and / or conditional relationship between the text information corresponding to the two nodes.

[0083] For example, in the directed edge generation stage, explicit directed relationships are captured based on causal templates and predictive statement templates; directed relationships in the node list formed in the previous step are directly extracted using a natural language model (for example, the prompt could be "The existence of directed edges between nodes means that there is a causal or conditional relationship between two nodes, please extract all directed edges in the node list"); the directed edges obtained by the first two methods are confirmed and supplemented based on human knowledge.

[0084] Step 102: Using a heuristic approach, the multiple dialogue texts are processed using a natural language model to assess each participant's comprehension ability, and the problem-solving ability and creative thinking of each participant are assessed based on the influence graph.

[0085] For example, in the embodiments of this application, a heuristic method can be used to evaluate the participants' three levels of comprehension ability based on the output of the natural language model and to construct a scoring system.

[0086] Specifically, step 102 above, which involves processing the multiple dialogue texts using a natural language model and evaluating the comprehension ability of each participant, may further include step 102a:

[0087] Step 102a: Preprocess the dialogue text of each participant and input the processed dialogue text into the natural language model. Use the guide words corresponding to the dialogue topic to control the natural language model to evaluate the three-level comprehension ability of each participant.

[0088] The assessment of comprehension ability is divided into three levels: low-level comprehension ability is assessed by the number of key points appearing in the dialogue text; intermediate-level comprehension ability is assessed by the number of conceptual statements appearing in the dialogue text; and advanced-level comprehension ability is assessed by the number of specific examples in the dialogue text that combine historical and current realities.

[0089] For example, taking the understanding of a certain topic as an example, the basic level of comprehension is measured by the key points in the input text, with each key point worth 1 point; the intermediate level of comprehension is measured by the conceptual explanations in the input text, with each explanation worth 2 points; and the advanced level of comprehension is measured by specific examples in the input text that combine historical and current realities, with each example worth 3 points. The sum of these scores constitutes the participant's assessment of their comprehension and explanatory abilities.

[0090] For example, in the embodiments of this application, problem-solving ability includes at least one of the following: observation ability, comprehension ability, and reasoning ability.

[0091] Specifically, based on the constructed influence map, step 102 above, which involves evaluating each participant's problem-solving ability and creative thinking based on the influence map, may further include at least one of the following steps 102a1 to 102a3:

[0092] 102a1. The observation ability of each participant is evaluated based on the number of opportunity nodes hit by each participant in the influence graph.

[0093] 102a2. Each participant’s comprehension ability is assessed based on the number of utility nodes hit in the influence graph.

[0094] 102a3. Assess each participant's comprehension ability based on the frequency of occurrence of each type of reasoning pattern in each participant's multi-type reasoning patterns.

[0095] The various reasoning patterns include: syllogism, relational reasoning, inductive reasoning, and abductive reasoning; the syllogism pattern in the influence graph is that abstract nodes point to fact nodes; the relational reasoning pattern in the influence graph is that there are a predetermined number of directed edges connecting fact nodes sequentially; the inductive reasoning pattern in the influence graph is that abstract nodes point to abstract nodes, or fact nodes point to abstract nodes; the abductive reasoning pattern in the influence graph is that abstract nodes point to fact nodes, or fact nodes point to abstract nodes, and there is a causal relationship; the fact nodes are nodes in the influence graph that contain factual information; the abstract nodes are nodes in the influence graph that contain abstract information.

[0096] An assessment was conducted on each participant's problem-solving abilities and creative thinking, specifically including:

[0097] 1. Observational ability assessment

[0098] Participants' observational skills are reflected in their comprehensive understanding and detailed examination of the information within the problem, i.e., the comprehensiveness of information consideration. The comprehensiveness of information consideration is measured by the number of opportunity nodes hit in the influence graph.

[0099] 2. Comprehension Ability Assessment

[0100] Participants' comprehension ability is reflected in their grasp of the core values ​​and key pain points involved in the problem, i.e., the comprehensiveness of value consideration. The comprehensiveness of value consideration is measured by the number of utility nodes hit in the influence diagram.

[0101] 3. Reasoning ability assessment

[0102] Participants' reasoning ability is demonstrated by their use of four reasoning models to deeply analyze and explore the problem domain, connecting ideas to form a chain of thought. Reasoning ability is measured by the frequency of the four reasoning models appearing in the discussion text.

[0103] Specifically, based on the constructed influence map, step 102 above, which involves evaluating the creative thinking of each participant based on the influence map, may further include at least one of the following steps 102b1 to 102b3:

[0104] Step 102b1: Based on the dialogue text of each participant, determine the mention order and hit rate of each participant in the opportunity node, decision node, and utility node of the influence graph, and evaluate the inspiration and transfer ability of each participant based on the mention order and the hit rate.

[0105] Step 102b2: Based on the semantic features and mention frequency of the decision nodes corresponding to the dialogue text of each participant in the influence graph, evaluate the innovation and imagination of each participant.

[0106] Step 102b3: Based on the dialogue text of each participant, determine the target number of times each participant will inspire and transfer ideas between value-based decisions and solutions, and evaluate the thinking characteristics of each participant in generating innovative solutions based on the target number of times.

[0107] For example, in the embodiments of this application, the participants' inspiration and transfer abilities can be evaluated by using the mention order and proportion of the three types of nodes in the influence graph, the participants' innovativeness and imagination can be evaluated by using the semantic features and mention frequency of decision nodes, and the participants' thinking characteristics in generating innovative solutions can be evaluated by using the occurrence of value-based decisions and associations between solutions.

[0108] It should be noted that value-based decision-making in the corpus is manifested by the simultaneous appearance of utility nodes and decision nodes (the order is not required), and the pattern in the influence graph is that utility nodes point to decision nodes or decision nodes point to utility nodes; the association between solutions in the corpus is manifested by the simultaneous appearance of two or more decision nodes that are related (semantic similarity is greater than the similarity threshold).

[0109] Specifically, the assessment of each participant's thinking characteristics in generating innovative solutions includes:

[0110] 1. Visualization of inspirational and transferable abilities

[0111] Participants' inspirational and transferable abilities are reflected in their ability to generate high-quality and efficient solutions (decision nodes) in the solution domain based on their observation (opportunity nodes) and understanding (utility nodes) of the problem domain. The order and proportion of mention of the three types of nodes are visualized through timelines and pie charts (ggplot).

[0112] 2. Evaluation of the innovativeness of the solution

[0113] The innovativeness of a participant's solution is reflected in the maximum irrelevance (semantic distance) of their solution, i.e., the breadth of thought. Innovativeness is measured by the ratio of the maximum semantic distance between the hit decision nodes to the maximum semantic distance between all decision nodes.

[0114] 3. Evaluation of the characteristics of the solution generation process

[0115] The characteristics of the participants' solution generation process are the number of times they use two relatively ideal thinking modes (value-based decision-making and association between solutions) for inspiration and transfer.

[0116] 4. Imagination Assessment

[0117] A participant's imagination is reflected in the number of novel, creative solutions they propose that other participants have not proposed or have proposed less of. Imagination is measured by the number of novel (less than any other solution proposed by all participants) solutions mentioned.

[0118] Optionally, in the embodiments of this application, the overall performance of the group can also be measured by the product of the node relevance, the number of hit nodes, and the average degree of high-connection nodes in the discussion text of each group, and the scores of each participant can be normalized based on the overall performance of the group and calibrated between groups.

[0119] For example, in this embodiment of the application, the plurality of dialogue texts are dialogue texts of participants in multiple groups.

[0120] For example, after step 102 above, the method for evaluating the understanding and innovation inspiration ability based on multi-person discussion text provided in this application embodiment may further include the following steps 103 and 104:

[0121] Step 103: Based on the product of the node relevance, number of hit nodes, and average degree of high-connection nodes of the dialogue text of each participant in each of the multiple groups, a comprehensive evaluation is performed on each group to obtain the comprehensive evaluation result of each group.

[0122] Step 104: Perform component calibration based on the overall evaluation results of each group to assess the performance of participants in groups with different dialogue topics.

[0123] For example, node relevance refers to the average similarity between the nodes mentioned by the group and the text in the domain knowledge base, calculated using a pre-trained model (SimBERT); hit node count refers to the total number of nodes mentioned in the group discussion text; and high-connectivity node average refers to the average number of edges connected to (including pointing and pointing) highly connected nodes (those with a large number of connected edges in the influence graph). Inter-group calibration methods can be used to compare the performance of participants in groups with not entirely identical discussion topics.

[0124] The method for evaluating comprehension and innovative inspiration capabilities based on multi-person discussion texts provided in this application first obtains multiple dialogue texts obtained from multi-person discussions, and constructs an influence graph representing the problem domain and solution domain based on these multiple dialogue texts; one dialogue text corresponds to one participant; then, a heuristic method is used to process the multiple dialogue texts using a natural language model to evaluate each participant's comprehension ability, and to evaluate each participant's problem-solving ability and creative thinking based on the influence graph. In this way, a comprehensive and accurate evaluation of the comprehension and innovative inspiration capabilities of each participant in a multi-person discussion scenario can be achieved.

[0125] It should be noted that the method for evaluating the comprehension and innovation inspiration ability based on multi-person discussion text provided in this application embodiment can be executed by a device for evaluating the comprehension and innovation inspiration ability based on multi-person discussion text, or by a control module within that device for executing the method. This application embodiment uses the execution of the method by a device for evaluating the comprehension and innovation inspiration ability based on multi-person discussion text as an example to illustrate the device provided in this application embodiment for evaluating the comprehension and innovation inspiration ability based on multi-person discussion text.

[0126] It should be noted that, in the embodiments of this application, the methods illustrated in the accompanying drawings are all illustrative examples of the methods for assessing the understanding and innovation inspiration capabilities based on multi-person discussion texts, using one accompanying drawing from one of the embodiments of this application as an example. In specific implementations, the methods for assessing the understanding and innovation inspiration capabilities based on multi-person discussion texts illustrated in the accompanying drawings can also be implemented in conjunction with any other accompanying drawings shown in the above embodiments, which will not be elaborated here.

[0127] The following describes the apparatus for assessing the ability to understand and innovate based on multi-person discussion texts provided in this application. The description below corresponds to the method for assessing the ability to understand and innovate based on multi-person discussion texts described above.

[0128] Figure 2 A schematic diagram of the structure of the device for assessing the understanding and creative inspiration ability based on multi-person discussion text provided in this application embodiment is shown below. Figure 2 As shown, it specifically includes:

[0129] The acquisition module 201 is used to acquire multiple dialogue texts obtained from multi-person discussions; the construction module 202 is used to construct an influence map representing the problem domain and solution domain based on the multiple dialogue texts; one participant corresponds to one dialogue text; the evaluation module 203 is used to use heuristic methods and natural language models to process the multiple dialogue texts, evaluate the comprehension ability of each participant, and evaluate the problem-solving ability and creative thinking of each participant based on the influence map.

[0130] Optionally, the construction module 202 is specifically used to construct a corpus based on a domain knowledge base related to the dialogue topic and the multiple dialogue texts, and to construct an influence graph representing the problem domain and the solution domain based on the corpus; wherein, the nodes of the influence graph include: opportunity nodes, decision nodes and utility nodes; the nodes of the influence graph are used to represent the information involved in the dialogue texts; the directed edges of the influence graph are used to represent the correlation between information.

[0131] Optionally, the construction module 202 is specifically used to generate multiple candidate nodes by employing a preset word processing method and combining a stop word dictionary with domain knowledge related to the dialogue topic; the construction module 202 is further used to extract noun phrases from the corpus using a natural language model as the filling content for each candidate node among the multiple candidate nodes, and to merge semantically similar nodes to obtain multiple nodes to be classified; the construction module 202 is further used to perform a classification operation on the multiple nodes to be classified using a natural language model to obtain a classification result; wherein, the classification operation is used to divide the nodes of the influence graph into: opportunity nodes, decision nodes, and utility nodes; the preset word processing method includes at least one of the following: word segmentation method and syntactic analysis method.

[0132] Optionally, the construction module 202 is specifically used to extract explicit directed relations from the multiple dialogue texts based on causal templates and statement templates; the construction module 202 is also specifically used to extract related relations from multiple nodes of the influence graph using a natural language model to obtain multiple directed edges of the influence graph; wherein, the directed edge between any two nodes is obtained based on the causal relationship and / or conditional relationship between the text information corresponding to the two nodes.

[0133] Optionally, the problem-solving ability includes at least one of the following: observation ability, comprehension ability, and reasoning ability; the evaluation module 203 is specifically used to evaluate each participant's observation ability based on the number of opportunity nodes hit in the influence graph; the evaluation module 203 is further used to evaluate each participant's comprehension ability based on the number of utility nodes hit in the influence graph; the evaluation module 203 is further used to evaluate each participant's comprehension ability based on the frequency of occurrence of each type of reasoning pattern in each participant's multi-type reasoning patterns; wherein, the multi-type reasoning patterns include: syllogism, relational reasoning, etc. Inductive reasoning and abductive reasoning; the syllogism in the influence graph is in the form of abstract nodes pointing to fact nodes; the relational reasoning in the influence graph is in the form of directed edges connecting a predetermined number of fact nodes in sequence; the inductive reasoning in the influence graph is in the form of abstract nodes pointing to abstract nodes, or fact nodes pointing to abstract nodes; the abductive reasoning in the influence graph is in the form of abstract nodes pointing to fact nodes, or fact nodes pointing to abstract nodes, and there is a causal relationship; the fact node is: a node in the influence graph that contains factual information; the abstract node is: a node in the influence graph that contains abstract information.

[0134] Optionally, the creative thinking includes at least one of the following: inspiration and transferability, innovation and imagination, and thinking characteristics for generating innovative solutions; the evaluation module 203 is specifically used to determine the mention order and hit ratio of each participant's opportunity nodes, decision nodes, and utility nodes in the influence diagram based on each participant's dialogue text, and to evaluate each participant's inspiration and transferability based on the mention order and hit ratio; the evaluation module 203 is also specifically used to evaluate each participant's innovation and imagination based on the semantic features and mention frequency of the decision nodes corresponding to each participant's dialogue text in the influence diagram; the evaluation module 203 is also specifically used to determine the target number of times each participant inspires and transfers ideas between value-based decisions and solutions based on each participant's dialogue text, and to evaluate each participant's thinking characteristics for generating innovative solutions based on the target number of times.

[0135] Optionally, the evaluation module 203 is specifically used to preprocess the dialogue text of each participant and input the processed dialogue text into a natural language model. The natural language model is controlled by guide words corresponding to the dialogue topic to evaluate the three levels of comprehension ability of each participant. The evaluation of the three levels of comprehension ability includes: evaluating the low-level comprehension ability by the number of key points appearing in the dialogue text, evaluating the intermediate-level comprehension ability by the number of conceptual statements appearing in the dialogue text, and evaluating the high-level comprehension ability by the number of specific examples combining historical and current realities appearing in the dialogue text.

[0136] Optionally, the plurality of dialogue texts are dialogue texts of participants in multiple groups; the evaluation module 203 is further configured to perform a comprehensive evaluation of each group based on the product of the node relevance, the number of hit nodes, and the average degree of high-connection nodes of the dialogue texts of each participant in each group, to obtain a comprehensive evaluation result for each group; the evaluation module 203 is further configured to perform component calibration based on the comprehensive evaluation result of each group, so as to evaluate the performance of participants in groups with different dialogue topics.

[0137] The device for evaluating comprehension and innovative inspiration capabilities based on multi-person discussion texts provided in this application first acquires multiple dialogue texts obtained from multi-person discussions, and constructs an influence graph representing the problem domain and solution domain based on these multiple dialogue texts; one dialogue text corresponds to one participant; then, using a heuristic method, the multiple dialogue texts are processed using a natural language model to evaluate each participant's comprehension ability, and based on the influence graph, each participant's problem-solving ability and creative thinking are evaluated. In this way, a comprehensive and accurate evaluation of the comprehension and innovative inspiration capabilities of each participant in a multi-person discussion scenario can be achieved.

[0138] Figure 3 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 3As shown, the electronic device may include a processor 310, a communications interface 320, a memory 330, and a communication bus 340, wherein the processor 310, communications interface 320, and memory 330 communicate with each other via the communication bus 340. The processor 310 can invoke logical instructions in the memory 330 to execute a method for evaluating comprehension and innovation-inspiring abilities based on multi-person discussion texts. This method includes: acquiring multiple dialogue texts obtained from multi-person discussions, and constructing an influence graph representing the problem domain and solution domain based on the multiple dialogue texts; one dialogue text per participant; using a heuristic method and a natural language model to process the multiple dialogue texts, evaluating each participant's comprehension ability, and evaluating each participant's problem-solving ability and creative thinking based on the influence graph.

[0139] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0140] On the other hand, this application also provides a computer program product, which includes a computer program stored on a computer-readable storage medium. The computer program includes program instructions, and when the program instructions are executed by a computer, the computer is able to execute the above-described methods for evaluating the understanding and innovation inspiration capabilities based on multi-person discussion texts. The method includes: acquiring multiple dialogue texts obtained from multi-person discussions, and constructing an influence map representing the problem domain and solution domain based on the multiple dialogue texts; one dialogue text per participant; using a heuristic method and a natural language model to process the multiple dialogue texts, evaluating the understanding ability of each participant, and evaluating the problem-solving ability and creative thinking of each participant based on the influence map.

[0141] In another aspect, this application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, is implemented to perform the aforementioned methods for evaluating comprehension and innovation-inspiring abilities based on multi-person discussion texts. The method includes: acquiring multiple dialogue texts obtained from multi-person discussions, and constructing an influence graph representing a problem domain and a solution domain based on the multiple dialogue texts; one dialogue text per participant; using a heuristic method and a natural language model to process the multiple dialogue texts, evaluating the comprehension ability of each participant, and evaluating each participant's problem-solving ability and creative thinking based on the influence graph.

[0142] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0143] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0144] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for assessing comprehension and creative inspiration abilities based on multi-person discussion texts, characterized in that, include: Obtain multiple dialogue texts based on multi-person discussions, and construct an influence graph representing the problem domain and solution domain based on the multiple dialogue texts; one participant corresponds to one dialogue text; A heuristic approach was adopted, using a natural language model to process the multiple dialogue texts, to assess each participant's comprehension ability, and to evaluate each participant's problem-solving ability and creative thinking based on the influence graph. The problem-solving ability includes: reasoning ability; The assessment of each participant's problem-solving ability based on the influence diagram includes: Each participant's reasoning ability is assessed based on the frequency of each reasoning pattern appearing in the multi-type reasoning patterns of each participant; The various reasoning patterns include: syllogism, relational reasoning, inductive reasoning, and abductive reasoning; the syllogism pattern in the influence graph is that abstract nodes point to fact nodes; the relational reasoning pattern in the influence graph is that there are a predetermined number of directed edges connecting fact nodes sequentially; the inductive reasoning pattern in the influence graph is that abstract nodes point to abstract nodes, or fact nodes point to abstract nodes; the abductive reasoning pattern in the influence graph is that abstract nodes point to fact nodes, or fact nodes point to abstract nodes, and there is a causal relationship; the fact nodes are nodes in the influence graph that contain factual information; the abstract nodes are nodes in the influence graph that contain abstract information. The construction of an influence graph representing the problem domain and solution domain based on the multiple dialogue texts includes: A corpus is constructed based on a domain knowledge base related to the dialogue topic and the multiple dialogue texts, and an influence graph representing the problem domain and solution domain is constructed based on the corpus. The nodes of the influence graph include: opportunity nodes, decision nodes, and utility nodes; the nodes of the influence graph are used to represent the information involved in the dialogue text; the directed edges of the influence graph are used to represent the correlation between information.

2. The method according to claim 1, characterized in that, The construction of the influence graph representing the problem domain and solution domain based on the corpus includes: A pre-defined word processing method is used, combined with a stop word dictionary that has domain knowledge related to the dialogue topic, to generate multiple candidate nodes; Natural language models are used to extract noun phrases from the corpus as the filling content for each candidate node among the multiple candidate nodes, and semantically similar nodes are merged to obtain multiple nodes to be classified. A natural language model is used to perform a classification operation on the multiple nodes to be classified, and the classification result is obtained. The classification operation is used to divide the nodes of the influence graph into: opportunity nodes, decision nodes, and utility nodes; the preset word processing method includes at least one of the following: word segmentation method and syntactic analysis method.

3. The method according to claim 1 or 2, characterized in that, The construction of the influence graph representing the problem domain and solution domain based on the corpus includes: Explicit directed relationships are extracted from the multiple dialogue texts based on causal templates and statement templates; By using a natural language model, correlations are extracted from multiple nodes of the influence graph to obtain multiple directed edges of the influence graph; The directed edge between any two nodes is obtained based on the causal relationship and / or conditional relationship between the text information corresponding to the two nodes.

4. The method according to claim 3, characterized in that, The problem-solving ability also includes: observation ability, and / or comprehension ability; The assessment of each participant's problem-solving ability based on the influence diagram includes: Each participant's observation ability is evaluated based on the number of opportunity nodes hit in the influence graph. And / or, Each participant's comprehension ability is assessed based on the number of utility nodes hit in the influence graph.

5. The method according to claim 3, characterized in that, The creative thinking mentioned includes at least one of the following: the ability to inspire and transfer ideas, innovation and imagination, and the thinking characteristics of generating innovative solutions; The assessment of each participant's creative thinking based on the influence map includes: Based on each participant's dialogue text, the order and hit rate of each participant's mention of opportunity nodes, decision nodes, and utility nodes in the influence graph are determined, and the inspiration and transferability of each participant are evaluated based on the mention order and the hit rate. And / or, Based on the semantic features and mention frequency of the decision nodes corresponding to the dialogue text of each participant in the influence graph, the innovativeness and imagination of each participant are evaluated. And / or, Based on each participant's dialogue text, the target number of times each participant inspires and transfers value-based decisions and associations between solutions is determined, and the thinking characteristics of each participant in generating innovative solutions are evaluated based on the target number of times.

6. The method according to claim 1, characterized in that, The method employs a heuristic approach, utilizing a natural language model to process the multiple dialogue texts and assess each participant's comprehension ability, including: The dialogue text of each participant is preprocessed and then input into a natural language model. Guide words corresponding to the dialogue topic are used to control the natural language model to evaluate the three-level comprehension ability of each participant. The assessment of comprehension ability is divided into three levels: low-level comprehension ability is assessed by the number of key points appearing in the dialogue text; intermediate-level comprehension ability is assessed by the number of conceptual statements appearing in the dialogue text; and high-level comprehension ability is assessed by the number of specific examples in the dialogue text that combine historical and current realities.

7. The method according to claim 1, characterized in that, The multiple dialogue texts are the dialogue texts of participants from multiple groups; The method, which employs a heuristic approach, utilizes a natural language model to process the multiple dialogue texts, assesses each participant's comprehension ability, and evaluates each participant's problem-solving ability and creative thinking based on the influence graph, further includes: Based on the product of the node relevance, number of hit nodes, and average degree of high-connection nodes of the dialogue text of each participant in each of the multiple groups, a comprehensive evaluation is performed on each group to obtain the comprehensive evaluation result of each group. Component calibration is performed based on the overall evaluation results of each group to assess the performance of participants in groups with different dialogue topics.

8. A device for assessing comprehension and creative inspiration abilities based on multi-person discussion texts, characterized in that, The device includes: The acquisition module is used to acquire multiple dialogue texts obtained based on multi-person discussions; The construction module is used to build an influence graph representing the problem domain and solution domain based on the multiple dialogue texts; one participant corresponds to one dialogue text; The evaluation module is used to process the multiple dialogue texts using a heuristic method and a natural language model to evaluate each participant's comprehension ability, and to evaluate each participant's problem-solving ability and creative thinking based on the influence graph; the problem-solving ability includes reasoning ability. The evaluation module is also specifically used to evaluate each participant's reasoning ability based on the number of times each type of reasoning pattern appears in each participant's multi-type reasoning patterns. The various reasoning patterns include: syllogism, relational reasoning, inductive reasoning, and abductive reasoning; the syllogism pattern in the influence graph is that abstract nodes point to fact nodes; the relational reasoning pattern in the influence graph is that there are a predetermined number of directed edges connecting fact nodes sequentially; the inductive reasoning pattern in the influence graph is that abstract nodes point to abstract nodes, or fact nodes point to abstract nodes; the abductive reasoning pattern in the influence graph is that abstract nodes point to fact nodes, or fact nodes point to abstract nodes, and there is a causal relationship; the fact nodes are nodes in the influence graph that contain factual information; the abstract nodes are nodes in the influence graph that contain abstract information. The construction module is specifically used to extract explicit directed relationships from the multiple dialogue texts based on causal templates and statement templates; The construction module is further configured to use a natural language model to extract correlations from multiple nodes of the influence graph to obtain multiple directed edges of the influence graph. The directed edge between any two nodes is obtained based on the causal relationship and / or conditional relationship between the text information corresponding to the two nodes.

9. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method for assessing comprehension and innovation inspiration capabilities based on multi-person discussion text as described in any one of claims 1 to 7.

Images