Conversation analysis system and method

Abstract

The invention provides a dialogue analysis system and method, and the system comprises an obtaining module which is used for obtaining a text sequence; wherein the text sequence is obtained by performing real-time text conversion on classroom dialogues; the matching module is used for searching whether a target sequence matched with the text sequence exists or not through a large language model in a pre-established rule base; wherein the rule base comprises classification rule sequences of multiple dialogue types and / or reference text sequences corresponding to each dialogue type; and the analysis module is used for performing analysis according to the target sequence matched by the matching module to obtain a target dialogue type corresponding to the classroom dialogue. According to the method, the rule base and the large language model can be combined, usability and flexibility of classroom dialogue analysis are achieved, deep integration of the dialogue analysis system on the education theory is ensured, complex dialogue sequences can be processed through the large language model, and accurate dialogue classification and analysis are achieved.

Description

Technical Field

[0001] The present invention relates to the field of natural language processing, and in particular to a conversation analysis system and method. Background Art

[0002] In the field of educational technology, the quality of classroom conversations is an important factor affecting student learning outcomes. High-quality classroom conversations can promote students’ critical thinking, deep learning, and collaboration.

[0003] However, the process of systematically analyzing these conversations is extremely complex and time-consuming, especially the analysis of classroom conversation sequences. Traditional analysis methods mainly rely on manual coding, which requires a lot of manual participation, is time-consuming and easily affected by subjective factors.

[0004] In addition, existing automated analysis tools are typically based on machine learning and natural language processing techniques (such as convolutional neural networks and long short-term memory networks). Although these technologies have made progress in dialogue pattern recognition, they often require a large amount of data for training, are difficult to adapt to changes in different educational scenarios, and lack in-depth integration with teaching theories. Summary of the invention

[0005] In view of this, this specification provides a conversation analysis system and method.

[0006] According to a first aspect of an embodiment of this specification, a conversation analysis system is provided, the system comprising:

[0007] An acquisition module, used for acquiring a text sequence; wherein the text sequence is obtained by performing real-time text conversion on classroom conversations;

[0008] A matching module, used to search in a pre-established rule base, through a large language model, whether there is a target sequence matching the text sequence; wherein the rule base includes classification rule sequences of multiple dialogue types and / or reference text sequences corresponding to each dialogue type;

[0009] The analysis module is used to analyze and obtain the target dialogue type corresponding to the classroom dialogue according to the target sequence matched by the matching module.

[0010] According to a second aspect of an embodiment of this specification, a conversation analysis method is provided, the method comprising:

[0011] Acquire a text sequence; wherein the text sequence is obtained by performing real-time text conversion on classroom conversations;

[0012] In a pre-established rule base, searching through a large language model whether there is a target sequence matching the text sequence; wherein the rule base includes classification rule sequences of multiple dialogue types and / or reference text sequences corresponding to each dialogue type;

[0013] According to the matched target sequence, the target dialogue type corresponding to the classroom dialogue is analyzed and obtained.

[0014] According to a third aspect of the embodiments of this specification, there is provided a conversation analysis device, the device comprising: one or more processors; wherein the processor is used to execute the steps of the method as described in any one of the second aspects.

[0015] According to a fourth aspect of the embodiments of this specification, a computer-readable storage medium is provided, on which computer instructions are stored, and when the instructions are executed by a processor, the steps of the method described in any one of the second aspects are implemented.

[0016] According to a fifth aspect of the embodiments of this specification, a computer program product is provided, comprising a computer program / instruction, which, when executed by a processor, implements the steps of the method described in any one of the second aspects.

[0017] The technical solutions provided by the embodiments of this specification may have the following beneficial effects:

[0018] In the embodiments of this specification, the rule library can be combined with a large language model to achieve the availability and flexibility of classroom dialogue analysis, which not only ensures the deep integration of the dialogue analysis system into educational theory, but also can process complex dialogue sequences through the large language model to achieve accurate dialogue classification and analysis.

[0019] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present specification. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the specification and, together with the description, serve to explain the principles of the specification.

[0021] Figure 1 This is one of the structural diagrams of a conversation analysis system provided by an exemplary embodiment.

[0022] Figure 2 This is the second structural diagram of a conversation analysis system provided by an exemplary embodiment.

[0023] Figure 3 This is the third structural diagram of a conversation analysis system provided by an exemplary embodiment.

[0024] Figure 4 This is the fourth structural diagram of a conversation analysis system provided by an exemplary embodiment.

[0025] Figure 5 This is one of the flowcharts of a conversation analysis method provided by an exemplary embodiment.

[0026] Figure 6 This is the second flowchart of a conversation analysis method provided by an exemplary embodiment.

[0027] Figure 7 This is the third flowchart of a conversation analysis method provided by an exemplary embodiment.

[0028] Figure 8 This is the fifth structural diagram of a conversation analysis system provided by an exemplary embodiment.

[0029] Fig. 9 It is a structural diagram of a server provided by an exemplary embodiment. DETAILED DESCRIPTION

[0030] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with this specification. Instead, they are merely examples of devices and methods consistent with some aspects of this specification as detailed in the appended claims.

[0031] The terms used in this specification are for the purpose of describing specific embodiments only and are not intended to limit this specification. The singular forms "a", "the" and "the" used in this specification and the appended claims are also intended to include plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0032] It should be understood that although the terms first, second, third, etc. may be used in this specification to describe various information, this information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of this specification, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0033] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0034] Currently, language understanding and conversation analysis can be performed in the following ways:

[0035] Method 1: Manual coding.

[0036] The earliest conversation analysis relied on manual coding analysis by human experts. Although this method has certain flexibility and accuracy, it becomes increasingly unfeasible as class size and conversation complexity increase.

[0037] Method 2: Rule-based Artificial Intelligence (AI).

[0038] Early AI systems relied on predefined rules to operate. These systems can work effectively in relatively closed environments, but their performance is limited in dynamic, complex classroom environments because they lack the ability to adapt to new situations.

[0039] Method 3: the introduction of machine learning and large language models.

[0040] With the rise of machine learning and deep learning, the field of education has also begun to introduce data-driven analysis models, such as Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM).

[0041] The specific implementation method may include the following steps:

[0042] Dataset creation: First, we record classroom conversations and transcribe them to generate a conversation dataset for training the model.

[0043] Model training: Use these datasets to train CNN or LSTM, and the model performs classification by learning patterns in the conversation.

[0044] Automatic coding: After training, the model can automatically encode and classify newly input classroom conversations, thereby generating an analysis report on the structure of classroom conversations.

[0045] Although these models can handle large amounts of data, they have several major drawbacks:

[0046] Insufficient integration of educational theory: Most automated dialogue analysis technologies rely on data-driven learning and lack in-depth integration with educational theory. Therefore, although these systems can provide analysis of dialogue patterns in general, they cannot conduct in-depth analysis based on specific teaching objectives and classroom needs, which limits their application in educational scenarios.

[0047] Insufficient ability to identify dialogue sequences: It focuses more on the classification of single dialogue segments and cannot effectively analyze complex dialogue sequences. Dialogue sequences are crucial in education because students' thinking development is often reflected in the process of continuous dialogue, and it is impossible to identify and classify these continuous dialogue patterns.

[0048] Model dependency and flexibility issues: Since automated analysis techniques are highly dependent on pre-trained models, these models usually require a large amount of labeled data to be effectively trained. However, in the field of education, the acquisition cost of labeled data is high, and this data is often context-dependent. The model performs well in one scenario, but has poor adaptability in different classroom scenarios.

[0049] Lack of real-time feedback: Machine learning models usually run in an offline environment, and the analysis results often need to be generated after the class ends. They cannot provide teachers with real-time classroom feedback, limiting their application value in actual teaching.

[0050] In order to automatically analyze classroom dialogue sequences and provide real-time theoretical support feedback, thereby providing effective support for improving the quality of classroom teaching. This specification proposes a dialogue analysis system and method, which can achieve the usability and flexibility of classroom dialogue analysis based on a system combining a rule base and a large language model, which not only ensures the deep integration of the dialogue analysis system with educational theory, but also can process complex dialogue sequences through a large language model to achieve accurate dialogue classification and analysis.

[0051] The system and method provided in this specification can be applied to various educational scenarios, especially in primary and secondary school classrooms, to analyze the conversations between students and teachers, and the discussions between students. In addition, it can also be widely used in teacher professional development projects to help teachers obtain real-time feedback on their classroom conversations, thereby improving teaching effectiveness.

[0052] It is understandable that the system and method provided in this specification can also be applied to other scenarios, including but not limited to corporate training, psychological counseling, medical counseling, social media analysis, customer service, etc. Any solution that combines the rule base with the large language model for dialogue analysis should fall within the scope of protection of this disclosure.

[0053] The following is an introduction to the conversation analysis system provided by the present disclosure.

[0054] Figure 1 FIG. 1 is a schematic diagram of a conversation analysis system provided by an exemplary embodiment. Figure 1 As shown, the system may include:

[0055] Acquisition module 101;

[0056] Matching module 102;

[0057] Analysis module 103.

[0058] In some embodiments, the acquisition module 101 may be used to acquire a text sequence, wherein the text sequence is obtained by performing real-time text conversion on classroom conversations.

[0059] In one example, the acquisition module 101 may be connected to an audio acquisition device, such as a microphone, to acquire audio signals of classroom conversations through the audio acquisition device, and convert the acquired audio signals into text sequences in real time.

[0060] In some embodiments, the matching module 102 may be used to search a pre-established rule base through a large language model to see whether there is a target sequence matching the text sequence.

[0061] In one example, a large language model (LLM) refers to a deep learning model trained with a large amount of text data that can generate natural language text or understand the meaning of language text. The large language model can handle a variety of natural language tasks, such as text classification, question answering, and dialogue. In this specification, the large language model can be used to perform text analysis on a text sequence and find a matching target sequence in the rule base. The large language model can also be used to generate and output a dialogue analysis report later.

[0062] In one example, the conversation analysis system may collect a large amount of literature related to educational theory, analyze and summarize the literature, and thus create the rule base.

[0063] Exemplarily, the rule base may include classification rule sequences for multiple dialog types and / or reference text sequences corresponding to each dialog type.

[0064] Exemplarily, the types of dialogue involved in this specification include, but are not limited to, at least one of the following: critical inquiry dialogue; knowledge collaborative construction dialogue; guiding dialogue; supportive dialogue; reflective dialogue; and metacognitive dialogue.

[0065] Among them, critical inquiry dialogue can refer to a dialogue in which participants constructively challenge and counter-challenge each other's views during the discussion or learning process to promote in-depth understanding and resolution of problems. It emphasizes that participants question existing views or assumptions, and through rational dialogue and reflection, critically examine and evaluate various arguments, thereby promoting the collaborative construction and innovation of knowledge.

[0066] Among them, knowledge collaborative construction dialogue can refer to a dialogue in which participants use knowledge from multiple disciplines to collaboratively solve the same problem during the discussion or learning process.

[0067] Among them, guiding dialogue can refer to a guiding dialogue that can help participants think of solutions to problems or even reflect on them.

[0068] Among them, supportive dialogue can refer to a dialogue in which when at least one participant proposes a theory, other participants support their theory by summarizing the arguments, providing evidence, etc.

[0069] Among them, reflective dialogue can refer to a dialogue in which participants think about the impact of the problem on other people or other areas while solving the problem.

[0070] Among them, metacognitive dialogue can refer to the dialogue on "metacognition", among which metacognition refers to the cognition of cognition. Specifically for learning, metacognition refers to the recognition and control of thinking and learning activities, including metacognitive knowledge and metacognitive control.

[0071] The above description is merely an example, and the conversation type may also include other types, which are not limited in this specification.

[0072] Exemplarily, the classification rule sequence in this specification may refer to the rules that the conversation complies with when classifying the conversation. For example, when a keyword matching a certain conversation type A appears in the conversation, it can be determined that the conversation belongs to conversation type A.

[0073] For example, when guiding keywords appear in a conversation, the conversation can be classified as a guiding conversation. Among them, guiding keywords include "Can you give an example?" and "Is there anything else to add?"

[0074] For another example, when a supportive keyword appears in a conversation, the conversation can be classified as a supportive conversation, wherein the supportive keyword includes, for example, "agree", "approve", "that's right", etc.

[0075] For another example, when critical keywords appear in a conversation, the conversation can be classified as a critical inquiry conversation. Among them, critical keywords include "this approach is wrong", "this idea is wrong", "this brings negative effects", etc.

[0076] Exemplarily, the classification rule sequence may be used to indicate at least one of the following: a minimum number of dialogue turns corresponding to each dialogue type; and a dialogue structure corresponding to each dialogue type.

[0077] For example, the minimum number of dialogue turns for a critical inquiry dialogue is 3, and the corresponding dialogue structure includes: (Reasoning Invitation (REI), Elaboration Invitation (ELI)) and argumentation (Reasoning (RE), Elaboration (EL)), and includes querying (Q) behavior.

[0078] For another example, the minimum number of dialogue turns corresponding to a reflective dialogue is 2, and the corresponding dialogue structure includes: (Reasoning (RE), Agreement (A)), and may include Querying (Q) behavior.

[0079] The above description is merely an exemplary description, and the present disclosure does not limit the classification rules of conversation types.

[0080] Exemplarily, each dialogue type may correspond to at least one reference text sequence, which may also be referred to as an "ideal text sequence", and is the most appropriate and reasonable text sequence obtained by performing sequence prediction based on the connotation interpretation of the dialogue type classification and the encoding of the dialogue function.

[0081] For example,

[0082] The reference text sequence corresponding to a certain dialogue type may be at least one of the following:

[0083] REI→RE→Q: Invitation to reason, followed by reasoning, and then questioning of the reasoning.

[0084] Q→RE→REI: Question a statement, follow up with reasoning, and then invite further reasoning.

[0085] Co-ordination Invitation (CI) → Q → RE: Co-ordination Invitation, followed by Question, and then Reasoning.

[0086] The above description is merely an exemplary description, and the present disclosure does not limit the specific content included in the rule base.

[0087] In one example, the matching module 102 may first annotate the text content of the text sequence through a large language model to obtain an annotation result. Further, in the rule base, the large language model may be used to search whether there is a target sequence that matches the annotation result.

[0088] Exemplarily, the text sequence converted from the classroom conversation may include at least one text subsequence, and each text subsequence corresponds to a round of conversation in the classroom conversation.

[0089] Exemplarily, the large language model annotates the text content for each text subsequence (i.e., each round of dialogue). Specifically, the large language model can encode the text content of each text subsequence and convert the text subsequence into codewords that the large language model can understand. These codewords constitute the annotation results.

[0090] Among them, the large language model can obtain the above-mentioned annotation results according to a predefined dialogue encoding table (a mapping relationship table between text and codewords).

[0091] Exemplarily, if the annotation result matches the minimum number of dialogue turns corresponding to at least one dialogue type, it can be determined that the target sequence exists.

[0092] Exemplarily, if the annotation result matches the dialogue structure corresponding to at least one dialogue type, it can be determined that the target sequence exists.

[0093] Exemplarily, if the annotation result matches the minimum dialogue turns and / or dialogue structure corresponding to at least one dialogue type, it can be determined that the target sequence exists.

[0094] Exemplarily, if the annotation result matches a reference text sequence corresponding to at least one dialogue type, it can be determined that the target sequence exists.

[0095] Exemplarily, if the annotation result matches the minimum number of dialogue turns corresponding to at least one dialogue type and the reference text sequence, it can be determined that the target sequence exists.

[0096] Exemplarily, if the annotation result matches the dialogue structure corresponding to at least one dialogue type and the reference text sequence, it can be determined that the target sequence exists.

[0097] Exemplarily, if the annotation result matches the minimum dialogue turns, dialogue structure, and reference text sequence corresponding to at least one dialogue type, it can be determined that the target sequence exists. In some embodiments, if the matching module 102 matches at least one target sequence, the analysis module 103 can analyze and obtain the target dialogue type to which the classroom dialogue belongs based on the target sequence.

[0098] For example, the classroom conversation includes three rounds of conversations, and the target sequence matched by the matching module 102 is REI→RE→Q, then the analysis module 103 can determine that the classroom conversation belongs to a critical inquiry conversation.

[0099] It is understandable that in this specification, the large language model no longer needs to train the sample data set in order to achieve the classification of classroom dialogues, that is, there is no need to manually annotate the sample dialogue text set, saving labor costs. In this specification, the large language model can directly match the sequence in the rule library according to the standard results to determine the target dialogue type corresponding to the classroom dialogue, thereby achieving accurate dialogue classification and analysis.

[0100] In the above embodiment, the rule library can be combined with the large language model to achieve the availability and flexibility of classroom dialogue analysis, which not only ensures the deep integration of the dialogue analysis system into educational theory, but also can process complex dialogue sequences through the large language model to achieve accurate dialogue classification and analysis.

[0101] In some embodiments, Figure 2 is based on Figure 1 The structural diagram of another conversation analysis system provided by the embodiment shown is as follows. Figure 2 As shown, the system may also include:

[0102] Processing module 104.

[0103] In one example, the processing module 104 may be used to preprocess the converted text sequence, where the preprocessing at least includes text proofreading, which includes but is not limited to proofreading text content, text structure, and text format.

[0104] In the above embodiment, the system may further include a processing module to pre-process the text sequence, thereby improving the reliability of classroom dialogue analysis.

[0105] In some embodiments, Figure 3 is based on Figure 1 The structural diagram of another conversation analysis system provided by the embodiment shown is as follows. Figure 3 As shown, the system may also include:

[0106] Output module 105.

[0107] In an example, the output module 105 may be configured to generate and output a conversation analysis report based on the target conversation type using a large language model.

[0108] Exemplarily, the conversation analysis report includes but is not limited to at least one of the following:

[0109] The target conversation type to which each conversation round in the classroom conversation belongs;

[0110] the proportion of each target conversation type;

[0111] Results of a content assessment of the classroom conversation.

[0112] The content evaluation results of the classroom conversation may include but are not limited to the advantages of the classroom conversation and / or areas that need improvement.

[0113] In the above embodiment, a real-time analysis report of classroom conversations can be provided in real time, which is helpful for understanding classroom content and promoting interaction between teachers and students, and has high usability.

[0114] In some embodiments, Figure 4 is based on Figure 1 The structural diagram of another conversation analysis system provided by the embodiment shown is as follows. Figure 4 As shown, the system may also include:

[0115] Processing module 104;

[0116] Output module 105.

[0117] In the above embodiment, the conversation analysis system can process and analyze conversation data in real time during the class, and generate a conversation analysis report to help teachers adjust teaching strategies in a timely manner. This real-time feedback function greatly improves the practicality of the system and enhances the interactivity and flexibility of classroom teaching.

[0118] Figure 5 FIG. 1 is a flow chart of a conversation analysis method provided by an exemplary embodiment. Figure 5 As shown, the method may include the following steps:

[0119] Step S501, obtaining a text sequence.

[0120] In some embodiments, the text sequence is obtained by real-time text conversion of classroom conversations. The process of obtaining the text sequence is similar to the action performed by the acquisition module 101, and will not be repeated here.

[0121] Step S503: searching in a pre-established rule base using a large language model whether there is a target sequence matching the text sequence.

[0122] In some embodiments, the rule base includes classification rule sequences of multiple dialogue types and / or reference text sequences corresponding to each dialogue type. The process of matching the target sequence is similar to the action performed by the matching module 102 and will not be repeated here.

[0123] Step S504: Analyze and obtain the target dialogue type corresponding to the classroom dialogue according to the matched target sequence.

[0124] In some embodiments, the process of analyzing and obtaining the target conversation type is similar to the action performed by the analysis module 103 and will not be described in detail here.

[0125] In the above embodiment, the rule library can be combined with the large language model to achieve the availability and flexibility of classroom dialogue analysis, which not only ensures the deep integration of the dialogue analysis system into educational theory, but also can process complex dialogue sequences through the large language model to achieve accurate dialogue classification and analysis.

[0126] Figure 6 is based on Figure 5 The embodiment shown is a flow chart of another method for conversation analysis. Figure 6 As shown, the method may include the following steps:

[0127] Step S502: preprocess the text sequence.

[0128] In some embodiments, the preprocessing includes at least text proofreading. The preprocessing process is similar to the actions performed by the processing module 104 and will not be described in detail here.

[0129] In the above embodiment, the text sequence can be preprocessed to improve the reliability of classroom conversation analysis.

[0130] Figure 7 is based on Figure 6 The embodiment shown is a flow chart of another method for conversation analysis. Figure 7 As shown, the method may include the following steps:

[0131] Step S505: Generate and output a conversation analysis report based on the target conversation type using the large language model.

[0132] In some embodiments, the process of outputting the conversation analysis report is similar to the action performed by the output module 105 and will not be described in detail here.

[0133] In the above embodiment, the conversation data can be processed and analyzed in real time during the class, and a conversation analysis report can be generated to help teachers adjust teaching strategies in a timely manner. This real-time feedback function greatly improves the practicality of the system and enhances the interactivity and flexibility of classroom teaching.

[0134] The above process is further illustrated as follows.

[0135] This specification provides a classroom dialogue analysis system based on (expert) rule base and large language model. By combining the knowledge of education experts with the natural language processing capabilities of large language models, the system can automatically identify and analyze classroom dialogue sequences and provide real-time feedback. The specific technical implementation scheme is as follows:

[0136] Creation of expert rule base:

[0137] The system first creates an expert rule base based on a large amount of educational theory and existing literature. The rule base contains classification criteria for different types of dialogues and ideal patterns for dialogue sequences. These rules are defined according to educational goals, such as exploratory dialogue, collaborative knowledge construction, reflective dialogue, etc., to ensure that the system's analysis is consistent with educational theory.

[0138] Introduction of large language model:

[0139] Large language models (such as GPT) are used to understand the natural language in classroom conversations and combined with expert knowledge in the rule base. This combination enables the system to automatically classify classroom conversations using both symbolic reasoning and flexible language understanding.

[0140] The dialogue sequence analysis process includes:

[0141] Step 1, data collection and preprocessing: obtain classroom conversation data through classroom recordings or classroom conversation transcriptions, and perform text preprocessing to ensure the data is structured.

[0142] Step 2: Identify the type of conversation: Use a large language model to process the conversation data, identify the type of conversation (such as exploratory conversation, collaborative conversation, etc.), and match it with the rules in the expert rule base.

[0143] Step 3, sequence pattern analysis: Through the dialogue sequence definition in the rule base, the system classifies continuous dialogue segments into sequences and identifies complex dialogue patterns, such as teacher-student questions and answers, discussions between students, deep reasoning, etc.

[0144] Step 4, real-time feedback: Based on the analysis results of the conversation, the system provides teachers with a real-time conversation analysis report, pointing out the strengths and areas for improvement in the classroom conversation.

[0145] System architecture and implementation:

[0146] This system, for example Figure 8 As shown, it mainly includes a data processing module 801, an expert rule base module 802, a natural language processing module 803, and a feedback generation module 804. Each module cooperates with each other to realize the whole process processing from data collection, analysis to feedback.

[0147] The data processing module 801 may convert the audio signal of the classroom conversation into a text sequence and pre-process the text sequence.

[0148] The expert rule base module 802 may store classification rule sequences of multiple dialog types and / or an ideal dialog sequence corresponding to each dialog type.

[0149] Among them, the natural language processing module 803 can be a large language model, which can determine the target sequence that matches the text sequence in the sequence provided by the expert rule base module 802, and determine the target dialogue type corresponding to the classroom dialogue based on the target sequence to achieve classification of complex dialogues.

[0150] The feedback generation module 804 may generate and output a conversation analysis report.

[0151] Algorithm implementation:

[0152] The rule base module uses a symbolic reasoning mechanism based on expert knowledge and combines it with a large language model to perform semantic analysis on the conversation content. The system uses a hierarchical rule matching algorithm to screen and classify conversation sequences step by step to ensure the accuracy of classification and the rationality of sequence recognition.

[0153] In the above embodiment, the accuracy and flexibility of dialogue analysis are improved. By combining the symbolic expert rule base with the large language model, the system can accurately identify complex sequences in classroom dialogues and classify and analyze them according to specific educational goals. Compared with the existing purely data-driven model, this system has better adaptability and generalizability in different classroom situations. In addition, the dependence on labeled data is reduced: the existing machine learning models usually rely on a large amount of labeled data, while the system reduces the dependence on large-scale training data by introducing the expert rule base, and can perform effective dialogue analysis when the data is limited. And real-time feedback is provided. The system can process classroom dialogue data in real time and generate instant dialogue analysis reports to help teachers adjust teaching strategies according to students' dialogue performance and improve the quality of classroom teaching. In addition, it supports the deep integration of educational theories. By combining the expert rule base of educational theories, the present invention can not only perform dialogue analysis, but also provide theoretical support according to educational goals, and provide a basis for teachers' teaching decisions.

[0154] In some embodiments, the present specification also provides a conversation analysis system based on full machine learning.

[0155] In this solution, the system relies entirely on machine learning models rather than combining expert rule bases. Although this solution can reduce human intervention, it does not rely on the knowledge of education experts, resulting in weak accuracy and theoretical support for dialogue analysis. At the same time, the model requires a lot of training time and annotated data, making it difficult to adapt to different educational situations.

[0156] In some embodiments, this specification also provides a conversation analysis system based on a predefined rule set. In this solution, the system relies entirely on a predefined rule set, does not use a large language model, and the conversation analysis report can be manually generated. Although this solution can provide relatively accurate analysis in specific scenarios, it has poor flexibility and is difficult to handle complex and changeable classroom conversations, and is not adaptable enough to new educational situations.

[0157] In this manual, it is necessary to explain:

[0158] (1) Combination of expert rule base and large language model.

[0159] The core innovation of this invention is to combine the symbolic expert rule base with the large language model to achieve flexible and theoretically supported classroom dialogue analysis. This innovation not only ensures the deep integration of educational theory into the system, but also can process complex dialogue sequences through the large language model to achieve accurate classification and analysis.

[0160] (2) Real-time feedback mechanism.

[0161] The system can process and analyze conversation data in real time during class and generate conversation analysis reports to help teachers adjust teaching strategies in a timely manner. This real-time feedback function greatly improves the practicality of the system and enhances the interactivity and flexibility of classroom teaching.

[0162] (3) Automatic classification and recognition of dialogue sequences.

[0163] Through the expert rule base, the system can effectively identify and classify continuous dialogue sequences, capturing the subtle changes in students' thinking development and classroom interaction. This function overcomes the shortcomings of insufficient dialogue sequence recognition ability.

[0164] Example 1, scenario: primary school mathematics class, theme is "Preliminary understanding of fractions".

[0165] Teacher: Today we are going to learn a new mathematical concept - fraction. Do you know what a fraction is?

[0166] Student A: A fraction is the number by which something is divided. For example, if an apple is divided into two halves, it is one-half.

[0167] Student B: I know that fractions can also be used to represent how many pieces a pie is cut into and how much of the whole pie each piece occupies.

[0168] Teacher: Very good, the students have a good understanding of fractions. So, what are the characteristics of fractions?

[0169] Student C: A fraction consists of a numerator and a denominator. The numerator represents the part to be divided, and the denominator represents the number of parts the whole is divided into.

[0170] Student D: I think fractions can be compared, for example, one half is larger than one third.

[0171] Teacher: The students have said very well. Next, let's practice how to write fractions and how to compare the size of fractions.

[0172] System analysis results:

[0173] Dialogue Type: Guidance Dialogue

[0174] Discourse roles: teacher (guide), student A (statement maker), student B (statement maker), student C (statement maker), student D (statement maker).

[0175] Discourse function: Teacher: ask questions, guide students to think, and organize exercises; Students A, B, C, D: state their views, express their ideas, and participate in exercises.

[0176] Sequential mode: teacher-student question and answer, student discussion, teacher-organized exercises

[0177] The conversation analysis report includes the following:

[0178] 1. Classroom dialogue analysis

[0179] The main type of classroom dialogue is exploratory dialogue. Teachers guide students to think and express themselves through questions. Students actively participate in discussions and the classroom atmosphere is lively.

[0180] Students are able to understand the concept of fractions and give examples to illustrate the application of fractions, which reflects the extent to which they have mastered the new knowledge.

[0181] Students are able to state the characteristics of fractions, such as that fractions are composed of numerators and denominators, and that fractions can be compared in size, which reflects students' logical thinking ability.

[0182] 2. Student Engagement Analysis

[0183] Students are highly involved, speak actively, express themselves fluently, and are able to use rich vocabulary and sentences.

[0184] Students are able to actively participate in exercises and are able to write fractions correctly and compare the size of fractions.

[0185] 3. Analysis of students’ thinking development

[0186] Students can think about problems from different perspectives, such as understanding the concept of fractions from the perspectives of physical division and graphic division.

[0187] Students are able to perform simple reasoning and judgement, such as comparing fractions.

[0188] 4. Analysis of teachers’ teaching behavior

[0189] Teachers can effectively guide students' thinking and expression, provide appropriate feedback, and encourage students to actively participate in discussions.

[0190] Teachers can use a variety of teaching methods, such as questioning, discussion, inspiration, practice, etc. to stimulate students' interest in learning.

[0191] 5. Classroom improvement suggestions

[0192] Students can be further guided to conduct in-depth thinking and discussion, such as thinking about the application of fractions in real life and how to solve practical problems related to fractions.

[0193] Students can be encouraged to do writing exercises, such as keeping a math journal about fractions.

[0194] Multimedia technologies, such as animation and games, can be used to enrich classroom teaching content and enhance students' interest in learning.

[0195] Example 2, scenario: junior high school English class, theme is "environmental protection"

[0196] Teacher: Today we learned about environmental protection. What do you think about environmental protection?

[0197] Student A: I think environmental protection is very important because environmental problems can affect our lives.

[0198] Student B: I agree. We should start with ourselves to reduce waste and save resources.

[0199] Student C: I think environmental protection is not just an individual behavior, it also requires the joint efforts of the government and all sectors of society.

[0200] Teacher: So what do you think we should do to better protect the environment?

[0201] Student D: We can participate in environmental volunteer activities, such as planting trees, cleaning up garbage, etc.

[0202] Student E: We can also advocate green travel, such as riding bicycles, taking public transportation, etc.

[0203] Teacher: Very good, everyone has made many good suggestions. So, in real life, have you ever encountered any environmental problems?

[0204] Student F: I find that some people litter in public places, which is very uncivilized.

[0205] Student G: I think some people over-consume and waste a lot of resources.

[0206] Teacher: So how do you think we should solve these problems?

[0207] Student H: We can raise environmental awareness and let more people understand the importance of environmental protection.

[0208] Student I: We can formulate relevant laws and regulations to punish acts that damage the environment.

[0209] Teacher: Very good, everyone has put forward many good suggestions. I hope that through our joint efforts, we can protect our environment.

[0210] System analysis results:

[0211] Type of dialogue: Reflective dialogue

[0212] Discourse roles: teacher (guide), student A (statement), student B (statement), student C (statement), student D (statement), student E (statement), student F (statement), student G (statement), student H (statement), student I (statement)

[0213] Discourse function: Teacher: ask questions and guide students to think; Student AI: state opinions, express ideas, and reflect on their own behavior.

[0214] Sequential mode: teacher-student question-and-answer, student-student discussion, reflective dialogue

[0215] The conversation analysis report includes the following:

[0216] 1. Classroom dialogue analysis

[0217] The main type of classroom dialogue is reflective dialogue, where teachers guide students to reflect on their own behavior and think about how to better protect the environment.

[0218] Students were able to actively participate in discussions and put forward some practical environmental protection suggestions, which reflected their concern and thinking about environmental issues.

[0219] Students are able to reflect on their own behaviour, such as littering, over-consumption, etc., and propose measures for improvement.

[0220] 2. Student Engagement Analysis

[0221] Students are highly involved, speak actively, express themselves fluently, and are able to use rich vocabulary and sentences.

[0222] Students are able to actively participate in reflective dialogue and are able to raise some valuable points.

[0223] 3. Analysis of students’ thinking development

[0224] Students can think about problems from different perspectives, such as environmental issues from the perspectives of personal behavior, social phenomena, laws and regulations, etc.

[0225] Students are able to engage in in-depth reflection, for example reflecting on the impact of their own actions on the environment and being able to propose measures for improvement.

[0226] 4. Analysis of teachers’ teaching behavior

[0227] Teachers can effectively guide students' thinking and expression, provide appropriate feedback, and encourage students to actively participate in discussions.

[0228] Teachers can use a variety of teaching methods, such as questioning, discussion, inspiration, etc., to stimulate students' interest in learning and guide students to reflect.

[0229] 5. Classroom improvement suggestions

[0230] Students can be further guided to conduct in-depth reflection, such as thinking about the relationship between environmental protection and economic development, and how to balance environmental protection and human needs.

[0231] Students can be encouraged to engage in writing exercises, such as writing a reflective essay on environmental protection.

[0232] Students can be organized to participate in environmental protection practice activities, such as visiting environmental protection bases, participating in environmental protection volunteer activities, etc.

[0233] In some embodiments, the present specification also provides a conversation analysis device, which can be deployed on a server of a conversation analysis system. The device may include: one or more processors; wherein the processor is used to execute the steps of any of the methods described above.

[0234] Fig. 9 This is a schematic diagram of a server of a conversation analysis system provided by an exemplary embodiment. Fig. 9 At the hardware level, the device includes a processor 902, an internal bus 904, a network interface 906, a memory 808, and a non-volatile memory 910, and may also include hardware required for other functions. One or more embodiments of this specification can be implemented based on software. For example, a large model is deployed on the server, and the processor 902 reads the corresponding computer program from the non-volatile memory 910 to the memory 908 and then runs it. Of course, in addition to the software implementation, one or more embodiments of this specification do not exclude other implementations, such as logic devices or a combination of software and hardware, etc., that is, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0235] Based on the same concept as the above method, this specification also provides a computer-readable storage medium on which computer instructions are stored. When the instructions are executed by a processor, the steps of the method described in any of the above embodiments are implemented.

[0236] Based on the same concept as the above method, this specification also provides a computer program product, including a computer program / instruction, which implements the steps of the method described in any of the above embodiments when executed by a processor.

[0237] In addition, the present disclosure also provides a database operation and maintenance system, which may include a database; a large language model; wherein the model may execute the steps of the method described in any of the above embodiments.

[0238] Those skilled in the art will readily appreciate other embodiments of the specification after considering the specification and practicing the invention claimed herein. The specification is intended to cover any variations, uses or adaptations of the specification that follow the general principles of the specification and include common knowledge or customary techniques in the art that are not claimed in the specification. The specification and examples are to be considered exemplary only, and the true scope and spirit of the specification are indicated by the following claims.

[0239] It should be understood that the present description is not limited to the precise structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present description is limited only by the appended claims.

[0240] The above description is only a preferred embodiment of this specification and is not intended to limit this specification. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of this specification should be included in the scope of protection of this specification.

Claims

1. A conversation analysis system, characterized in that: The system comprises: An acquisition module, used for acquiring a text sequence; wherein the text sequence is obtained by performing real-time text conversion on classroom conversations; A matching module, used to search in a pre-established rule base, through a large language model, whether there is a target sequence matching the text sequence; wherein the rule base includes classification rule sequences of multiple dialogue types and / or reference text sequences corresponding to each dialogue type; The analysis module is used to analyze and obtain the target dialogue type corresponding to the classroom dialogue according to the target sequence matched by the matching module.

2. The system according to claim 1, characterized in that The system further comprises: The processing module is used to pre-process the text sequence, and the pre-processing at least includes text proofreading.

3. The system according to claim 1, characterized in that The classification rule sequence is used to indicate at least one of the following: The minimum number of conversation turns for each conversation type; The dialogue structure corresponding to each dialogue type.

4. The system according to claim 3, characterized in that The matching module is also used for: Performing text content annotation on the text sequence by using a large language model; In the rule base, searching through the large language model whether there is the target sequence matching the annotation result; If the annotation result matches the minimum dialogue turns and / or dialogue structure corresponding to at least one dialogue type, it is determined that the target sequence exists, and / or if the annotation result matches the reference text sequence corresponding to at least one dialogue type, it is determined that the target sequence exists.

5. The system according to claim 1, characterized in that The conversation type includes at least one of the following: critical inquiry dialogue; Knowledge collaborative construction dialogue; Guidance dialogue; supportive conversations; reflective dialogue; Metacognitive dialogue.

6. The system according to any one of claims 1 to 5, characterized in that: The system further comprises: The output module is used to generate and output a dialogue analysis report through a large language model according to the target dialogue type.

7. The system according to claim 6, characterized in that The conversation analysis report includes at least one of the following: The target dialogue type corresponding to each round of dialogue in the classroom dialogue; the proportion of each target conversation type; Results of a content assessment of the classroom conversation.

8. A conversation analysis method, characterized in that: The method comprises: Acquire a text sequence; wherein the text sequence is obtained by performing real-time text conversion on classroom conversations; In a pre-established rule base, searching through a large language model whether there is a target sequence matching the text sequence; wherein the rule base includes classification rule sequences of multiple dialogue types and / or reference text sequences corresponding to each dialogue type; According to the matched target sequence, the target dialogue type corresponding to the classroom dialogue is analyzed and obtained.

9. The method according to claim 8, characterized in that The step of searching, in a pre-established rule base, whether there is a target sequence matching the text sequence by using a large language model comprises: Performing text content annotation on the text sequence by using a large language model; In the rule base, searching through the large language model whether there is the target sequence matching the annotation result; If the annotation result matches the minimum dialogue turns and / or dialogue structure corresponding to at least one dialogue type, it is determined that the target sequence exists, and / or if the annotation result matches the reference text sequence corresponding to at least one dialogue type, it is determined that the target sequence exists.

10. The method according to claim 8 or 9, characterized in that: The method further comprises: According to the target dialogue type, a dialogue analysis report is generated and outputted through a large language model.

Images