Natural language based computer aided engineering software intelligent agent human-machine interaction system

Abstract

The application discloses a kind of natural language-based computer aided engineering software intelligent agent human-computer interaction system, and voice instruction is obtained by voice instruction acquisition module;First text information is obtained by STT function module;Second text information is obtained by semantic correction module;Second text information is input to intent recognition module by compliance analysis module;Model operation task list is obtained by intent recognition module;CAE intelligent execution module reads and runs model operation task in model operation task list;Language paradigm text description is obtained by natural language conversion module;Language paradigm text description is converted into second voice information by TTS function module;Second voice information is output by voice output module, and simultaneously, result feedback module is called to visualize CAE model result after executing model operation task.The problems that existing method relies on manual operation point, line, surface and other bottom geometric elements, and there are cumbersome and inefficient steps are solved.

Description

Technical Field

[0001] This invention relates to the field of computer-aided engineering software technology, and in particular to a human-computer interaction system for computer-aided engineering software intelligent agents based on natural language. Background Technology

[0002] While intelligent agent technology in the CAE field is still in its nascent stage, it can only perform a series of simple dialogue functions. The actual operation of CAE-based M&S (modeling and simulation) involves complex geometric and mesh operations based on graphical user interfaces. For example, during model assembly, to assemble a part in the correct position, it is necessary to move and rotate the geometry of that part; in setting physical boundaries and loads, it is necessary to select specific geometric surfaces, lines, and points of the part and set physical boundaries and loads. Existing computer-aided engineering software mainly uses keyboard and mouse operations to complete human-computer interaction, such as selecting parts, surfaces, curves, and geometric points by clicking the left mouse button; through... Rotate the view using the left mouse button; Use the middle mouse button to pan the view; (using...) The right mouse button allows for view zooming and other operations; or the left mouse button allows for geometric feature selection, the middle mouse button allows for view zooming, and the right mouse button allows for view rotation.

[0003] However, the above methods are applicable to users operating the keyboard and mouse, and still rely on manually manipulating underlying geometric elements such as points, lines, and surfaces, which results in cumbersome steps and low efficiency. Summary of the Invention

[0004] This invention provides a human-computer interaction system for computer-aided engineering software intelligent agents based on natural language, in order to overcome the above-mentioned technical problems.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows:

[0006] A computer-aided engineering software intelligent agent human-computer interaction system based on natural language includes a voice command acquisition module, a STT function module, a semantic correction module, a compliance analysis module, an intent recognition module, a CAE intelligent execution module, a natural language conversion module, a TTS function module, a voice output module, and a result feedback module.

[0007] The voice command acquisition module is used to acquire voice commands through a preset voice acquisition device;

[0008] The STT function module is used to convert voice commands into first text information;

[0009] The semantic correction module is used to perform semantic correction operations on the first text information based on the preset first large language model LLM to obtain the second text information; and the semantic correction operations include, but are not limited to, word missing completion, homophone and misspelling correction and punctuation correction.

[0010] The compliance analysis module is used to confirm whether the second text information contains words or phrases that do not comply with the compliance analysis rules based on the compliance analysis rules. If it does, the third text information is output, and the TTS function module is called to convert the third text information into the first voice information. The first voice information is then output through a preset sound output device, and the voice command acquisition module is called again to collect the voice command. If it does not, the second text information is directly input to the intent recognition module.

[0011] The intent recognition module is used to identify one or more user intents in the second text information through a preset second large language model LLM, and to perform correlation analysis between the user intents and the model operation text descriptions in the constructed model operation knowledge base to obtain the model operation results of intent recognition. At the same time, it obtains the model operation task list corresponding to the user intents in the second text information based on the model operation results.

[0012] Furthermore, the model operation task list includes one or more model operation tasks arranged in chronological order;

[0013] The CAE intelligent execution module is used to read and run the model operation tasks in the model operation task list one by one in sequence, so as to perform model operation tasks on the current CAE model and return the text results after the model operation tasks are executed.

[0014] The natural language conversion module is used to perform natural language conversion on the text results using a pre-set third major language model (LLM) to obtain a language paradigm text description containing the text results.

[0015] The TTS function module is used to convert the text description of the language paradigm into second speech information;

[0016] The voice output module outputs second voice information through a preset sound output device, and at the same time calls the result feedback module to visualize the CAE model results after performing the model operation task.

[0017] Furthermore, it also includes a labeling module for CAE models;

[0018] The marking module is used to obtain the part name of each part in the CAE model, generate a unique identifier number for each part based on a preset numbering rule, encode each surface, curve or geometric point included in each part to obtain a unique identifier number, and associate the unique identifier number with the corresponding part to obtain an association table. The association table is used to call the corresponding part when performing model operation tasks on the current CAE model.

[0019] Furthermore, the compliance analysis rules are as follows:

[0020] S100: Define and obtain the set of sensitive word vectors;

[0021] S101: Perform serialization on the text in the second text information to obtain a text sequence; simultaneously call the Chinese word segmentation toolkit to perform word segmentation on the text sequence to obtain a word segmentation set; then perform vectorization on the word segments in the word segmentation set to obtain a word segmentation vector set;

[0022] S102: Obtain the Euclidean distance similarity between each word segmentation vector in the word segmentation vector set and the sensitive word vector in the sensitive word vector set. Based on the preset similarity threshold, determine whether there are sensitive words in the word segmentation vector set according to the Euclidean distance similarity.

[0023] If there are sensitive words in the word segmentation vector set, it is confirmed that there are words or phrases in the second text information that do not comply with the compliance analysis rules; if there are no sensitive words in the word segmentation vector set, it is confirmed that there are no words or phrases in the second text information that do not comply with the compliance analysis rules.

[0024] Furthermore, the method for obtaining the model operation result of the intent recognition specifically includes the following steps:

[0025] Step S1: Identify and obtain one or more user intentions from the second text information through the pre-set second large language model LLM, and configure timestamps for each user intention to obtain the user intention time sequence; and the user intention is represented in the form of word and sentence vectors, which are used to characterize the semantic content corresponding to the model operation task performed on the CAE model;

[0026] Step S2: Construct a knowledge base for model operations:

[0027] Step S21: Preset a set of model operation task categories, which includes part display, part selection, view control, geometric element display, and geometric element selection;

[0028] Step S22: By calling the preset matching module, the preset model operation functions are mapped to the corresponding model operation text descriptions in the model operation task category set, thereby obtaining the model operation knowledge base;

[0029] Step S3: Traverse each user intent in the user intent time sequence; and based on the model operation knowledge base, perform correlation analysis on each user intent and the model operation text description to obtain a correlation score;

[0030] Step S4: Obtain a descending sorted sequence list of model operation text descriptions corresponding to user intents based on relevance scores, and take the model operation text description with the highest relevance score as the final model operation result of intent recognition.

[0031] Furthermore, it also includes a part number management module and a geometric element numbering module;

[0032] The part numbering management module is configured to automatically add or modify part numbers based on the current model status by calling the tagging module, and to control the display or hiding of part numbers in response to natural language commands;

[0033] The geometric element numbering module is configured to automatically add or modify numbers for the surfaces, curves, and geometric points of the currently displayed parts by calling the marking module, and respond to natural language commands to control the display or hiding of the corresponding numbers.

[0034] The natural language instructions are the voice instructions obtained by the voice instruction acquisition module, which are then passed through the STT function module, semantic correction module, compliance analysis module, and intent recognition module to output the task instructions corresponding to the model operation task.

[0035] Beneficial effects: This invention provides a human-computer interaction system for computer-aided engineering software intelligent agents based on natural language, which solves the real-time interaction barrier for model operations in natural language-based CAE Agent applications. By replacing the traditional mouse-based operation logic for models, parts, surfaces, curves, and points with natural language descriptions, it improves the user's work efficiency based on CAEAgent. Attached Figure Description

[0036] To more clearly illustrate the technical solutions in the embodiments of the present invention 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 the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1 This is a schematic diagram of the human-computer interaction system of the computer-aided engineering software intelligent agent based on natural language according to the present invention;

[0038] Figure 2 This is a schematic diagram of the part numbers in the CAE model of this embodiment;

[0039] Figure 3 This is a schematic diagram of the part surface of the CAE model in this embodiment;

[0040] Figure 4 This is a schematic diagram of the part curves in the CAE model of this embodiment;

[0041] Figure 5 This is a schematic diagram of the geometric points of the part in the CAE model in this embodiment. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] This embodiment provides a human-computer interaction system for computer-aided engineering software intelligent agents based on natural language, such as... Figure 1 As shown, it includes a voice command acquisition module, an STT function module, a semantic correction module, a compliance analysis module, an intent recognition module, a CAE intelligent execution module, a natural language conversion module, a TTS function module, a voice output module, and a result feedback module.

[0044] The voice command acquisition module is used to acquire voice commands through a preset voice acquisition device (e.g., a microphone);

[0045] The STT function module is used to convert voice commands into first text information;

[0046] The semantic correction module is used to perform semantic correction operations on the first text information based on the preset first large language model LLM to obtain the second text information; and the semantic correction operations include, but are not limited to, word missing completion, homophone and misspelling correction and punctuation correction.

[0047] The compliance analysis module is used to confirm whether the second text information contains any words or phrases that do not comply with the compliance analysis rules, based on the compliance analysis rules. If so, it outputs preset third text information, then calls the TTS function module to convert the third text information into first voice information, outputs the first voice information through a preset sound output device, and calls the voice command acquisition module again to collect voice commands. If not, it directly inputs the second text information into the intent recognition module. Specifically, the compliance analysis rules in this embodiment are:

[0048] S100: Define and obtain the set of sensitive word vectors;

[0049] The sensitive words are text words and phrases that are unrelated to the operation of the intelligent agent in the auxiliary engineering software, as defined by empirical values.

[0050] S101: Perform serialization on the text in the second text information to obtain the text sequence; simultaneously call the Chinese word segmentation toolkit. Perform word segmentation on the text sequence to obtain a word segmentation set; then use natural language processing (NLP) technology to vectorize the word segments in the word segmentation set to obtain a word segmentation vector set;

[0051] S102: Obtain the Euclidean distance similarity between each word segmentation vector in the word segmentation vector set and the sensitive word vector in the sensitive word vector set. Based on the preset similarity threshold, determine whether there are sensitive words in the word segmentation vector set according to the Euclidean distance similarity.

[0052] If there are sensitive words in the word segmentation vector set, it is confirmed that there are words or phrases in the second text information that do not comply with the compliance analysis rules; if there are no sensitive words in the word segmentation vector set, it is confirmed that there are no words or phrases in the second text information that do not comply with the compliance analysis rules.

[0053] Or the compliance analysis rules include:

[0054] S100: Based on expert experience, define a set of text phrases related to the operation of computer-aided engineering software intelligent agents, and set a language text specification format sequence vector according to the set of text phrases: "A+B,C+...+D,E+...+F", where A represents the language specification start text; B and C represent verb phrases; C and E represent noun phrases; and F represents the language specification end text; for example: "Please (A) display (B) part number n (C), rotate (D) part number n (E)... end (F)";

[0055] S101: Extract a set of keywords representing language norms from the second text information using natural language processing technology; the keywords are verb phrases and noun phrases used to represent language norms;

[0056] It obtains the Euclidean distance similarity between each keyword in the keyword set and the text phrases based on the text phrase set; at the same time, based on the preset first similarity threshold range, it confirms whether each keyword is a phrase that conforms to language norms according to the Euclidean distance similarity of the word vectors.

[0057] If the Euclidean distance similarity of the word vectors meets the first similarity threshold range, then continue to execute S102; otherwise, confirm that the keyword is a word or phrase that does not conform to the language norms, that is, confirm that it does not conform to the compliance analysis rules.

[0058] S102: Use natural language processing technology to extract the keyword sequence representing language norms from the second text information. The keyword sequence is the conditional statement used to represent the keyword sequence. Then, convert the conditional statement into the form of a statement sequence vector to obtain the conditional statement sequence vector.

[0059] Based on the language text standard format sequence vector and the conditional statement sequence vector, the part-of-speech sequence Euclidean distance similarity of the statement part-of-speech sequence in the sequence vector is obtained; at the same time, based on the preset second similarity threshold range, the keyword sequence is confirmed to conform to the language text standard format according to the part-of-speech sequence Euclidean distance similarity.

[0060] If the Euclidean distance similarity of the part-of-speech sequence meets the second similarity threshold range, then the second text information is confirmed to comply with the compliance analysis rules and is deemed compliant; otherwise, the second text information is confirmed to not comply with the compliance analysis rules and is deemed non-compliant.

[0061] The intent recognition module is used to identify one or more user intents in the second text information through a preset second large language model LLM, and to perform correlation analysis between the user intents and the model operation text descriptions in the constructed model operation knowledge base to obtain the model operation results of intent recognition.

[0062] The method for obtaining the model operation result of intent recognition described in this embodiment specifically includes the following steps:

[0063] Step S1: Identify and obtain one or more user intentions from the second text information through the pre-set second large language model LLM, and configure timestamps for each user intention to obtain the user intention time sequence; and the user intention is represented in the form of word and sentence vectors, which are used to characterize the semantic content corresponding to the model operation task performed on the CAE model;

[0064] Step S2: Construct a knowledge base for model operations:

[0065] Step S21: Preset a set of model operation task categories, which includes part display, part selection, view control, geometric element display, and geometric element selection;

[0066] Specifically, a set of model operation task categories is defined for performing model operation tasks on CAE models; the set of model operation task categories includes, but is not limited to, displaying part numbers; selecting part number *; rotating the viewpoint; translating the viewpoint; displaying surface or curve or part geometric point numbers; and selecting surface or curve or geometric point number *.

[0067] Step S22: By calling the preset matching module, the preset model operation functions are mapped to the corresponding model operation text descriptions in the model operation task category set, thereby obtaining the model operation knowledge base;

[0068] Step S3: Traverse each user intent in the user intent time sequence; and based on the model operation knowledge base, perform correlation analysis on each user intent and the model operation text description to obtain a correlation score;

[0069] Specifically, this includes: obtaining the target embedding vector of the target text through a pre-set embedding unit; and the target embedding vector is used to represent the first embedding vector of the text information features corresponding to the user intent, and also to represent the second embedding vector of the text information features corresponding to the model operation text description; obtaining the correlation index between the first embedding vector and the second embedding vector through the Pearson correlation coefficient, and using the correlation index as the correlation score, i.e., the correlation analysis result;

[0070] Step S4: Obtain a descending sorted sequence list of model operation text descriptions corresponding to user intents based on relevance scores, and take the model operation text description with the highest relevance score in the descending sorted sequence list as the final model operation result of intent recognition;

[0071] Simultaneously, based on the model operation results, a list of model operation tasks corresponding to the user intent in the second text information is obtained, and the list of model operation tasks includes one or more model operation tasks arranged in chronological order.

[0072] The CAE intelligent execution module is used to read and run the model operation tasks in the model operation task list one by one in sequence, so as to perform model operation tasks on the current CAE model and return the text results after the model operation tasks are executed.

[0073] The natural language conversion module is used to perform natural language conversion on the text results using a pre-set third major language model (LLM) to obtain a language paradigm text description containing the text results.

[0074] The TTS function module is used to convert the text description of the language paradigm into second speech information;

[0075] The voice output module outputs second voice information through a preset sound output device, and at the same time calls the result feedback module to visualize the CAE model results after performing the model operation task.

[0076] In a specific embodiment, the system described in this embodiment also includes a CAE model marking module, a part number management module, and a geometric element numbering module;

[0077] The tagging module is used to obtain the part name of each part in the CAE model, such as... Figures 2 to 5As shown, part1 represents part 1; part2 represents part 2; surf1 represents surface 1; surf2 represents surface 2; surf3 represents surface 3; edge1 represents curve 1; edge2 represents curve 2; edge3 represents curve 3; edge4 represents curve 4; edge5 represents curve 5; edge6 represents curve 6; edge7 represents curve 7; point1 represents geometric point 1; point2 represents geometric point 2; point3 represents geometric point 3; point4 represents geometric point 4. A unique identifier is generated for each part based on a preset numbering rule. Simultaneously, each surface, curve, or geometric point included in each part is encoded to obtain a unique identifier. The unique identifier is then associated with the corresponding part to obtain an association table. This association table is used to call the corresponding part when performing model operation tasks on the current CAE model.

[0078] The part number management module is configured to automatically add or modify part numbers by calling the tagging module based on the current model state, and to control the display or hiding of part numbers in response to natural language commands;

[0079] The geometric element numbering module is configured to automatically add or modify numbers for the surfaces, curves, and geometric points of the currently displayed part by calling the marking module, and to control the display or hiding of the corresponding numbers in response to natural language commands. The natural language commands are the voice commands acquired by the voice command acquisition module, which are then processed sequentially by the STT function module, semantic correction module, compliance analysis module, and intent recognition module to output the task commands corresponding to the model operation task.

[0080] The system described in this embodiment proposes a CAE intelligent interactive system based on automatic numbering. Based on the current state of the model, it can automatically add part numbers to parts, surfaces, curves, and points through an algorithm module and display (or hide) them. The selected parts, surfaces, curves, and points are determined by the numbering information in natural language, and then operations are performed on the parts (e.g., moving part number *, rotating part number *), and operations are performed on the surfaces, curves, and points (e.g., selecting surface number *, selecting surface number *, selecting geometric point number *, etc.). This solves the real-time interaction barrier of CAE Agent applications based on natural language for model operations. By replacing the traditional mouse-based operation logic of models, parts, surfaces, curves, and points with natural language descriptions, it improves the user's work efficiency based on CAEAgent.

[0081] The specific example in this embodiment is as follows:

[0082] S1: The user inputs a voice command via microphone (e.g., voice command "Display part number").

[0083] S2: The input voice is converted into text information TEXT01 through the STT function, such as: text information [display zero minus edited]};

[0084] S3: Text information TEXT01 undergoes semantic correction via LLM to correct potential problems such as missing words, homophones, and punctuation marks that may exist in the STT speech-to-text process, transforming it into text information TEXT02 that is more grammatically correct, such as: text information [Display part number];

[0085] S4: Input TEXT02 into the compliance analysis module. If TEXT02 violates regulations, output TEXT03, such as the text message "[Sorry, please enter your requirements]". Then, use the TTS function to convert TEXT03 into the voice message SOUND01, such as the voice message "[Sorry, please enter your requirements]". Then, output AUDIO01 through the speaker, such as the audio message "[Sorry, please enter your requirements]", and end this round of interaction. If TEXT02 is compliant, output TEXT04, such as the text message "[Display part number]".

[0086] S5: Input TEXT04 into the user intent recognition module (the user intent module is implemented based on the mapping between LLM+ model operation functions and function descriptions, such as: text information [display part number -- show_part_number()], and output the task list LIST01, such as: text information [step01: function_call=show_part_number()];

[0087] S6: Input LIST01 into the intelligent execution module, run each subtask in LIST01 one by one (e.g., display part number; select part number *; rotate part view; translate part view; display surface, display curve, display geometric point number; select surface number *, select surface number *, select geometric point number *, etc.), and return the text result RES01, such as: text information [task='Display part number', function_call=show_part_number(), return=0];

[0088] S7: Input RES01 into LLM for natural language conversion, so that the output content contains both the key command execution results and conforms to the language paradigm of human-to-human dialogue; finally, it is converted into TEXT04 that conforms to the grammatical expression, such as: text information [Part number has been successfully displayed, clear selected parts];

[0089] S8: Input TEXT04 to the STT function to convert it into an audio file SOUND02, such as: voice message [Part number has been successfully displayed, clear selected parts];

[0090] S9: Output AUDIO2 via the speaker in the form of voice, such as sound wave information [Part number has been successfully displayed, clear the selected part]; through the above S1~S9, a closed loop of intelligent human-computer interaction is completed, from user voice input to CAE intelligent agent response and execution, and then the execution result is fed back to the user.

[0091] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention 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 or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A computer-aided engineering software intelligent agent human-computer interaction system based on natural language, characterized in that, It includes a voice command acquisition module, an STT function module, a semantic correction module, a compliance analysis module, an intent recognition module, a CAE intelligent execution module, a natural language conversion module, a TTS function module, a voice output module, and a result feedback module; The voice command acquisition module is used to acquire voice commands through a preset voice acquisition device; The STT function module is used to convert voice commands into first text information; The semantic correction module is used to perform semantic correction operations on the first text information based on the preset first large language model LLM to obtain the second text information; and the semantic correction operations include word missing completion, homophone and typo correction and punctuation correction. The compliance analysis module is used to confirm, based on compliance analysis rules, whether the second text information contains any words or phrases that do not comply with the compliance analysis rules; If it exists, the third text information is output, and the TTS function module is called to convert the third text information into the first voice information. The first voice information is then output through the preset sound output device, and the voice command acquisition module is called again to collect the voice command. If it does not exist, the second text information is directly input into the intent recognition module. The intent recognition module is used to identify one or more user intents in the second text information through a preset second large language model LLM, and to perform correlation analysis between the user intents and the model operation text descriptions in the constructed model operation knowledge base to obtain the model operation results of intent recognition. At the same time, it obtains the model operation task list corresponding to the user intents in the second text information based on the model operation results. Furthermore, the model operation task list includes one or more model operation tasks arranged in chronological order; The CAE intelligent execution module is used to read and run the model operation tasks in the model operation task list one by one in sequence, so as to perform model operation tasks on the current CAE model and return the text results after the model operation tasks are executed. The natural language conversion module is used to perform natural language conversion on the text results using a pre-set third major language model (LLM) to obtain a language paradigm text description containing the text results. The TTS function module is also used to convert the text description of the language paradigm into second speech information; The voice output module outputs second voice information through a preset sound output device, and at the same time calls the result feedback module to visualize the CAE model results after performing the model operation task.

2. The computer-aided engineering software intelligent agent human-computer interaction system based on natural language according to claim 1, characterized in that, It also includes a tagging module for CAE models; The marking module is used to obtain the part name of each part in the CAE model, generate a unique identifier number for each part based on a preset numbering rule, encode each surface, curve or geometric point included in each part to obtain a unique identifier number, and associate the unique identifier number with the corresponding part to obtain an association table. The association table is used to call the corresponding part when performing model operation tasks on the current CAE model.

3. The computer-aided engineering software intelligent agent human-computer interaction system based on natural language according to claim 2, characterized in that, The compliance analysis rules are as follows: S100: Define and obtain the set of sensitive word vectors; S101: Perform serialization on the text in the second text information to obtain a text sequence; simultaneously call the Chinese word segmentation toolkit to perform word segmentation on the text sequence to obtain a word segmentation set; then perform vectorization on the word segments in the word segmentation set to obtain a word segmentation vector set; S102: Obtain the Euclidean distance similarity between each word segmentation vector in the word segmentation vector set and the sensitive word vector in the sensitive word vector set. Based on the preset similarity threshold, determine whether there are sensitive words in the word segmentation vector set according to the Euclidean distance similarity. If there are sensitive words in the word segmentation vector set, it is confirmed that there are words or phrases in the second text information that do not comply with the compliance analysis rules; if there are no sensitive words in the word segmentation vector set, it is confirmed that there are no words or phrases in the second text information that do not comply with the compliance analysis rules.

4. The computer-aided engineering software intelligent agent human-computer interaction system based on natural language according to claim 3, characterized in that, The method for obtaining the model operation result of the intent recognition specifically includes the following steps: Step S1: Identify and obtain one or more user intentions from the second text information through the pre-set second large language model LLM, and configure timestamps for each user intention to obtain the user intention time sequence; and the user intention is represented in the form of word and sentence vectors, which are used to characterize the semantic content corresponding to the model operation task performed on the CAE model; Step S2: Construct a knowledge base for model operations: Step S21: Preset a set of model operation task categories, which includes part display, part selection, view control, geometric element display, and geometric element selection; Step S22: By calling the preset matching module, the preset model operation functions are mapped to the corresponding model operation text descriptions in the model operation task category set, thereby obtaining the model operation knowledge base; Step S3: Traverse each user intent in the user intent time sequence; Based on the model operation knowledge base, a correlation analysis is performed on the correlation between each user's intent and the text description of the model operation to obtain a correlation score; Step S4: Obtain a descending sorted sequence list of model operation text descriptions corresponding to user intents based on relevance scores, and take the model operation text description with the highest relevance score as the final model operation result of intent recognition.

5. The computer-aided engineering software intelligent agent human-computer interaction system based on natural language according to claim 4, characterized in that, It also includes a parts number management module and a geometric element numbering module; The part numbering management module is configured to automatically add or modify part numbers based on the current model status by calling the tagging module, and to control the display or hiding of part numbers in response to natural language commands; The geometric element numbering module is configured to automatically add or modify numbers for the surfaces, curves, and geometric points of the currently displayed parts by calling the marking module, and respond to natural language commands to control the display or hiding of the corresponding numbers. The natural language instructions are the voice instructions obtained by the voice instruction acquisition module, which are then passed through the STT function module, semantic correction module, compliance analysis module, and intent recognition module to output the task instructions corresponding to the model operation task.

Images