Information interaction method and system for whole-process engineering consulting projects

Abstract

This invention discloses a method and system for information interaction in engineering consulting projects throughout the entire process, belonging to the field of information interaction technology. The method specifically includes: identifying user roles in the engineering consulting project; collecting historical data of the target role within the project; constructing a corresponding role corpus; adjusting parameters based on a general pre-trained language model using the role corpus; generating a role-customized interaction model; and when a user initiates an interaction request, generating and outputting response content based on the user's role identity using the role-customized interaction model. This application significantly improves the professionalism, semantic matching accuracy, and multi-role adaptability in the information interaction process, effectively solving the problems of information misunderstanding and low communication efficiency caused by numerous participants and significant contextual differences in engineering consulting projects.

Description

Technical Field

[0001] This invention belongs to the field of information interaction technology, specifically a method and system for information interaction in whole-process engineering consulting projects. Background Technology

[0002] Full-process engineering consulting refers to providing owners or project management units with integrated consulting services covering all stages of an engineering construction project's lifecycle, including feasibility studies, surveying and design, bidding and procurement, construction implementation, final acceptance, and operation and maintenance. This model involves numerous stakeholders, typically including owners, design firms, construction companies, supervision units, cost consulting agencies, and operation and maintenance management teams.

[0003] In actual engineering projects, the responsibilities, task boundaries, and information needs of different participants vary significantly. With the development of artificial intelligence, natural language processing, and large language models, some engineering management systems have begun to introduce AI-based question-and-answer interaction modules to assist project participants in obtaining information. However, most of the commonly used AI question-and-answer or knowledge graph systems in the current technology are built on general question-and-answer engines or unified corpus models, which are difficult to accurately cover the professional terminology, expression styles, and question backgrounds of various roles. This often results in answers that deviate from the role's needs, lack professionalism, or are inconsistent with the context.

[0004] Therefore, there is an urgent need for a method for information exchange in engineering consulting projects throughout the entire process. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention proposes a method and system for information exchange in full-process engineering consulting projects.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] Information exchange methods for whole-process engineering consulting projects include:

[0008] Identify user roles in engineering consulting projects;

[0009] For the target role, collect its historical data in the project and build a corresponding role corpus;

[0010] Based on a general pre-trained language model, the character corpus is used to perform parameter-oriented adjustments to generate a character-customized interaction model.

[0011] When a user initiates an interaction request, the system uses a role-customized interaction model to generate and output response content based on the user's role.

[0012] Specifically, for the target role, historical data of that role in the project is collected to construct a corresponding role corpus, including:

[0013] Collect historical data on engineering projects related to each role;

[0014] Semantic segmentation, named entity recognition, and intent classification are performed on the collected historical data of engineering projects to extract data related to the characters.

[0015] Based on cluster analysis, the extracted results are categorized by subject to generate a preliminary corpus set;

[0016] Based on expert knowledge graphs, semantic enhancement and hyponym / hypernym expansion are performed on the initial corpus to construct a structured role corpus.

[0017] Specifically, the step of generating a character-customized interaction model based on a general pre-trained language model and using the character corpus for parameter-oriented adjustment includes:

[0018] Based on the aforementioned character corpus, a character task guidance set is constructed, which includes questions, scenarios, semantic information paths, and semantic tags.

[0019] The system adapts to different roles by assigning dynamic weights to training samples based on semantic tags in the role task guidance set during the training of the language model.

[0020] The language model was tested using evaluation samples containing role identity cues to examine its matching degree to the roles in the output content;

[0021] Based on the matching degree, the parameters of the language model are adjusted in a targeted manner to obtain a trained language model, namely the role-customized interaction model.

[0022] Specifically, based on the aforementioned character corpus, a character task guidance set is constructed, including:

[0023] The character corpus is clustered into task clusters.

[0024] By combining the knowledge graph of the engineering industry, the semantic path relationship between entity words in the role corpus and task nodes is matched;

[0025] Extract corpus fragments whose path weights in semantic path relationships are higher than a preset weight threshold, wherein the corpus fragments contain entity words and task node content;

[0026] The extracted corpus fragments are used to perform role duplication and task intent annotation to generate a structured role-task guidance set.

[0027] Specifically, the adaptation of roles, based on the semantic tags in the role task guidance set, involves assigning dynamic weights to training samples during the training of the language model, including:

[0028] Based on the corpus features of the target character, extract the character's language style from the character task guidance set;

[0029] Each sentence in the training sample is matched with a predefined semantic label to construct a label matching vector. The predefined semantic labels include task type labels, role language style labels, term density labels, and context complexity labels.

[0030] The initial semantic importance score of each training sample is obtained by performing a dot product operation between the label matching vector and the predefined label weight level vector.

[0031] The overlap between the semantic paths in the training samples and the semantic paths in the role-task guidance set is calculated, and the initial semantic importance score is adjusted by multiplication; thus, a weighted semantic importance score is obtained.

[0032] During the language model training process, the weights of the training samples are dynamically adjusted based on the real-time loss convergence speed or gradient change trend, combined with the weighted semantic importance score, to obtain the dynamic weights of the training samples.

[0033] Specifically, the process of testing the language model using evaluation samples containing role identity cues to detect its matching degree with the role in the output content, and then adjusting the parameters of the language model based on the matching degree, includes:

[0034] Construct an evaluation sample, wherein each input statement in the evaluation sample contains explicit role identity prompts;

[0035] Using a trained language model, input evaluation samples are used to perform reasoning and generate response text;

[0036] Perform role semantic matching analysis on the generated response text;

[0037] The matching analysis results are converted into a role consistency score and compared with a preset role semantic baseline. If the score is lower than the threshold, the corpus is adjusted or incremental training is performed.

[0038] Specifically, when a user initiates an interaction request, the system calls a role-customized interaction model based on the user's role identity to generate and output response content, including:

[0039] Receive user interaction requests and parse their identity information, determine the user's role category based on the identity information or historical behavior characteristics, and generate a role identification code;

[0040] Input the user's input request and role identification code into the role-customized interaction model to generate the current user's response text;

[0041] Perform role consistency review on the response text. Based on the role semantic review results, output the response text to the user terminal or perform response correction processing and output the corrected response text to the user terminal.

[0042] Specifically, the user roles identified in the engineering consulting project include:

[0043] Receive user access requests and extract the identity information associated with the user access requests, including user account information, access terminal information, or historical behavior records;

[0044] Based on preset role recognition rules, the matching relationship between the identity information and the predefined role set is analyzed;

[0045] Users are categorized into role types corresponding to their identity information, and role tags representing these role types are generated.

[0046] The whole-process engineering consulting project information exchange system is used to implement the above-mentioned whole-process engineering consulting project information exchange method, including: role recognition module, corpus construction module, model generation module and content output module;

[0047] The role recognition module is used to identify user roles in engineering consulting projects;

[0048] The corpus construction module is used to collect historical data of a target role in the project and construct a corresponding role corpus.

[0049] The model generation module is used to generate a customized interactive model for a character by adjusting parameters based on a general pre-trained language model and the character corpus.

[0050] The content output module is used to generate and output response content based on the user's role identity by calling the role-customized interaction model when the user initiates an interaction request.

[0051] Specifically, the model generation module includes: a task guidance set construction unit, a weight allocation unit, and a training and orientation adjustment unit;

[0052] The task guidance set construction unit is used to construct a role task guidance set based on the role corpus.

[0053] The weight allocation unit is used to adapt the role and allocate dynamic weights to the training samples during the training of the language model based on the semantic tags in the role task guidance set.

[0054] The training and orientation adjustment unit is used to assign dynamic weights to training samples and train them, use the output results of evaluation samples to detect the matching degree, and adjust the parameters of the language model according to the matching degree.

[0055] Compared with the prior art, the beneficial effects of the present invention are:

[0056] This invention proposes a method and system for information exchange in engineering consulting projects throughout the entire process. Based on user identity information, role categories are determined, and a role task guidance set containing semantic tags such as task type and language style is constructed. The general language model is fine-tuned for role customization using sample weighted training and dynamic parameter tuning, and role consistency evaluation feedback is introduced to optimize the model. This significantly improves the professionalism, semantic matching degree, and multi-role adaptability in the information exchange process, effectively solving the problems of information misunderstanding and low communication efficiency caused by multiple participants and large differences in context in engineering consulting projects. Attached Figure Description

[0057] Figure 1 Flowchart of the whole-process engineering consulting project information interaction method provided by the present invention;

[0058] Figure 2 Flowchart for generating a character-customized interaction model provided by this invention;

[0059] Figure 3 The text generation flowchart provided by this invention;

[0060] Figure 4 The architecture diagram of the full-process engineering consulting project information interaction system provided by this invention. Detailed Implementation

[0061] The present application will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present application, but do not limit the present application in any way. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present application. These all fall within the protection scope of the present application.

[0062] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0063] It should be noted that, unless there is a conflict, the various features in the embodiments of this application can be combined with each other, all of which are within the protection scope of this application. Furthermore, although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than the module division in the device or the order in the flowchart. In addition, the terms "first," "second," and "third" used in this application do not limit the data or execution order, but only distinguish identical or similar items with essentially the same function and effect.

[0064] Unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The term "and / or" as used in this specification includes any and all combinations of one or more of the associated listed items.

[0065] Example 1

[0066] Please see Figures 1-3 The present invention provides an embodiment of a method for information exchange in a full-process engineering consulting project, comprising the following specific steps:

[0067] Step S1: Identify user roles in the engineering consulting project.

[0068] BIM (Building Information Modeling) is one of the data sources for engineering projects. By extracting structured information such as component attributes, construction progress, and spatial relationships contained in the BIM model into semantic fragments, and mapping them to corresponding task nodes and terminology tags in the role task guidance set, it serves as an important supplement to the construction of role corpus. For example, for the construction unit role, information such as floor, component, and construction status can be extracted from the BIM model to form query expressions related to site management and resource scheduling. This enhances the model's ability to understand spatial semantics and task status, and improves the practicality and responsiveness of the role-customized interaction model in engineering scenarios.

[0069] In actual deployment, when a user initiates an information request through the project management platform, the access interceptor module or the front-end access gateway is first triggered to obtain the basic information bound to the request.

[0070] The specific steps of step S1 are as follows:

[0071] Step S101: Receive a user access request and extract the identity information associated with the user access request, including user account information, access terminal information, or historical behavior records.

[0072] In this embodiment, by parsing the user's Token or Session information, the user's account ID is extracted, and then the user information table is called to quickly locate the user's basic identity information, such as role code, department, project ID, etc. If the account information is insufficient to clarify the role affiliation, the terminal features and the user's behavioral trajectory are further extracted to construct a set of behavioral feature vectors.

[0073] Step S102: Based on preset role recognition rules, analyze the matching relationship between the identity information and the predefined role set.

[0074] In this embodiment, a role mapping rule table is preset. Based on the extracted user identity information, each item is compared with the above rule triggering conditions. The comparison adopts methods such as Boolean matching, range matching or fuzzy matching.

[0075] Step S103: Divide users into role categories corresponding to their identity information and generate role tags representing role categories.

[0076] In this embodiment, if a user has multiple features that match multiple roles, a weighted scoring mechanism is used to assign different weight scores to each rule item. The final score is used as the role fit, the role with the highest score is identified, and it is classified as the user's main role. The role determination result is then converted into a structured label format.

[0077] Step S2: For the target role, collect its historical data in the project and build a corresponding role corpus.

[0078] The specific steps of step S2 are as follows:

[0079] Step S201: Collect historical data of engineering projects related to each role.

[0080] Step S202: Perform semantic segmentation, named entity recognition, and intent classification on the collected historical data of engineering projects to extract data related to the characters.

[0081] In this embodiment, cost data is used as one of the key role corpora, spanning all stages from project initiation, design, bidding, construction to final settlement. By extracting information such as bill of quantities, quota data, change records, payment nodes, and settlement logic, semantic paths and task nodes are linked and incorporated into the role task guidance set. At the same time, semantic tags such as cost control, budget deviation, and change pricing are combined to enhance the model's understanding of cost-related tasks, achieving deep and penetrating utilization of cost corpora throughout the entire process, thereby improving the professional response effect for roles such as owners and cost consultants in cost management-related Q&A.

[0082] Specifically, semantic segmentation is first performed using sentence segmentation rules and engineering terminology segmentation algorithms. Based on the engineering context dictionary, patterns such as "construction status as follows" and "acceptance opinions as follows" are identified, and logical blocks are divided. Each text segment is processed by a pre-trained NER model to identify entities related to the engineering task. The segmented logical blocks are then fed into an intent classifier for label determination. By comparing entity type, intent category, and role responsibility, role labels are applied to the logical blocks, and only paragraphs associated with roles are retained as training samples.

[0083] Step S203: Based on cluster analysis, the extracted results are categorized by subject to generate a preliminary corpus set.

[0084] In this embodiment, the extracted data related to the characters is converted into semantic vectors through an embedding model. These vectors reflect the contextual meaning, tone, entity words, and other information of the data related to the characters. Clustering is performed based on the Euclidean distance or cosine similarity between the semantic vectors, and a label is generated for each cluster center, such as risk warning, personnel scheduling, design change response, etc. This label is used as the semantic subject classification label of the corpus. Each type of data and its cluster label are packaged and stored to obtain a preliminary corpus set. The fields include {character ID, semantic topic, original data, intent classification, named entity, etc.}.

[0085] Step S204: Based on the expert knowledge graph, perform semantic enhancement and hyponym expansion on the preliminary corpus to construct a structured role corpus.

[0086] In this embodiment, the initial corpus has been clustered and has semantic topic tags, but the following problems still exist: the terminology in the corpus is not uniform, some are very detailed and some are too broad, such as cold-drawn threaded steel bars, such as structural construction; the same corpus block may not be clustered into the same topic due to different expression methods.

[0087] Specifically, the process begins by scanning all technical terms appearing in the initial corpus. Through term segmentation and entity recognition, these terms are mapped to nodes in the expert knowledge graph. For example, steel welding is mapped to the welding process node. If a term is a subordinate node, its superior term is retrieved from the graph and inserted as a supplementary field into the original corpus. The original fragment is then added to the graph's supplementary content to form structural fields, such as {original text, main term, superior term, subordinate term, role, task type, graph path}. This elevates the corpus from fragmented text to semantic objects.

[0088] Step S3: Based on the general pre-trained language model, the parameters are adjusted using the character corpus to generate a customized interaction model for the character.

[0089] The specific steps of step S3 are as follows:

[0090] Step S301: Based on the character corpus, construct a character task guidance set, which includes questions, scenarios, semantic information paths, and semantic tags.

[0091] The specific steps of step S301 are as follows:

[0092] Step S3011: Perform task clustering on the character corpus and divide it into task clusters.

[0093] In this embodiment, each piece of corpus, i.e. a sentence or a paragraph, is converted into a semantic vector by a general language encoder. The semantic distance between different corpora is quantified. An unsupervised clustering algorithm is used to cluster the corpus into several groups based on the similarity between semantic vectors. Each group is a task cluster, representing a specific role business task semantic domain. Each cluster automatically generates task labels by analyzing high-frequency keywords, entity categories and intent types in its central corpus.

[0094] Step S3012: Combine the engineering industry knowledge graph to match the semantic path relationship between entity words in the role corpus and task nodes.

[0095] In this embodiment, representative professional terms are extracted from the entities in each character's corpus. Node matching is then performed in the knowledge graph to identify whether synonymous or near-synonymous nodes exist. For each entity word's matched graph node, its graph path is traversed upwards or downwards to find connection paths with predefined task nodes. For example, template support can be traced upwards to concrete pouring preparation and then to the structural construction stage. The number of hops, relationship type, and node semantic similarity of each path are used to calculate the path's semantic weight.

[0096] Step S3013: Extract corpus fragments whose path weights in semantic path relationships are higher than a preset weight threshold. The corpus fragments contain entity words and task node content.

[0097] In this embodiment, several semantic path scoring rules are set. For example, if a path consists of only one hop, it is set to high weight; if a path has more than three hops and involves abstract nodes in the middle, it is set to low weight. In addition, the relationship type involved in the path also participates in the weighting, and finally a weight score between 0 and 1 is generated for each semantic path. A weight threshold is set, and only the corpus fragments corresponding to semantic paths with weights greater than this threshold are retained.

[0098] Specifically, the preset weight thresholds are determined by those skilled in the art through cross-validation or simulation experiments.

[0099] Step S3014: Perform role duplication and task intent annotation on the extracted corpus fragments to generate a structured role-task guidance set.

[0100] In this embodiment, the original fragments extracted from the corpus are often expressed from only one role's perspective. By introducing a rule base, the extracted corpus fragments are copied into semantically equivalent expressions for other roles. For example, the construction party's suggestion to rectify is changed to the supervisor issuing a rectification notice or the owner paying attention to current construction hazards. Combining the previous stage's graph path, keywords, and template library, the extracted corpus fragments are automatically tagged with task intent labels. Each generated corpus is transformed into a structured record, resulting in a structured role task guidance set.

[0101] Step S302: Adapt the role and assign dynamic weights to the training samples during the training of the language model based on the semantic tags in the role task guidance set.

[0102] The specific steps of step S302 are as follows:

[0103] Step S3021: Extract the character's language style from the character task guidance set based on the target character's corpus features.

[0104] In this embodiment, language feature analysis is performed on each corpus to extract: keyword preference, sentence pattern, distribution of professional terminology, and tone intensity and tendency. The above multi-dimensional features are transformed into a unified style vector to represent the average expressive characteristics of a certain role's language. For example, the construction party's expression style is: verb-dominated + operational terminology + clear time plan, while the supervisor's style is biased towards judgment + risk warning + suggestion tone. A general expression template is generated based on the clustering results of the role style vector.

[0105] Step S3022: Match each sentence of the training sample with a predefined semantic label to construct a label matching vector. The predefined semantic labels include task type label, role language style label, term density label and context complexity label.

[0106] In this embodiment, a predefined list of semantic tags is set by experts or a rule base, such as a task type tag set, a style tag set, a term density level, and a context complexity level. Each sentence in the training sample is matched according to the predefined semantic tags, and each corpus is mapped to a semantic tag vector.

[0107] Step S3023: Perform a dot product operation between the label matching vector and the predefined label weight level vector to obtain the initial semantic importance score for each training sample.

[0108] In this embodiment, for each training sample, there is already a label matching vector. Those skilled in the art set a label weight level vector according to the training objective, and then summed the corresponding dimension products of the two vectors to obtain the initial semantic importance score of the training sample.

[0109] Step S3024: Calculate the overlap between the semantic paths in the training samples and the semantic paths in the role task guidance set, and adjust the initial semantic importance score by multiplication; obtain the weighted semantic importance score.

[0110] In this embodiment, for each training sample, the main entity word and context task verb are identified, and the semantic path of the training sample is constructed, for example: construction elevator malfunction → transportation interruption → affecting progress; it is compared with the typical path set in the role task guidance set, for example, in the construction elevator related path, there is already construction elevator → temporary vertical transportation → operation coordination task, and there are two overlapping nodes in the two. The path matching algorithm is used to obtain an overlap score of 0-1. The overlap score is multiplied by the original importance score to obtain the weighted semantic importance score.

[0111] Step S3025: During the language model training process, based on the real-time loss convergence speed or gradient change trend, and combined with the weighted semantic importance score, the weights of the training samples are dynamically adjusted to obtain the dynamic weights of the training samples.

[0112] In this embodiment, the loss decline curve of each training sample during training iterations is recorded. If the decline slows down or fluctuates repeatedly, it indicates that the model has not fully learned the training sample. By tracking the magnitude of parameter gradient changes, it is determined whether the sample continues to affect the model. If the gradient tends to zero for a long time, it indicates that the training sample has been learned. Dynamic weight adjustment logic is constructed: high semantic importance + poor learning → increase weight; high semantic importance + learning too fast → moderately weaken; low semantic importance + well learned → set weight to a low value. Finally, the dynamic weight of the training sample is output.

[0113] The benefits of this step are as follows: By extracting the character's language style from the character task guidance set and constructing a label matching vector by combining multi-dimensional semantic labels such as task type, language style, terminology density, and contextual complexity, the semantic features of each training sample can be accurately identified. Then, by performing a dot product operation with the label weight level, an initial importance score is obtained. Combined with the structural overlap between the sample and the character task semantic path, the semantic value of the sample is further enhanced. On this basis, the loss convergence speed and gradient change trend during the training process are introduced to achieve dynamic adjustment of the training sample weights. This makes the training process pay more attention to samples that are difficult to master but strongly related to the character, effectively improving the model's performance in terms of character language expression and task context.

[0114] Step S303: Test the language model using evaluation samples containing role identity cues to detect its matching degree with the role in the output content.

[0115] The specific steps of step S303 are as follows:

[0116] Step S3031: Construct an evaluation sample, wherein each input statement in the evaluation sample contains explicit role identity prompts;

[0117] In this embodiment, key participating roles, such as owners, construction parties, design units, supervision units, and third-party auditors, are identified from the whole-process engineering consulting scenario. Based on the previously collected corpus and semantic task templates, the issues or expressions that each role often focuses on in actual engineering are extracted, and several language templates are designed for each role to ensure that the role identity is clearly indicated in the input. The above prompt words are embedded in typical expressions to form an evaluation sample.

[0118] Step S3032: Using the trained language model, input the evaluation sample for reasoning and generate response text;

[0119] Step S3033: Perform role semantic matching degree analysis on the generated response text.

[0120] It should be noted that the matching degree is calculated using a weighted scoring method.

[0121] The benefits of this step are as follows: by embedding role identity cues in the evaluation samples, we can accurately test whether the model can understand and reflect the target role's language style, terminology habits, and task focus during the generation process; combined with the role semantic matching degree analysis, the model output is compared with the role's preset semantic baseline and quantified into a role consistency score.

[0122] Step S304: Adjust the parameters of the language model according to the matching degree to obtain the trained language model, i.e., the role-customized interaction model;

[0123] The matching analysis results are converted into a role consistency score and compared with a preset role semantic baseline. If the score is lower than the threshold, the corpus is adjusted or incremental training is performed.

[0124] In this embodiment, by comparing the vocabulary, sentence structure, intent, and other dimensions in the output statement, a score is given to determine whether the output is consistent with the input role, which serves as the role consistency score. Each role corresponds to a set of semantic baseline vectors, including common terms, task focus, and expression style. If these elements are lacking in the output, or if other role-specific language appears, the consistency score is low. If a role's consistency score is consistently low, the corpus is analyzed to determine if there are issues such as scarce samples of that role or incomplete terminology coverage. The corpus is then adjusted or incremental training is performed on the role.

[0125] The benefits of this step are as follows: The process of building a role-customized interaction model based on a general pre-trained language model, by introducing a role corpus and task guidance set, not only achieves in-depth modeling of different engineering roles (such as owners, construction, and supervisors) in terms of language expression, task focus, and terminology usage, but also realizes sample differentiation learning and knowledge focus during model training through semantic labels and dynamic weight mechanisms, significantly improving the model's sensitivity to specific role task contexts. At the same time, by designing evaluation samples through role identity prompts, the accuracy of the model's role matching can be quantified, and the model parameters can be fine-tuned based on matching feedback, so that the final generated role-customized interaction model has role recognition ability and language style adaptation ability.

[0126] Step S4: When a user initiates an interaction request, the response content is generated and output based on the user's role identity by calling the role-customized interaction model.

[0127] The specific steps of step S4 are as follows:

[0128] Step S401: Receive user interaction requests and parse their identity information, determine the user's role category based on the identity information or historical behavior characteristics, and generate a role identification code.

[0129] Step S402: Input the user input request and role identification code into the role customization interaction model to generate the current user's response text.

[0130] Step S403: Perform role consistency review on the response text. Based on the role semantic review result, output the response text to the user terminal or perform response correction processing and output the corrected response text to the user terminal.

[0131] The benefits of this step are as follows: By combining user identification, role-customized model invocation, and response consistency review, personalized and professional Q&A output can be achieved for multiple stakeholders. It can accurately identify user roles such as owners, designers, construction workers, and supervisors, and dynamically invoke the corresponding language models to ensure that the output content is consistent with the role requirements in terms of terminology, task focus, and language style. At the same time, the introduction of role consistency review avoids generating answers that are inconsistent with the role identity, thereby improving the accuracy of interaction and role adaptability.

[0132] Example 2

[0133] Please see Figure 4 Another embodiment of the present invention provides: a full-process engineering consulting project information interaction system, including: a role recognition module, a corpus construction module, a model generation module, and a content output module;

[0134] The role recognition module is used to identify user roles in engineering consulting projects;

[0135] The corpus construction module is used to collect historical data of a target role in the project and construct a corresponding role corpus.

[0136] The model generation module is used to generate a customized interactive model for a character by adjusting parameters based on a general pre-trained language model and the character corpus.

[0137] The content output module is used to generate and output response content based on the user's role identity by calling the role-customized interaction model when the user initiates an interaction request.

[0138] The model generation module includes: a task guidance set construction unit, a weight allocation unit, and a training and orientation adjustment unit;

[0139] The task guidance set construction unit is used to construct a role task guidance set based on the role corpus.

[0140] The weight allocation unit is used to adapt the role and allocate dynamic weights to the training samples during the training of the language model based on the semantic tags in the role task guidance set.

[0141] The training and orientation adjustment unit is used to assign dynamic weights to training samples and train them, use the output results of evaluation samples to detect the matching degree, and adjust the parameters of the language model according to the matching degree.

[0142] In addition, the parts of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of the corresponding technical solutions in the prior art have not been described in detail, so as to avoid excessive elaboration.

[0143] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for information exchange in whole-process engineering consulting projects, characterized in that: include: Identify user roles in engineering consulting projects; For the target role, collect its historical data in the project and build a corresponding role corpus; Based on a general pre-trained language model, the parameter orientation adjustment is performed using the character corpus to generate a character-customized interaction model. The parameter orientation adjustment is based on the matching degree of the character in the output content, and the corpus is adjusted or incrementally trained. When a user initiates an interaction request, the system uses the user's role-customized interaction model to generate and output the response content. The process of generating a character-customized interaction model based on a general pre-trained language model and using the character corpus for parameter-oriented adjustment includes: Based on the aforementioned character corpus, a character task guidance set is constructed, which includes questions, scenarios, semantic information paths, and semantic tags. To adapt to the role, dynamic weights are assigned to the training samples during the training of the language model based on the semantic tags in the role task guidance set. The language model was tested using evaluation samples containing role identity cues to examine its matching degree to the roles in the output content; Based on the matching degree, the parameters of the language model are adjusted in a targeted manner to obtain a trained language model, namely the role-customized interaction model; Based on the aforementioned character corpus, a character task guidance set is constructed, including: The character corpus is clustered into task clusters. By combining the knowledge graph of the engineering industry, the semantic path relationship between entity words in the role corpus and task nodes is matched; Extract corpus fragments whose path weights in semantic path relationships are higher than a preset weight threshold, wherein the corpus fragments contain entity words and task node content; The extracted corpus fragments are used to perform role duplication and task intent annotation to generate a structured role-task guidance set; The adaptation of roles, based on the semantic tags in the role task guidance set, involves assigning dynamic weights to training samples during the training of the language model, including: Based on the corpus features of the target character, extract the character's language style from the character task guidance set; Each sentence in the training sample is matched with a predefined semantic label to construct a label matching vector. The predefined semantic labels include task type labels, role language style labels, term density labels, and context complexity labels. The initial semantic importance score of each training sample is obtained by performing a dot product operation between the label matching vector and the predefined label weight level vector. The overlap between the semantic paths in the training samples and the semantic paths in the role-task guidance set is calculated, and the initial semantic importance score is adjusted by multiplication; thus, a weighted semantic importance score is obtained. During the language model training process, the weights of the training samples are dynamically adjusted based on the real-time loss convergence speed or gradient change trend, combined with the weighted semantic importance score, to obtain the dynamic weights of the training samples. The process involves testing the language model using evaluation samples containing role identity cues to detect its matching degree with the role in the output content, and then adjusting the language model's parameters based on the matching degree, including: Construct an evaluation sample, wherein each input statement in the evaluation sample contains explicit role identity prompts; Using a trained language model, input evaluation samples are used to perform reasoning and generate response text; Perform role semantic matching analysis on the generated response text; The matching analysis results are converted into a role consistency score and compared with a preset role semantic baseline. If the score is lower than the threshold, the corpus is adjusted or incremental training is performed.

2. The information exchange method for whole-process engineering consulting projects as described in claim 1, characterized in that, The process involves collecting historical data about the target role within the project and constructing a corresponding role corpus, including: Collect historical data on engineering projects related to each role; Semantic segmentation, named entity recognition, and intent classification are performed on the collected historical data of engineering projects to extract data related to the characters. Based on cluster analysis, the extracted results are categorized by subject to generate a preliminary corpus set; Based on expert knowledge graphs, semantic enhancement and hyponym / hypernym expansion are performed on the initial corpus to construct a structured role corpus.

3. The information exchange method for whole-process engineering consulting projects as described in claim 2, characterized in that, When a user initiates an interaction request, the system calls a role-customized interaction model based on the user's role identity to generate and output response content, including: Receive user interaction requests and parse their identity information, determine the user's role category based on the identity information or historical behavior characteristics, and generate a role identification code; Input the user's input request and role identification code into the role-customized interaction model to generate the current user's response text; Perform role consistency review on the response text. Based on the role semantic review results, output the response text to the user terminal or perform response correction processing and output the corrected response text to the user terminal.

4. The information exchange method for whole-process engineering consulting projects as described in claim 3, characterized in that, The user roles identified in the engineering consulting project include: Receive user access requests and extract the identity information associated with the user access requests, including user account information, access terminal information, or historical behavior records; Based on preset role recognition rules, the matching relationship between the identity information and the predefined role set is analyzed; Users are categorized into role types corresponding to their identity information, and role tags representing these role types are generated.

5. A whole-process engineering consulting project information exchange system, used to implement the whole-process engineering consulting project information exchange method as described in any one of claims 1-4, characterized in that, include: The module includes a role recognition module, a corpus construction module, a model generation module, and a content output module. The role recognition module is used to identify user roles in engineering consulting projects; The corpus construction module is used to collect historical data of a target role in the project and construct a corresponding role corpus. The model generation module is used to generate a customized interactive model for a character by adjusting parameters based on a general pre-trained language model and the character corpus. The content output module is used to generate and output response content based on the user's role identity by calling the role-customized interaction model when the user initiates an interaction request.

6. The information exchange system for whole-process engineering consulting projects as described in claim 5, characterized in that, The model generation module includes: a task guidance set construction unit, a weight allocation unit, and a training and orientation adjustment unit; The task guidance set construction unit is used to construct a role task guidance set based on the role corpus. The weight allocation unit is used to adapt the role and allocate dynamic weights to the training samples during the training of the language model based on the semantic tags in the role task guidance set. The training and orientation adjustment unit is used to assign dynamic weights to training samples and train them, use the output results of evaluation samples to detect the matching degree, and adjust the parameters of the language model according to the matching degree.

Images