Engineering management report generation method and system based on natural language processing

Abstract

The invention relates to the technical field of text processing, and provides an engineering management report generation method and system based on natural language processing, which are used for realizing deep analysis of construction logs and providing a comprehensive and accurate decision basis for engineering management. The method comprises the steps that a construction log text set generated in the engineering construction process is acquired, natural language processing is conducted on the construction log text set, key entities and progress events in each construction stage description unit are extracted, and a construction feature set containing entity identification information and event time sequence information is obtained; performing semantic alignment processing on the construction feature set and a preset engineering specification knowledge base to generate specification association information between the construction stage description unit and an engineering specification requirement; and performing structured integration processing on the standard associated information based on a hierarchical template matching algorithm to generate a structured report text meeting the engineering management requirements.

Description

Technical Field

[0001] The present invention belongs to the technical field of text processing, and particularly relates to a method and system for generating an engineering management report based on natural language processing. Background Art

[0002] In the field of engineering management, the processing of construction logs mostly stays at the level of simple text records and manual access. In the traditional way, the recording formats of construction logs are not unified, the content is complex and lacks systematic arrangement, making it difficult to quickly and accurately obtain key information from them. When analyzing the compliance of construction conditions with engineering specifications, it often relies on manual experience comparison, which is inefficient and prone to omissions. At the same time, the generated reports are mostly scattered and unstructured texts, lacking comprehensive integration of construction progress, compliance with specifications, and targeted suggestions, and cannot provide effective decision-making support for engineering management. It can be seen that the prior art fails to fully deeply mine and analyze construction logs, resulting in unsmooth information and lack of scientific decision-making in the process of engineering management, and it is difficult to ensure the efficient progress and standardized implementation of engineering projects. Summary of the Invention

[0003] The present invention provides a method and system for generating an engineering management report based on natural language processing to achieve in-depth analysis of construction logs and provide comprehensive and accurate decision-making basis for engineering management.

[0004] In a first aspect, an embodiment of the present invention provides a method for generating an engineering management report based on natural language processing, which is applied to an engineering management report generation system. The method includes: Obtaining a set of construction log text generated during the process of engineering construction, where the set of construction log text includes multiple continuously recorded construction stage description units, and each construction stage description unit includes at least one construction activity description statement and a corresponding on-site status description statement; Performing natural language processing on the set of construction log text, extracting key entities and progress events in each construction stage description unit, and obtaining a set of construction features including entity identification information and event time sequence information; Performing semantic alignment processing on the set of construction features and a pre-set engineering specification knowledge base to generate specification association information between the construction stage description unit and the engineering specification requirements; Performing structured integration processing on the specification association information based on a hierarchical template matching algorithm to generate a structured report text that meets the requirements of engineering management, where the structured report text includes construction progress description, specification compliance analysis, and recommended content.

[0005] In a second aspect, an embodiment of the present invention provides an engineering management report generation system, which includes a processor and a memory. Among them, the memory stores a computer program, and when the computer program is executed by the processor, the processor is caused to execute the steps of the above method.

[0006] In a third aspect, an embodiment of the present invention provides a computer-readable storage medium, which includes a computer program. When the computer program runs on the engineering management report generation system, the computer program is used to cause the engineering management report generation system to execute the steps of the above method.

[0007] By obtaining a construction log text set containing detailed construction stage description units and combining natural language processing technology for mining the construction log text set, the embodiment of the present invention can accurately extract key entities and progress events to form a construction feature set, breaking through the limitations of traditional manual analysis and greatly improving the accuracy and efficiency of information extraction; the semantic alignment processing of the construction feature set and the engineering specification knowledge base realizes the deep association between the actual construction and the specification requirements, and can timely discover potential non-compliance problems; based on the structured integration of the hierarchical template matching algorithm, the generated structured report text comprehensively covers construction progress, specification compliance and recommended content, providing a systematic, clear and targeted decision-making basis for engineering management, enhancing the scientificity, standardization and foresight of engineering management, and effectively promoting the efficient and compliant development of engineering construction projects. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 It is a schematic flowchart of a method for generating an engineering management report based on natural language processing provided by an embodiment of the present invention.

[0009] Figure 2 It is a schematic structural diagram of an engineering management report generation system provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0010] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Based on the embodiments described in this document of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of the technical solutions of the present invention.

[0011] See Figure 1 , which is a method for generating an engineering management report based on natural language processing provided in an embodiment of the present invention. This method can be applied to an engineering management report generation system, and the specific process is as shown in steps 110 - step 140.

[0012] Step 110: Obtain a collection of construction log texts generated during the project construction process. The collection of construction log texts contains multiple continuously recorded construction stage description units, and each construction stage description unit includes at least one construction activity description statement and a corresponding on-site status description statement.

[0013] For ease of understanding the overall technical solution, an embodiment of the present invention will be described in detail by taking a large commercial building project as an example. This project covers multiple construction links, from infrastructure construction to main structure erection, and then to interior decoration, etc. The construction log is an important document recording the entire construction process, and construction personnel record construction activities and on-site status every day. In this project, the construction stage description units in the collection of construction log texts record the construction conditions at different stages in detail. For example, during the foundation construction stage, the construction activity description statement may record "Conduct foundation excavation operations, using excavators and loaders", and the corresponding on-site status description statement may be "The on-site ground is leveled, and warning signs are set up around", and these description units are continuously recorded in chronological order, forming a complete construction record chain.

[0014] As an implementation method, the obtaining of the collection of construction log texts generated during the project construction process includes: Step 111: Collect the original construction log data from the project management information system. The original construction log data contains multiple segments of unstructured text records indexed by timestamps.

[0015] In this commercial building project, a set of project management information system is used to record the construction situation, and this system allows construction personnel to upload construction logs at any time. The original construction log data is indexed by timestamps, which is convenient for searching and sorting in chronological order. For example, in the record of a certain day, the timestamp shows "9:00 am, September 15, 2023", and the corresponding unstructured text record may be "Today, the steel bar binding work for the basement wall starts, and the on-site workers are operating in an orderly manner. The weather is sunny, and the construction equipment is running normally". Although these text records have no fixed format, they contain rich construction information, and the chronological order of construction can be clearly understood through the timestamps.

[0016] Step 112: Perform data cleaning processing on the original construction log data, filter out text segments containing duplicate content, punctuation marks with format errors, and redundant content irrelevant to construction, and obtain the candidate construction log texts after preliminary cleaning.

[0017] It is understandable that there may be some problems with the original construction log data and it needs to be cleaned. For example, in this project, sometimes construction workers may, due to negligence, duplicate some records. When cleaning the data, first, perform duplicate text detection on the original construction log data. By calculating the text similarity value between adjacent text segments, the text segments with similarity values exceeding the preset threshold are determined as duplicate content and only one of them is retained. For example, for two text segments "Concrete pouring work was carried out today, and the on-site workers were skilled" and "Concrete pouring operation was carried out today, and the on-site workers were skilled", after calculating, the similarity value exceeds the preset threshold, so only one of them is retained.

[0018] At the same time, perform format verification on the punctuation marks in the original construction log data. In project management, punctuation marks also need to be unified and standardized. Unify full-width symbols and half-width symbols into the corresponding full-width punctuation format for project management, and correct multiple consecutive punctuation marks into a single standard punctuation mark. For example, correct "Today's work was very smooth!!! " to "Today's work was very smooth! ".

[0019] In addition, perform topic relevance analysis on the text content in the original construction log data. Extract the keyword set of the text content, match the keyword set with the preset construction topic keyword library, and screen out the text segments with keyword matching degrees exceeding the preset threshold as valid content related to construction. For example, the preset construction topic keyword library includes keywords such as "building materials", "construction equipment", "construction technology", etc. If a text segment mentions "New building materials were used for wall masonry", after keyword matching, this text segment belongs to the valid content related to construction.

[0020] Finally, merge the detected non-duplicate content text segments, the text segments with corrected formats, and the valid content text segments related to construction to generate a candidate text for the construction log after preliminary cleaning. This candidate text for the construction log does not contain duplicate content, punctuation marks with format errors, and redundant content unrelated to construction.

[0021] In a preferred embodiment, the data cleaning process for the original construction log data, filtering out text segments containing duplicate content, punctuation marks with format errors, and redundant content unrelated to construction, to obtain a candidate text for the construction log after preliminary cleaning, includes: Step 1120: Perform duplicate text detection processing on the original construction log data, calculate the text similarity values between adjacent text segments, determine the text segments with similarity values exceeding the preset threshold as duplicate content and retain only one of them; perform format verification processing on the punctuation marks in the original construction log data, unify full-width symbols and half-width symbols into the full-width punctuation format corresponding to project management, and correct multiple consecutive punctuation marks into a single standard punctuation mark; perform theme relevance analysis processing on the text content in the original construction log data, extract the keyword set of the text content, match the keyword set with the preset construction theme keyword library, and screen out the text segments with keyword matching degrees exceeding the preset threshold as valid content related to construction; merge the detected non-duplicate content text segments, the text segments with corrected formats, and the valid content text segments related to construction to generate a candidate text for the construction log after preliminary cleaning, and the candidate text for the construction log does not contain duplicate content, punctuation marks with format errors, and redundant content unrelated to construction.

[0022] Continuing with this commercial building project as an example, in the duplicate text detection processing, the similarity between every two adjacent text segments is calculated through a text similarity calculation algorithm. For example, one text segment describes "Some doors and windows were installed today, and the workers operated according to the specifications", and another text segment is "Part of the door and window installation work was carried out today, and the workers' operations met the specifications". After calculation, their similarity value exceeds the preset threshold. At this time, these two text segments will be determined as duplicate content, and only one of them will be retained.

[0023] In the punctuation mark format verification processing, carefully check the punctuation marks in the original construction log data. For some cases of mixed use of full-width and half-width punctuation marks, uniformly convert them to the full-width punctuation format stipulated by project management. For example, convert "Today's work is completed, but there are some problems?" to "Today's work is completed, but there are some problems?", and correct multiple consecutive punctuation marks, such as "!!!", into a single standard punctuation mark "!".

[0024] In the theme relevance analysis processing, extract the keyword set from each text segment of the original construction log data. For example, for a text segment "At the construction site, a crane was used to lift building materials, and at the same time, the wall painting work was carried out", the extracted keyword set may be "crane", "building materials", "wall painting". Then match these keyword sets with the preset construction theme keyword library. Only when the keyword matching degree exceeds the preset threshold, the text segment is considered as valid content related to construction. Finally, merge the text segments after these processes to obtain a candidate text for the construction log after preliminary cleaning, ensuring that the text does not contain duplicate content, punctuation marks with format errors, and redundant content unrelated to construction.

[0025] Step 113: Segment the candidate construction log text at preset time intervals, and merge the text content describing the same construction stage within a continuous time range into a construction stage description unit. The time range of the construction stage description unit is consistent with the stage division rules in the project construction plan.

[0026] In this commercial building project, the project construction plan divides the entire project into multiple construction stages, such as the foundation construction stage, the main structure construction stage, the interior decoration stage, etc. The preset time interval can be one week or one month, etc. For example, taking one week as the time interval, segment the candidate construction log text after preliminary cleaning. During a certain week, the construction log records the work situation from Monday to Friday. From Monday to Wednesday, the main work is the binding of foundation steel bars, and on Thursday and Friday, the installation of foundation formwork begins. Then, the text content describing the binding of foundation steel bars and the text content describing the installation of foundation formwork are merged into a construction stage description unit. The corresponding construction stage of this unit is the foundation construction stage, and its time range is consistent with the stage division rules in the project construction plan, which is convenient for subsequent systematic analysis and summary of the construction situation.

[0027] Step 114: Split the sentences of each construction stage description unit to separate the construction activity description sentences describing construction activities and the on-site status description sentences recording the on-site environment and equipment status, so that each construction stage description unit contains at least one construction activity description sentence and one on-site status description sentence.

[0028] Optionally, perform sentence splitting on the above-mentioned merged foundation construction stage description unit. For example, a paragraph of text in the construction stage description unit is "In the foundation construction area, from Monday to Wednesday, workers use a steel bar bender to bind the foundation steel bars, the on-site environment is clean, and the equipment is operating normally". Through analysis, the construction activity description sentence "Workers use a steel bar bender to bind the foundation steel bars" and the on-site status description sentence "The on-site environment is clean, and the equipment is operating normally" can be separated. Thus, each construction stage description unit can clearly contain information on both construction activities and on-site status, providing more accurate data for subsequent analysis.

[0029] Step 115: Check the text integrity of the split construction activity description sentences and on-site status description sentences, and generate a set of construction stage description units containing complete semantics as the construction log text set.

[0030] Further, perform text integrity verification on the separated construction activity description statements and on-site status description statements. For example, for the construction activity description statement "Workers use a steel bar bender to tie the foundation steel bars", check whether it contains key information such as the construction subject (workers), construction equipment (steel bar bender), construction object (foundation steel bars), and construction behavior (tying work). For the on-site status description statement "The on-site environment is clean and the equipment is operating normally", check whether it clearly describes the on-site environment and equipment status. Only when these statements all contain complete semantics will they be recombined into construction stage description units, and finally a set of construction stage description units containing complete semantics will be generated as the construction log text set.

[0031] Step 120: Perform natural language processing on the construction log text set, extract the key entities and progress events in each construction stage description unit, and obtain a construction feature set containing entity identification information and event time sequence information.

[0032] In this commercial building project, perform natural language processing on the generated construction log text set. For example, in a construction stage description unit in the foundation construction stage, the text is "On Tuesday, workers use a concrete mixer to mix concrete in the foundation construction area, the on-site ground has no ponding, and the equipment is operating normally". Through natural language processing technology, key entities and progress events need to be extracted from this text.

[0033] In one implementation manner, the performing natural language processing on the construction log text set, extracting the key entities and progress events in each construction stage description unit, and obtaining a construction feature set containing entity identification information and event time sequence information includes: Step 121: Perform word segmentation processing on the construction activity description statements and on-site status description statements in the construction log text set to obtain a word segmentation result set composed of basic vocabulary units.

[0034] For the construction activity description statement "Workers use a concrete mixer to mix concrete in the foundation construction area" and the on-site status description statement "The on-site ground has no ponding, and the equipment is operating normally" in the above construction stage description unit, perform word segmentation processing. The word segmentation algorithm will split these statements into basic vocabulary units. For example, the construction activity description statement may be segmented into "workers", "in", "foundation construction area", "use", "concrete mixer", "mix", "concrete", and the on-site status description statement may be segmented into "on-site", "ground", "no ponding", "equipment", "operating normally". These basic vocabulary units form the word segmentation result set.

[0035] Step 122: Invoke the pre-trained BERT language model to perform context semantic encoding processing on the tokenized result set, generating context embedding vectors for each basic lexical unit, where the context embedding vectors contain semantic association information of the words in the corresponding statements.

[0036] In this step, the tokenized result set is input into the pre-trained BERT language model. In the context of this commercial building project, the BERT language model will generate context embedding vectors for each basic lexical unit according to the context information of the words in the statements. For example, for the word "concrete mixer", the model will consider its context in the sentence "Workers use a concrete mixer to mix concrete in the foundation construction area", and generate a context embedding vector containing its semantic association information. This vector not only contains the semantics of "concrete mixer" itself, but also contains its semantic relationships with other words such as "workers", "mix", and "concrete".

[0037] Step 123: Based on the context embedding vectors, perform entity recognition processing to identify construction object entities, construction resource entities, and construction technology entities in the construction activity description statements, and identify environmental state entities, equipment state entities, and personnel state entities in the on-site status description statements, obtaining an entity identification set containing entity type and entity location information.

[0038] Optionally, in the construction activity description statement, through entity recognition algorithms, identify the construction object entity "concrete", the construction resource entities "concrete mixer" and "workers", and the construction technology entity "mix". In the on-site status description statement, identify the environmental state entity "site ground", the equipment state entity "equipment", and the personnel state entity (not explicitly mentioned in this statement, but may appear in other relevant statements, such as "site management personnel", etc.). At the same time, record the location information of each entity in the statement, such as the start and end positions of "concrete mixer" in the construction activity description statement, and finally obtain an entity identification set containing entity type and entity location information.

[0039] As a design idea, the performing entity recognition processing based on the context embedding vectors to identify construction object entities, construction resource entities, and construction technology entities in the construction activity description statements, and identify environmental state entities, equipment state entities, and personnel state entities in the on-site status description statements, obtaining an entity identification set containing entity type and entity location information includes: Step 1230: perform sequence labeling on the context embedding vector, label each basic vocabulary unit with an entity type label, the entity type label includes a construction object entity label, a construction resource entity label, a construction technology entity label, an environmental status entity label, an equipment status entity label and a personnel status entity label; perform entity boundary recognition on the labeled vocabulary units, merge vocabulary units that are continuously labeled with the same entity type label to form a complete entity name expression; record the starting position and ending position of each entity name expression in the construction activity description statement or the site status description statement, and generate a character index range representing the entity position; perform semantic consistency verification on the identified entity name expressions, and unify different expression forms of the same entity into a standard entity name in the engineering specification knowledge base; integrate the entity type label, standard entity name and character index range to generate an entity identifier set containing entity type and entity location information.

[0040] In this commercial construction project, context embedding vectors were sequentially labeled. For example, for lexical units in a construction activity description, "concrete" was labeled with a construction object entity label, "concrete mixer" was labeled with a construction resource entity label, and "mixing" was labeled with a construction technology entity label, based on their semantics and role in the sentence. Entity boundaries were then identified for the labeled lexical units, and consecutive lexical units labeled with the same entity type label were merged. For example, "concrete mixer" was considered a single construction resource entity. Next, the starting and ending positions of each entity name in the sentence were recorded, such as the character index range of "concrete mixer" in the sentence. Semantic consistency checks were also performed on the identified entity name representations. For example, if "concrete mixer" ("concrete" is short for "concrete") appears in other sentences, it is unified with the standard entity name "concrete mixer" in the engineering specification knowledge base. Finally, the entity type label, standard entity name, and character index range are combined to form an entity identifier set containing entity type and location information.

[0041] Step 124: Perform temporal relationship analysis on the time adverbials and sequential conjunctions in the construction phase description unit to determine the execution order between the construction activity description statements and the temporal correspondence between the on-site status description statements and the construction activities, and generate event timing information representing the execution order of the construction activities and the timing of status changes.

[0042] In the construction phase description unit during the foundation construction phase, analyze the time adverbials and sequential conjunctions. For example, in the construction phase description unit, there is a description like "On Tuesday, the workers first carry out the steel bar binding work, and then start installing the formwork on Wednesday". By analyzing the time adverbials and sequential conjunctions such as "Tuesday", "Wednesday", "first", and "then", it is determined that the steel bar binding work is carried out on Tuesday, the formwork installation work is carried out on Wednesday, and the steel bar binding work precedes the formwork installation work. At the same time, for the on-site status description statement "The on-site ground is level on Tuesday, and there is a small amount of materials piled up on the site during the formwork installation on Wednesday", determine the time correspondence relationship between the on-site status and the construction activities. Finally, generate event time series information representing the execution order of construction activities and the timing of status changes, clarifying the sequence of construction activities and the changes in on-site status over time.

[0043] Step 125: Perform an association and integration process on the set of entity identifiers and the event time series information to generate a construction feature set containing entity identifier information and event time series information. The entity identifier information includes entity type, entity name, and statement position, and the event time series information includes event execution order and status change time points.

[0044] Optionally, perform an association and integration on the previously obtained set of entity identifiers and event time series information. In this commercial building project, the set of entity identifiers contains information on construction resource entities such as "concrete mixer", and the event time series information includes information such as "The steel bar binding work is carried out on Tuesday, and the formwork installation work is carried out on Wednesday". After integrating them, the generated construction feature set not only knows about the construction resource entity "concrete mixer", but also knows that it may be used in the steel bar binding work on Tuesday or the formwork installation work on Wednesday, clarifying the temporal association between the entity and the event. Thus, the construction feature set contains complete entity identifier information and event time series information.

[0045] Step 130: Perform a semantic alignment process on the construction feature set and a pre-set engineering specification knowledge base to generate specification association information between the construction phase description unit and the engineering specification requirements.

[0046] In a commercial building project, the engineering specification knowledge base contains various construction specifications and standards. The construction feature set contains entity and event information extracted from the construction log, and the two need to be semantically aligned to determine the association between the construction phase description unit and the engineering specification requirements.

[0047] In a preferred embodiment, the performing a semantic alignment process on the construction feature set and a pre-set engineering specification knowledge base to generate specification association information between the construction phase description unit and the engineering specification requirements includes: Step 131: Extract a set of specification entries related to the construction object entity and the construction resource entity from the engineering specification knowledge base. The set of specification entries includes construction technical requirements, resource allocation standards, and environmental safety regulations.

[0048] From the engineering specification knowledge base, for the foundation construction stage of a commercial building project, extract a set of specification entries related to construction object entities such as "concrete" and construction resource entities such as "concrete mixer". For example, for "concrete", the specification entries may include construction technical requirements such as the mix ratio requirements and pouring temperature requirements of concrete; for "concrete mixer", the specification entries may involve resource allocation standards such as the regular maintenance standards of the equipment and the qualification requirements of the operators, as well as environmental safety regulations such as the safety protection regulations during equipment operation.

[0049] Step 132: Perform semantic feature extraction processing on the entity identification information in the construction feature set to generate an entity semantic vector representing the entity type and entity attributes.

[0050] Furthermore, perform semantic feature extraction processing on the entity identification information in the construction feature set. For example, for the entity identification information of "concrete mixer", analyze its entity type (construction resource entity) and its attributes, such as the model and power of the equipment. Encode this information to generate an entity semantic vector representing the entity type and entity attributes. This vector can represent the semantic features of "concrete mixer" in the construction feature set and provide a basis for subsequent comparison with the specification semantic vector.

[0051] Step 133: Perform semantic feature extraction processing on the specification content in the set of specification entries to generate a specification semantic vector representing the specification requirements and applicable conditions.

[0052] Optionally, for the set of specification entries extracted from the engineering specification knowledge base, perform semantic feature extraction on the specification content therein. Taking the specification entries of "concrete mixer" as an example, the specification requirements may include that the equipment needs to be comprehensively inspected regularly and the operators need to have corresponding operation certificates, etc.; the applicable conditions may be in the corresponding construction environment and construction tasks. Analyze and encode these specification requirements and applicable conditions to generate a specification semantic vector representing the specification requirements and applicable conditions. This vector reflects the semantic features of the specification entries and is convenient for matching and comparing with the entity semantic vector.

[0053] Step 134: Calculate the semantic similarity value between the entity semantic vector and the specification semantic vector, and filter out the specification entries whose semantic similarity value with the entity identification information in the construction feature set exceeds a preset threshold as associated specification entries.

[0054] In this step, the semantic similarity value between the entity semantic vector and the canonical semantic vector can be calculated by the cosine similarity algorithm. In this commercial building project, taking the entity semantic vector of "concrete mixer" and its corresponding canonical semantic vector as an example, the similarity between the two is calculated. After the calculation, a preset threshold is set, and the canonical entries with semantic similarity values greater than this preset threshold are selected as candidate associated canonical entries. For example, if the similarity value between the semantic vector of the canonical entry "The concrete mixer needs to be comprehensively maintained once every two weeks" and the entity semantic vector of "concrete mixer" exceeds the preset threshold, then this canonical entry becomes a candidate associated canonical entry.

[0055] In a preferred embodiment, calculating the semantic similarity value between the entity semantic vector and the canonical semantic vector, and selecting the canonical entries with semantic similarity values exceeding the preset threshold with respect to the entity identification information in the construction feature set as associated canonical entries includes: Step 1340: Perform dimension alignment processing on the entity semantic vector and the canonical semantic vector; calculate the cosine value of the angle between the entity semantic vector and the canonical semantic vector after dimension alignment processing as the semantic similarity value; perform normalization processing on the semantic similarity value to obtain a normalized similarity value; set a semantic similarity threshold, and select the canonical entries with normalized similarity values greater than the semantic similarity threshold as candidate associated canonical entries; perform redundancy screening processing on the candidate associated canonical entries, remove the duplicate canonical entries associated with the same entity identification information, and retain the canonical entry with the highest semantic similarity value as the associated canonical entry.

[0056] In a commercial building project, first, dimension alignment processing is performed on the entity semantic vector of "concrete mixer" and the relevant canonical semantic vector to ensure that the two vectors are consistent in dimension for subsequent calculation. Then, the cosine value of the angle is calculated to determine the semantic similarity value between them. For example, the cosine value of the angle between the entity semantic vector and the canonical semantic vector is calculated, and this value reflects the similarity degree of the two vectors in direction, thus representing the semantic similarity degree.

[0057] Then, normalization processing is performed on the obtained semantic similarity value to map it to a set range (such as [0, 1]) to obtain a normalized similarity value, which is convenient for comparison and threshold setting. Set a suitable semantic similarity threshold, such as 0.7. Select the canonical entries with normalized similarity values greater than 0.7 as candidate associated canonical entries.

[0058] After obtaining the candidate associated specification entries, there may be some duplicate or redundant specification entries. For example, for a "concrete mixer", there may be multiple specification entries related to equipment maintenance, but the expressions are slightly different. At this time, redundant screening is performed on the candidate associated specification entries to remove the duplicate specification entries associated with the same entity identification information, and only the specification entry with the highest semantic similarity value is retained as the associated specification entry, which can ensure that the associated specification entries are both accurate and concise, providing more effective information for subsequent analysis.

[0059] Step 135: Combine the event time sequence information in the construction feature set, analyze the applicable time range of the associated specification entry in the construction stage description unit, determine the time matching relationship between the execution order of construction activities and the state change time point and the specification requirements, and generate specification association information including the associated specification entry and the time matching relationship.

[0060] It can be understood that in terms of the event time sequence information in the construction feature set, in the foundation construction stage of a commercial building project, the event time sequence information of "steel bar binding work is carried out on Tuesday, formwork installation work is carried out on Wednesday, and a concrete mixer may be used in these two works" and the associated specification entry of "the concrete mixer needs to be comprehensively maintained every two weeks" are known. Analyze the applicable time range of this associated specification entry in the construction stage description unit, and judge whether it is time to comprehensively maintain the concrete mixer during the steel bar binding work on Tuesday and the formwork installation work on Wednesday. Determine the time matching relationship between the execution order of construction activities and the state change time point and the specification requirements. For example, if the specification requires maintenance after the equipment has been continuously used for one week, and in the current construction plan, the concrete mixer has been used from last week to this Wednesday and is close to one week, then there is a time matching relationship. Finally, generate specification association information including the associated specification entry and the time matching relationship.

[0061] Step 140: Perform structured integration processing on the specification association information based on the hierarchical template matching algorithm to generate a structured report text that meets the requirements of project management. The structured report text includes a construction progress description, a specification compliance analysis, and recommended content.

[0062] In a commercial building project, the hierarchical template matching algorithm is used to process the specification association information to generate a structured report text that meets the requirements of project management, providing clear and accurate information for project management.

[0063] In an alternative embodiment, the structured integration processing of the specification association information based on the hierarchical template matching algorithm to generate a structured report text that meets the requirements of project management, the structured report text includes a construction progress description, a specification compliance analysis, and recommended content, including: Step 141: Select a hierarchical template structure from a preset project management report template library. The hierarchical template structure includes a first-level heading layer, a second-level content layer, and a third-level detail layer. The first-level heading layer corresponds to the construction progress description, specification compliance analysis, and recommendation content.

[0064] In a commercial building project, there are various template structures available in the preset project management report template library. Select the hierarchical template structure, where the first-level heading layer clearly corresponds to the three key parts: construction progress description, specification compliance analysis, and recommendation content. For example, in this commercial building project, the construction progress description part is used to clearly show the progress of each construction stage; the specification compliance analysis part is used to judge whether the construction activities meet the project specification requirements; the recommendation content part provides targeted improvement or optimization suggestions based on the analysis results. The second-level content layer and the third-level detail layer provide more detailed elaboration and supporting information for these main parts.

[0065] Step 142: Map the event time sequence information in the construction feature set to the construction progress description template area of the first-level heading layer, and organize the timeline description content of the construction activities according to the event execution order and the status change time points.

[0066] Specifically, map the event time sequence information in the construction feature set, such as "Reinforcement binding work was carried out on Tuesday, and formwork installation work was carried out on Wednesday", to the construction progress description template area of the first-level heading layer. Organize the timeline description content of the construction activities in the template area according to the event execution order and the status change time points. In the report of the commercial building project, it may be presented as "In the foundation construction stage, reinforcement binding work was carried out on Tuesday, and the construction status was normal; formwork installation work was carried out on Wednesday, and the on-site construction progressed smoothly". In this way, the time sequence and progress of the construction activities are clearly shown, enabling managers to quickly understand the construction progress of the project.

[0067] Step 143: Map the associated specification items and time matching relationships in the specification association information to the specification compliance analysis template area of the first-level heading layer, compare the time matching relationship between the construction activity execution situation and the specification requirements, and generate an analysis conclusion on whether the construction activities meet the specification requirements.

[0068] Specifically, map the associated specification entries and time matching relationships in the specification association information, such as "The concrete mixer needs to be comprehensively maintained every two weeks. In the current construction, the concrete mixer has been continuously used for one week, and it is expected to need maintenance next week", to the specification compliance analysis template area at the first-level heading layer. In this area, compare the time matching relationship between the implementation of construction activities and the specification requirements. For the foundation construction stage in a commercial building project, if the specification requires the concrete mixer to be inspected after being continuously used for one week, but the inspection has not been carried out yet when this time has been reached in actual construction, then the analysis conclusion is that the current construction activities do not meet the specification requirements in terms of the maintenance of the concrete mixer. Through the above comparison and analysis, generate a clear analysis conclusion on whether the construction activities meet the specification requirements, providing an important decision-making basis for project management.

[0069] Step 144: Based on the analysis conclusion in the specification compliance analysis template area, extract improvement measure suggestions for non-compliance situations and optimization suggestions for compliance situations from the engineering specification knowledge base, and map them to the suggestion content template area at the first-level heading layer.

[0070] Optionally, according to the analysis conclusion obtained from the specification compliance analysis template area, in a commercial building project, if the analysis conclusion shows that the maintenance of the concrete mixer does not meet the specification requirements, extract improvement measure suggestions for this non-compliance situation from the engineering specification knowledge base, such as "Immediately arrange a comprehensive maintenance of the concrete mixer and establish a regular maintenance reminder mechanism". If the construction activities meet the specification requirements, also extract optimization suggestions from the knowledge base, such as "Consider upgrading the equipment during maintenance to improve construction efficiency". Map these suggestions to the suggestion content template area at the first-level heading layer to provide specific directions and measures for the improvement and optimization of the project.

[0071] Step 145: Optimize the semantic description of the second-level content layer and the third-level detail layer. Adopt a transfer learning strategy to adjust the general description statements in the template into optimized description statements that conform to the semantic style of the construction log text set, so that the language expression of the structured report text is consistent with the semantic output state of the original construction log, and generate a structured report text including the construction progress description, specification compliance analysis, and suggestion content.

[0072] Specifically, the optimization of the semantic description of the second-level content layer and the third-level detail layer, adopting a transfer learning strategy to adjust the general description statements in the template into optimized description statements that conform to the semantic style of the construction log text set, so that the language expression of the structured report text is consistent with the semantic output state of the original construction log, includes: Step 1450: Extract the set of frequent item words and key sentence structures from the construction log text collection to generate a language feature template representing the semantic style of the original construction log; Select a pre-trained transfer learning model, use the language feature template as the target style for style transfer, and use the general description statements in the language feature template as the input text; Fine-tune the transfer learning model so that the transfer learning model learns to convert the general description statements into optimized description statements that conform to the language feature template; Input the general description statements in the secondary content layer and the tertiary detail layer into the fine-tuned transfer learning model to generate optimized description statements after style transfer; Check the grammatical correctness and semantic coherence of the optimized description statements, and correct the grammatical errors and semantic ambiguities generated during the transfer process to obtain optimized description statements that are consistent with the semantic output state of the original construction log.

[0073] In a commercial building project, extract the set of frequent item words, such as words like "carry out", "use", "site", etc., and key sentence structures, such as "In... area, ... carry out... work", etc. from the construction log text collection. Integrate this information to generate a language feature template representing the semantic style of the original construction log.

[0074] Then, select a pre-trained transfer learning model, use the generated language feature template as the target style for style transfer, and use general description statements in the language feature template, such as "Carry out specific work in a specific area" as the input text. Fine-tune the transfer learning model to let the model learn how to convert such general description statements into optimized description statements that conform to the language feature template, for example, convert it to "At the construction site, workers carry out specific construction work".

[0075] Furthermore, input the general description statements in the secondary content layer and the tertiary detail layer into the fine-tuned transfer learning model. For example, the description of construction equipment in the secondary content layer "The equipment is used in the site", after being processed by the model, generates an optimized description statement after style transfer "The equipment is put into use at the construction site".

[0076] Next, check the grammatical correctness and semantic coherence of the generated optimized description statements. Check whether there are grammatical errors in the statements, such as improper subject-predicate-object collocations, etc., and at the same time check whether the semantics is coherent and whether it conforms to the logic of the construction scenario. Correct the statements with grammatical errors or semantic ambiguities, and finally obtain optimized description statements that are consistent with the semantic output state of the original construction log, so as to generate a structured report text containing construction progress descriptions, specification compliance analysis, and recommended content.

[0077] In an alternative embodiment, after generating the structured report text that meets the requirements of project management, it further includes: Step 210: Check the content integrity of the construction progress description report area in the structured report text to see if it includes the event execution order and status change time points of all construction stage description units.

[0078] After generating the structured report text for a commercial building project, check the construction progress description report area therein. For example, in the foundation construction stage, the construction progress description should cover the event execution order of each construction activity from foundation excavation, steel bar binding, formwork installation to concrete pouring, as well as the status change time points corresponding to each activity, such as start time, end time, and whether it is successfully completed. Check whether these information are completely recorded in the report. If there is an omission of a construction activity or the record of its time point is incomplete, it is marked as incomplete content and needs to be further supplemented and improved.

[0079] Step 220: Check the logical rationality of the specification compliance analysis report area in the structured report text to see if the analysis conclusion is drawn based on the associated specification items and time matching relationships in the specification association information.

[0080] For the specification compliance analysis report area, check the logical rationality of the analysis conclusion. In a commercial building project, if the analysis conclusion indicates that a certain construction activity does not meet the specification requirements, check whether this conclusion is drawn based on the associated specification items and time matching relationships in the specification association information. For example, if the conclusion is that the concrete pouring temperature does not meet the specification requirements, then it is necessary to confirm whether there are corresponding specification items stipulating the temperature range for concrete pouring, and whether the concrete pouring time and actual temperature recorded in the construction log are correctly compared and analyzed with the specification requirements. If the analysis conclusion is not reasonably drawn based on the specification association information, then the analysis process needs to be reexamined and corrected.

[0081] Step 230: Check the feasibility of the suggestions in the suggestion content report area of the structured report text to see if there are corresponding specification bases in the engineering specification knowledge base for the improvement measure suggestions and optimization suggestions.

[0082] In a commercial building project, check the suggestion content report area. Check whether there are corresponding specification bases in the engineering specification knowledge base for the proposed improvement measure suggestions and optimization suggestions. For example, if it is suggested to improve the construction technology to enhance the construction quality, it is necessary to check whether there are relevant specifications and standards in the engineering specification knowledge base to support the improvement of this construction technology. If the suggestion has no corresponding specification basis, then the feasibility of this suggestion is questionable, and the suggestion content needs to be re-evaluated and adjusted to ensure that the suggestion has practical operability and normativeness.

[0083] Step 240: Correct the report areas with missing content, logical contradictions, or unfounded suggestions found during the verification process, supplement the missing progress information, adjust the contradictory analysis conclusions, and replace the unfounded suggestion content.

[0084] According to the verification results, correct the structured report text. For the missing progress information in the construction progress description report area, such as the omission of the time record of a certain construction link, search and supplement it from the original construction log. For the analysis conclusions with logical contradictions in the specification compliance analysis report area, re-examine the specification-related information and the actual construction situation, and adjust the analysis conclusions to make them logical. For the unfounded suggestion content in the suggestion content report area, find appropriate basis from the engineering specification knowledge base or re-propose well-founded suggestions to replace the original unfounded suggestions to ensure the accuracy and reliability of the report text.

[0085] Step 250: Generate an updated structured report text that has undergone content integrity verification and correction processing.

[0086] After the above verification and correction processing, integrate the corrected content to generate an updated structured report text. This updated report text is more accurate, complete, and reasonable in terms of construction progress description, specification compliance analysis, and suggestion content, and can provide more effective decision-making support and reference basis for the project management of commercial building projects.

[0087] As a non-limiting embodiment, after generating the structured report text that meets the requirements of project management, it further includes: Obtain the semantic fluency evaluation requirements of the structured report text, where the semantic fluency evaluation requirements include the formality requirements and expression coherence standards of engineering management documents; Call a pre-trained NLP generation optimization model to perform language style conversion processing on the structured report text, and the NLP generation optimization model is trained based on the standard document corpus in the field of engineering management; Through the NLP generation optimization model, perform written conversion processing on the colloquial expressions in the structured report text, merge the repeated texts, and supplement connecting words for the logically jumping sentences; Extract the set of professional terms in the structured report text, perform consistency matching processing on the set of professional terms with the standard term library in the engineering specification knowledge base, and replace the non-standard terms in the set of professional terms with standard terms; Generate a semantic optimization report text that has undergone language style conversion, repeated text merging, logical connection supplement, and term consistency correction.

[0088] It can be understood that in commercial building projects, the requirements for evaluating the semantic fluency of structured report texts are first clarified. These requirements include the formality requirements of engineering management documents, such as rigorous and standardized language expressions, avoiding colloquial and casual expressions; and the coherence criteria of expressions, that is, there should be reasonable logical connections between sentences, avoiding logical leaps.

[0089] Call a pre-trained NLP generation optimization model, which is trained based on a large corpus of standard documents in the field of engineering management, and perform language style conversion processing on the structured report texts of commercial building projects. For example, convert the colloquial expression "start working today" in the report text to the written form "start construction work today".

[0090] Merge the repeated texts in the report text. For example, if "concrete pouring work is carried out at the construction site" is mentioned multiple times, it is merged into "concrete pouring work is carried out at the construction site". For sentences with logical leaps, add connecting words to make them logically coherent. For example, "The construction workers carried out steel bar binding, and then started formwork installation" becomes "After the construction workers completed the steel bar binding work, they immediately started the formwork installation work" after adding connecting words.

[0091] Extract the set of professional terms in the structured report text, such as "concrete" (the Chinese character for concrete), "scaffold", etc., and perform consistency matching processing on this set of professional terms with the standard term library in the engineering specification knowledge base. If it is found that the professional term "concrete" (the Chinese character for concrete) in the report does not conform to "concrete" in the standard term library, it is replaced with the standard term. After the above processing, a semantically optimized report text is generated, making the report more in line with the requirements of engineering management documents in terms of language style, expression coherence, and the use of professional terms.

[0092] As a non-limiting embodiment, after generating the structured report text that meets the requirements of engineering management, it further includes: Extract the event timing information of the construction progress description text set, the associated specification entries of the specification compliance analysis text set, and the improvement measure suggestions of the suggestion content text set from the structured report text to form a set of report content elements; Perform logical association analysis processing on the set of report content elements to identify whether there are contradictions in the event execution order in the construction progress description and the time matching relationship in the specification compliance analysis, and check whether there is a logical break between the improvement measure suggestions in the suggestion content and the conclusions of the specification compliance analysis; For the identified contradictions or break points, trace back the entity identification information and event timing information in the construction feature set to verify whether the contradiction points are caused by the original record deviation of the construction log text set or the information mapping error of the hierarchical template matching algorithm, and obtain the traceback verification result; Perform local content correction processing on the structured report text according to the backtracking verification results, adjust the expression of the event execution order to match the time matching relationship, and supplement the logical derivation process between the recommended content and the analysis conclusion; Generate a logical verification report text that has undergone logical association analysis and local content correction.

[0093] Specifically, in a commercial building project, extract the event time sequence information of the construction progress description text set from the structured report text, such as "Reinforcement binding is carried out on Tuesday and formwork installation is carried out on Wednesday", the associated specification items of the specification compliance analysis text set, such as "Concrete mixers need to be maintained regularly", and the improvement measure suggestions of the recommended content text set, such as "The daily inspection frequency of concrete mixers should be increased", and integrate this information to form a report content element set.

[0094] Perform logical association analysis processing on the report content element set. In this commercial building project, focus on identifying whether there are contradictions between the event execution order in the construction progress description and the time matching relationship in the specification compliance analysis. For example, the construction progress description shows that the concrete mixer has been in use for construction during a certain period, but the associated specification item in the specification compliance analysis requires that the concrete mixer should be maintained during this period, yet there is no maintenance-related content in the actual records, then there is a contradiction in the time matching relationship. At the same time, check whether there is a logical break between the improvement measure suggestions in the recommended content and the conclusion of the specification compliance analysis. For example, the conclusion of the specification compliance analysis shows that the concrete pouring temperature does not meet the specification requirements, but the recommended content proposes to strengthen the concrete mixing intensity, which is obviously not directly logically related to the analysis conclusion and belongs to the situation of logical break.

[0095] For the identified contradictions or breakpoints, backtrack the entity identification information and event time sequence information in the construction feature set. For example, in the above example of the concrete mixer, view the entity identification information about the concrete mixer in the construction feature set, including equipment model, usage records, etc., and the event time sequence information to determine whether there are deviations in the original records of the construction log text set. It is possible that the construction personnel made a recording error, and the actual maintenance should have been carried out but was misrecorded as normal use; or check whether it is an information mapping error of the hierarchical template matching algorithm that caused it. For example, when mapping the construction feature information to the specification analysis template area, there may be errors in time or entity association. After detailed inspection and verification, obtain the backtracking verification results.

[0096] Perform local content correction processing on the structured report text according to the backtracking verification results. If the contradiction is caused by the deviation of the original records in the construction log text set, such as a recording error being found, then adjust the description of the event execution order to match the time matching relationship, and clearly record the actual use and maintenance of the concrete mixer in the report. If there is a logical break between the recommended content and the analysis conclusion, supplement the logical derivation process between the recommended content and the analysis conclusion. For example, on the basis of recommending to strengthen the concrete mixing intensity, further explain the relationship between the mixing intensity and the concrete temperature, and how to affect the concrete pouring temperature by adjusting the mixing intensity to meet the specification requirements.

[0097] Through the above operations, a logical verification report text with logical correlation analysis and local content correction is generated. This report text establishes a closer and more reasonable logical connection among the construction progress, specification compliance, and recommended content, improving the accuracy and reliability of the report, and providing more valuable reference information for the project management of commercial building projects.

[0098] As a non-limiting embodiment, after generating the structured report text that meets the requirements of project management, it further includes: Obtain the role identification information of the target reading user, and the role identification information includes the management user role, the technical execution layer user role, and the external supervision user role; Extract the set of information focus points corresponding to the role from the preset user requirement feature library according to the role identification information, and the set of information focus points includes the summary of key indicators concerned by the management layer, the description of operation details concerned by the technical execution layer, and the description of specification basis concerned by the external supervision user; Perform content priority sorting processing on the structured report text, perform information focusing processing on the sorted content, generate an overview report containing the summary of key indicators and conclusion overview for the management user role, generate a marked report containing the description of operation details and step instructions for the technical execution layer user role, and generate a compliance report containing the description of specification basis and compliance certificate for the external supervision user role; Generate multi-version adapted report texts for different target reading user roles, and the content detail level and information focusing direction of the multi-version adapted report texts match the set of information focus points of the corresponding roles.

[0099] In a commercial building project, first obtain the role identification information of the target reading user. Different user roles have different requirements and concerns for the report. The embodiments of the present invention include the management user role, the technical execution layer user role, and the external supervision user role.

[0100] Among them, the management is more concerned about the overall progress of the project and key indicators in order to make decisions. According to the role identification information, the set of information focus points concerned by the management is extracted from the preset user demand feature library, which includes a summary of key indicators, such as the project schedule completion rate, cost control situation, quality compliance rate, etc. For commercial building projects, the management may be concerned about whether the foundation construction stage is completed as planned, whether the cost is within the budget, and whether the quality of key construction links such as concrete pouring meets the standards.

[0101] In addition, the technical implementation layer mainly focuses on the details and steps of construction operations to ensure that construction tasks can be executed accurately. The set of information focus points concerned by the technical implementation layer is extracted from the user demand feature library, which includes descriptions of operation details, such as the specific operation process of the concrete mixer, the spacing requirements for steel bar binding, etc. In commercial building projects, the personnel of the technical implementation layer need to understand the specific operation requirements of each construction link in detail in order to guide on-site construction.

[0102] Furthermore, external regulatory users mainly focus on whether the project complies with relevant specifications and standards, and a clear specification basis description is required. The set of information focus points concerned by external regulatory users is extracted from the user demand feature library, including the specification basis description, such as the engineering specification clauses and safety standards based on which each construction activity is carried out. For commercial building projects, external regulatory users will check whether the foundation construction complies with specification requirements such as the Construction Quality Acceptance Standard for Building Foundation Engineering.

[0103] Next, the content priority sorting process is carried out on the structured report text. According to the information focus points of different user roles, the content in the report is sorted by importance. For example, for management users, the content related to key indicators is ranked at the top; for users of the technical implementation layer, the priority of the content related to operation details is increased; for external regulatory users, the content related to the specification basis is highlighted.

[0104] Perform information focusing processing on the sorted content. For the management user role, extract information related to the key metric summary and conclusion overview from the structured report text, and generate an overview report containing this content. For example, concisely present key information such as the progress completion, cost deviation, and overall quality evaluation during the foundation construction stage in the overview report, enabling management to quickly understand the core situation of the project. For the technical implementation user role, extract content related to the operation detail description and step instructions from the report, and generate a marked report. For example, in the marked report, describe in detail the startup, operation process, and shutdown steps of the concrete mixer, as well as operation details such as the specific spacing and tying method of steel bar binding, providing detailed construction guidance for technical implementation personnel. For the external supervision user role, extract content related to the specification basis description and compliance certificate, and generate a compliance report. For example, clearly indicate in the compliance report the engineering specification clauses relied on for each link of the foundation construction, and by what detection means it is proved that the construction activities comply with these specification requirements, providing a clear compliance certificate for external supervision users.

[0105] Through the above processing, multi-version adapted report texts for different target reading user roles are finally generated. These report texts closely match the information focus of the corresponding roles in terms of content detail and information focus direction, meeting the diverse needs of different user roles in the engineering management of commercial building projects, helping to improve the efficiency of project management and the scientific nature of decision-making, and ensuring the smooth progress of the project and compliance with relevant specification standards.

[0106] In the actual application process, for text similarity calculation, it can be achieved by introducing ready-made similarity functions (such as cosine similarity or Jaccard similarity) in NLP toolkits (such as spaCy or scikit-learn), calculating the text segment similarity value by calling their interfaces and applying a preset threshold to filter duplicate content; for word segmentation processing and context semantic encoding, directly implement word segmentation and embedding vector generation based on the interfaces of pre-trained BERT models (such as those provided by the HuggingFace library), without the need for self-training or in-depth study of the structure; for entity recognition processing, complete entity type annotation and boundary merging with the help of a standard sequence annotation model library (such as the entity recognition module of spaCy or BERT-based NER), and use the default trained model to process text data; for vector alignment and cosine similarity in semantic similarity calculation, directly use the vector similarity calculation function in the NLP library (such as the cosine similarity implementation of NumPy) to perform dimension and value processing.

[0107] More specifically, for the optimization process of transfer learning, a transfer learning framework (such as TensorFlowHub or the pre-trained T5 model) is used for fine-tuning training. By loading the base model and inputting the style feature template, the general statements are adjusted. For rules such as time interval division, the intervals are automatically set with reference to the engineering management standard rule library (such as the Primavera project management rule set). For the knowledge base matching rules, a SQL-like rule engine (such as Drools) is used to process the screening of specification entries.

[0108] In addition, the logical relevance and hierarchy between the input construction log text set and the output structured report text are described as follows: The input is the original construction log text set (including construction activity description statements and on-site status description statements). First, through the data cleaning processing layer, redundant and incorrect content is filtered to form the preliminary cleaned text. This layer focuses on the normalization and integrity of the original data. Then, through the segmentation processing layer, the text is merged into construction stage description units, and the time dimension structure is added according to the time interval rule. Then it enters the natural language processing layer. The tokenization results generate context embedding vectors through the BERT model, providing a semantic basis for entity recognition and time series analysis. This layer introduces the entity identification set and event time series information as intermediate outputs, constituting the core elements of the construction feature set. Subsequently, through the semantic alignment processing layer, the construction feature set is matched with the engineering specification knowledge base to generate specification association information, adding a new information layer of specification requirements and time matching relationships. Finally, through the hierarchical template matching processing layer, the information is integrated into the structured report text. The construction progress description, specification compliance analysis, and suggestion content are incorporated into the template mapping. Each processing layer deepens or expands the input data. The output text not only inherits the semantic style of the input log but also realizes information refinement and normalization through the multi-layer stacked structured process, reflecting the bottom-up hierarchical logic.

[0109] In summary, the embodiment of the present invention can accurately extract key entities and progress events to form a construction feature set by obtaining the construction log text set containing detailed construction stage description units and combining natural language processing technology, breaking the limitations of traditional manual analysis and greatly improving the accuracy and efficiency of information extraction. The semantic alignment processing of the construction feature set and the engineering specification knowledge base realizes the deep association between the construction reality and the specification requirements, and can timely discover potential specification non-compliance problems. The structured integration based on the hierarchical template matching algorithm generates a structured report text that comprehensively covers construction progress, specification compliance, and suggestion content, providing a systematic, clear, and targeted decision-making basis for project management, improving the scientificity, standardization, and forward-looking of project management, and effectively promoting the efficient and compliant development of engineering construction projects.

[0110] Based on the same inventive concept, the embodiment of the present invention also provides an engineering management report generation system. Refer toFigure 2 As shown, it is a schematic structural diagram of a possible engineering management report generation system provided in an embodiment of the present invention. Figure 2 In this, the engineering management report generation system 200 includes: a processor 210 and a memory 220. Among them, the memory 220 stores a computer program executable by the processor 210. By executing the instructions stored in the memory 220, the processor 210 can execute the steps of the above-mentioned engineering management report generation method based on natural language processing.

[0111] Based on the same inventive concept, an embodiment of the present invention provides a computer-readable storage medium, which includes a computer program. When the computer program runs on the engineering management report generation system, the computer program is used to cause the engineering management report generation system to execute the steps of the above-mentioned engineering management report generation method based on natural language processing. In some possible implementation manners, various aspects of the engineering management report generation method provided by the present invention can also be implemented in the form of a program product, which includes a computer program. When the program product runs on the engineering management report generation system, the computer program is used to cause the engineering management report generation system to execute the steps in the above-mentioned engineering management report generation method based on natural language processing. For example, the engineering management report generation system can execute steps as Figure 1 shown in.

[0112] In the technical solutions involved in the above embodiments of the present invention, whether it is for calculating the comparison of multi-dimensional features or constructing composite parameters, when there are problems caused by significant differences in the number of dimensions, dimension units, and semantic meanings of different features, those skilled in the art, based on their professional knowledge and past practical experience, can fully understand that these differences need to be properly processed so that the calculation results are accurate and comparable, and to avoid situations such as logical confusion and unclear mathematical meanings.

[0113] Specifically, when faced with features having different numbers of dimensions, in order to accurately calculate the similarity, matching degree, or feature distance between different features, those skilled in the art can use various strategies, including but not limited to feature selection, feature extraction, and kernel function processing.

[0114] When dealing with the comparison of multi-dimensional features, in order to achieve comparable alignment of the feature space, those skilled in the art can adopt a variety of existing general technical means, including but not limited to standardization preprocessing, mapping transformation, and space projection.

[0115] In the construction process of composite parameters (such as loss function values), different parameter items often have different dimensions. Those skilled in the art can adopt normalization processing or an adaptive weight allocation mechanism based on distribution characteristics.

[0116] The above general technical means for solving the problems of feature matching and loss balance all belong to the common general knowledge in this field. These technical means have been fully verified and widely used in a large number of practical applications. When faced with similar dimensional difference problems, those skilled in the art can proficiently and flexibly use these methods to handle them.

[0117] In the embodiments of the present invention, the formulas and calculation processes involved, whether for multi-dimensional feature comparison or composite loss function construction, strictly follow the principle of dimension correspondence. Each variable in the formula has a clear and definite physical meaning, and its operation logic fully conforms to the basic mathematical and physical logics. The operation result is necessarily a reasonable result expected by the present invention. Those skilled in the art are capable of comprehensively using the above general technical means according to the specific data situation and business requirements to effectively solve various problems brought about by the number of dimensions, dimensional differences, etc. in the multi-dimensional feature comparison calculation and composite loss function construction in the embodiments, and ensure the accuracy, reliability, and feasibility of the technical solution of the present invention.

Claims

1. A method for generating an engineering management report based on natural language processing, characterized in that The method includes: Obtaining a collection of construction log texts generated during the engineering construction process, where the collection of construction log texts contains multiple continuously recorded construction stage description units, and each construction stage description unit includes at least one construction activity description statement and a corresponding on-site status description statement; Performing natural language processing on the collection of construction log texts, extracting key entities and progress events in each construction stage description unit, and obtaining a construction feature set containing entity identification information and event timing information; Performing semantic alignment processing on the construction feature set with a pre-set engineering specification knowledge base to generate specification association information between the construction stage description unit and the engineering specification requirements; Performing structured integration processing on the specification association information based on a hierarchical template matching algorithm to generate a structured report text that meets the requirements of project management, and the structured report text includes construction progress description, specification compliance analysis, and suggestion content.

2. The method for generating an engineering management report based on natural language processing according to claim 1, wherein The obtaining of the collection of construction log texts generated during the engineering construction process includes: Collecting original construction log data from an engineering management information system, where the original construction log data contains multiple segments of unstructured text records indexed by timestamps; Performing data cleaning processing on the original construction log data, filtering out text segments containing duplicate content, punctuation marks with format errors, and redundant content irrelevant to construction, and obtaining a candidate construction log text after preliminary cleaning; Performing segmentation processing on the candidate construction log text at a pre-set time interval, and merging text content describing the same construction stage within a continuous time range into a construction stage description unit, where the time range of the construction stage description unit is consistent with the stage division rule in the engineering construction plan; Performing statement splitting processing on each construction stage description unit, separating the construction activity description statement describing the construction activity and the on-site status description statement recording the on-site environment and equipment status, so that each construction stage description unit includes at least one construction activity description statement and one on-site status description statement; Performing text integrity verification on the split construction activity description statement and on-site status description statement, and generating a set of construction stage description units containing complete semantics as the collection of construction log texts.

3. The method for generating an engineering management report based on natural language processing according to claim 1, characterized in that, The performing of natural language processing on the collection of construction log texts, extracting key entities and progress events in each construction stage description unit, and obtaining a construction feature set containing entity identification information and event timing information includes: Performing word segmentation processing on the construction activity description statements and on-site status description statements in the collection of construction log texts to obtain a set of word segmentation results composed of basic vocabulary units; Invoking a pre-trained BERT language model to perform context semantic encoding processing on the set of word segmentation results, generating context embedding vectors for each basic vocabulary unit, and the context embedding vectors contain semantic association information of the vocabulary in the corresponding statement; Perform entity recognition processing based on the context embedding vectors to identify construction object entities, construction resource entities, and construction technology entities in the construction activity description statements, and identify environmental state entities, equipment state entities, and personnel state entities in the on-site state description statements, obtaining an entity identification set containing entity type and entity location information; Perform temporal relationship analysis on the time adverbs and sequential conjunctions in the construction stage description unit to determine the sequential execution order between construction activity description statements and the time correspondence relationship between on-site state description statements and construction activities, generating event temporal information representing the execution order of construction activities and the timing of state changes; Perform associated integration processing on the entity identification set and the event temporal information to generate a construction feature set containing entity identification information and event temporal information, where the entity identification information includes entity type, entity name, and statement location, and the event temporal information includes event execution order and state change time points.

4. The method for generating an engineering management report based on natural language processing according to claim 1, wherein Perform semantic alignment processing on the construction feature set and a pre-set engineering specification knowledge base to generate specification association information between the construction stage description unit and engineering specification requirements, including: Extract a set of specification entries related to construction object entities and construction resource entities from the engineering specification knowledge base, where the set of specification entries includes construction technology requirements, resource allocation standards, and environmental safety regulations; Perform semantic feature extraction processing on the entity identification information in the construction feature set to generate entity semantic vectors representing entity types and entity attributes; Perform semantic feature extraction processing on the specification content in the set of specification entries to generate specification semantic vectors representing specification requirements and applicable conditions; Calculate the semantic similarity value between the entity semantic vectors and the specification semantic vectors, and filter out the specification entries with semantic similarity values exceeding the preset threshold with respect to the entity identification information in the construction feature set as associated specification entries; Combined with the event temporal information in the construction feature set, analyze the applicable time range of the associated specification entries in the construction stage description unit, determine the time matching relationship between the execution order of construction activities and the state change time points and the specification requirements, generating specification association information containing associated specification entries and time matching relationships.

5. The method for generating an engineering management report based on natural language processing according to claim 1, wherein Perform structured integration processing on the specification association information based on a hierarchical template matching algorithm to generate a structured report text that meets the requirements of project management, where the structured report text includes construction progress description, specification compliance analysis, and suggestion content, including: Select a hierarchical template structure from a pre-set project management report template library, where the hierarchical template structure includes a first-level title layer, a second-level content layer, and a third-level detail layer, and the first-level title layer corresponds to construction progress description, specification compliance analysis, and suggestion content; Map the event temporal information in the construction feature set to the construction progress description template area of the first-level title layer, and organize the timeline description content of construction activities according to the event execution order and state change time points; Map the associated specification entries and time matching relationships in the said specification association information to the specification compliance analysis template area at the first-level heading layer, compare the time matching relationship between the implementation of construction activities and the specification requirements, and generate an analysis conclusion on whether the construction activities meet the specification requirements; Based on the analysis conclusion in the said specification compliance analysis template area, extract improvement measure suggestions for non-compliance situations and optimization suggestions for compliance situations from the said engineering specification knowledge base, and map them to the suggestion content template area at the first-level heading layer; Optimize the semantic description of the second-level content layer and the third-level detail layer, and adopt a transfer learning strategy to adjust the general description statements in the template into optimized description statements that conform to the semantic style of the construction log text set, so that the language expression of the structured report text is consistent with the semantic output state of the original construction log, and generate a structured report text including construction progress description, specification compliance analysis and suggestion content; The said optimizing the semantic description of the second-level content layer and the third-level detail layer, and adopting a transfer learning strategy to adjust the general description statements in the template into optimized description statements that conform to the semantic style of the construction log text set, so that the language expression of the structured report text is consistent with the semantic output state of the original construction log, includes: Extract the frequent item vocabulary set and key sentence structures from the said construction log text set, and generate a language feature template representing the semantic style of the original construction log; Select a pre-trained transfer learning model, use the said language feature template as the target style of style transfer, and use the general description statements in the said language feature template as the input text; Fine-tune and train the said transfer learning model so that the said transfer learning model learns to convert general description statements into optimized description statements that conform to the said language feature template; Input the general description statements in the second-level content layer and the third-level detail layer into the fine-tuned transfer learning model to generate optimized description statements after style transfer; Check the grammatical correctness and semantic coherence of the said optimized description statements, and correct the grammatical errors and semantic ambiguities generated during the transfer process to obtain optimized description statements that are consistent with the semantic output state of the original construction log.

6. The method for generating an engineering management report based on natural language processing according to claim 2, wherein The said data cleaning process for the said original construction log data, filtering text segments containing duplicate content, punctuation marks with format errors and redundant content irrelevant to construction, to obtain candidate construction log texts after preliminary cleaning, includes: Perform duplicate text detection on the said original construction log data, calculate the text similarity value between adjacent text segments, and determine the text segments with similarity values exceeding the preset threshold as duplicate content and retain one of them; Perform format verification on the punctuation marks in the said original construction log data, unify full-width symbols and half-width symbols into the full-width punctuation format corresponding to project management, and correct multiple consecutive punctuation marks into a single standard punctuation mark; Perform a theme relevance analysis on the text content in the original construction log data, extract the keyword set of the text content, match the keyword set with the preset construction theme keyword library, and screen out the text segments with keyword matching degree exceeding the preset threshold as the effective content related to construction; Merge the detected non-duplicate content text segments, the text segments after format correction, and the effective content text segments related to construction to generate a candidate text for the construction log after preliminary cleaning, where the candidate text for the construction log does not contain duplicate content, format error punctuation, and redundant content unrelated to construction.

7. The method for generating an engineering management report based on natural language processing according to claim 3, characterized in that, Perform entity recognition processing based on the context embedding vector, recognize the construction object entity, construction resource entity, and construction technology entity in the construction activity description statement, and recognize the environmental state entity, equipment state entity, and personnel state entity in the on-site state description statement to obtain an entity identification set including entity type and entity location information, including: Perform sequence annotation processing on the context embedding vector, and label each basic vocabulary unit with an entity type label, where the entity type label includes a construction object entity label, a construction resource entity label, a construction technology entity label, an environmental state entity label, an equipment state entity label, and a personnel state entity label; Perform entity boundary recognition processing on the labeled vocabulary units, and merge the vocabulary units continuously labeled with the same entity type label to form a complete entity name expression; Record the start position and end position of each entity name expression in the construction activity description statement or the on-site state description statement to generate a character index range representing the entity location; Perform semantic consistency verification processing on the recognized entity name expressions, and unify different expression forms of the same entity into the standard entity name in the engineering specification knowledge base; Integrate the entity type label, the standard entity name, and the character index range to generate an entity identification set including entity type and entity location information.

8. The method for generating an engineering management report based on natural language processing according to claim 4, wherein Calculate the semantic similarity value between the entity semantic vector and the specification semantic vector, and screen out the specification entries with the semantic similarity value exceeding the preset threshold in the entity identification information of the construction feature set as the associated specification entries, including: Perform dimension alignment processing on the entity semantic vector and the specification semantic vector; Calculate the cosine value of the angle between the entity semantic vector and the specification semantic vector after dimension alignment processing as the semantic similarity value; Perform normalization processing on the semantic similarity value to obtain a normalized similarity value; Set a semantic similarity threshold, and screen out the specification entries with the normalized similarity value greater than the semantic similarity threshold as candidate associated specification entries; Perform redundant screening processing on the candidate associated specification entries, remove the duplicate specification entries associated with the same entity identification information, and retain the specification entry with the highest semantic similarity value as the associated specification entry.

9. The method for generating an engineering management report based on natural language processing according to claim 1, wherein, After generating the structured report text that meets the requirements of project management, it further includes: Perform content integrity verification on the construction progress description report area in the structured report text to check whether it includes the event execution order and status change time points of all construction stage description units; Performing a logical rationality check on the specification compliance analysis report area in the structured report text to check whether the analysis conclusion is based on the associated specification entries and time matching relationship in the specification association information; Performing feasibility verification on the suggestion content report area in the structured report text to check whether the improvement measure suggestions and optimization suggestions have corresponding specification basis in the engineering specification knowledge base; Correct report areas found to have missing content, logical contradictions, or unfounded suggestions during the verification process, supplement missing progress information, adjust contradictory analysis conclusions, and replace unfounded suggestions; Generates updated structured report text that has been verified and corrected for content integrity.

10. An engineering management report generation system, characterized in that, The method comprises a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor is enabled to perform the steps of any one of the methods of claims 1 to 9.

Images