Knowledge graph construction method for intelligent construction of tunnel through drilling and blasting method

Abstract

The embodiment of the invention relates to the field of knowledge maps and data processing, and provides a knowledge map construction method for drilling and blasting method tunnel intelligent construction, and the method comprises the steps: constructing a field corpus applied to drilling and blasting method tunnel intelligent construction; performing keyword extraction combined with word vector analysis on the data in the field corpus to obtain a keyword set; performing entity extraction on the word vector analysis result to obtain an entity extraction result; performing relation extraction on entities obtained in the entity extraction result to obtain an entity relation extraction result; and performing knowledge graph construction according to the entity relationship extraction result to obtain a target knowledge graph, and performing knowledge graph construction according to the entity relationship extraction result extracted from the constructed field corpus to obtain the target knowledge graph, so that the association relationship between the entities can be displayed more intuitively, and the knowledge graph construction efficiency is improved. And the logic embedding expression effect of the data is improved.

Description

Technical Field

[0001] The present application relates to the fields of knowledge graphs and data processing technology, and specifically to a method for constructing a knowledge graph for intelligent construction of tunnels using drill-and-blast methods. Background Art

[0002] Most of the highway and railway tunnels currently under construction and planned in my country are located in the mountainous areas of western China. Tunnel engineering is gradually evolving towards longer, larger, deeper, and clustered tunnels. However, the challenges of exceptional length, high altitude, deep burial depth, and unfavorable geological conditions present significant challenges and significantly increase the complexity of construction. In response to this construction context and industry demand, unmanned construction has become a key area of ​​transformation and upgrading in the tunnel engineering sector. Through the deep integration of automation and intelligent technologies, unmanned construction not only effectively improves construction efficiency and reduces labor costs, but also plays an increasingly important role in ensuring construction safety and improving project quality, becoming a key driver of future construction.

[0003] The particularity of engineering knowledge: Engineering knowledge is multidimensional and heterogeneous, and is scattered across various sources such as specifications, academic papers, and professional books. Knowledge exists in different forms such as text, tables, pictures, and formulas, making the knowledge acquisition process extremely complex. Engineering concepts and parameters are highly specialized, forming professional barriers and requiring professional knowledge understanding. There are hierarchical and multidimensional relationships between engineering elements, which are difficult to adapt to with a single extraction model.

[0004] Limitations of existing methods: Traditional rule-based methods are highly interpretable but struggle to cope with the diversity and complexity of knowledge in the engineering field; machine learning-based methods are inadequate in handling the semantic understanding of professional terminology and cross-text knowledge associations; joint extraction methods perform poorly in the engineering field, with an accuracy rate of only approximately 20% and unstable training curves; although large language models have powerful semantic understanding capabilities, they face challenges in knowledge accuracy when applied in professional fields.

[0005] The special requirements of tunnel drilling and blasting construction scenarios: involving complex geological conditions, dynamic surrounding rock characteristics and changeable construction conditions; including multi-level knowledge dimensions such as drilling layout, drilling process, blasting process, etc.; requiring comprehensive knowledge support across specifications, standards, books and academic papers.

[0006] Under the current research status and industry needs, it is urgent to focus on existing industry specifications, technical standards, and professional documents, extract semantic information and logical associations, accurately refine and integrate professional knowledge and field experience in the field of tunnel construction, improve the informatization dimension of tunnel construction, enhance the level of integration of field data and existing information, lower the application threshold of small sample models in engineering, and at the same time form an empirical logical embedding expression. Summary of the Invention

[0007] An embodiment of the present application provides a knowledge graph construction method for intelligent construction of tunnels using the drill-and-blast method. The method can construct a knowledge graph based on the entity relationship extraction results extracted from a constructed domain corpus to obtain a target knowledge graph, which can more intuitively display the association relationship between entities and improve the logical embedding expression effect of the data.

[0008] A first aspect of an embodiment of the present application provides a method for constructing a knowledge graph for intelligent construction of a tunnel using a drill-and-blast method, the method comprising:

[0009] Construct a domain corpus for intelligent tunnel construction using drill-and-blast method;

[0010] Perform keyword extraction combined with word vector analysis on the data in the domain corpus to obtain a keyword set;

[0011] Perform entity extraction on the word vector analysis results to obtain entity extraction results;

[0012] Performing relationship extraction on the entities obtained in the entity extraction result to obtain an entity relationship extraction result;

[0013] A knowledge graph is constructed based on the entity relationship extraction results to obtain a target knowledge graph.

[0014] In one possible implementation, the keyword extraction combined with word vector analysis on the data in the domain corpus to obtain a keyword set includes:

[0015] Performing word document importance scoring calculation on the data in the domain corpus to obtain an importance score value set, wherein the obtained importance score value set includes the importance score value of each word in the domain corpus;

[0016] Performing word vector semantic analysis on the data in the domain corpus to obtain a set of semantic relevance score values, wherein the set of semantic relevance score values ​​includes a semantic relevance score value for each word in the domain corpus;

[0017] Determining a keyword score value for each word in a domain corpus according to the importance score value set and the semantic relevance score value set to obtain a keyword score value set;

[0018] The keyword set is determined according to the keyword score value set and preset keyword score values.

[0019] In one possible implementation, the entity relationship extraction result includes entity relationship triples, and constructing a knowledge graph based on the entity relationship extraction result to obtain a target knowledge graph includes:

[0020] Performing a preliminary screening on the entity relationship triples to obtain a first screening result;

[0021] Obtaining semantic matching between entities in the entity relationship triples in the first screening result to obtain a semantic matching result;

[0022] Performing a secondary screening on the preliminary screening result according to the semantic matching result and a preset semantic matching degree to obtain a second screening result;

[0023] Obtaining target knowledge contribution of the entity relationship triples in the second screening result;

[0024] Performing screening according to the target knowledge contribution and the preset knowledge contribution to obtain a third screening result;

[0025] A knowledge graph is constructed based on the third screening result to obtain the target knowledge graph.

[0026] In one possible implementation, performing entity extraction on the word vector analysis result to obtain the entity extraction result includes:

[0027] Extracting an entity extraction model, wherein the entity extraction model includes a Bert-BiLSTM-CRF model;

[0028] The Bert-BiLSTM-CRF model is used to perform entity extraction on the word vector analysis results to obtain entity extraction results.

[0029] In one possible implementation, a domain corpus for intelligent tunnel construction using drill-and-blast methods is constructed, including:

[0030] Extract relevant material data for intelligent construction of drill-and-blast tunnels;

[0031] Performing text recognition on the relevant material data to obtain a text recognition result;

[0032] Normalizing the text recognition results to obtain initial corpus data;

[0033] The initial corpus data is used to construct a domain corpus to obtain the domain corpus applied to intelligent construction of tunnels using the drill and blast method.

[0034] A second aspect of an embodiment of the present application provides a knowledge graph construction device for intelligent tunnel construction using drill-and-blast method, the device comprising:

[0035] The first construction unit is used to construct a domain corpus applied to intelligent tunnel construction using drill and blast method;

[0036] A processing unit is configured to perform keyword extraction in combination with word vector analysis on the data in the domain corpus to obtain a keyword set; perform entity extraction on the word vector analysis results to obtain an entity extraction result; and perform relationship extraction on the entities obtained in the entity extraction results to obtain an entity relationship extraction result;

[0037] The second construction unit is used to construct a knowledge graph based on the entity relationship extraction results to obtain a target knowledge graph.

[0038] In one possible implementation, in extracting keywords in combination with word vector analysis on the data in the domain corpus to obtain a keyword set, the processing unit is configured to:

[0039] Performing word document importance scoring calculation on the data in the domain corpus to obtain an importance score value set, wherein the obtained importance score value set includes the importance score value of each word in the domain corpus;

[0040] Performing word vector semantic analysis on the data in the domain corpus to obtain a set of semantic relevance score values, wherein the set of semantic relevance score values ​​includes a semantic relevance score value for each word in the domain corpus;

[0041] Determining a keyword score value for each word in a domain corpus according to the importance score value set and the semantic relevance score value set to obtain a keyword score value set;

[0042] The keyword set is determined according to the keyword score value set and preset keyword score values.

[0043] In one possible implementation, the entity relationship extraction result includes entity relationship triples, and in terms of constructing a knowledge graph based on the entity relationship extraction result to obtain a target knowledge graph, the processing unit is configured to:

[0044] Performing a preliminary screening on the entity relationship triples to obtain a first screening result;

[0045] Obtaining semantic matching between entities in the entity relationship triples in the first screening result to obtain a semantic matching result;

[0046] Performing a secondary screening on the preliminary screening result according to the semantic matching result and a preset semantic matching degree to obtain a second screening result;

[0047] Obtaining target knowledge contribution of the entity relationship triples in the second screening result;

[0048] Performing screening according to the target knowledge contribution and the preset knowledge contribution to obtain a third screening result;

[0049] A knowledge graph is constructed based on the third screening result to obtain the target knowledge graph.

[0050] In one possible implementation, in performing entity extraction on the word vector analysis result to obtain the entity extraction result, the processing unit is specifically configured to:

[0051] Extracting an entity extraction model, wherein the entity extraction model includes a Bert-BiLSTM-CRF model;

[0052] The Bert-BiLSTM-CRF model is used to perform entity extraction on the word vector analysis results to obtain entity extraction results.

[0053] In one possible implementation, the building block is used to:

[0054] Extract relevant material data for intelligent construction of drill-and-blast tunnels;

[0055] Performing text recognition on the relevant material data to obtain a text recognition result;

[0056] Normalizing the text recognition results to obtain initial corpus data;

[0057] The initial corpus data is used to construct a domain corpus to obtain the domain corpus applied to intelligent construction of tunnels using the drill and blast method.

[0058] A third aspect of an embodiment of the present application provides a terminal, comprising a processor, an input device, an output device, and a memory, wherein the processor, input device, output device, and memory are interconnected, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to call the program instructions and execute the step instructions in the first aspect of the embodiment of the present application.

[0059] The fourth aspect of the embodiments of the present application provides a computer-readable storage medium, wherein the above-mentioned computer-readable storage medium stores a computer program for electronic data exchange, wherein the above-mentioned computer program enables a computer to execute some or all of the steps described in the first aspect of the embodiments of the present application.

[0060] A fifth aspect of the embodiments of the present application provides a computer program product, wherein the computer program product includes a non-transitory computer-readable storage medium storing a computer program, wherein the computer program is operable to cause a computer to perform some or all of the steps described in the first aspect of the embodiments of the present application. The computer program product may be a software installation package.

[0061] The implementation of the embodiments of the present application has the following beneficial effects:

[0062] By constructing a domain corpus applied to intelligent construction of tunnels using the drill-and-blast method, keyword extraction combined with word vector analysis is performed on the data in the domain corpus to obtain a keyword set, entity extraction is performed on the word vector analysis results to obtain an entity extraction result, relationship extraction is performed on the entities obtained in the entity extraction result to obtain an entity relationship extraction result, and a knowledge graph is constructed based on the entity relationship extraction result to obtain a target knowledge graph. Therefore, a knowledge graph can be constructed based on the entity relationship extraction results extracted from the constructed domain corpus to obtain a target knowledge graph, which can more intuitively display the association relationship between entities and improve the logical embedding expression effect of the data. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0064] Figure 1 A flowchart of a method for constructing a knowledge graph for intelligent tunnel construction using the drill-and-blast method is provided for an embodiment of the present application.

[0065] Figure 2 A screening schematic diagram is provided for the embodiment of the present application;

[0066] Figure 3 A schematic diagram of a Bert-BiLSTM-CRF model is provided for an embodiment of the present application;

[0067] Figure 4 A schematic diagram of a relationship extraction template is provided for an embodiment of the present application;

[0068] Figure 5 A schematic diagram of a screening result is provided for the embodiment of the present application;

[0069] Figure 6 A schematic diagram of a target knowledge graph is provided for an embodiment of the present application;

[0070] Figure 7 A schematic diagram of the structure of a terminal provided in an embodiment of the present application;

[0071] Figure 8 A structural schematic diagram of a knowledge graph construction device for intelligent construction of tunnels using drill-and-blast method is provided for an embodiment of the present application. DETAILED DESCRIPTION

[0072] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0073] The terms "first," "second," and the like in the specification and claims of this application and the accompanying drawings are used to distinguish between different objects, not to describe a particular order. Furthermore, the terms "including," "having," and any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus comprising a series of steps or elements is not limited to the listed steps or elements but may optionally include steps or elements not listed, or may optionally include other steps or elements inherent to the process, method, product, or apparatus.

[0074] References to "embodiments" in this application mean that a particular feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment of the application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described in this application may be combined with other embodiments.

[0075] See also Figure 1 , Figure 1 The present invention provides a flow chart of a method for constructing a knowledge graph for intelligent tunnel construction using drill and blast method. Figure 1 As shown, the method includes:

[0076] 101. Construct a domain corpus for intelligent tunnel construction using drill and blast method.

[0077] Among them, a method for constructing a domain corpus applied to intelligent construction of tunnels using the drill and blast method includes: extracting relevant material data for intelligent construction of tunnels using the drill and blast method; performing text recognition on the relevant material data to obtain text recognition results; normalizing the text recognition results to obtain initial corpus data; and using the initial corpus data to construct a domain corpus to obtain the domain corpus applied to intelligent construction of tunnels using the drill and blast method.

[0078] Specifically, the relevant material data includes 13 relevant specifications and standards, 4 relevant books, and 363 relevant Chinese papers. The specifications cover key processes such as drilling and blasting design, drilling, and blasting. The books primarily cover intelligent construction topics, which are less covered in the specifications and standards. The papers serve as cutting-edge knowledge to supplement the knowledge graph database. Common text recognition and normalization methods can be used to perform relevant processing to obtain the initial expected database data. Finally, a common domain corpus construction method is used to construct a domain corpus for intelligent tunnel construction using drilling and blasting methods.

[0079] 102. Perform keyword extraction combined with word vector analysis on the data in the domain corpus to obtain a keyword set.

[0080] Among them, a method for extracting keywords from the data in the domain corpus in combination with word vector analysis to obtain a keyword set includes: performing word document importance score calculation on the data in the domain corpus to obtain an importance score value set, and the importance score value set obtained includes the importance score value of each word in the domain corpus; performing word vector semantic analysis on the data in the domain corpus to obtain a semantic relevance score value set, and the semantic relevance score value set includes the semantic relevance score value of each word in the domain corpus; determining the keyword score value of each word in the domain corpus based on the importance score value set and the semantic relevance score value set to obtain a keyword score value set; determining the keyword set based on the keyword score value set and the preset keyword score value.

[0081] Specifically, when calculating the importance score value set, the details are as follows:

[0082] First, we use the TF-IDF keyword extraction method to calculate the importance score value set. Specifically, we calculate the term frequency (TF) between words. Term frequency is used to measure the importance of words in a single document. The calculation formula is as follows:

[0083]

[0084] Among them, n i,j It represents the number of times word i appears in document j, and the denominator represents the sum of the number of times all words appear in document j. Calculating in this way can eliminate the impact of document length on word frequency statistics.

[0085] Calculate the inverse document frequency (IDF) of a word. The IDF is used to measure the general importance of a word in the entire document collection. The calculation formula is as follows:

[0086]

[0087] Where |D| refers to the total number of documents in the corpus, |{j:t i ∈d j}|refers to the word t i If the word is not in the corpus, the item is assigned a value of 1+|{j:t i ∈d j}|.

[0088] The TF-IDF algorithm, combining formulas (1) and (2), can be expressed as formula (3). A high word frequency within a specific file and a low file frequency of that word in the entire file collection can produce a high-weighted TF-IDF. Therefore, TF-IDF tends to filter out common words and retain important words.

[0089]

[0090] in, is the importance score of word i.

[0091] To improve the accuracy and completeness of keyword extraction, this application provides an extraction strategy that integrates statistical and semantic features. This strategy can more comprehensively identify and evaluate the importance of professional terms through a combination of multi-dimensional feature analysis.

[0092] First, a threshold-based multi-constraint screening mechanism is introduced, through the word frequency threshold (TF min ) and document frequency threshold (IDF min ) to preliminarily screen the candidate words. The candidate words must satisfy formula (4) at the same time:

[0093]

[0094] Among them, TF min The setting can filter out rare words, IDF min It is used to eliminate overly common words. The specific values ​​of the two thresholds can be adjusted according to the characteristics of the actual corpus.

[0095] Secondly, considering that statistical features may not fully reflect the semantic association between words, a semantic analysis method based on word vectors is introduced. This method uses the Skip-gram model to train word vectors, and its objective function is defined as formula (5):

[0096]

[0097] Where W represents the term, C(w) represents the context window of word w, and p(c│w) represents the conditional probability of context c appearing given word w. The semantic similarity between words is calculated by vector cosine, as shown in formula (6):

[0098]

[0099] Where, v i and v j Respectively represent the word w i and the word w j The vector representation of , ||·|| represents the Euclidean norm of the vector.

[0100] Finally, in order to achieve the effective fusion of statistical features and semantic features, a weighted hybrid ranking strategy is designed by integrating formulas (4)-(6), as shown in formula (7):

[0101] Score(w)=α·TF IDF(w) +(1―α)·Semantic(w)#(7)

[0102] In the formula, Score(w) is the keyword score value of w, α is the weight coefficient, TF IDF(w) represents the normalized TF-IDF score of word w, and Semantic(w) represents the semantic relevance score calculated based on word vectors (the score calculated by the aforementioned Skip-gram model). This fusion strategy, by comprehensively considering the statistical significance and semantic relevance of words, can more accurately assess the importance of professional terms, providing a reliable terminology foundation for subsequent knowledge graph construction.

[0103] Use TF-IDF extraction results to sort semantic relevance and filter the results manually, such as Figure 2 As shown in the figure, synonyms are extracted by taking "landslide" and "mudstone" as examples. Although the relevance of synonyms of "landslide" is lower than that of "limestone", it is difficult to replace "limestone" with other lithologies in engineering, so it is artificially determined that they are not synonyms. For "landslide", similar risk problems have little difference in the entire research topic, so some of them can be selected as synonyms.

[0104] In this example, TF-IDF and word vector analysis techniques are used to process a large sample of paper documents, and stop words are set to limit the expression of noise. In this embodiment, the large number of Chinese papers already in the corpus have significant statistical analysis value. However, some characters unrelated to the research topic are present in the typesetting of Chinese papers, such as characters and numbers generated by DOI, China Library Classification Number, article number, English that appears in the English abstract and the main body of the paper, and irrelevant words that are inconsistent with the research direction. These numbers, symbols, English, and redundant words are not part of the content filtered by TF-IDF, so stop words need to be set before statistics are performed to systematically clean up the irrelevant words and symbols that are prevalent in Chinese literature. The set stop words are shown in Table 1 below.

[0105] Table 1 Stop word settings in TF-IDF

[0106]

[0107] Keywords corresponding to keyword scoring values ​​higher than a preset keyword scoring value in the keyword scoring value set may be determined as elements in the keyword set.

[0108] 103. Perform entity extraction on the word vector analysis results to obtain entity extraction results.

[0109] Among them, the entity extraction is performed on the word vector analysis result to obtain the entity extraction result, including: extracting an entity extraction model, the entity extraction model includes a Bert-BiLSTM-CRF model; using the Bert-BiLSTM-CRF model to perform entity extraction on the word vector analysis result to obtain the entity extraction result.

[0110] Specifically, such as Figure 3 The Bert-BiLSTM-CRF model consists of a Bert encoding layer, a BiLSTM feature extraction layer, and a CRF sequence labeling layer from bottom to top. Through layer-by-layer feature transformation, it achieves accurate mapping from raw text to entity labels.

[0111] During the forward computation of the model, the input text is first converted into a character sequence and special tags [CLS] and [SEP] are added. The Bert layer encodes the input sequence through a multi-head self-attention mechanism to obtain a semantic representation that takes into account the global context, as shown in formula (17):

[0112] H=Bert(X)#(17)

[0113] The BiLSTM layer takes the output features of Bert as input and further extracts the contextual dependencies of the sequence through two LSTM networks, forward and backward.

[0114] Finally, the CRF layer performs label prediction based on the feature sequence extracted by BiLSTM. The CRF layer uses the BIO labeling scheme to label each character with the corresponding entity type label (e.g., B-explosion, I-explosion, O). By considering the transfer constraints between labels, CRF can generate a reasonable label sequence at the overall level. The loss function of the model is defined as the negative logarithm of the conditional probability, see formula (18):

[0115] Loss=―logP(Y│X)#(18)

[0116] This hierarchical model structure fully leverages the strengths of each component: the BERT layer provides rich pre-trained semantic knowledge, the BiLSTM layer enhances local context modeling, and the CRF layer ensures overall consistency in the annotation results. The organic combination of these three components forms an end-to-end entity recognition solution. In practice, the model employs a differentiated learning rate strategy: a smaller learning rate for the BERT layer to preserve pre-trained knowledge, and a larger learning rate for the BiLSTM and CRF layers to accelerate optimization convergence.

[0117] In this example, a labeled dataset is formed by manual annotation for pre-training, and the pre-trained model is used to automatically extract knowledge and realize the automatic generation of large-scale triples. When labeling, it is necessary to pre-set the labeling scope and pre-define the entity information involved. The pre-defined entity information is shown in Table 2 below.

[0118] Table 2 Predefined entities for tunnel drilling and blasting construction

[0119]

[0120]

[0121] In this example, for intelligent tunnel construction using the drill-and-blast method, the annotation scopes for face drilling and blasting design, drilling construction procedures, and blasting construction procedures were defined. Ten entity forms were identified, and detailed pre-defined definitions and annotation examples were provided for each form. Manual annotation was then performed based on these pre-defined entities. Based on the annotation results, the Bert-BiLSTM-CRF model was selected as the language pre-training model for training.

[0122] 104. Perform relationship extraction on the entities obtained in the entity extraction result to obtain an entity relationship extraction result.

[0123] This method achieves comprehensive discovery and accurate identification of relationships between entities by combining cluster analysis of unsupervised learning and the semantic understanding capabilities of large language models.

[0124] A1. Relational Reasoning Based on Hierarchical Clustering

[0125] In the field of tunnel drilling and blasting, the relationships between entities often manifest as diverse semantic association patterns. This study uses a hierarchical clustering method to efficiently mine potential relationship types between entities by analyzing their co-occurrence characteristics and semantic distribution in text corpora.

[0126] To achieve vectorized representation of the document corpus, we first construct a feature matrix for N text documents. For each of the M different words in the document collection, we create an N×M matrix where each row represents the feature vector of a document. The vector representation of document i is defined as formula (19):

[0127] d i =[x i1 ,x i2 ,…,x iM ]#(19)

[0128] Where x ij Represents the frequency weight of the jth word in document i. All document vectors are combined into a feature matrix, as shown in formula (20):

[0129] D=[d1,d2,…,d N ] T #(20)

[0130] In the calculation of entity similarity, Euclidean distance is used as the metric for the similarity between words. For two feature vectors x and y, their similarity calculation is shown in formula (21):

[0131]

[0132] Where x and y are two feature vectors, and n is the dimension of the vector. The hierarchical clustering process determines the merging strategy by calculating the similarity between clusters. For two clusters C1 and C2, their similarity is defined as formula (22):

[0133]

[0134] Where V(·) represents the variance of the cluster, and |C1| and |C2| represent the number of entities in clusters C1 and C2, respectively. This calculation method takes into account the compactness within the cluster and the separation between clusters.

[0135] By iteratively merging the most similar clusters, the hierarchical clustering algorithm gradually constructs hierarchical relationships between entities. At each level, entities within the same cluster may share similar semantic features or functional attributes. For example, entities such as "drilling," "charging," and "filling" may be clustered together due to their sequential association within the blasting process. Similarly, entities such as "overbreak" and "underbreak" may be grouped together due to their correlation in describing the blasting effects. This semantically related clustering effectively reflects the inherent connections between entities during the drilling and blasting construction process.

[0136] A2. Relation Extraction Method Based on GPT-PROMPT

[0137] To fully leverage the advantages of large language models in semantic understanding and knowledge reasoning, this study designed a relation extraction method based on the ChatGPT model. This method achieves cross-text entity relationship topology through a "role construction" approach.

[0138] In the process of relation extraction, a standardized PROMPT template is first constructed. Figure 4 As shown in the figure, the model is guided by open-ended instructions such as "Given the text content, infer and list the relationships between entities." This prompts the model to perform deep semantic analysis of the input text. The model output is in the form of "entity + relationship type + entity" triples, which facilitates subsequent normalization processing.

[0139] PROMPT's design includes three key elements: role definition, skill description, and constraints. This model is positioned as an expert in knowledge graph construction, with the specialized ability to extract key entities and relationships from various sentences.

[0140] In the skill description section, the model is explicitly required to extract triples based on a predefined context. By setting specific "Triple Extraction" task requirements, the model is guided to automatically filter out irrelevant numerical and symbolic information and focus on discovering semantic relationships. This structured output format not only facilitates subsequent data processing but also improves the accuracy and consistency of relation extraction. This GPT-PROMPT-based relation extraction method has the following advantages:

[0141] a) Enhance the model’s understanding of professional fields through role positioning;

[0142] b) Standardize the expression of relationships through structured output formats;

[0143] c) Use constraints to ensure the professionalism and standardization of extraction results;

[0144] d) Support batch processing of large-scale texts to improve extraction efficiency.

[0145] A3. Relationship extraction results

[0146] In this example, several entities were extracted using the entity extraction techniques described in the previous examples. Entities extracted using the TF-IDF method were analyzed for keyword similarity, and entities extracted using the Bert-BiLSTM-CRF algorithm were classified and clustered using a predefined method. First, relationships were predefined (as shown in Table 3 below). Based on this, expert evaluation was used to infer relationships between entities in each category. These relationships were then annotated and expressed using manual chimera, forming triples for intelligent tunnel drilling and blasting construction.

[0147] Table 3 Predefined table of intelligent construction relations for drill-and-blast tunnels

[0148]

[0149] Relational reasoning based on entity clustering and similarity metrics can connect existing entities, but it lacks the ability to accurately model implicit semantic relationships and cross-text associations. To extract relational information within long texts and achieve cross-text relational links, this embodiment addresses this deficiency by employing a GPT-PROMT model. The GPT-PROMT model uses a "role creation" approach to drive a large language model, constructing a vertical domain knowledge extraction model. This model significantly enhances cross-context semantic extraction and logical associations, and can be used to address missing associations across text corpora. Because the research focus is on the drill-and-blast construction process, to maintain the rigor of the knowledge graph, it is necessary to manually filter out triples that are inconsistent with the research perspective. Specifically, the model is first fed with the long text to be analyzed, and an open-ended instruction, such as "Given the text content, infer and list the relationships between entities," is used to perform preliminary relationship identification. The model output is in the form of "entity + relationship type + entity" triples, facilitating subsequent normalization. To ensure the reliability of the extraction results, this method adopts a human-assisted review mechanism, where experts can evaluate and screen the relationships output by the model, such as Figure 5 Some triplets focus on tunnel support and have little relevance to drilling and blasting construction, so they are artificially screened out.

[0150] 105. Construct a knowledge graph based on the entity relationship extraction results to obtain a target knowledge graph.

[0151] In one possible implementation, the entity relationship extraction result includes entity relationship triples, and constructing a knowledge graph based on the entity relationship extraction result to obtain a target knowledge graph includes:

[0152] B1. Preliminarily screening the entity relationship triples to obtain a first screening result;

[0153] B2. Obtaining semantic matching between entities in the entity relationship triples in the first screening result to obtain a semantic matching result;

[0154] B3, performing a secondary screening on the preliminary screening result according to the semantic matching result and a preset semantic matching degree to obtain a second screening result;

[0155] B4. Obtain target knowledge contribution of the entity relationship triples in the second screening result;

[0156] B5. Perform screening based on the target knowledge contribution and the preset knowledge contribution to obtain a third screening result;

[0157] B6. Construct a knowledge graph based on the third screening result to obtain the target knowledge graph.

[0158] Among them, during the preliminary screening, entity relationship triplets that meet the following conditions are retained: triplets that must contain at least one core entity category (design parameters, drilling parameters, blasting parameters, excavation methods) are retained first; triplets that only contain auxiliary entity categories (equipment, construction activities) and are directly related to the drilling and blasting method are retained second priority; triplets whose end entities do not belong to the predefined categories in Table 2 are automatically eliminated.

[0159] After the initial screening, the entity-relationship triples in the first screening results can be verified for compliance, and non-compliant entity-relationship triples can be screened out. Specifically, the relationship types in the triples can be checked to see if they conform to the predefined relationships in Table 3. For non-predefined relationship types generated by GPT, semantic mapping is used to convert them to the closest predefined relationship class. Relationship types that cannot be reasonably mapped are marked as "pending relationship" and submitted to expert review. The semantic matching between the entities and relationships in the entity-relationship triples that meet compliance is then extracted. The matching meets the following rules: Strong match: The relationship type highly matches the semantic characteristics of the entities at both ends, such as (drill hole spacing, impact relationship, blasting effect); Weak match: The relationship type and entity semantics have a reasonable but non-optimal connection, requiring further verification; False match: The relationship type clearly does not match the entity semantics, such as (drilling rig, causal relationship, tunnel length).

[0160] The preset semantic matching degree is determined by empirical values ​​or historical data. When screening, entity relationship triples with semantic matching results higher than the preset matching degree are retained.

[0161] The knowledge contribution of entity-relationship triples to intelligent construction using drilling and blasting can also be extracted, as follows:

[0162] Knowledge novelty assessment:

[0163] Knowledge that is repeated or highly similar to existing triples is eliminated;

[0164] Merge triplets with different expressions but the same semantics and retain the more precise version;

[0165] For contradictory triples, retain the version with higher source reliability (norms > professional books > academic papers > model reasoning).

[0166] Construction practicality assessment:

[0167] Priority will be given to retaining knowledge that has direct guiding significance for intelligent drilling and blasting construction;

[0168] Knowledge with early warning value for safety risk prevention and control will be retained with lower priority;

[0169] Triplet groups that are purely theoretical knowledge and difficult to convert into construction guidance are eliminated.

[0170] The knowledge contribution is calculated by using the knowledge novelty evaluation and practicality evaluation standards to obtain the target knowledge contribution. Specifically, the novelty evaluation and practicality evaluation standards can be quantified by rule scoring, and can be processed by a general quantification method. When performing the screening process, the entity relationship triples with a degree lower than the preset knowledge contribution can be eliminated to obtain the third screening result. Finally, the general knowledge graph wrapping method is used to construct the knowledge graph to obtain the target knowledge graph. A schematic diagram of a target knowledge graph is shown below. Figure 6 shown.

[0171] For the same example as above, please refer to Figure 7 , Figure 7 A schematic diagram of the structure of a terminal provided in an embodiment of the present application is shown in FIG. Figure 7 As shown, the system comprises a processor, an input device, an output device and a memory, which are connected to each other, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to call the program instructions, and the program includes instructions for executing the following steps;

[0172] Construct a domain corpus for intelligent tunnel construction using drill-and-blast method;

[0173] Perform keyword extraction combined with word vector analysis on the data in the domain corpus to obtain a keyword set;

[0174] Perform entity extraction on the word vector analysis results to obtain entity extraction results;

[0175] Performing relationship extraction on the entities obtained in the entity extraction result to obtain an entity relationship extraction result;

[0176] A knowledge graph is constructed based on the entity relationship extraction results to obtain a target knowledge graph.

[0177] In this example, by obtaining the user's product demand information, one or more functional modules that match the product demand information are determined in the functional module group, and the target business product is generated based on the one or more functional modules. In this way, the target business product can be automatically generated through the product demand information, thereby improving the efficiency of determining the target business product.

[0178] The above mainly introduces the scheme of the embodiment of the present application from the perspective of the execution process on the method side. It is understandable that, in order to implement the above functions, the terminal includes hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should easily realize that, in combination with the units and algorithm steps of the various examples described in the embodiments provided herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in a hardware or computer software driven hardware manner depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0179] The embodiment of the present application can divide the terminal into functional units according to the above method example. For example, each functional unit can be divided according to each function, or two or more functions can be integrated into one processing unit. The above integrated unit can be implemented in the form of hardware or in the form of software functional units. It should be noted that the division of units in the embodiment of the present application is schematic and is only a logical functional division. In actual implementation, there may be other division methods.

[0180] In line with the above, please see Figure 8 , Figure 8 The present invention provides a schematic diagram of a knowledge graph for intelligent tunnel construction using drill and blast method. Figure 8 As shown, the device includes:

[0181] The first construction unit 301 is used to construct a domain corpus applied to intelligent tunnel construction using drill and blast method;

[0182] Processing unit 302 is configured to perform keyword extraction combined with word vector analysis on the data in the domain corpus to obtain a keyword set; perform entity extraction on the word vector analysis results to obtain an entity extraction result; and perform relationship extraction on the entities obtained in the entity extraction results to obtain an entity relationship extraction result.

[0183] The second construction unit 303 is used to construct a knowledge graph based on the entity relationship extraction results to obtain a target knowledge graph.

[0184] In one possible implementation, in extracting keywords in combination with word vector analysis on the data in the domain corpus to obtain a keyword set, the processing unit is configured to:

[0185] Calculate the word document importance score for the data in the domain corpus to obtain the importance score value of each word in the domain corpus and obtain a set of importance score values;

[0186] Perform word vector semantic analysis on the data in the domain corpus to obtain a semantic relevance score value for each word in the domain corpus, thereby obtaining a set of semantic relevance score values;

[0187] Determining a keyword score value for each word in a domain corpus according to the importance score value set and the semantic relevance score value set to obtain a keyword score value set;

[0188] The keyword set is determined according to the keyword score value set and preset keyword score values.

[0189] In one possible implementation, the entity relationship extraction result includes entity relationship triples, and in terms of constructing a knowledge graph based on the entity relationship extraction result to obtain a target knowledge graph, the processing unit is configured to:

[0190] Performing a preliminary screening on the entity relationship triples to obtain a first screening result;

[0191] Obtaining semantic matching between entities in the entity relationship triples in the first screening result to obtain a semantic matching result;

[0192] Performing a secondary screening on the preliminary screening result according to the semantic matching result and a preset semantic matching degree to obtain a second screening result;

[0193] Obtaining target knowledge contribution of the entity relationship triples in the second screening result;

[0194] Performing screening according to the target knowledge contribution and the preset knowledge contribution to obtain a third screening result;

[0195] A knowledge graph is constructed based on the third screening result to obtain the target knowledge graph.

[0196] In one possible implementation, in performing entity extraction on the word vector analysis result to obtain the entity extraction result, the processing unit is specifically configured to:

[0197] Extracting an entity extraction model, wherein the entity extraction model includes a Bert-BiLSTM-CRF model;

[0198] The Bert-BiLSTM-CRF model is used to perform entity extraction on the word vector analysis results to obtain entity extraction results.

[0199] In one possible implementation, the building block is used to:

[0200] Extract relevant material data for intelligent construction of drill-and-blast tunnels;

[0201] Performing text recognition on the relevant material data to obtain a text recognition result;

[0202] Normalizing the text recognition results to obtain initial corpus data;

[0203] The initial corpus data is used to construct a domain corpus to obtain the domain corpus applied to intelligent construction of tunnels using the drill and blast method.

[0204] An embodiment of the present application also provides a computer storage medium, wherein the computer storage medium stores a computer program for electronic data exchange, and the computer program enables a computer to execute part or all of the steps of any one of the knowledge graph construction methods for intelligent construction of tunnels using the drill and blast method as described in the above method embodiments.

[0205] An embodiment of the present application also provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program enables a computer to execute part or all of the steps of any one of the knowledge graph construction methods for intelligent construction of tunnels using the drill-and-blast method as described in the above method embodiments.

[0206] It should be noted that for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that this application is not limited by the order of the actions described, because according to this application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by this application.

[0207] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0208] In the several embodiments provided in this application, it should be understood that the disclosed devices can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection of devices or units can be electrical or other forms.

[0209] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0210] In addition, the functional units in the various embodiments of the application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software program modules.

[0211] If the integrated unit is implemented in the form of a software program module and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a memory, including a number of instructions for enabling a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned memory includes: various media that can store program codes, such as a U disk, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk.

[0212] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable memory, which may include: a flash drive, a read-only memory, a random access memory, a magnetic disk or an optical disk, etc.

[0213] The above is a detailed introduction to the embodiments of the present application. Specific examples are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core idea of ​​the present application. At the same time, for those skilled in the art, according to the idea of ​​the present application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A knowledge graph construction method for intelligent tunnel construction using drill and blast method, characterized in that: The method comprises: Construct a domain corpus for intelligent tunnel construction using drill-and-blast method; Perform keyword extraction combined with word vector analysis on the data in the domain corpus to obtain a keyword set; Perform entity extraction on the word vector analysis results to obtain entity extraction results; Performing relationship extraction on the entities obtained in the entity extraction result to obtain an entity relationship extraction result; A knowledge graph is constructed based on the entity relationship extraction results to obtain a target knowledge graph.

2. The knowledge graph construction method for intelligent construction of drill-and-blast tunnels according to claim 1 is characterized in that: The keyword extraction combined with word vector analysis on the data in the domain corpus to obtain a keyword set includes: Performing word document importance scoring calculation on the data in the domain corpus to obtain an importance score value set, wherein the obtained importance score value set includes the importance score value of each word in the domain corpus; Performing word vector semantic analysis on the data in the domain corpus to obtain a set of semantic relevance score values, wherein the set of semantic relevance score values ​​includes a semantic relevance score value for each word in the domain corpus; Determining a keyword score value for each word in a domain corpus according to the importance score value set and the semantic relevance score value set to obtain a keyword score value set; The keyword set is determined according to the keyword score value set and preset keyword score values.

3. The knowledge graph construction method for intelligent construction of drill-and-blast tunnels according to claim 2 is characterized in that: The entity relationship extraction result includes entity relationship triples, and the knowledge graph is constructed based on the entity relationship extraction result to obtain a target knowledge graph, including: Performing a preliminary screening on the entity relationship triples to obtain a first screening result; Obtaining semantic matching between entities in the entity relationship triples in the first screening result to obtain a semantic matching result; Performing a secondary screening on the preliminary screening result according to the semantic matching result and a preset semantic matching degree to obtain a second screening result; Obtaining target knowledge contribution of the entity relationship triples in the second screening result; Performing screening according to the target knowledge contribution and the preset knowledge contribution to obtain a third screening result; A knowledge graph is constructed based on the third screening result to obtain the target knowledge graph.

4. The knowledge graph construction method for intelligent tunnel construction using the drill and blast method according to claim 3 is characterized in that: The entity extraction is performed on the word vector analysis result to obtain the entity extraction result, including: Extracting an entity extraction model, wherein the entity extraction model includes a Bert-BiLSTM-CRF model; The Bert-BiLSTM-CRF model is used to perform entity extraction on the word vector analysis results to obtain entity extraction results.

5. The knowledge graph construction method for intelligent construction of drill-and-blast tunnels according to claim 3 or 4, characterized in that: Construct a domain corpus for intelligent tunnel construction using drill-and-blast methods, including: Extract relevant material data for intelligent construction of drill-and-blast tunnels; Performing text recognition on the relevant material data to obtain a text recognition result; Normalizing the text recognition results to obtain initial corpus data; The initial corpus data is used to construct a domain corpus to obtain the domain corpus applied to intelligent construction of tunnels using the drill and blast method.

6. A knowledge graph construction device for intelligent tunnel construction using drill and blast method, characterized in that: The device comprises: Construction unit, used to build a domain corpus applied to intelligent construction of tunnels using drill and blast method; A processing unit is configured to perform keyword extraction in combination with word vector analysis on the data in the domain corpus to obtain a keyword set; perform entity extraction on the word vector analysis results to obtain an entity extraction result; and perform relationship extraction on the entities obtained in the entity extraction results to obtain an entity relationship extraction result; A construction unit is used to construct a knowledge graph based on the entity relationship extraction results to obtain a target knowledge graph.

7. The knowledge graph construction device for intelligent tunnel construction using drill and blast method according to claim 6, characterized in that: In the aspect of extracting keywords in combination with word vector analysis on the data in the domain corpus to obtain a keyword set, the processing unit is used to: Performing word document importance scoring calculation on the data in the domain corpus to obtain an importance score value set, wherein the obtained importance score value set includes the importance score value of each word in the domain corpus; Performing word vector semantic analysis on the data in the domain corpus to obtain a set of semantic relevance score values, wherein the set of semantic relevance score values ​​includes a semantic relevance score value for each word in the domain corpus; Determining a keyword score value for each word in a domain corpus according to the importance score value set and the semantic relevance score value set to obtain a keyword score value set; The keyword set is determined according to the keyword score value set and preset keyword score values.

8. The knowledge graph construction device for intelligent tunnel construction using drill and blast method according to claim 7, characterized in that: In the aspect that the entity relationship extraction result includes entity relationship triples and the knowledge graph is constructed according to the entity relationship extraction result to obtain a target knowledge graph, the processing unit is configured to: Performing a preliminary screening on the entity relationship triples to obtain a first screening result; Obtaining semantic matching between entities in the entity relationship triples in the first screening result to obtain a semantic matching result; Performing a secondary screening on the preliminary screening result according to the semantic matching result and a preset semantic matching degree to obtain a second screening result; Obtaining target knowledge contribution of the entity relationship triples in the second screening result; Performing screening according to the target knowledge contribution and the preset knowledge contribution to obtain a third screening result; A knowledge graph is constructed based on the third screening result to obtain the target knowledge graph.

9. A terminal, characterized in that: The method comprises a processor, an input device, an output device and a memory, wherein the processor, the input device, the output device and the memory are interconnected, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to call the program instructions to execute the knowledge graph construction method for intelligent construction of drill and blast tunnels according to any one of claims 1 to 5.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which includes program instructions. When the program instructions are executed by a processor, the processor executes the knowledge graph construction method for intelligent construction of drill-and-blast tunnels as described in any one of claims 1 to 5.

Images