Underground cavern support design knowledge element extraction method and device

Abstract

The invention relates to the technical field of underground engineering, and discloses an underground cavern support design knowledge element extraction method and device. The method comprises the following steps: acquiring underground cavern support design data; extracting a plurality of entities in the underground cavern support design data; the entities comprise one or more of a first entity, a second entity and a third entity; the first entity meets a preset form rule in the underground cavern support design data; statistical data of the second entity in the underground cavern support design data meet a preset condition; the third entity is obtained based on a pre-constructed entity extraction model; and determining knowledge elements of each entity in the underground cavern support design data. According to the method and the device, the knowledge elements are extracted from the multi-source support design data by utilizing various modes such as preset rules, statistics and models.

Description

Technical Field

[0001] The present invention relates to the field of underground engineering technology, and in particular to a method and device for extracting knowledge elements from underground cavern support design. Background Art

[0002] Underground engineering support design is a critical step in ensuring project stability and safety. During the current support design process, designers need to refer to numerous specifications, standards, and literature. Manual analysis and extraction of design knowledge elements is not only inefficient but also difficult to ensure accuracy. Furthermore, support design knowledge is complex, involving multi-dimensional data and relational expressions. Traditional design methods are unable to effectively address the need for automatic extraction and updating of design knowledge elements, resulting in inefficient knowledge utilization and restricting the intelligent generation of design solutions. Therefore, how to implement knowledge elements in underground engineering support design data is a current focus. Summary of the Invention

[0003] In view of this, the present invention provides a method and device for extracting knowledge elements of underground cavern support design to solve the problem of extracting knowledge elements of underground cavern support design.

[0004] In a first aspect, the present invention provides a method for extracting knowledge elements from underground cavern support design, the method comprising:

[0005] Obtain underground cavern support design data;

[0006] Extracting multiple entities from underground cavern support design data; the entity includes one or more of a first entity, a second entity, and a third entity; the first entity satisfies a preset form rule in the underground cavern support design data; statistical data of the second entity in the underground cavern support design data satisfies a preset condition; and the third entity is obtained based on a pre-built entity extraction model;

[0007] Determine the knowledge elements of each entity in the underground cavern support design data.

[0008] Through the method provided in this embodiment, preset formal rules, statistical data and entity extraction models are combined to ensure that entities from different sources can be effectively obtained, avoiding information omissions caused by a single method. Specifically, the preset formal rules can ensure the standardized extraction of structured underground cavern support design data (such as underground cavern support design specifications), statistical data can extract key entities in underground cavern support design data, and the entity extraction model can deeply mine semantic entities, especially unstructured underground cavern support design data. The embodiment of this application complements and improves the accuracy of entity extraction through three ways.

[0009] In an optional embodiment, the underground cavern support design data includes a plurality of design documents; the statistical data of the second entity characterizing the second entity can be used to distinguish between the design documents.

[0010] Through the above implementation, the second entity (such as high-frequency process terms) screened out through statistical data can be used to distinguish multiple documents and accurately reflect the differences between different documents, which is of great significance for achieving rapid classification and accurate retrieval of design documents.

[0011] In an optional embodiment, the statistical data is determined based on the number of times the entity appears in each design document and the number of design documents containing the entity.

[0012] Through the above-mentioned implementation mode, by counting the number of occurrences of an entity in each design document, the importance or frequency of use of the entity in the support design can be quantified. For example, the more times an entity appears, the more critical or commonly used the entity is in the design. The number of design documents containing the entity also reflects the popularity or scope of application of the entity. For example, the more documents containing the entity, the more applications the entity has in multiple design scenarios or projects. In an embodiment of the present application, based on statistical data, the key entity in the underground cavern support design data, i.e., the second entity, can be extracted.

[0013] In an optional embodiment, the entities in the underground cavern support design data correspond to at least one category; the entity extraction model includes a bidirectional long short-term memory network and a conditional random field model, and the method further includes:

[0014] Get a training dataset; the training dataset contains multiple entities and the category corresponding to each entity;

[0015] Based on the training data set, the initial entity extraction model is trained to obtain the entity extraction model.

[0016] Through the above implementation, the bidirectional long short-term memory network can simultaneously capture both forward and backward information of the input sequence, thereby better understanding the context of entities in the text. This helps accurately identify entities in underground cavern support design data and reduce false positives and missed positives. In sequence labeling tasks, the conditional random field model can consider the dependencies between labels and more accurately delineate entity boundaries. Combined with the output of the bidirectional long short-term memory network, the conditional random field model can further improve the accuracy of entity extraction.

[0017] In an optional embodiment, the underground cavern support design data includes underground cavern support design specifications; the preset form rules include entity number, entity name, and entity definition; the entity number is used to indicate the position of the entity in the underground cavern support design specifications.

[0018] Through the above implementation method, specific words, terms and phrases in underground cavern support design data are extracted using predefined formal rules, which is suitable for entity extraction of underground cavern support design data with relatively clear structured information.

[0019] In an optional embodiment, the knowledge element includes the relationship between entities; determining the knowledge element of each entity in the underground cavern support design data includes:

[0020] Each preset word in the underground cavern support design data is regarded as an independent cluster, and the inter-cluster similarity between independent clusters is determined;

[0021] Select the two clusters with the smallest inter-cluster similarity to merge;

[0022] Update the inter-cluster similarity between each cluster, return to the step of selecting the two clusters with the smallest inter-cluster similarity to merge, until the preset iteration condition is met and at least one final cluster is obtained;

[0023] Based on entities being in the same final cluster, the relationship between entities is determined.

[0024] Through the above implementation, inter-cluster similarity is used to characterize the semantic or contextual similarity between words. By iteratively merging clusters with the smallest similarity and updating the inter-cluster similarity, the recognition results of entity relationships are continuously optimized. This not only identifies obvious entity relationships in underground cavern support design data, but also discovers complex relationships hidden in the data by gradually merging clusters, thereby improving the accuracy of extracting relationships between entities.

[0025] In an optional embodiment, the underground cavern support design data includes a plurality of design documents; each preset word in the underground cavern support design data is treated as an independent cluster, and the inter-cluster similarity between the independent clusters is determined, including:

[0026] Determining a vector representation of each preset word based on the number of times each preset word appears in each design document;

[0027] Each preset word is regarded as an independent cluster, and the inter-cluster similarity between the independent clusters is determined based on the vector representation of each preset word.

[0028] Through the above implementation, the importance or significance of each word can be quantified by counting the number of times a preset word appears in the design document. The vector representation not only reflects the individual characteristics of the word, but also implies the correlation between words. By calculating the similarity between vectors, it is possible to reveal which words often appear together in the support design data, thereby inferring the possible semantic or functional connections between them, and thus mining the hidden knowledge in the support design data, such as the type of support result, material combination method, etc., which is of great significance for optimizing the support design scheme.

[0029] In a second aspect, the present invention provides a knowledge element extraction device for underground cavern support design, the device comprising:

[0030] The first acquisition module is used to obtain underground cavern support design data;

[0031] An extraction module is used to extract multiple entities from the underground cavern support design data; the entity includes one or more of a first entity, a second entity, and a third entity; the first entity satisfies a preset form rule in the underground cavern support design data; the statistical data of the second entity in the underground cavern support design data satisfies a preset condition; and the third entity is obtained based on a pre-built entity extraction model;

[0032] The determination module is used to determine the knowledge elements of each entity in the underground cavern support design data.

[0033] Through the device provided by this embodiment, preset formal rules, statistical data and entity extraction models are combined to ensure that entities from different sources can be effectively obtained, avoiding information omissions caused by a single method. Specifically, the preset formal rules can ensure the standardized extraction of structured underground cavern support design data (such as underground cavern support design specifications), statistical data can extract key entities in underground cavern support design data, and the entity extraction model can deeply mine semantic entities, especially unstructured underground cavern support design data. The embodiment of this application complements and improves the accuracy of entity extraction through three ways.

[0034] In a third aspect, the present invention provides a computer device comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to execute the underground cavern support design knowledge element extraction method of the above-mentioned first aspect or any corresponding embodiment thereof.

[0035] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the underground cavern support design knowledge element extraction method of the above-mentioned first aspect or any corresponding embodiment thereof.

[0036] In a fifth aspect, the present invention provides a computer program product comprising computer instructions for enabling a computer to execute the method for extracting knowledge elements from underground cavern support design according to the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0038] Figure 1 is a flow chart of a method for extracting knowledge elements from underground cavern support design according to an embodiment of the present invention;

[0039] Figure 2 This is a diagram of an application scenario using a BiLSTM model and a CRF model according to an embodiment of the present invention;

[0040] Figure 3 This is a structural block diagram of a knowledge element extraction device for underground cavern support design according to an embodiment of the present invention;

[0041] Figure 4 The figure is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0042] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative efforts shall fall within the scope of protection of the present invention.

[0043] First, the application scenarios of the embodiments of the present application are exemplarily introduced.

[0044] Underground engineering support design, a core component of ensuring project stability and safety, requires designers to thoroughly study extensive specifications, standards, and literature. Currently, manual analysis and extraction of design knowledge elements is inefficient and difficult to ensure accuracy. Given the complexity of support design knowledge points and the multidimensional nature of data and relationship expression, traditional entity extraction methods are unable to meet the urgent need for automatic extraction and dynamic updating of design knowledge elements. This leads to inefficient knowledge utilization and severely restricts the advancement of the intelligent generation process for design solutions. Therefore, how to efficiently and automatically extract and effectively manage knowledge elements from underground engineering support design data has become a focus of current research.

[0045] In view of this, an embodiment of the present application provides a method for extracting knowledge elements from underground cavern support design, which uses preset rules, statistics, models and other methods to extract knowledge elements from multi-source underground cavern support design data. While improving the efficiency of support design knowledge extraction, it also provides a basis for knowledge-based design solutions. It should be noted that the execution subject of the method for extracting knowledge elements from underground cavern support design provided by the embodiment of the present invention can be a device for extracting knowledge elements from underground cavern support design. The device for extracting knowledge elements from underground cavern support design can be implemented as part or all of an electronic device through software, hardware or a combination of software and hardware, wherein the electronic device can be a server or a terminal, wherein the server in the embodiment of the present application can be a single server or a server cluster composed of multiple servers, and the terminal in the embodiment of the present application can be a smart phone, personal computer, tablet computer, wearable device, smart robot and other smart hardware devices. In the following method embodiments, the execution subject is an electronic device as an example for explanation.

[0046] According to an embodiment of the present invention, an embodiment of a method for extracting knowledge elements from underground cavern support design is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0047] In this embodiment, a method for extracting knowledge elements from underground cavern support design is provided, which can be used in the above-mentioned electronic devices, such as servers, etc. Figure 1 FIG. 1 is a flow chart of a method for extracting knowledge elements from underground cavern support design according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:

[0048] S101: Obtain underground cavern support design data.

[0049] In one possible implementation, underground cavern support design data refers to all information carriers related to the support design of underground caverns (such as tunnels, chambers, etc.), covering multi-dimensional content such as specifications, parameters, processes, cases, etc. involved in the design process.

[0050] For example, underground cavern support design data includes but is not limited to geological conditions (such as surrounding rock grade, groundwater distribution, ground stress state, etc.), support structures (such as types and parameters of anchor rods, anchor cables, shotcrete, steel arch frames, etc.), construction technology (such as excavation methods (such as full-section method, micro-step method), support timing, etc.), and specifications and standards (such as industry specifications, design verification formulas, etc.).

[0051] In one possible implementation, the underground cavern support design data may be derived from standards, scientific literature, and support design plans, etc., covering the technical requirements and quality standards of the support design. This application does not specifically limit the source of the underground cavern support design data.

[0052] In a possible implementation, the underground cavern support design data is pre-processed data, which includes but is not limited to deduplication, error correction, missing value filling, data conversion, etc.

[0053] S102: Extract multiple entities from underground cavern support design data.

[0054] Among them, the entity includes one or more of the first entity, the second entity, and the third entity; the first entity meets the preset formal rules in the underground cavern support design data; the statistical data of the second entity in the underground cavern support design data meets the preset conditions; the third entity is obtained based on a pre-built entity extraction model.

[0055] In one possible implementation, underground cavern support design data is used to describe various types of information of the entire process of underground cavern support design, including geological survey data, support structure parameters, calculation books, drawings, monitoring data, etc.

[0056] In one possible implementation, the entities in the underground cavern support design data can be entities related to the geological environment, such as surrounding rock grade, groundwater status, etc., and can also be entities related to the support structure, such as anchor rods, shotcrete, steel frame support, etc., and can also be entities related to the construction process, such as advanced support, excavation method, etc., and can also be entities related to design parameters, such as support thickness, anchor rod spacing, etc.

[0057] In a possible implementation, the preset formal rule can be understood as a structured rule, a regular expression, etc., which usually has a clear format or a fixed expression. The first entity can be extracted from underground cavern support design data with relatively clear structured information.

[0058] It should be noted that the specific implementation method of extracting the first entity based on the preset formal rule will be described in subsequent embodiments and will not be repeated here.

[0059] In one possible implementation, the statistical data of an entity may be term frequency, document coverage, term frequency-inverse document frequency (TF-IDF) value, etc., to reflect the importance and discrimination of the entity in the underground cavern support design data. Accordingly, the preset conditions may be that the term frequency of the entity is greater than a preset term frequency threshold, the document coverage of the entity is greater than a preset document coverage, the TF-IDF value of the entity is greater than a preset TF-IDF value, etc.

[0060] It should be noted that the specific implementation method of extracting the second reality based on entity statistical data will be described in subsequent embodiments and will not be repeated here.

[0061] In one possible implementation, the entity extraction model can be a machine learning model, a deep learning model, or the like. The entities identified by the pre-built entity extraction model can capture the implicit semantics or complex relationships of the underground cavern support design. Therefore, the third entity can be extracted from unstructured underground cavern support design data.

[0062] It should be noted that the specific implementation method of extracting the third entity from the underground cavern support design data based on the pre-built entity extraction model will be described in subsequent embodiments and will not be repeated here.

[0063] S103: Determine the knowledge element of each entity in the underground cavern support design data.

[0064] In one possible implementation, a knowledge element refers to the smallest indivisible unit in a knowledge system, which is used to describe the attributes, relationships, etc. of an entity.

[0065] Entity attributes refer to the characteristics, properties, or characteristics of an entity. These can include the characteristics, performance, and parameters of entities such as the underground cavern itself, surrounding rock, and support structures. For example, in the case of support structures such as anchors and steel arches, entity attributes can include material properties, structural layout properties, and mechanical performance properties.

[0066] The relationship between entities refers to the relationship between multiple entities. The relationship between entities includes classification relationships and non-classification relationships. Classification relationships refer to hierarchical relationships between entities, such as "belong to", "similar to", "part of", etc. For example, the classification relationships between entities can be extracted through methods based on preset rules and machine learning methods. Non-classification relationships are not limited to predefined relationship categories and are intended to discover potential relationships between entities in underground cavern support design data, such as influence relationships, determination relationships, causal relationships, spatial relationships, semantic relationships, etc.

[0067] For example, a knowledge element can be represented in the form of a triple, such as (entity, predicate, entity / value). For example, the entity is "anchor," the predicate is "length is," and the value is "attribute value, such as 3.5 meters." Another example is the entity is "surrounding rock," the predicate is "belongs to," and the entity is "shotcrete."

[0068] In one possible implementation, after determining the entities in the underground cavern support design data, the relationships between the entities can be further extracted through manual labeling, supervised learning, unsupervised learning, etc.

[0069] Optionally, the classification relationship between entities is obtained by extracting preset keywords. Specifically, by establishing a keyword triggering mechanism, rules for identifying each classification relationship are defined. For example, if an entity is followed by words such as "type", "category" or "belongs to", it may indicate a subordinate relationship between a superordinate and a subordinate, for example, sprayed anchor support is a type of support structure; if an entity is followed by words such as "part", "composed of", "includes", etc., it may indicate a relationship between a whole and a part, for example, a steel mesh is part of a lining structure; if there are expressions such as "same as", "equivalent to", "abbreviated as" or "same as..." between two entities, it may indicate a relationship between different expressions of an entity, for example, sprayed concrete and shotcrete are different names for the same thing.

[0070] Optionally, non-classification relationships between entities are obtained through machine learning. It should be noted that the use of machine learning to extract non-classification relationships between entities will be described in subsequent embodiments and will not be repeated here.

[0071] Through the method provided in this embodiment, preset formal rules, statistical data and entity extraction models are combined to ensure that entities from different sources can be effectively obtained, avoiding information omissions caused by a single method. Specifically, the preset formal rules can ensure the standardized extraction of structured underground cavern support design data (such as underground cavern support design specifications), statistical data can extract key entities in underground cavern support design data, and the entity extraction model can deeply mine semantic entities, especially unstructured underground cavern support design data. The embodiment of this application complements and improves the accuracy of entity extraction through three ways.

[0072] In some embodiments, the underground cavern support design data includes a plurality of design documents; the statistical data representing the second entity can be used to distinguish between the design documents.

[0073] In a possible implementation, the design documents in the underground cavern support design data may be a geological survey report, a support structure design manual, a construction drawing, and the like.

[0074] In one possible implementation, the entity statistics can be the number of times the entity appears in a design document, or the output volume of design documents containing the entity. Using statistical data to filter out the second entity distinguishes multiple documents and accurately reflects the differences between them, which is crucial for enabling rapid classification and accurate retrieval of design documents.

[0075] Optionally, the statistics are determined based on the number of times the entity appears in each design document and the number of design documents containing the entity.

[0076] In one possible case, the TF-IDF value of the entity is used to obtain the second entity in the underground cavern support design data. For example, the TF-IDF value of the entity is determined by the following formula:

[0077] TF-IDF(t,d)=TF(t,d)×IDF(t)

[0078] Here, TF-IDF(t,d) refers to the TF-IDF value of entity t in design document d, TF(t,d) is the number of times the term entity t appears in design document d, and IDF(t) is the inverse document frequency of the term entity t.

[0079] IDF(t) is calculated as follows:

[0080]

[0081] Where N is the total number of design documents and DF(t) is the number of documents containing entity t.

[0082] In an embodiment of the present application, before determining the second entity in the underground cavern support design data, natural language processing, image recognition and other technologies are first used to preprocess the data source, including but not limited to Chinese word segmentation, stop word removal and vocabulary extraction.

[0083] In an embodiment of the present application, by counting the number of occurrences of an entity in each design document, the importance or frequency of use of the entity in the support design can be quantified. For example, the more times an entity appears, the more critical or commonly used the entity is in the design. The number of design documents that contain the entity also reflects the popularity or scope of application of the entity. For example, the more documents that contain the entity, the more applications the entity has in multiple design scenarios or projects. In an embodiment of the present application, based on statistical data, the key entities in the underground cavern support design data, i.e., the second entity, can be extracted.

[0084] In some embodiments, the entities in the underground cavern support design data correspond to at least one category; the entity extraction model includes a bidirectional long short-term memory network (BiLSTM) and a conditional random field (CRF) model.

[0085] The BiLSTM model helps understand entity boundaries and context, and can capture the contextual information of the input sequence. The CRF model uses the hidden state of each time step provided by the BiLSTM model to predict the final label sequence by considering the constraints and interdependencies of the label sequence.

[0086] Figure 2 This is an application scenario diagram using the BiLSTM model and the CRF model. In the figure, the text representation (i.e., text vector) corresponding to the vocabulary in the document is input into the BiLSTM layer (i.e., the BiLSTM model), and the context information of the input vocabulary is captured by the BiLSTM layer. The CRF layer (i.e., the CRF model) extracts the entities in the document based on the context information of the input vocabulary, i.e., Figure 2 The geological entity is severe rock burst, and the support entity is pot support.

[0087] The method provided in the embodiment of the present application further includes constructing an entity extraction model, which specifically includes the following steps a1-a2:

[0088] a1: Get the training dataset.

[0089] The training data set contains multiple entities and the category corresponding to each entity.

[0090] Exemplarily, the categories corresponding to entities include but are not limited to construction (activity), such as construction process, construction behavior, etc.; feedback (reaction), that is, the feedback of structure or rock mass to load, such as deformation and stress distribution; structure (structure), such as lining, anchor frame and other supporting structure bodies; support (support), such as shotcrete support, steel arch support, etc.; geology (geology), such as surrounding rock grade and fault; tunnel (tunnel), that is, the main body of underground caverns, such as main tunnel and connecting channel; parameter (value), that is, design indicators or quantitative data, such as anchor length 3m; surrounding environment (surrounding), that is, the rock mass, soil or external conditions around the tunnel, such as surrounding rock pressure and groundwater; construction method (method), such as drilling and blasting method; support material (material), such as steel bar and shotcrete; equipment (facility), such as grouting pump and displacement meter.

[0091] In one possible implementation, the training dataset is in a format that the entity extraction model can understand. Exemplarily, the training dataset includes text vectors corresponding to entities and text vectors for categories corresponding to the entities. Exemplarily, the text vectors corresponding to the entities and the text vectors for categories corresponding to the entities are obtained using one-hot encoding.

[0092] a2: Based on the training dataset, the initial entity extraction model is trained to obtain the entity extraction model.

[0093] In one possible implementation, the training process involves adjusting model parameters (such as weights) to minimize the difference between the predicted label and the true label. Here, the label is used to represent the category of the entity.

[0094] For example, the objective function can be expressed as:

[0095]

[0096] Among them, L is the loss function, N is the number of samples, P(y i |x i ; θ) is the model parameter θ, input x i When , the model predicts the output label y i The conditional probability of .

[0097] In one possible implementation, the performance of the entity extraction model is evaluated on a test dataset. For example, the performance of the entity extraction model is tested using four metrics: accuracy, precision, recall, and F1 value.

[0098] During the testing process, there are four prediction situations: TP (True Positives) is the number of samples correctly labeled as positive by the entity extraction model, TN (True Negatives) is the number of samples correctly labeled as negative by the entity extraction model, FP (False Positives) is the number of samples incorrectly labeled as positive by the entity extraction model, and FN (False Negatives) is the number of samples incorrectly labeled as negative by the entity extraction model.

[0099] Accuracy is the ratio of the number of samples correctly predicted by the entity extraction model to the total number of samples, and is expressed as follows:

[0100]

[0101] Precision is the ratio of the number of samples correctly identified as positive by the entity extraction model to the total number of samples identified as positive by the model. The expression is as follows:

[0102]

[0103] Recall is the ratio of the number of samples correctly identified as positive by the entity extraction model to the total number of samples that are actually positive. The expression is as follows:

[0104]

[0105] F1 score (F1 score ) is the harmonic mean of precision and recall, considering both Precision and Recall, the expression is as follows:

[0106]

[0107] Among them, β is the weight coefficient. When β=1, the weights of Precision and Recall are the same.

[0108] In an embodiment of the present application, the bidirectional long short-term memory network can simultaneously capture the forward and backward information of the input sequence, thereby better understanding the context of the entity in the text. This helps to accurately identify entities in underground cavern support design data and reduce misjudgments and missed judgments. In sequence labeling tasks, the conditional random field model can consider the dependencies between labels and make more accurate divisions of entity boundaries. Combined with the output of the bidirectional long short-term memory network, the conditional random field model can further improve the accuracy of entity extraction.

[0109] In some embodiments, the underground cavern support design data includes underground cavern support design specifications; the preset form rules include entity number, entity name, and entity definition; the entity number is used to indicate the position of the entity in the underground cavern support design specifications.

[0110] In one possible implementation, underground cavern support design specifications are authoritative technical documents developed by industry authorities or standardization organizations. They define the principles, methods, parameter values, and technical requirements for underground cavern support design and serve as mandatory or recommended standards for design work. These specifications can help mitigate safety hazards caused by variability in human experience. For example, the specifications specify minimum support thicknesses for different surrounding rock grades.

[0111] In one possible implementation, the entity number is used to uniquely identify the entity in the specification. For example, the entity number can be the clause number in the underground cavern support design specification, and the clause number can be used to quickly locate the entity in the underground cavern support design specification.

[0112] In a possible implementation, the entity name may be the Chinese name of the entity, or the English name of the entity, etc., which is not specifically limited in this application.

[0113] In one possible implementation, the entity definition is a description and explanation of the essential characteristics of the entity, including but not limited to the entity's connotation, scope of use, and technical requirements.

[0114] For example, for documents such as "Terms", "Terms and Symbols" in standard specifications in the design data after the underground cave, the first entity is extracted using preset formal rules.

[0115] In the implementation of this application, the preset form rules include multiple implementation methods. The following is an illustrative explanation using Rule 1 and Rule 2.

[0116] Form Rule 1: <Article Number><Substantial Chinese><Substantial English><Explanation>

[0117] The following is part of the text from the underground cavern support design specification:

[0118] “2.1.27 Invert

[0119] An inverted arch lining structure set at the bottom of the tunnel.

[0120] 2.1.28 shoterete and rockbohs lining

[0121] Tunnel surrounding rock support structures using shotcrete, anchor rods, steel mesh and steel frames alone or in combination.

[0122] 2.1.29 Monolithic lining

[0123] A tunnel lining structure constructed with cast-in-place concrete or masonry after tunnel excavation.

[0124] 2.1.30 Composite lining

[0125] A composite lining structure consisting of shotcrete lining, waterproof layer and cast-in-place concrete lining.

[0126] 2.1.31 Advancing Drift

[0127] A small-section tunnel excavated ahead of the tunnel face within the tunnel excavation section.

[0128] In the above-mentioned underground cavern support design specifications, taking the inverted arch entity as an example, the clause number of this entity is 2.1.27, the Chinese word for the entity is inverted arch, and the English word for the entity is invert. The explanation of the entity is the reverse arch lining structure set at the bottom of the tunnel.

[0129] Form Rule 2: <Article Number><Substantial Chinese><Explanation>

[0130] The following is another part of the text content in the underground cavern support design specification:

[0131] "The long bench method is to divide the section into two sections, the upper and lower sections, for excavation. The upper and lower sections are far apart, and the upper bench should be more than 50m ahead or more than five times the tunnel span. The upper and lower sections can be operated in parallel. When the tunnel length is short, the upper section can be fully excavated first before the lower section is constructed. This is the half-section method."

[0132] In the above-mentioned underground cavern support design specification, the entity is the long step method, the entity is numbered 1, the entity name is the long step method, and the entity explanation is a description of the specific implementation process of the long step method.

[0133] In an embodiment of the present application, predefined formal rules are used to extract specific words, terms and phrases from underground cavern support design data, which is suitable for entity extraction of underground cavern support design data with relatively clear structured information.

[0134] In some embodiments, the knowledge element includes the relationship between entities, wherein the relationship between entities can be the above-mentioned classification relationship or a non-classification relationship.

[0135] In the above S103, the knowledge element of each entity in the underground cavern support design data is determined through the following steps b1-b4:

[0136] b1: Treat each preset word in the underground cavern support design data as an independent cluster and determine the inter-cluster similarity between independent clusters.

[0137] In a possible implementation, the underground cavern support design data includes multiple design documents; in the above b1, the inter-cluster similarity between independent clusters is determined by the following method:

[0138] First, based on the number of times each preset word appears in each design document, a vector representation of each preset word is determined.

[0139] For example, the bag-of-words model is used to obtain the vector representation of vocabulary, which regards the text as a set of unordered words, ignores the order and structure of the vocabulary in the text, and only considers the frequency of vocabulary appearance. Assume that there are N design documents and M preset vocabulary. For each design document i, an M-dimensional vector is created, where the jth element represents the number of times the jth word in the preset vocabulary appears in the design document i. The vector representation of design document i is i=[x i1 ,x i2 ,…,x iM ], where x ij Indicates the number of times the jth preset word appears in design document i.

[0140] Continuing with N design documents and M preset words as an example, the vector representation of the preset word j can be j=[y j1 ,y j2 ,…,y jN ], where y ij Indicates the number of times the preset word j appears in the design document i.

[0141] Then, each preset word is regarded as an independent cluster, and the inter-cluster similarity between the independent clusters is determined based on the vector representation of each preset word.

[0142] For example, the inter-cluster similarity between independent clusters can be expressed by the Euclidean distance:

[0143]

[0144] Among them, Concept Similarity(x,y) is the inter-cluster similarity between independent clusters x and y, x i and y i are the elements in the vector representation of the independent clusters, and n is the dimension of the vector representation.

[0145] In this implementation, the importance or significance of each word can be quantified by counting the number of times a preset word appears in the design document. The vector representation not only reflects the individual characteristics of the word, but also implies the correlation between words. By calculating the similarity between vectors, it is possible to reveal which words frequently appear together in the support design data, thereby inferring the possible semantic or functional connections between them, and thus mining the hidden knowledge in the support design data, such as the support result type and material combination method, which is of great significance for optimizing the support design scheme.

[0146] b2: Select the two clusters with the smallest inter-cluster similarity to merge.

[0147] In a possible implementation, a vector representation of the merged cluster is generated by performing a mathematical operation, such as addition, on corresponding elements of the vector representations of the two clusters.

[0148] b3: Update the inter-cluster similarity between clusters and return to the step of selecting the two clusters with the smallest inter-cluster similarity to merge until the preset iteration conditions are met and at least one final cluster is obtained.

[0149] In a possible implementation, inter-cluster similarity is used to measure the similarity between different clusters, so as to determine which clusters to merge together based on the inter-cluster similarity.

[0150] For example, assuming there are two clusters C1 and C2, the similarity between clusters is calculated as follows:

[0151]

[0152] Among them, Cluster Similarity(C1,C2) is the inter-cluster similarity of clusters C1 and C2, V(C1∪C2) represents the variance of the merged cluster C1∪C2, V(C1) and V(C2) represent the variances of clusters C1 and C2 respectively, and |C1| and |C2| represent the number of words in clusters C1 and C2 respectively.

[0153] In one possible implementation, the preset iteration condition may be stopping when the number of clusters decreases to a preset number threshold, stopping when the minimum inter-cluster distance exceeds a preset distance threshold, or stopping when a maximum number of iterations is reached or when the runtime has expired. This application does not impose any specific restrictions on the preset iteration condition and may be set based on actual circumstances.

[0154] b4: Determine the relationship between entities based on the entities in the same final cluster.

[0155] In one possible implementation, entities belonging to the same final cluster are considered to have potential associations. By analyzing the semantic features of the entities within the cluster, the type of relationship between the two entities, such as inclusion relationship, causal relationship, etc., can be determined. For example, it is assumed that the entities in a certain final cluster are surrounding rock level, anchor length, shotcrete thickness, steel arch spacing, and support scheme. Among them, the surrounding rock level and anchor length can be determined as a causal relationship, and the shotcrete thickness and steel arch spacing can be determined as a collaborative relationship. For example, the smaller the steel arch spacing, the greater the shotcrete thickness is likely to be. The support scheme and (anchor, steel arch, shotcrete) are in a whole-part relationship, that is, the support scheme is composed of a combination of multiple support structures.

[0156] In one possible implementation, the association relationship between two entities can be determined based on the number of co-occurrences or mutual information between the entities in the same final cluster. For example, the mutual information between the entities in the same final cluster can be calculated to determine whether the two entities are associated. If the mutual information between two entities in the same final cluster exceeds a preset mutual information threshold, the two entities are determined to be associated. The higher the mutual information between the two entities, the stronger the association relationship between the entities.

[0157] In another possible implementation, the relationship between entities in the same final cluster is determined based on a machine learning method.

[0158] For example, machine learning based on knowledge graph embedding determines the relationships between entities within the same final cluster. During machine learning training, the training dataset contains multiple entities and relationships between them, such as determining relationships and influencing relationships. For example, the relationship between the two entities, steel arch spacing and shotcrete thickness, is an influencing relationship, meaning that the steel arch spacing affects the shotcrete thickness.

[0159] In yet another possible implementation, the relationship between entities is determined based on a preset mapping relationship in the underground hole support design.

[0160] Optionally, the preset mapping relationships include but are not limited to mapping between surrounding rock grade and support parameters, mapping between support structure and materials / processes, mapping between construction methods and support timing, etc.

[0161] For the mapping of surrounding rock grade and support parameters, the preset mapping relationship can be that when the surrounding rock grade is V, anchor bolts + steel arches are required. Therefore, if the same final cluster contains both V-grade surrounding rock and steel arches, the relationship between the two entities is defined as a determining relationship. Another preset mapping relationship can be that when groundwater pressure is greater than 0.5 MPa, a drain / waterproofing plate is required. If the same final cluster contains both groundwater pressure and a drain / waterproofing plate, the relationship between the two entities is defined as a determining relationship.

[0162] For the mapping between support structure and material / process, the preset mapping relationship can be that the material requirement for steel arch support is I-beam. In other words, the steel arch support and I-beam have a determining relationship.

[0163] In the embodiments of this application, inter-cluster similarity is used to characterize the semantic or contextual similarity between words. By iteratively merging clusters with the smallest similarity and updating the inter-cluster similarity, the recognition results of entity relationships are continuously optimized. This not only allows the identification of obvious entity relationships in underground cavern support design data, but also allows the gradual merging of clusters to discover complex relationships hidden in the data, thereby improving the accuracy of extracting relationships between entities.

[0164] The above mainly introduces the solution provided in the embodiment of the present application from the perspective of method.

[0165] In the embodiments of the present application, a device for extracting knowledge elements from underground cavern support design is also provided. This device is used to implement the above-mentioned embodiments and preferred implementation methods. Details that have already been described will not be repeated here. As used below, the term "module" may refer to a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation using hardware, or a combination of software and hardware, is also possible and contemplated.

[0166] This embodiment provides a knowledge element extraction device for underground cavern support design. Figure 3 As shown, the device includes:

[0167] The first acquisition module 301 is used to obtain underground cavern support design data;

[0168] Extraction module 302 is used to extract multiple entities from the underground cavern support design data; the entity includes one or more of a first entity, a second entity, and a third entity; the first entity satisfies a preset form rule in the underground cavern support design data; the statistical data of the second entity in the underground cavern support design data satisfies a preset condition; and the third entity is obtained based on a pre-built entity extraction model;

[0169] The determination module 303 is used to determine the knowledge element of each entity in the underground cavern support design data.

[0170] Through the device provided by this embodiment, preset formal rules, statistical data and entity extraction models are combined to ensure that entities from different sources can be effectively obtained, avoiding information omissions caused by a single method. Specifically, the preset formal rules can ensure the standardized extraction of structured underground cavern support design data (such as underground cavern support design specifications), statistical data can extract key entities in underground cavern support design data, and the entity extraction model can deeply mine semantic entities, especially unstructured underground cavern support design data. The embodiment of this application complements and improves the accuracy of entity extraction through three ways.

[0171] In some optional embodiments, in the device, the underground cavern support design data includes a plurality of design documents; the statistical data of the second entity characterizing the second entity can be used to distinguish between the design documents.

[0172] In some optional embodiments, in the device, the statistical data is determined based on the number of times the entity appears in each design document and the number of design documents containing the entity.

[0173] In some optional embodiments, the entities in the underground cavern support design data correspond to at least one category; the entity extraction model includes a bidirectional long short-term memory network and a conditional random field model, and the device further includes:

[0174] The second acquisition module is used to obtain a training data set; the training data set contains multiple entities and the category corresponding to each entity;

[0175] The training module is used to train the initial entity extraction model based on the training data set to obtain the entity extraction model.

[0176] In some optional embodiments, in the device, the underground cavern support design data includes underground cavern support design specifications; the preset form rules include entity number, entity name, and entity definition; the entity number is used to indicate the position of the entity in the underground cavern support design specifications.

[0177] In some optional embodiments, the knowledge element includes the relationship between entities; the determination module 303 is specifically used to treat each preset word in the underground cavern support design data as an independent cluster, determine the inter-cluster similarity between the independent clusters; select the two clusters with the smallest inter-cluster similarity to merge; the inter-cluster similarity between the new clusters, return to the step of selecting the two clusters with the smallest inter-cluster similarity to merge, until the preset iteration condition is met, and obtain at least one final cluster; based on the entities in the same final cluster, determine the relationship between the entities.

[0178] In some optional embodiments, the underground cavern support design data includes multiple design documents; the determination module 303 is specifically used to determine the vector representation of each preset word based on the number of times each preset word appears in each design document; each preset word is regarded as an independent cluster, and based on the vector representation of each preset word, the inter-cluster similarity between each independent cluster is determined.

[0179] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0180] The underground cavern support design knowledge element extraction device in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0181] The embodiment of the present invention also provides a computer device having the above Figure 3 The knowledge element extraction device for underground cavern support design is shown.

[0182] See also Figure 4 , Figure 4 is a structural diagram of a computer device provided by an optional embodiment of the present invention, such as Figure 4 As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components utilize different buses to communicate with each other and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of a GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories. Equally, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 4 A processor 10 is taken as an example.

[0183] The processor 10 may be a central processing unit, a network processor, or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit, a programmable logic device, or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic, or any combination thereof.

[0184] The memory 20 stores instructions that can be executed by at least one processor 10, so as to enable at least one processor 10 to execute the method shown in the above embodiment.

[0185] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created based on the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0186] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0187] The computer device further includes a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0188] The embodiment of the present invention also provides a computer-readable storage medium. The above-mentioned method according to the embodiment of the present invention can be implemented in hardware, firmware, or implemented as a computer code that can be recorded in a storage medium, or implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and downloaded through a network and will be stored in a local storage medium, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state drive, etc.; further, the storage medium can also include a combination of the above-mentioned types of memory. It can be understood that a computer, a processor, a microprocessor controller or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor or hardware, the method shown in the above embodiment is implemented.

[0189] A portion of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the form in which the computer program instruction exists in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc. Accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium that can be accessed by the computer.

[0190] Although the embodiments of the present invention have been described with reference to the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention. Such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A knowledge element extraction method for underground cavern support design, characterized in that: The method comprises: Obtain underground cavern support design data; Extracting multiple entities from the underground cavern support design data; the entities include one or more of a first entity, a second entity, and a third entity; the first entity satisfies a preset form rule in the underground cavern support design data; statistical data of the second entity in the underground cavern support design data satisfies a preset condition; and the third entity is obtained based on a pre-built entity extraction model; Determine the knowledge element of each entity in the underground cavern support design data.

2. The method according to claim 1, characterized in that The underground cavern support design data includes a plurality of design documents; the statistical data of the second entity represents that the second entity can be used to distinguish each of the design documents.

3. The method according to claim 2, characterized in that The statistical data is determined based on the number of times the entity appears in each of the design documents and the number of design documents containing the entity.

4. The method according to any one of claims 1 to 3, characterized in that The entities in the underground cavern support design data correspond to at least one category; the entity extraction model includes a bidirectional long short-term memory network and a conditional random field model, and the method further includes: Obtain a training data set; the training data set includes multiple entities and a category corresponding to each entity; Based on the training data set, an initial entity extraction model is trained to obtain the entity extraction model.

5. The method according to claim 1, characterized in that The underground cavern support design data includes underground cavern support design specifications; the preset format rules include entity number, entity name, and entity definition; the entity number is used to indicate the position of the entity in the underground cavern support design specifications.

6. The method according to claim 1 or 5, characterized in that The knowledge element includes the relationship between the entities; The determining of the knowledge element of each entity in the underground cavern support design data includes: Taking each preset word in the underground cavern support design data as an independent cluster, and determining the inter-cluster similarity between the independent clusters; Select the two clusters with the smallest inter-cluster similarity to merge; Updating the inter-cluster similarity between the clusters, returning to the step of selecting two clusters with the smallest inter-cluster similarity to merge, until a preset iteration condition is met, and at least one final cluster is obtained; Based on the entities being in the same final cluster, the relationship between the entities is determined.

7. The method according to claim 6, characterized in that The underground cavern support design data includes a plurality of design documents; taking each preset word in the underground cavern support design data as an independent cluster and determining the inter-cluster similarity between the independent clusters includes: Determining a vector representation of each of the preset words based on the number of times each of the preset words appears in each of the design documents; Each preset word is regarded as an independent cluster, and based on the vector representation of each preset word, the inter-cluster similarity between the independent clusters is determined.

8. A knowledge element extraction device for underground cavern support design, characterized in that: The device comprises: The first acquisition module is used to obtain underground cavern support design data; An extraction module is configured to extract multiple entities from the underground cavern support design data; the entities include one or more of a first entity, a second entity, and a third entity; the first entity satisfies a preset formal rule in the underground cavern support design data; the statistical data of the second entity in the underground cavern support design data satisfies a preset condition; and the third entity is obtained based on a pre-built entity extraction model; The determination module is used to determine the knowledge element of each entity in the underground cavern support design data.

9. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the underground cavern support design knowledge element extraction method according to any one of claims 1 to 7 by executing the computer instructions.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the underground cavern support design knowledge element extraction method according to any one of claims 1 to 7.

Images