Oil and gas business node identification matching method, device, storage medium and program product

Abstract

This invention relates to the field of oil and gas operations, and discloses a method, device, storage medium, and program product for identifying and matching oil and gas business nodes. The method includes: uniformly converting raw business problem information received from the oil and gas business system into business problem text with a consistent text format; parsing the business problem text using a pre-acquired general information extraction model to extract oil and gas business nodes from the business problems and outputting a node set; calling a semantic embedding vector model to convert each oil and gas business node in the node set into a high-dimensional business semantic vector with a vector dimension of at least a preset dimension, and performing matching calculations with the real business semantic vectors formed by a pre-constructed set of real business nodes, selecting the real business node that reaches the similarity threshold and has the highest similarity threshold as the matching node for the corresponding oil and gas business node. The method provided by this invention can improve the matching success rate and efficiency of oil and gas business nodes.

Description

Technical Field

[0001] This invention relates to the field of oil and gas business, and in particular to a method, device, storage medium and program product for identifying and matching oil and gas business nodes. Background Technology

[0002] Currently, the standardization of oil and gas business issues includes two methods: manual standardization and automated processing based on keyword matching. The method for manual standardization involves business personnel manually identifying the core business nodes (such as drilling, oil production, gathering and transportation, refining, equipment maintenance, etc.) in oil and gas business standard node dictionary, and then replacing the non-standard nodes in the problem with unified real business nodes, thus completing the problem standardization for subsequent work order allocation, intelligent Q&A and other business processes.

[0003] The automated processing method based on keyword matching is to pre-define a keyword library for oil and gas business nodes, use exact or fuzzy string matching to extract content that matches the keyword library from the problem as a business node, and then directly associate it with the standard node corresponding to the keyword library to achieve problem standardization.

[0004] Currently, both methods have the following problems: First, manual standardization is inefficient: The number of business issues generated daily in oil and gas operations (such as frontline operational inquiries, equipment malfunction feedback, and business process consultations) is enormous. Relying solely on manual processing would incur significant manpower costs and have long processing cycles, failing to meet the real-time response requirements of the business. Furthermore, manual identification is susceptible to subjective experience. For complex business issues (such as those involving multiple sub-nodes or non-standard node descriptions), errors in node identification and substitution are prone to occur, leading to poor accuracy in the standardization results.

[0005] Second, keyword matching methods lack semantic understanding: Traditional keyword matching relies solely on literal character comparison and fails to grasp the contextual semantics of oil and gas business issues. Due to the complexity of the oil and gas business node system, there are numerous synonyms (e.g., "drilling" vs. "exploration"), similar expressions (e.g., "oil production equipment maintenance" vs. "pumping unit overhaul"), and subdivided nodes (e.g., "drill bit replacement in drilling" vs. "wellhead pressure adjustment in oil production"). Relying solely on keyword matching cannot distinguish these semantics, leading to incomplete, missed, or misidentified nodes. Furthermore, when non-standard expressions (e.g., colloquialisms, abbreviations) exist, keyword matching cannot effectively identify them, further reducing the standardization success rate.

[0006] Third, the matching accuracy is insufficient and the standardization effect is poor: Keyword matching lacks semantic-level association calculation and relies solely on character similarity, failing to accurately associate the core semantics of nodes in the question with the actual business nodes. This results in inconsistencies in nodes and non-standardized expressions even after standardization, affecting the smooth progress of subsequent business processes. Summary of the Invention

[0007] This invention provides an oil and gas business node identification and matching method, device, storage medium, and program product to solve at least one of the above-mentioned problems.

[0008] In a first aspect, embodiments of the present invention provide an oil and gas business node identification and matching method, comprising: S1: uniformly converting the original business problem information received from the oil and gas business system into business problem text with a consistent text format; S2: parsing the business problem text through a pre-acquired general information extraction model, extracting oil and gas business nodes in the business problem, and outputting a node set formed by multiple oil and gas business nodes; S3: calling a pre-acquired semantic embedding vector model to convert each oil and gas business node in the node set into a high-dimensional business semantic vector with a vector dimension of more than a preset dimension; S4: performing matching calculations between the high-dimensional semantic vector and the real business semantic vector formed by a pre-constructed set of real business nodes, and selecting the real business node in the set of real business nodes that reaches the similarity threshold and has the highest similarity threshold as the matching node of the corresponding oil and gas business node.

[0009] Optionally, the general information extraction model is obtained through the following steps: the original general information extraction model is fine-tuned and trained using a pre-acquired oil and gas labeled dataset to obtain the fine-tuned general information extraction model; wherein, the oil and gas labeled dataset includes multiple pre-labeled text labeled nodes of oil and gas business scenarios, and the oil and gas labeled dataset includes multiple first-level nodes and multiple second-level sub-nodes, and each first-level node includes multiple second-level sub-nodes.

[0010] Optionally, the oil and gas business node identification and matching method also includes: adjusting the semantic embedding vector model by segmenting it using a pre-acquired oil and gas annotation dataset, so that the fine-tuned semantic embedding vector model can identify and retain oil and gas terms in the business question text.

[0011] Optionally, the default dimension is 768.

[0012] Optionally, the original business problem information includes business problem information in text format and business problem information in audio format.

[0013] Optionally, the steps of the general information extraction model parsing the business problem text, extracting oil and gas business nodes from the business problem, and outputting a node set formed by multiple oil and gas business nodes include: inputting the business problem text into the general information extraction model, parsing the semantics of the business problem text, extracting explicit and implicit oil and gas business nodes from the business problem text; and outputting all extracted explicit and implicit oil and gas business nodes in the form of a node set.

[0014] Optionally, the step of matching the high-dimensional semantic vector with the real business semantic vector formed by the pre-constructed set of real business nodes includes: constructing a set of real business nodes through a pre-acquired real node vector library of oil and gas business; converting the pre-acquired set of real business nodes into real business semantic vectors through a semantic embedding vector model, and constructing a real node vector library including all real business semantic vectors; establishing a vector index structure for indexing real business semantic vectors, and storing the vector index structure and the real node vector library in a distributed database; inputting the high-dimensional semantic vector into the real node vector library, and calculating the similarity between the high-dimensional semantic vector and each real business semantic vector through the vector index structure.

[0015] Optionally, the oil and gas business node identification and matching method further includes: when a new real business node is received, generating a real business semantic vector corresponding to the new real business node through a semantic embedding vector model, and inputting it into the real node vector library.

[0016] Optionally, after obtaining the real business node that matches the oil and gas business node, the oil and gas business node identification and matching method further includes: replacing the oil and gas business node in the original business problem information with the matched real business node; and generating business problem text based on the original business problem information after the node replacement.

[0017] Optionally, the oil and gas business node identification and matching method further includes: after obtaining the business problem text, inputting the business problem text into the pre-acquired binary classification verification model to determine the degree of matching between the text format of the business problem text and the standard text format of oil and gas business. If the degree of matching reaches the preset matching threshold, the business problem text is output; otherwise, return to step S2 to re-extract the oil and gas business nodes.

[0018] Optionally, after step S2 and before step S3, the oil and gas business node identification and matching method further includes: filtering the node set based on the pre-acquired list of oil and gas business nodes, removing erroneous nodes from the node set, and supplementing the missing fields of fuzzy nodes in the node set to generate a filtered node set.

[0019] Optionally, in step S4, the oil and gas business node identification and matching method further includes: if the similarity of the high-dimensional semantic vector with the highest similarity to the real business semantic vector is less than the similarity threshold, then the matching range is expanded to the semantic vectors of the associated nodes of the real business semantic vector, and the similarity between the high-dimensional semantic vector and the semantic vector of the associated node is calculated. If the similarity between the high-dimensional semantic vector and the semantic vector of the associated node still does not reach the similarity threshold, then the preset oil and gas business rule engine is invoked, and the high-dimensional semantic vector is forcibly matched with the real business semantic vector based on the oil and gas business rules to obtain the real business node that matches the oil and gas business node.

[0020] In a second aspect, embodiments of the present invention provide an electronic device, including: a processor and a memory, wherein the memory stores instructions; the processor invokes the instructions in the memory to cause the processor to execute the oil and gas business node identification and matching method of any of the foregoing embodiments of the first aspect of the present invention.

[0021] Thirdly, embodiments of the present invention provide a computer-readable storage medium storing instructions that, when executed by a processor, implement the oil and gas business node identification and matching method of any of the foregoing embodiments of the first aspect of the present invention.

[0022] Fourthly, embodiments of the present invention provide a computer program product, which includes a computer program that, when executed by a processor, implements the oil and gas business node identification and matching method of any of the foregoing embodiments of the first aspect of the present invention.

[0023] The above-mentioned technical solution of the present invention can accurately screen out the real business nodes that best match the oil and gas business nodes, thereby achieving accurate understanding of the semantics of business issues received by the oil and gas business system, improving the matching success rate and matching efficiency, and reducing business operation costs. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0025] Figure 1 This is a flowchart of one embodiment of the oil and gas business node identification and matching method of the present invention; Figure 2 This is a flowchart of step S120 in one embodiment of the oil and gas business node identification and matching method of the present invention; Figure 3This is a flowchart illustrating the construction of an oil and gas labeled dataset in one embodiment of the oil and gas business node identification and matching method of the present invention. Figure 4 This is a flowchart of training a general information extraction model in one embodiment of the oil and gas business node identification and matching method of the present invention. Figure 5 This is a flowchart of step S140 in one embodiment of the oil and gas business node identification and matching method of the present invention; Figure 6 A flowchart illustrating the optimized model parameters in one embodiment of the oil and gas business node identification and matching method of the present invention; Figure 7 This is a flowchart illustrating the construction of a real node vector library in one embodiment of the oil and gas business node identification and matching method of the present invention; Figure 8 This is an overall flowchart of one embodiment of the oil and gas business node identification and matching method of the present invention; Figure 9 This is a flowchart illustrating the calculation and verification of similarity in one embodiment of the oil and gas business node identification and matching method of the present invention. Figure 10 This is a structural block diagram of one embodiment of the electronic device of the present invention.

[0026] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0028] It should be noted that all directional indications in the embodiments of the present invention, such as up, down, left, right, front, back, etc., are only used to explain the relative positional relationship and movement of the components in a specific posture as shown in the attached figure. If the specific posture changes, the directional indication will also change accordingly.

[0029] Furthermore, the use of terms such as "first" and "second" in this invention is for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature. Additionally, the technical solutions of the various embodiments can be combined with each other, but only on the basis of being achievable by those skilled in the art. When the combination of technical solutions is contradictory or impossible to implement, such a combination of technical solutions should be considered non-existent and not within the scope of protection claimed by this invention.

[0030] For ease of understanding, the oil and gas business node identification and matching method of this invention is described below, such as... Figure 1 As shown, the oil and gas business node identification and matching method in this embodiment of the invention includes steps S110 to S140.

[0031] In step S110, the original business problem information received from the oil and gas business system is uniformly converted into business problem text with a consistent text format.

[0032] In some optional embodiments, the original business problem information includes business problem information in text format and business problem information in audio format.

[0033] For example, the original business questions in the oil and gas business system include text-based questions and audio-based questions. Text-based questions might be like "How to deal with severe drill bit wear during drilling?", while audio-based questions are sent voice data. After the audio-based questions are identified, they are converted into text-based questions, resulting in spoken text after speech recognition.

[0034] In step S120, the business problem text is parsed using a pre-acquired general information extraction model, oil and gas business nodes in the business problem are extracted, and a node set formed by multiple oil and gas business nodes is output.

[0035] like Figure 2 As shown, in some optional embodiments, step S120 includes steps S121 to S122.

[0036] In step S121, the business problem text is input into the general information extraction model, the semantics of the business problem text are parsed, and explicit oil and gas business nodes and implicit oil and gas business nodes in the business problem text are extracted.

[0037] In step S122, all the extracted explicit oil and gas business nodes and all the implicit oil and gas business nodes are output as a node set.

[0038] Specifically, the general information extraction model is obtained through the following steps: By using a pre-acquired oil and gas labeled dataset, the original general information extraction model is fine-tuned and trained to obtain a fine-tuned general information extraction model.

[0039] The oil and gas annotation dataset consists of multiple pre-annotated text annotation nodes for oil and gas business scenarios. The oil and gas annotation dataset also includes multiple first-level nodes and multiple second-level sub-nodes, with each first-level node containing multiple second-level sub-nodes.

[0040] For example, the UIE (Universal Information Extraction) model under the PaddlePaddle framework can be used as the base model for training and fine-tuning.

[0041] The UIE model adopts an instruction-based unified information extraction framework, which models tasks such as entity extraction, relation extraction, and event extraction as a sequence labeling problem: given an extraction instruction, output structured results. Its underlying layer uses a pre-trained language model for encoding and a pointer network to predict the start and end positions of the extracted segments.

[0042] In the specific training process, the training flow of the UIE model includes the following steps: Step 1: Task Instruction Construction Based on the requirements for identifying oil and gas business nodes, oil and gas business nodes at different levels, including primary business nodes and secondary sub-nodes, are uniformly abstracted into extraction instructions, such as extracting oil and gas business nodes from text. Drilling-related business nodes are extracted, and the business node types are mapped to the schema structure supported by the UIE model.

[0043] Step 2: Data Format Standardization The oil and gas business annotation dataset is converted into a unified JSON format required by the UIE model. Each data sample in the oil and gas business annotation dataset includes the input text, the corresponding extraction instructions, and the start and end positions of the labeled nodes, enabling the model to learn the boundaries and semantic features of nodes given the extraction target.

[0044] Step 3: Model Initialization and Parameter Loading The parameters of the pre-trained UIE basic model are initialized to enable the model to have general Chinese semantic understanding capabilities, and on this basis, data from the oil and gas business domain are introduced for transfer learning.

[0045] Step 4: Supervised Fine-tuning Training The UIE model is trained under supervision using an oil and gas business labeled dataset. By minimizing the loss function between predicted nodes and real labeled nodes, the model parameters are continuously optimized, enabling the model to gradually learn the node representation, contextual dependencies, and implicit semantic features in oil and gas business scenarios.

[0046] Step 5: Model Evaluation and Iterative Optimization During training, the model is evaluated using a validation set, and the accuracy, recall, and F1 score of business node extraction are calculated. Based on the evaluation results, the training rounds, learning rate, and schema design are adjusted to obtain a general information extraction model suitable for oil and gas business scenarios.

[0047] Through the above training process, the fine-tuned general information extraction model can significantly enhance its ability to recognize oil and gas business terminology, oil and gas industry semantics, and complex oil and gas business node combinations while maintaining its general information extraction capabilities.

[0048] like Figure 3 and Figure 4 As shown, in this embodiment, the system receives original business issues from the operation terminal, consultation platform, and equipment fault reporting platform of the oil and gas business system, converts them into a standard text format, and then proceeds to subsequent processing.

[0049] In this embodiment, the general information extraction model obtained above is fine-tuned using an oil and gas business corpus. The gas business annotation dataset is automatically constructed based on an oil and gas industry lexicon and rule scripts to increase the data scale and coverage, and reduce the cost of manual annotation.

[0050] Specifically, the fine-tuning process of the general information extraction model includes the following steps: First, the construction of a terminology system for the oil and gas industry.

[0051] First, based on oil and gas industry standard documents, business specifications, and historical business system data, a vocabulary system for the oil and gas business domain is constructed, including: Primary business node terms: such as drilling, oil production, gathering and transportation, oil refining, equipment maintenance, safety management, etc.

[0052] Secondary subdivided business node vocabulary: such as drilling - drill bit replacement, oil production - wellhead pressure regulation, gathering and transportation - pipeline inspection, etc.; as well as industry-specific terms, synonyms, abbreviations and colloquial expressions.

[0053] Second, data collection based on the OiO Data platform.

[0054] By using a pre-set data collection script, historical oil and gas business problem texts, work order descriptions, equipment operation and maintenance records, and business consultation content are automatically collected from the OiO Data platform to form a raw business corpus. This raw business corpus can cover a variety of business scenarios, multiple expressions, and problem descriptions of varying complexity, and can truly reflect the actual language characteristics in oil and gas business.

[0055] Third, automated statement generation and enhancement.

[0056] Based on the existing collection of original business corpora, the collected original corpora are automatically expanded through rule scripts to generate diverse business question sentence structures. These can include: recombination of different word order structures, replacement of synonyms and near-synonyms, conversion between colloquial and standardized expressions, and automatic splicing of questions combining multiple business nodes.

[0057] Using the above methods, comprehensive and diverse texts on oil and gas business issues can be automatically generated.

[0058] Fourth, automatic annotation and node alignment.

[0059] While generating business problem text, the system automatically labels oil and gas business nodes in the text using a business node vocabulary and rule templates. This labels the nodes with names, types, and their start and end positions within the text, enabling automatic labeling under weak supervision. High-value samples can be manually sampled and corrected to improve labeling accuracy.

[0060] Fifth, the training data format conversion for general information extraction models.

[0061] The automatically generated and labeled business corpus is converted into the data format required by the general information extraction model, including input text, extraction instructions, and corresponding node annotation results, thereby forming a standardized UIE training dataset that can be directly used for supervised fine-tuning training of the general information extraction model.

[0062] Sixth, model fine-tuning and continuous iteration.

[0063] Based on the oil and gas business training corpus constructed through the above steps, the general information extraction model is fine-tuned and trained, enabling it to gradually learn the node distribution patterns, contextual dependencies, and implicit semantic mapping relationships in the oil and gas business domain. Simultaneously, new problems from actual business operations and manual corrections are continuously fed back into the training dataset, achieving continuous model iteration and performance improvement.

[0064] Based on the collected oil and gas industry labeled dataset, in this embodiment, more than 100,000 real business questions labeled with business nodes are used as the labeled dataset to train the UIE model, so as to optimize the model's ability to understand the semantics of oil and gas business, enabling it to accurately identify business nodes in the question text and obtain a general information extraction model.

[0065] This labeled dataset covers all business scenarios of oil and gas operations, including drilling, oil production, gathering and transportation, refining, equipment maintenance, and safety supervision. The labeled nodes include more than 300 types of primary nodes and corresponding secondary sub-nodes.

[0066] The standardized original problem is input into a fine-tuned general information extraction model. The model uses semantic parsing to extract explicit and implicit nodes representing oil and gas business nodes within the problem. For example, in the text field "anomaly at the wellhead pressure," the explicit node "wellhead" and the implicit node "pressure regulation" are extracted. The resulting node set is S = {s1, s2, ..., s}. n}; where n is the number of nodes identified.

[0067] Alternatively, a BERT model fine-tuned from an oil and gas business corpus can be used to replace the UIE model for node identification. In some optional embodiments, after step S120 and before step S130, the oil and gas business node identification and matching method further includes: Based on the pre-acquired list of oil and gas business nodes, the node set is filtered to remove erroneous nodes and fill in the missing fields of ambiguous nodes, thus generating a filtered node set.

[0068] In this embodiment, by constructing an oil and gas business node dictionary containing all real nodes and a list of common erroneous nodes, the node set S identified in step S120 is filtered, and non-oil and gas business nodes, i.e. erroneous nodes, in the list of erroneous nodes are removed, while oil and gas business nodes are retained.

[0069] Meanwhile, for nodes with ambiguous meanings, such as those that only identify "equipment" but do not specify the specific equipment type, the context of the question text is used to complete the meaning. For example, if the question text contains "oil production equipment" or "oil production", "abnormal equipment pressure" is completed as "oil production equipment pressure regulation", and a filtered node set S' is generated to reduce invalid vector transformations and matching calculations and improve processing efficiency.

[0070] In step S130, the pre-acquired semantic embedding vector model is invoked to convert each oil and gas business node in the node set into a high-dimensional business semantic vector with a vector dimension of more than a preset dimension.

[0071] Specifically, the preset dimension in this embodiment is 768 dimensions. Before calling the semantic embedding vector model, the semantic embedding vector model is segmented and adjusted using a pre-acquired oil and gas annotation dataset, so that the fine-tuned semantic embedding vector model can recognize and retain oil and gas terms in the business question text.

[0072] The semantic embedding vector model in this embodiment is the Qwen3-embedding-0.6B vector model, which is obtained by extracting each business node s in step S120. i (i=1,2,...,n) is transformed into a high-dimensional semantic vector v with at least 768 dimensions. i This enables the precise capture of semantic features of business nodes through the vector, achieving semantic normalization of synonyms and similar expression nodes.

[0073] Alternatively, the Word2Vec or GloVe vector model can be used instead of the Qwen3-embedding-0.6B model for vector transformation.

[0074] In step S140, the high-dimensional semantic vector is matched with the real business semantic vector formed by the pre-constructed set of real business nodes. The real business nodes that reach the similarity threshold and have the highest similarity threshold in the set of real business nodes are selected as the matching nodes of the corresponding oil and gas business nodes.

[0075] like Figure 5 As shown, in some optional embodiments, the step of matching and calculating the high-dimensional semantic vector with the real business semantic vector formed by the pre-built set of real business nodes in step S140 includes steps S141 to S144.

[0076] In step S141, a set of real business nodes is constructed using the pre-acquired real node vector library of oil and gas business.

[0077] In step S142, the pre-acquired set of real business nodes is converted into real business semantic vectors through a semantic embedding vector model, and a real node vector library including all real business semantic vectors is constructed.

[0078] In step S143, a vector index structure for indexing real business semantic vectors is established, and the vector index structure and the real node vector library are stored in a distributed database.

[0079] In step S144, the high-dimensional semantic vector is input into the real node vector library, and the similarity between the high-dimensional semantic vector and each real business semantic vector is calculated by matching the vector index structure.

[0080] like Figure 6 and Figure 7As shown, in this embodiment, the steps for constructing the real node vector library for oil and gas operations are as follows: The obtained unified set of real business nodes T = {t1, t2, ..., t} m}; where m is the total number of actual nodes in the standard nodes of the full business scenario of oil and gas business.

[0081] The set T is transformed into the corresponding real business semantic vector u using the Qwen3-embedding-0.6B vector model. j (j=1,2,...,m), a vector index is built using the FAISS index structure to improve retrieval efficiency, and it is stored in a distributed database.

[0082] When performing semantic matching calculations, the high-dimensional semantic vector v obtained in step S130 is used... i Input a real node vector library and use the cosine similarity algorithm to calculate the high-dimensional semantic vector v. i Sim(v) is the similarity to all real business semantic vectors u. i ,u j The similarity threshold is set to θ, where the default θ = 0.85, and can be dynamically adjusted according to the business scenario to filter out Sim(v) i ,u j The true node t ≥ θ j As each business node s i The matching result, if there are multiple real nodes t j If the threshold requirement is met, then the real node t with the highest similarity is selected. j As the optimal matching node.

[0083] Alternatively, a hybrid approach combining oil and gas business rules and vector matching can be used. Explicit nodes in the question text can be extracted using regular expressions, and then matched using the Qwen3-embedding-0.6B model.

[0084] Furthermore, step S140 also includes step S145.

[0085] In step S145, if the similarity of the high-dimensional semantic vector with the highest similarity to the real business semantic vector is less than the similarity threshold, the matching range is expanded to the semantic vectors of the associated nodes of the real business semantic vector, and the similarity between the high-dimensional semantic vector and the semantic vectors of the associated nodes is calculated.

[0086] If the similarity between the high-dimensional semantic vector and the semantic vector of the associated node still does not reach the similarity threshold, the preset oil and gas business rule engine is invoked to forcibly match the high-dimensional semantic vector with the real business semantic vector based on the oil and gas business rules, so as to obtain the real business node that matches the oil and gas business node.

[0087] like Figure 8 and Figure 9 As shown, in this embodiment, if the service node s in step S140 i The highest similarity real node t j Sim (v i , u j If θ < θ, then multi-round matching is initiated: First, the similarity is recalculated by expanding the vector library search scope to the associated nodes of the real nodes; if the calculated real node t j If the similarity still does not meet the threshold requirement, the oil and gas business rule engine with built-in node association rules is invoked. For example, "drilling" is forcibly associated with "well drilling". The association rules are combined to perform forced matching to ensure that each valid node can be matched with the corresponding real node and avoid missed matching.

[0088] In some embodiments, after obtaining the real business node that matches the oil and gas business node, the oil and gas business node in the original business problem information is replaced with the matched real business node, and then the business problem text is generated based on the original business problem information after the node replacement.

[0089] Furthermore, after obtaining the business problem text, the business problem text is input into the pre-acquired binary classification verification model to determine the degree of matching between the text format of the business problem text and the standard text format of oil and gas business. If the degree of matching reaches the preset matching threshold, the business problem text is output; otherwise, the process returns to step S120 to re-extract the oil and gas business node.

[0090] In this embodiment, the standardized question is input into a binary classification model trained based on historical standardized correct data to determine whether the question conforms to the standardized format of oil and gas business. If the verification pass rate reaches a preset matching threshold, such as a verification pass rate ≥ 95%, the business question text is directly output; otherwise, the process returns to step S120 to re-identify the business question text node to achieve automatic verification.

[0091] For issues that fail automatic verification, they are pushed to the manual review terminal for manual correction. The correction results are then synchronously fed back to the general information extraction model and the real node vector library for iterative optimization of the model.

[0092] In some embodiments, when a new real business node is received, a real business semantic vector corresponding to the new real business node is generated through a semantic embedding vector model, and then input into the real node vector library.

[0093] like Figure 8 and Figure 9 As shown, in this embodiment, each business node s identified in the original question text is...i Replace with the real node t matched in step S140 j While keeping the rest of the semantics in the question text unchanged, a standardized oil and gas business question is generated. For example, the original question text "How to deal with severe drill bit wear during drilling?" is replaced with "How to deal with severe drill bit wear during drilling?".

[0094] Finally, the standardized question text is output to the oil and gas business system for further processing, such as work order allocation and intelligent Q&A. At the same time, log information such as the original node, matching node, and similarity value is recorded for subsequent traceability and model optimization.

[0095] The oil and gas business node identification and matching method in the above embodiments of this application adopts a two-layer labeling system of primary nodes and secondary sub-nodes, such as "drilling - drill bit replacement", "oil production - wellhead pressure regulation", and "refining - fractionation tower maintenance", to ensure that the accuracy of the labeling can adapt to business needs.

[0096] The model accurately identifies complex scenarios such as colloquial expressions, like "the drill pipe can't be turned," abbreviations like "the gathering and transportation pipe is leaking," and multi-node mixed expressions like "abnormal wellhead pressure was found during oil production equipment maintenance."

[0097] When training the model, a two-step approach of pre-training and fine-tuning is adopted. First, a general information extraction model is pre-trained using general text data and cross-validation is used to ensure the model's recognition accuracy. Then, oil and gas business labeled data is used for fine-tuning to optimize the model's attention mechanism, so that it focuses on oil and gas business-related terms and node features.

[0098] To address the terminology characteristics of oil and gas business nodes, the word segmentation strategy of the general information extraction model was adjusted, and custom word segmentation was performed on oil and gas industry-specific terms to avoid semantic loss caused by the splitting of oil and gas terms. The vector dimension was also adjusted to 768 dimensions to ensure semantic capture capabilities while balancing storage costs and retrieval efficiency.

[0099] Meanwhile, the vector library supports a real-time update mechanism. When a new node is added to the oil and gas business, a new vector is automatically generated through the general information extraction model and added to the vector library to ensure the timeliness of matching.

[0100] The oil and gas business node identification and matching method of the above embodiments of this application can more accurately understand the semantics of business problems compared with traditional keyword matching methods, effectively solve the problems of missed identification and misidentification, significantly improve processing efficiency, reduce manpower costs, meet the real-time response requirements of business, and has strong robustness. For complex problem scenarios, it can identify and match nodes in the entire oil and gas business scenario, with high consistency of standardized results. The better matching accuracy and processing efficiency also improve the user experience.

[0101] The oil and gas business node identification and matching method provided in this embodiment of the invention includes: uniformly converting the original business problem information received from the oil and gas business system into business problem text with a consistent text format; parsing the business problem text using a pre-acquired general information extraction model to extract oil and gas business nodes from the business problem and outputting a node set formed by multiple oil and gas business nodes; calling a pre-acquired semantic embedding vector model to convert each oil and gas business node in the node set into a high-dimensional business semantic vector with a vector dimension of more than a preset dimension; performing matching calculations between the high-dimensional semantic vector and the real business semantic vector formed by a pre-constructed set of real business nodes, and selecting the real business node in the set of real business nodes that reaches the similarity threshold and has the highest similarity threshold as the matching node of the corresponding oil and gas business node.

[0102] The oil and gas business node identification and matching method provided in this invention can accurately filter out the real business nodes that best match the oil and gas business nodes, thereby achieving accurate understanding of the semantics of business issues received by the oil and gas business system, improving the matching success rate and matching efficiency, and reducing business operation costs.

[0103] In addition to the above method embodiments, the present invention also provides, for example, Figure 10 An electronic device is shown, including a processor 201 and a memory 202, wherein the memory 202 stores instructions.

[0104] The processor 201 can call instructions in the memory 202 to execute the oil and gas business node identification and matching method of any of the above embodiments of the present invention.

[0105] The oil and gas business node identification and matching method provided in the above embodiments of the present invention includes: uniformly converting the original business problem information received from the oil and gas business system into business problem text with a consistent text format; parsing the business problem text through a pre-acquired general information extraction model to extract oil and gas business nodes in the business problem and outputting a node set formed by multiple oil and gas business nodes; calling a pre-acquired semantic embedding vector model to convert each oil and gas business node in the node set into a high-dimensional business semantic vector with a vector dimension of more than a preset dimension; performing matching calculations between the high-dimensional semantic vector and the real business semantic vector formed by a pre-constructed set of real business nodes, and selecting the real business node in the set of real business nodes that reaches the similarity threshold and has the highest similarity threshold as the matching node of the corresponding oil and gas business node.

[0106] The electronic device provided in this embodiment of the invention can accurately filter out the real business nodes that best match the oil and gas business nodes by implementing the above method, thereby achieving accurate understanding of the semantics of business issues received by the oil and gas business system, improving the matching success rate and matching efficiency, and reducing business operation costs.

[0107] Furthermore, the electronic device provided in the embodiments of the present invention may also include a communication interface 203 and a bus 204, wherein the processor 201, the memory 202 and the communication interface 203 are electrically connected through the bus 204.

[0108] The memory 202 may include high-speed random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 203 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc. The bus 204 can be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 10 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0109] Processor 201 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of processor 201 or by instructions in software form. The processor 201 can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly manifested as execution by a hardware decoding processor, or execution by a combination of hardware and software modules in the decoding processor. The software module can reside in a mature storage medium in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory 202. The processor 201 reads the information in memory 202 and, in conjunction with its hardware, completes the steps of the method described in the foregoing embodiments.

[0110] This invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. The computer-readable storage medium stores instructions that, when executed on a computer, cause the computer to perform the steps of the above-described oil and gas business node identification and matching method.

[0111] The computer-readable storage medium provided in this embodiment of the invention stores data and computer-executable instructions for the above-described oil and gas business node identification and matching method. The oil and gas business node identification and matching method provided in the above-described embodiment of the invention includes: uniformly converting the original business problem information received from the oil and gas business system into business problem text with a consistent text format; parsing the business problem text through a pre-acquired general information extraction model to extract oil and gas business nodes in the business problem and outputting a node set formed by multiple oil and gas business nodes; calling a pre-acquired semantic embedding vector model to convert each oil and gas business node in the node set into a high-dimensional business semantic vector with a vector dimension of more than a preset dimension; performing matching calculations between the high-dimensional semantic vector and the real business semantic vector formed by a pre-constructed set of real business nodes, and selecting the real business node in the set of real business nodes that reaches the similarity threshold and has the highest similarity threshold as the matching node of the corresponding oil and gas business node.

[0112] The computer-readable storage medium provided in this embodiment of the invention can accurately filter out the real business nodes that best match the oil and gas business nodes by implementing the above method, thereby achieving accurate understanding of the semantics of business issues received by the oil and gas business system, improving the matching success rate and matching efficiency, and reducing business operation costs.

[0113] This application also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program that initializes the following method steps: In step S110, the original business problem information received from the oil and gas business system is uniformly converted into business problem text with a consistent text format.

[0114] In step S120, the business problem text is parsed using a pre-acquired general information extraction model, oil and gas business nodes in the business problem are extracted, and a node set formed by multiple oil and gas business nodes is output.

[0115] In step S130, the pre-acquired semantic embedding vector model is invoked to convert each oil and gas business node in the node set into a high-dimensional business semantic vector with a vector dimension of more than a preset dimension.

[0116] In step S140, the high-dimensional semantic vector is matched with the real business semantic vector formed by the pre-constructed set of real business nodes. The real business nodes that reach the similarity threshold and have the highest similarity threshold in the set of real business nodes are selected as the matching nodes of the corresponding oil and gas business nodes.

[0117] The computer program product provided in this embodiment of the invention can accurately filter out the real business nodes that best match the oil and gas business nodes by implementing the above method, thereby achieving accurate understanding of the semantics of business issues received by the oil and gas business system, improving the matching success rate and matching efficiency, and reducing business operation costs.

[0118] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0119] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, 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. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0120] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for identifying and matching oil and gas business nodes, characterized in that, include: S1: Convert the raw business problem information received from the oil and gas business system into business problem text with a consistent text format; S2: Parse the business problem text using a pre-acquired general information extraction model, extract the oil and gas business nodes in the business problem, and output a node set formed by multiple oil and gas business nodes; S3: Invoke the pre-acquired semantic embedding vector model to convert each of the oil and gas business nodes in the node set into a high-dimensional business semantic vector with a vector dimension of more than a preset dimension; S4: Perform matching calculations between the high-dimensional semantic vector and the real business semantic vector formed by the pre-constructed set of real business nodes, and select the real business node in the set of real business nodes that reaches the similarity threshold and has the highest similarity threshold as the matching node of the corresponding oil and gas business node.

2. The oil and gas business node identification and matching method according to claim 1, characterized in that, The general information extraction model is obtained through the following steps: The original general information extraction model was fine-tuned and trained using a pre-acquired oil and gas labeled dataset to obtain the fine-tuned general information extraction model. The oil and gas labeled dataset is a dataset that includes multiple pre-labeled text labeling nodes for oil and gas business scenarios. The oil and gas labeled dataset is a dataset that includes multiple first-level nodes and multiple second-level sub-nodes, and each first-level node includes multiple second-level sub-nodes.

3. The oil and gas business node identification and matching method according to claim 2, characterized in that, The method further includes: The semantic embedding vector model is segmented and adjusted using a pre-acquired oil and gas labeled dataset. This enables the fine-tuned semantic embedding vector model to recognize and retain oil and gas terms in the business question text.

4. The oil and gas business node identification and matching method according to claim 3, characterized in that, The preset dimension is 768.

5. The oil and gas business node identification and matching method according to claim 1, characterized in that, The original business problem information includes business problem information in text format and business problem information in audio format.

6. The oil and gas business node identification and matching method according to claim 1, characterized in that, The steps of the general information extraction model parsing the business question text, extracting oil and gas business nodes from the business question, and outputting a node set formed by multiple oil and gas business nodes include: The business problem text is input into the general information extraction model to parse the semantics of the business problem text and extract the explicit oil and gas business nodes and implicit oil and gas business nodes in the business problem text. All the extracted explicit oil and gas business nodes and all the implicit oil and gas business nodes are output in the form of the node set.

7. The oil and gas business node identification and matching method according to claim 1, characterized in that, The steps for matching and calculating the high-dimensional semantic vector with the real business semantic vector formed by the pre-constructed set of real business nodes include: The set of real business nodes is constructed by pre-acquiring a real node vector library of oil and gas business. The semantic embedding vector model is used to convert the pre-acquired set of real business nodes into real business semantic vectors, and a real node vector library including all the real business semantic vectors is constructed. Establish a vector index structure for indexing the real business semantic vectors, and store the vector index structure and the real node vector library in a distributed database; The high-dimensional semantic vector is input into the real node vector library, and the similarity between the high-dimensional semantic vector and each real business semantic vector is calculated through the vector index structure matching.

8. The oil and gas business node identification and matching method according to claim 1, characterized in that, The method further includes: Upon receiving a newly added real business node, the semantic embedding vector model generates a real business semantic vector corresponding to the newly added real business node, and inputs it into the real node vector library.

9. The oil and gas business node identification and matching method according to claim 1, characterized in that, After obtaining the actual business node that matches the oil and gas business node, the method further includes: Replace the oil and gas business node in the original business problem information with the matched real business node; Based on the original business problem information after the node replacement, the business problem text is generated.

10. The oil and gas business node identification and matching method according to claim 9, characterized in that, The method further includes: After obtaining the business problem text, the business problem text is input into the pre-acquired binary classification verification model to determine the degree of matching between the text format of the business problem text and the standard text format of oil and gas business. If the degree of matching reaches the preset matching threshold, the business problem text is output; otherwise, the process returns to step S2 to re-extract the oil and gas business node.

11. The oil and gas business node identification and matching method according to claim 1, characterized in that, After step S2 and before step S3, the method further includes: Based on the pre-acquired list of oil and gas business nodes, the node set is filtered to remove erroneous nodes and fill in the missing fields of ambiguous nodes, thus generating the filtered node set.

12. The oil and gas business node identification and matching method according to claim 1, characterized in that, In step S4, the method further includes: If the similarity of the high-dimensional semantic vector that has the highest similarity to the real business semantic vector is less than the similarity threshold, then the matching range is expanded to the semantic vectors of the associated nodes of the real business semantic vector, and the similarity between the high-dimensional semantic vector and the semantic vectors of the associated nodes is calculated. If the similarity between the high-dimensional semantic vector and the semantic vector of the associated node still does not reach the similarity threshold, then a preset oil and gas business rule engine is invoked to forcibly match the high-dimensional semantic vector with the real business semantic vector based on the oil and gas business rules, thereby obtaining the real business node that matches the oil and gas business node.

13. An electronic device, characterized in that, The electronic device includes: a processor and a memory, wherein the memory stores instructions; The processor invokes the instructions in the memory to cause the electronic device to implement the oil and gas business node identification and matching method as described in any one of claims 1 to 12.

14. A computer-readable storage medium storing instructions thereon, characterized in that, When the instruction is executed by the processor, it implements the oil and gas business node identification and matching method as described in any one of claims 1 to 12.

15. A computer program product, characterized in that, It includes a computer program that, when executed by a processor, implements the oil and gas business node identification and matching method as described in any one of claims 1 to 12.

Images