A professional intelligent question-answering system and method for oil and gas geological survey

Abstract

The application discloses a professional intelligent question and answer system and method for oil and gas geological survey, receives a question input by a user, extracts professional vocabulary from the input question, interprets and translates the extracted professional vocabulary by using a translation model to form a popular explanation about the professional vocabulary, constructs a knowledge base in the field of oil and gas geological survey, extracts keywords in the popular explanation of the professional vocabulary, and matches related documents from the knowledge base by taking the keywords as retrieval conditions, extracts key information in the related documents, and feeds back the extracted key information to the user through a front-end interface, so that the system can automatically identify professional vocabulary in a question, expand the matching rate and matching accuracy of the output matching answer content by interpreting the professional vocabulary, and provide a seamless user experience, and both professional users and non-professional users can obtain accurate information.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of oil and gas geological survey question and answer system, and particularly relates to a professional intelligent question and answer system and method for oil and gas geological survey. BACKGROUND

[0002] In the oil and gas geological survey, professional terms are a highly specialized and technical field, involving a large number of professional vocabulary and terms, which are difficult for non-professionals to understand, affecting the efficiency of cross-disciplinary communication. With the increasing frequency of international exchanges, accurate translation and analysis of these professional vocabulary is particularly important. Therefore, the development of oil and gas geological survey professional vocabulary translation and analysis method technology is imminent, which improves the translation efficiency while ensuring accuracy.

[0003] In the field of oil and gas geological survey, there are currently various technical means to support the work of professionals, especially in information query and document management. Existing solutions mainly include:

[0004] 1. General language model direct dialogue + knowledge base query: after the user inputs the query request, the system directly dialogues through the general language model (such as BERT, GPT, etc.), extracts the keywords in the user input, and then searches the relevant documents from the knowledge base through keyword matching.

[0005] 2. Rule matching + keyword search: through pre-defined rules and keyword list, the keywords are extracted from the user input question, and then the relevant documents are searched from the knowledge base through keyword matching, which is suitable for query tasks with clear rules and fixed keywords.

[0006] Therefore, although the above solutions meet the needs of the oil and gas geological survey field to some extent, there are still some deficiencies, mainly in the following aspects:

[0007] (1) Low knowledge base matching rate: the field of oil and gas geological survey involves a large number of professional terms, which often have specific meanings and usage. For some non-professional users, they may input inaccurate keywords, and conventional language models are difficult to accurately fine-tune the keywords to form professional vocabulary. If the professional vocabulary recognition is not accurate, the subsequent translation and analysis will be wrong, affecting the overall performance of the system.

[0008] (2) Poor quality of professional vocabulary translation: due to the strong professionalism of the oil and gas geological survey field, many key vocabulary need to be further explained to be converted into a form that ordinary users can easily understand. Even if the professional vocabulary is identified, the existing system often lacks a "translation" mechanism for professional vocabulary, and may not be able to accurately translate these professional vocabulary into popular and easy-to-understand language. Moreover, professional vocabulary often has multiple meanings, and ambiguity often occurs in the translation process.

[0009] (3) User experience is poor: for non-professionals and part of the professionals, even if the system can provide relevant literature links or abstracts, but due to the lack of explanation and translation of professional terms, the effect of information acquisition is affected. SUMMARY

[0010] The purpose of the present application is to provide a professional intelligent question and answer system and method for oil and gas geological survey, to solve the technical problems in the prior art that it is difficult to accurately identify non-professional keywords, lack of professional vocabulary "translation" mechanism, and may not be able to accurately translate these professional vocabulary into popular and easy-to-understand language, due to the lack of explanation and translation of professional terms, users may still be difficult to understand the core content of the information, affecting the effect of information acquisition.

[0011] To solve the above technical problems, the present application specifically provides the following technical solutions:

[0012] A professional intelligent question and answer method for oil and gas geological survey, comprising the following steps:

[0013] Step 100, receiving a user input question, extracting professional vocabulary from the input question;

[0014] Step 200, using a translation model to explain and translate the extracted professional vocabulary, forming a popular explanation about the professional vocabulary;

[0015] Step 300, constructing a knowledge base in the field of oil and gas geological survey, extracting keywords in the popular explanation of the professional vocabulary, and matching related documents from the knowledge base with the keywords as the retrieval condition;

[0016] Step 400, extracting key information in the related documents, and feeding back the extracted key information to the user through the front-end interface.

[0017] As a preferred scheme of the present application, in the step 100, the implementation method of extracting professional vocabulary from the input question is:

[0018] Collecting related documents of oil and gas geological survey to establish a professional vocabulary library;

[0019] Capturing the context information and semantic features of the user input question, and combining with the self-defined rule base to identify the professional vocabulary in the user input question.

[0020] As a preferred scheme of the present application, the implementation step of establishing a professional vocabulary library is:

[0021] Collecting a large amount of document materials related to oil and gas geological survey, analyzing and processing the document materials, and constructing a professional vocabulary library containing oil and gas geological field specific terms and terminologies;

[0022] The explanation corresponding to each professional vocabulary is integrated into the professional vocabulary library, and the professional vocabulary in the professional vocabulary library is labeled to retain the professional vocabulary with consistent labeling in the professional vocabulary library and eliminate the professional vocabulary with inconsistent labeling.

[0023] As a preferred scheme of the present application, the implementation steps of capturing the professional vocabulary in the question input by the user in combination with the professional vocabulary library are as follows:

[0024] The question input by the user is subjected to word segmentation processing, and the question input by the user is disassembled into multiple independent words or phrases;

[0025] A stop word list is established to filter out the stop words in the question input by the user;

[0026] The result after word segmentation is subjected to part-of-speech tagging to distinguish the disassembled words into nouns, verbs or adjectives;

[0027] The words after word segmentation and part-of-speech tagging are matched with the pre-established professional vocabulary library, and the same professional vocabulary as in the professional vocabulary library is extracted from the words after word segmentation and part-of-speech tagging;

[0028] In combination with context analysis, two adjacent professional vocabularies are further combined to form an overall vocabulary, and the overall vocabulary is matched with the pre-established professional vocabulary library;

[0029] A self-defined rule library is used to further screen the extracted professional vocabulary same as in the professional vocabulary library to exclude the vocabulary belonging to the professional vocabulary but irrelevant to the current question.

[0030] As a preferred scheme of the present application, in the step 200, the translation model adopts an encoder-decoder structure, wherein the encoder encodes the input professional vocabulary into an intermediate representation, and the decoder decodes the intermediate representation into an easy-to-understand text, and the implementation manner of training the translation model by using the database is as follows:

[0031] The model of the encoder-decoder structure is selected as the translation model, and the vocabulary table in the database is divided into a training set and a validation set;

[0032] Each professional vocabulary and its popular explanation in the vocabulary table of the database are respectively converted into an input format acceptable to the translation model, and a loss function is used to calculate the difference between the output sequence of the translation model and the real sequence;

[0033] ;

[0034] wherein y i is the real popular explanation, p ian accuracy probability of the popular explanation generated by the translation model, N represents the number of samples;

[0035] The encoder and the decoder of the translation model are respectively fine-tuned based on the calculation result of the loss function until the loss value of the output sequence of the translation model meets the expectation.

[0036] As a preferred scheme of the present application, the specific implementation method of fine-tuning the encoder of the translation model based on the calculation result of the loss function is:

[0037] The professional vocabulary in the vocabulary table is input into the encoder of the translation model;

[0038] The encoder encodes the professional vocabulary into intermediate representation Z;

[0039] The intermediate representation Z generated by the encoder is compared with the expected intermediate representation Z true , and the loss value is calculated ;

[0040] Based on the calculated loss value, the parameters of the encoder are updated, .

[0041] As a preferred scheme of the present application, the specific implementation method of fine-tuning the decoder of the translation model based on the calculation result of the loss function is:

[0042] The intermediate representation generated by the encoder is input into the decoder;

[0043] The decoder generates a word sequence of the target language step by step until a complete popular explanation is generated;

[0044] The generated popular explanation y is compared with the expected popular explanation , and the loss value is calculated = , wherein y t is the real popular explanation, p t is an accuracy probability of the popular explanation generated by the translation model, and T represents the number of samples;

[0045] Based on the calculated loss value, the parameters of the decoder are updated, .

[0046] As a preferred scheme of the present application, in the step 300, the knowledge base is constructed, the explanation translation content and the document content are vectorized, the vector similarity matching algorithm is used to retrieve the most relevant document corresponding to the keyword in the professional vocabulary explanation translation content from the knowledge base, and the specific implementation method is:

[0047] Data entry: The collected data is entered into the knowledge base in a unified format, each collected data includes document title, author, abstract, keywords, full text link;

[0048] Data processing: using natural language processing techniques and machine learning algorithms to segment, extract key information and summarize the documents, and classify the collected data, add corresponding keyword tags to each data, and use keyword tags to index each document in the knowledge base;

[0049] Data storage: using word embedding technology to convert text data obtained by data processing into vector form and store in the knowledge base;

[0050] Matching query: using TF-IDF method to vectorize the document, and according to different similarity, retrieving the most relevant document corresponding to the keyword in the professional glossary explanation and translation content from the knowledge base.

[0051] As a preferred scheme of the present application, in the step 400, after retrieving the most relevant document from the knowledge base according to the user query, the implementation steps of extracting the key information in the most relevant document are:

[0052] Document preprocessing: decompose the text into words or phrases, convert to a unified text format, remove irrelevant format information, and label the part of speech of the decomposed single or short sentence;

[0053] Information recognition and extraction: using TF-IDF or TextRank to extract keywords in the most relevant document, wherein NER recognition technology parameters, device name entity are applied, the relationship between entities is analyzed, and data values and chart information in the extraction result are extracted;

[0054] Abstract generation: using extractive or generative summary method to generate document abstract;

[0055] Information structuring: organizing the extracted information into a structured form;

[0056] Information extraction: using entity recognition and relationship extraction techniques to extract key information.

[0057] In addition, the present application also provides a translation and analysis system for oil and gas geological survey professional glossary translation and analysis method, comprising:

[0058] User query operation interface: used for inputting user's questions, supporting text and voice input;

[0059] Professional glossary extraction unit, used for extracting professional glossary in the question asked by the user;

[0060] A professional vocabulary translation unit is configured to translate and interpret professional vocabulary into understandable content.

[0061] A knowledge base matching unit is configured to extract keywords from the translated content based on the professional vocabulary translation unit, vectorize the keywords and documents in the knowledge base content, and retrieve the most relevant documents from the knowledge base through vector similarity matching.

[0062] A key information extraction and generation module is configured to extract key information from the most relevant documents and generate coherent answers.

[0063] A user interface and feedback module is configured to provide an interactive interface and feedback channel for the user.

[0064] Compared with the prior art, the present application has the following beneficial effects:

[0065] (1) Improved professional term recognition and translation capabilities: By designing a professional vocabulary recognition algorithm and translation module (a fine-tuned oil and gas geological survey model), the system can accurately recognize and translate professional vocabulary in the oil and gas geological field, converting complex professional terms into understandable language, so that non-professionals can also understand the core content of the information, thereby improving the intelligibility and accuracy of the information and promoting effective communication and cooperation between different disciplines.

[0066] (2) Optimized knowledge base matching process: The system can more accurately match relevant documents from the knowledge base through keyword fine-tuning and professional vocabulary translation, form concise and clear answers, reduce mis-matching and missing-matching situations, ensure that the user obtains the most relevant and accurate information, and improve user satisfaction. BRIEF DESCRIPTION OF DRAWINGS

[0067] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only exemplary, and for those skilled in the art, other drawings can be obtained from the provided drawings without creative labor.

[0068] Figure 1 A flowchart of the professional intelligent question and answer method of the embodiments of the present application; DETAILED DESCRIPTION

[0069] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0070] As shown in Figure 1 The present application provides a professional intelligent question-answering method for oil and gas geological survey, comprising the following steps:

[0071] Step 100, receiving a user input question, and extracting professional vocabulary from the input question.

[0072] In the step 100, the implementation method of extracting professional vocabulary from the input question is as follows:

[0073] Step 101, collecting relevant documents of oil and gas geological survey to establish a professional vocabulary library.

[0074] Step 102, capturing context information and semantic features of the user input question, and combining a pre-defined rule library to identify professional vocabulary in the user input question.

[0075] In the step 102, the user inputs a question through a user inquiry operation interface, supporting text and voice input. The user inquiry operation interface transmits the user input signal to a processor (Chat model). The Chat model calls a keyword extraction unit to extract professional vocabulary from the user input question.

[0076] In order to extract professional vocabulary from the user input question, the implementation method of establishing a professional vocabulary library is as follows:

[0077] A large number of relevant documents of oil and gas geological survey are collected, including but not limited to literature library, research report, academic paper, etc. Through analysis and processing of these documents, a professional vocabulary library containing oil and gas geological field specific terms and terminologies is constructed.

[0078] In order to further realize the translation and explanation of each professional vocabulary in the next step, the explanation and translation content corresponding to each professional vocabulary is integrated into the professional vocabulary library in this step. The professional vocabulary in the professional vocabulary library is labeled to retain the professional vocabulary with consistent labels in the professional vocabulary library and eliminate the professional vocabulary with inconsistent labels.

[0079] After integrating the corresponding explanation and translation content of each professional vocabulary into the professional vocabulary library, a vocabulary table and a relationship table are formed, wherein the vocabulary table stores each professional vocabulary and the corresponding explanation and translation content of each professional vocabulary, and the relationship table stores the association relationship (including synonyms, antonyms and hyponym-hypernym relationship) between different professional vocabularies.

[0080] After constructing the professional vocabulary library, invite experts in the field to label the professional vocabularies in the vocabulary library to ensure the accuracy and consistency of the labeling, and then make a professional judgment on the professional vocabularies in the professional vocabulary library, including calculating the word frequency of each professional vocabulary, part-of-speech tagging, and extracting context information.

[0081] When capturing and identifying professional vocabularies in the user's input in combination with the professional vocabulary library, cleaning is a very important step, which ensures the accuracy and efficiency of subsequent processing. The specific implementation method of cleaning is as follows:

[0082] Segmentation: Perform segmentation on the user's input to break it down into individual words or phrases. This step is particularly important for Chinese, as Chinese does not have clear word boundaries. Common Chinese segmentation tools include jieba, HanLP, etc.

[0083] Remove stop words: Establish a list of stop words to filter out these words. Stop words are those that frequently appear in text but do not help understand the content of the text, such as "of", "is", "in", etc. These words are usually not kept as keywords.

[0084] Part-of-speech tagging: Perform part-of-speech tagging on the segmented results to better understand the role of each word in the sentence. Common part-of-speech tagging tools include Stanford POS Tagger, NLTK, spaCy, etc.

[0085] Match professional vocabulary library: Match the results of segmentation and part-of-speech tagging with the pre-established professional vocabulary library. Only those words that exist in the professional vocabulary library will be retained, and this step ensures that only professional vocabularies related to oil and gas geological survey are retained.

[0086] Context analysis: Even if some words exist in the professional vocabulary library, further consideration of their meaning in specific context is needed. For example, the word "pressure" may have different meanings in different contexts. Through context analysis, it can be more accurately determined whether a word should be retained.

[0087] Apply rule base: Combine with pre-defined rule base to further filter the remaining vocabulary. For example, a rule can be set to exclude those words that are part of the professional vocabulary but are not relevant to the current question.

[0088] The example process is as follows:

[0089] Assume the user input question is: "How to evaluate the reservoir characteristics of oil and gas fields?";

[0090] Segmentation: How / evaluate / oil and gas fields / reservoir / characteristics;

[0091] Remove stop words: evaluate / oil and gas fields / reservoir / characteristics;

[0092] Part-of-speech tagging: verb / noun / noun / noun;

[0093] Professional vocabulary library matching: Assuming "evaluate", "oil and gas fields", "reservoir", and "characteristics" are all in the professional vocabulary library, they are all retained.

[0094] Context analysis: Confirm that "evaluate", "oil and gas fields", "reservoir", and "characteristics" are all meaningful in the current question.

[0095] Apply rule base verification: Assuming there are no further rules in the rule base to exclude these words, the final retained words are "evaluate", "oil and gas fields", "reservoir", and "characteristics".

[0096] Through the above steps, the key words can be effectively extracted from the user input question, ensuring the accuracy and efficiency of subsequent processing.

[0097] When identifying professional vocabulary in user input questions, a suitable Chat model needs to be selected for data learning. For example, using rule matching method, professional vocabulary is identified according to pre-defined rule base. Using supervised (machine) learning method, such as support vector machine (SVM), random forest (Random Forest), etc., a classifier is trained to identify professional vocabulary. Using deep learning models such as Bi-LSTM, Transformer, etc., to capture the context information and semantic features of the vocabulary, improve the recognition accuracy.

[0098] Further, based on the pre-trained general language model, the extracted professional vocabulary is fine-tuned to form the same professional vocabulary as the vocabulary table in the professional vocabulary library. The general language model is trained through a large amount of professional literature and labeled data to ensure the accuracy of the professional vocabulary formed by fine-tuning in a specific field.

[0099] Specifically, a pre-trained language model (such as BERT, RoBERTa, T5, etc.) is selected, which has been pre-trained on a large corpus and has strong language understanding and generation capabilities. On the basis of the pre-trained language model, the keywords extracted from the user input question are fine-tuned using professional data in the oil and gas geological survey field, so that they become professional vocabulary.

[0100] After selecting the Chat model that identifies professional vocabulary in the user input question, the user input question is preprocessed, including tokenization, stop word removal, and stemming operations, to ensure data cleanliness and consistency. The frequency, part of speech, and context information of the professional vocabulary in the user input question are extracted to construct a feature vector. The user input question is used to train the model, optimize the model parameters, and improve the recognition accuracy.

[0101] To verify the impact of using the Chat model to identify professional vocabulary in the user input question on the accuracy of the entire question and answer system, a test set containing 1000 oil and gas geological survey documents was used to evaluate the system without using the professional vocabulary identification algorithm and the system using the professional vocabulary identification algorithm. The experimental results show that: by using the professional vocabulary identification algorithm, the performance of the system in identifying professional vocabulary has been significantly improved, with an accuracy rate and recall rate increase of 20% and 20%, respectively, and an F1 score increase of 20%.

[0102] Without using the professional vocabulary identification algorithm: the accuracy rate of professional vocabulary identification is 70%, the recall rate is 65%, and the F1 score is 0.67. Using the professional vocabulary identification algorithm: the accuracy rate of professional vocabulary identification is 90%, the recall rate is 85%, and the F1 score is 0.87.

[0103] Step 200: Use the translation model to interpret and translate the extracted professional vocabulary to form a popular explanation of the professional vocabulary.

[0104] The explanation of the professional vocabulary is collected from user feedback, expert knowledge, academic papers, textbooks, technical reports, and journal reports, etc. These data all contain a professional vocabulary and its corresponding popular explanation, for example, "reservoir" corresponds to the popular explanation "a formation that stores oil and gas"; "geological structure" corresponds to the explanation "the structure and form of the Earth's interior".

[0105] In the step 200, the translation model uses an encoder-decoder structure, where the encoder encodes the input professional vocabulary into an intermediate representation, and the decoder decodes the intermediate representation into understandable text. The database is used to train the translation model in the following way:

[0106] A model of a selected encoder-decoder structure is selected as a translation model, and the vocabulary in the database is divided into a training set and a validation set;

[0107] Each professional vocabulary and its popular explanation in the vocabulary of the database are respectively converted into an input format acceptable by the translation model, and a loss function is used to calculate the difference between the output sequence of the translation model and the real sequence;

[0108] ;

[0109] Wherein yi is the real popular explanation, pi is the accurate probability of the popular explanation generated by the translation model, and N represents the number of samples;

[0110] Based on the calculation result of the loss function, the parameters of the encoder and the decoder of the translation model are respectively fine-tuned until the loss value of the output sequence of the translation model meets the expectation.

[0111] The specific implementation method for fine-tuning the encoder of the translation model based on the calculation result of the loss function is:

[0112] The professional vocabulary in the vocabulary is input into the encoder of the translation model;

[0113] The encoder encodes the professional vocabulary into an intermediate representation Z;

[0114] The intermediate representation Z generated by the encoder is compared with the expected intermediate representation Z true The loss value is calculated ;

[0115] Based on the calculated loss value, the parameters of the encoder are updated, .

[0116] The specific implementation method for fine-tuning the decoder of the translation model based on the calculation result of the loss function is:

[0117] The intermediate representation generated by the encoder is input into the decoder;

[0118] The decoder generates a word sequence in the target language step by step until a complete popular explanation is generated;

[0119] The generated popular explanation y is compared with the expected popular explanation The loss value is calculated = , wherein y t is the real popular explanation, p t is the accurate probability of the popular explanation generated by the translation model, and T represents the number of samples;

[0120] update parameters of the decoder based on the computed loss value, .

[0121] Based on the above, in the experimental operation, by translating the professional vocabulary into popular and easy-to-understand words, and based on the translated words, the matching document key content is retrieved from various documents to output, which can improve the accuracy of professional vocabulary translation. A test set containing 1000 oil and gas geological survey professional vocabularies was used to evaluate the system without using the professional vocabulary explanation model and the system using the professional vocabulary explanation model. Experimental results: By using the professional vocabulary explanation model, the performance of the system in professional vocabulary explanation is significantly improved, the accuracy is improved by 25%, the BLEU score and the ROUGE score are increased by 20% and 20% respectively.

[0122] Without using the professional vocabulary explanation model: the accuracy of professional vocabulary explanation is 60%, the BLEU score is 0.65, and the ROUGE score is 0.60. Using the professional vocabulary explanation model: the accuracy of professional vocabulary explanation is 85%, the BLEU score is 0.85, and the ROUGE score is 0.80.

[0123] Through step 200, the embodiment can improve the professional term recognition and translation ability. By designing a professional vocabulary recognition algorithm and a translation module (a fine-tuned oil and gas geological survey model), the system can accurately recognize and translate professional vocabulary in the field of oil and gas geology, and convert complex professional terms into popular and easy-to-understand language, so that non-professionals can also understand the core content of the information, thereby improving the intelligibility and accuracy of the information, and promoting effective communication and cooperation between different disciplines.

[0124] Step 300, construct a knowledge base in the field of oil and gas geological survey, extract keywords in the popular explanation of professional vocabulary, and match related documents from the knowledge base with the keywords as the retrieval condition.

[0125] In step 300, after the professional vocabulary is explained as popular and easy-to-understand text, a paragraph of text is formed. In order to realize the matching retrieval of the corresponding analysis document from the knowledge base, the keywords used for matching retrieval need to be selected from the explained popular and easy-to-understand text, so as to ensure the relevance, representativeness and uniqueness of the keywords, and improve the accuracy and coverage of the retrieval.

[0126] The screening principles include the following contents:

[0127] Relevance: The keywords should be closely related to the core meaning of the professional vocabulary, and can accurately reflect the meaning of the professional vocabulary. For example, the keywords extracted for "reservoir" include "reservoir", "oil and gas", "stratum", etc. These keywords directly reflect the definition of the reservoir.

[0128] Representativeness: The keywords should have high representativeness and be able to summarize the main content of the paragraph. For example, the keywords extracted from "lithology" include "lithology", "rock", "property", "type", and "mineral", which cover the main characteristics of lithology.

[0129] Uniqueness: The keywords should have a certain degree of uniqueness to avoid too general words and improve the accuracy of retrieval. The keywords extracted from "geological structure" include "geological structure", "crust", "structure", "form", and "surface", which have high uniqueness and avoid general words such as "is" and "of".

[0130] Frequency: The keywords appear frequently in the paragraph, which helps to improve the coverage of retrieval. For example, the keywords "reservoir", "oil and gas", "stratum", "porosity", and "rock" extracted from "reservoir" appear multiple times in the paragraph, ensuring the coverage of retrieval.

[0131] Based on the above, the specific implementation steps and methods of extracting keywords from the explanation sentences of professional vocabulary are as follows:

[0132] (1) Preprocessing: Perform preprocessing operations such as word segmentation, stop word removal, and stem extraction on the paragraph to ensure data consistency and reduce inventory.

[0133] (2) Extract keywords;

[0134] Method 1: TF-IDF (Term Frequency-Inverse Document Frequency): a statistical method for evaluating the importance of a word in a document or corpus.

[0135] Steps: Calculate the term frequency (TF) of each word in the paragraph; calculate the inverse document frequency (IDF) of each word in the entire knowledge base; calculate the TF-IDF value of each word; select several words with the highest TF-IDF value as keywords.

[0136] Method 2: TextRank: a graph-based ranking algorithm for extracting keywords from text.

[0137] Steps: Construct a word co-occurrence graph, with nodes representing words and edge weights representing the co-occurrence frequency between words; use the PageRank algorithm to calculate the weight of each node; select several words with the highest weight as keywords.

[0138] Method 3: Rule-based method: manually select keywords according to pre-defined rules.

[0139] Step: Define the part of speech of the keywords (e.g., noun, verb, adjective, etc.); use natural language processing tools (e.g., NLTK, spaCy) to extract the part-of-speech tags in the passage; select the words that meet the rules as keywords.

[0140] Method 4: Model-based method: use machine learning or deep learning models to automatically extract keywords.

[0141] Step: Prepare a labeled dataset containing passages and corresponding keywords; train a classification model (e.g., SVM, BERT, etc.) to classify each word as a keyword or non-keyword; use the trained model to extract keywords from new passages.

[0142] Further, in the step 300, a comprehensive, accurate, and easy-to-retrieve professional vocabulary library in the field of oil and gas geological survey is constructed, providing efficient information support for researchers and engineers.

[0143] The construction process of the knowledge base:

[0144] Determine the scope of the knowledge base: clearly define the scope of the knowledge base, such as exploration technology, geological modeling, reservoir management, etc.

[0145] Collect information: collect relevant information from academic journals, conference papers, industry reports, technical manuals, etc.

[0146] Data entry: enter the collected information into the knowledge base according to a unified format, including document title, author, abstract, keywords, full-text link, etc.

[0147] Data processing: index each document in the knowledge base to ensure quick retrieval by keywords. Classify documents and add appropriate labels to facilitate user searches based on different topics or types. Use natural language processing techniques and machine learning algorithms to segment, extract key information, and summarize documents, generating structured data in the form of questions and replies;

[0148] Review and proofread: have experts in the field review the entered documents to ensure the accuracy and reliability of the information.

[0149] Data storage: use word embedding technology to convert text data into vector form and store it in a vector database.

[0150] Continuous update: regularly update the knowledge base by adding new research results and industry trends to ensure its timeliness and completeness.

[0151] User feedback: collect user feedback on the use of the knowledge base and continuously optimize the system's functionality and user experience.

[0152] Matching algorithm: Use vector similarity matching algorithm to retrieve the most relevant documents from the knowledge base. Glossary: A dynamically updated glossary or database that stores the relationship between professional terms and their corresponding explanations.

[0153] Through experiments, it is found that the use of a glossary translation database significantly improves the system's performance in glossary translation, with an accuracy rate increase of 20%, a recall rate increase of 20%, and a precision rate increase of 15%.

[0154] This embodiment can more accurately match relevant documents from the knowledge base through keyword fine-tuning and professional glossary translation, form concise and clear answers, reduce mis-matching and missing-matching situations, ensure that users obtain the most relevant and accurate information, and improve user satisfaction.

[0155] Accurate term translation reduces decision-making errors caused by term understanding errors, avoids unnecessary rework and cost increases. By automatically extracting and summarizing non-standardized documents, the integrity and accuracy of the knowledge base are ensured, thereby improving the reliability of decisions based on these data.

[0156] Step 400: Extract key information from relevant documents and feed the extracted key information to the user through the front-end interface.

[0157] (1) Key information extraction: Use natural language technology (publicly known NLP technology) to extract key information from relevant documents, including technical parameters, experimental results, and experimental conclusions. NLP technology can extract core ideas or conclusions, abstracts, keywords, and topics from documents. Implementation steps:

[0158] Document preprocessing: Break down the text into words or phrases, convert to a unified text format, remove irrelevant format information, and label the word class, such as "evaluation / oil and gas field / reservoir", word class labeling: verb / noun / noun.

[0159] Information recognition and extraction: Use TF-IDF or TextRank to extract keywords. Apply NER recognition technology to identify technical parameters, device names, etc. Analyze the relationship between entities. Extract data values and chart information from experimental results.

[0160] Abstract generation: Use extractive or generative summary methods to generate document abstracts. Ensure that the abstract contains key technical parameters, experimental results, and conclusions.

[0161] Information structuring: Organize the extracted information into structured forms such as tables and lists.

[0162] Front-end interface display: Design a user-friendly front-end interface. Highlight key information and provide options to view detailed content.

[0163] (2) Vector Similarity Matching: Vectorize the documents using the TF-IDF method and correct according to different similarities to retrieve the most relevant documents from the knowledge base.

[0164] (3) Information Extraction: Use entity recognition and relation extraction techniques to extract key information.

[0165] Named Entity Recognition (NER) and Relation Extraction (RE) techniques are publicly known technologies that have been widely researched and applied in academia and industry, and there are many open source tools and frameworks available, such as spaCy, Stanford NLP, BERT, etc. The working principle: based on predefined rules and patterns to identify entities; use machine learning algorithms (such as CRF, HMM, SVM, etc.) to train models, learn entity features through labeled data sets; use neural networks (such as BiLSTM-CRF, BERT, etc.) for entity recognition, relation extraction, these models can capture more complex context information.

[0166] Implementation steps: first, perform word segmentation, stop word removal, stem extraction, and morphological restoration on the document; use NER and RE models to identify entities (technical parameters, device names, etc.) and relationships (causal relationships, comparative relationships, etc.) between entities in the document; then organize the extracted information into a structured form, such as tables, lists, etc.; finally, display the information to the user through the front-end interface.

[0167] (4) Generate Answers: Generate coherent answers based on the extracted key information to ensure the accuracy and readability of the answers. Use natural language generation techniques to generate user-friendly answers.

[0168] (5) User Interface Feedback: Feedback the generated answers to the user through the front-end interface. Provide user feedback channels to collect user opinions and continuously optimize system performance.

[0169] Among them, NLP technology and entity recognition and relation extraction technology are two different methods, but they can be used complementarily to ensure that the extracted information is both comprehensive and accurate. The key information extracted by these two technologies may overlap, but may also cover different aspects: NLP technology is mainly used to understand and process natural language text, and can extract more generalized information such as main points, summaries, sentiment analysis, etc. Entity recognition and relation extraction technology focuses on identifying specific entities in the text and the relationships between these entities.

[0170] The complementarity is that they can both extract some same key words or entities, but the NLP technology focuses more on overall understanding and summary generation, providing overall understanding of the text; while the entity recognition and relationship extraction technology focuses more on specific details and relationship identification, which can extract specific technical parameters and experimental results mentioned in the summary.

[0171] In the user satisfaction survey, the user satisfaction of using the system of the application is improved by 25%. The user's use frequency is significantly increased. In a one-month observation period, the average daily use of the system of the application is 150 times, and the average daily use of the prior art is 100 times.

[0172] The translation system is integrated with the knowledge base, quickly and accurately parses professional terms, reduces the time spent on understanding and interpreting terms, so that the team can reach consensus faster and promote the project, and improves the understanding threshold of non-professionals and the user-friendliness of the system.

[0173] In addition, the application also provides a translation and analysis system for oil and gas geological survey professional vocabulary translation and analysis method, comprising:

[0174] A user inquiry operation interface is used to input user questions, supporting text and voice input;

[0175] A professional vocabulary extraction unit is used to extract professional vocabulary in the user's question;

[0176] A professional vocabulary translation unit is used to translate and explain the professional vocabulary into understandable content;

[0177] A knowledge base matching unit extracts keywords based on the content translated and explained by the professional vocabulary translation unit, vectorizes the keywords and the documents of the knowledge base content, and retrieves the most relevant documents from the knowledge base through vector similarity matching;

[0178] A key information extraction and generation module extracts key information in the most relevant documents to the user query and generates a coherent answer;

[0179] A user interface and feedback module provides a user interaction interface and a feedback channel.

[0180] The embodiment provides an efficient and coherent translation process, including user input processing, professional vocabulary recognition, vocabulary translation, document matching, and result generation. When the user asks a question, the system can automatically identify the professional vocabulary in the question, and expand the matching rate and accuracy of the output matching answer content by explaining the professional vocabulary, providing a seamless user experience. Both professional users and non-professional users can obtain accurate information.

[0181] Therefore, the present embodiment can achieve the following technical effects:

[0182] (1) Improve the ability to identify and translate professional terms: By designing a professional vocabulary recognition algorithm and translation module (fine-tuned oil and gas geological survey model), the system can accurately identify and translate professional terms in the field of oil and gas geology, converting complex professional terms into simple and understandable language, so that non-professionals can also understand the core content of the information, thereby improving the intelligibility and accuracy of the information, and promoting effective communication and cooperation between different disciplines.

[0183] (2) Optimize the knowledge base matching process: The system can more accurately match relevant documents from the knowledge base through keyword fine-tuning and professional vocabulary translation, form concise and clear answers, reduce mis-matching and missing-matching situations, and ensure that users obtain the most relevant and accurate information, thereby improving user satisfaction.

[0184] (3) Reduce error rate and improve work efficiency: Accurate term translation reduces decision-making errors caused by term understanding errors, avoids unnecessary rework and cost increases. By automatically extracting and summarizing non-standardized documents, the integrity and accuracy of the knowledge base are ensured, thereby improving the reliability of decisions based on these data.

[0185] (4) Improve user experience: The translation system is integrated with the knowledge base, quickly and accurately analyzing professional terms, reducing the time spent on understanding and interpreting terms, so that the team can reach a consensus faster and promote the project, and improve the understanding threshold of non-professionals and the user-friendliness of the system.

[0186] The above examples are only exemplary embodiments of the present application and are not intended to limit the present application, and the protection scope of the present application is defined by the claims. Those skilled in the art can make various modifications or equivalent replacements to the present application within the spirit and protection scope of the present application, and such modifications or equivalent replacements shall also be considered to fall within the protection scope of the present application.

Claims

1. A professional intelligent question-answering method for oil and gas geological survey, characterized in that, The method comprises the following steps: Step 100, receiving a user input question, and extracting a professional vocabulary from the input question; Step 200, using a translation model to interpret and translate the extracted professional vocabulary, and forming a popular interpretation about the professional vocabulary; In the step 200, the translation model adopts an encoder-decoder structure, wherein the encoder encodes the input professional vocabulary into an intermediate representation, and the decoder decodes the intermediate representation into popular and easy-to-understand text; The implementation of training the translation model using a database is as follows: selecting a model with an encoder-decoder structure as the translation model, dividing the vocabulary table in the database into a training set and a validation set; converting each professional vocabulary and its popular interpretation in the vocabulary table of the database into an input format that can be accepted by the translation model respectively, calculating the difference between the output sequence of the translation model and the real sequence using a loss function; ； where y i is the true plain explanation, p i is the accuracy probability of the plain explanation generated by the translation model, and N represents the number of samples. based on the calculation result of the loss function, respectively fine-tuning the parameters of the encoder and the decoder of the translation model until the loss value of the output sequence of the translation model meets the expectation; Step 300, constructing a knowledge base in the field of oil and gas geological survey, extracting keywords in the popular interpretation of the professional vocabulary, and matching related documents from the knowledge base using the keywords as retrieval conditions; In the step 300, the knowledge base is constructed, the interpretation and translation content is vectorized with the document content, and the vector similarity matching algorithm is used to retrieve the most relevant documents corresponding to the keywords in the professional vocabulary interpretation and translation content from the knowledge base, and the specific implementation is as follows: Data entry: the collected materials are entered into the knowledge base according to a unified format, and each collected material includes a document title, an author, an abstract, keywords, and a full-text link; Data processing: using natural language processing technology and machine learning algorithms to segment, extract key information and summarize the documents, and classifying the collected materials, adding keyword tags corresponding to each material, and using the keyword tags to index each document in the knowledge base; Data storage: using word embedding technology to convert the text data obtained by data processing into vector form and store in the knowledge base; Matching query: using the TF-IDF method to vectorize the documents, and according to different similarities, retrieving the most relevant documents corresponding to the keywords in the professional vocabulary interpretation and translation content from the knowledge base; Step 400, extracting key information from the related documents, and feeding back the extracted key information to the user through the front-end interface.

2. The professional intelligent question-answering method for oil and gas geological survey according to claim 1, characterized in that, In the step 100, the implementation method of extracting professional vocabulary from the input question is as follows: collecting related documents of oil and gas geological survey to establish a professional vocabulary library; capturing the context information and semantic features of the user input question, and combining with a self-defined rule library to identify the professional vocabulary in the user input question.

3. The professional intelligent question-answering method for oil and gas geological survey according to claim 2, characterized in that, the implementation steps of establishing the professional vocabulary library are as follows: A large number of documents related to oil and gas geological survey are collected, and the documents are analyzed and processed to build a professional vocabulary library containing oil and gas geological field-specific terms and terminologies; The corresponding explanation translation content of each professional vocabulary is integrated into the professional vocabulary library, and the professional vocabulary in the professional vocabulary library is labeled to retain the labeled consistent professional vocabulary in the professional vocabulary library and exclude the labeled inconsistent professional vocabulary.

4. The professional intelligent question answering method for oil and gas geological survey according to claim 2 or 3, characterized in that, The implementation steps of capturing and identifying professional vocabulary in the user input question in combination with the professional vocabulary library are: Carrying out word segmentation processing on the user input question to disassemble the user input question into multiple independent words or phrases; Establishing a stop word list to filter out stop words in the user input question; Carrying out part-of-speech tagging on the segmented results to distinguish the disassembled words into nouns, verbs, or adjectives; Matching the segmented and part-of-speech tagged words with the pre-established professional vocabulary library to extract the same professional vocabulary from the segmented and part-of-speech tagged words in the professional vocabulary library; Combining the context analysis to further combine two adjacent professional vocabulary into an overall vocabulary, and matching the overall vocabulary with the pre-established professional vocabulary library; Self-defining a rule library to further screen the extracted same professional vocabulary in the professional vocabulary library to exclude the vocabulary that belongs to the professional vocabulary but is irrelevant to the current question.

5. The professional intelligent question answering method for oil and gas geological survey according to claim 1, characterized in that, The specific implementation method of fine-tuning the encoder of the translation model based on the calculation result of the loss function is: Inputting the professional vocabulary in the vocabulary table into the encoder of the translation model; The encoder encodes the professional vocabulary into intermediate representation Z; comparing the intermediate representation Z generated by the encoder with the desired intermediate representation Z true performing the comparison, calculating a loss value ; updating parameters of the encoder based on the computed loss value, .

6. The professional intelligent question answering method for oil and gas geological survey according to claim 1, characterized in that, The specific implementation method of fine-tuning the decoder of the translation model based on the calculation result of the loss function is: Inputting the intermediate representation generated by the encoder into the decoder; The decoder generates a word sequence in the target language step by step until a complete popular explanation is generated; The generated popular explanation y is compared with the expected popular explanation The loss value is calculated = where y t is the true popular explanation, p t is the accurate probability of the popular explanation generated by the translation model, and T represents the number of samples. updating parameters of the decoder based on the computed loss value, .

7. The professional intelligent question answering method for oil and gas geological survey according to claim 1, characterized in that, In the step 400, after retrieving the most relevant documents from the knowledge base in response to the user query, the implementation steps of extracting key information in the most relevant documents are: Document preprocessing: decomposing the text into words or phrases, converting to a unified text format, removing irrelevant format information, and labeling the part-of-speech of the decomposed single or short sentence; Information recognition and extraction: using TF-IDF or TextRank to extract keywords in the most relevant documents, wherein NER recognition technology parameters and device name entities are applied to analyze the relationship between entities to make the data values and chart information in the extraction results; Abstract generation: using an extractive or generative abstract method to generate a document abstract; Information structuring: Organizing extracted information into structured form; Information extraction: Using entity recognition and relation extraction techniques to extract key information.

8. A translation and analysis system based on the translation and analysis method of the professional vocabulary of oil and gas geological survey according to any one of claims 1-7, characterized in that, Comprise: User query operation interface: for inputting user's questions, supporting text and voice input; Professional vocabulary extraction unit for extracting professional vocabulary in user's questions; Professional vocabulary translation unit for translating professional vocabulary into understandable content; Knowledge base matching unit based on the content explained by the professional vocabulary translation unit, extracting keywords, vectorizing the documents in the knowledge base content, and retrieving the most relevant documents from the knowledge base through vector similarity matching; Key information extraction and generation module: extract key information from the most relevant documents to the user query and generate coherent answers; User interface and feedback module: provides user interaction interface and feedback channel.

Images