29 results about "Edit distance" patented technology

The invention discloses an equipment fault diagnosis method and device based on a dynamic knowledge graph and a large model fine tuning technology. The method comprises the following steps: firstly, identifying a core entity from multi-source heterogeneous equipment fault data through a named entity identification model for fine tuning of domain data and a relation extraction model for special fine tuning of a fault diagnosis domain corpus, mining deep semantic association, and injecting the deep semantic association into a graph database after cleaning to form an initial knowledge graph; receiving user natural language fault description, realizing term and standard entity linking through editing distance fuzzy matching and Sension-BERT semantic vector similarity calculation, and combining bidirectional retrieval and attention mechanism fusion to obtain an enhanced context; and finally, generating a structured diagnosis report containing thinking chain reasoning based on an enhanced context by utilizing a specialized fine-tuning fault diagnosis large language model. According to the method, the limitation of a traditional diagnosis method is effectively solved, high-precision and interpretable equipment fault diagnosis is realized, and the diagnosis efficiency and reliability are improved.

The invention discloses a text-to-instruction system based on fuzzy pinyin matching and parameter robust analysis. The text-to-instruction system comprises an input preprocessing module, a text-to-instruction conversion module and a text-to-instruction conversion module, the fuzzy pinyin matching module is used for uniformly transferring the standardized text and a pre-stored instruction template into a silent pinyin sequence, and constructing a weighted editing distance matrix based on a preset pinyin character replacement cost; the parameter alignment search module is used for executing beam search on the weighted editing distance matrix and outputting a plurality of candidate matching paths with the minimum cost and a corresponding parameter fragment set; the parameter robust distribution module is used for performing legality and consistency checking on parameters in the candidate matching paths, completing parameter standardization and performing error correction in combination with a parameter candidate dictionary; the instruction output module is used for encoding the template identifier and the parameter key value pair obtained through robust distribution into a structured instruction and outputting the structured instruction; according to the method, text-to-instruction conversion can be completed in real time.

This application relates to the field of fintech and discloses a dynamic hybrid parsing and deterministic verification system and method for fuzzy bond instructions. The system constructs structured data through an instruction preprocessing module; a routing decision module calculates complexity scores based on structure, semantics, and business characteristics, and generates path selection instructions using a resource-aware dual-threshold strategy; a hybrid parsing module responds to instructions, employing a rule engine to handle low-complexity tasks and a large-scale model with enhanced retrieval to handle highly ambiguous tasks; a fallback verification module performs confidence-weighted integration and logical consistency checks, and repairs anomalies through issuer set operations and edit distance matching; a mapping generation module constructs standard query statements using an abstract syntax tree; and a closed-loop optimization module dynamically updates parameters based on user interaction feedback. This invention effectively balances computational efficiency and parsing accuracy, eliminates model illusion risks, and achieves adaptive evolution of the system.

This invention relates to the fields of audio signal processing and artificial intelligence, specifically to a lyrics alignment method and electronic device based on automatic speech recognition. The lyrics alignment method based on automatic speech recognition in this application uses a pre-trained automatic speech recognition model to recognize audio samples and generate recognized text. The edit distance sequence between the recognized text and the lyrics text is calculated, and spoken segments are determined by analyzing the distance change pattern through a sliding window. These spoken segments are then replaced with silent segments, effectively eliminating spoken segments and avoiding interference from non-lyric content in the alignment process, thus laying the foundation for subsequent alignment.

The invention relates to a code error feedback method combining static analysis and dynamic analysis, and mainly solves the problems of insufficient error feedback and low efficiency in an Online Judge system: reconstructing codes of correct and error versions, including alignment and renaming of variables, and rearranging statements on the premise of not influencing semantics; static and dynamic multi-level alignment is carried out, in the static alignment, an editing script with the minimum cost is calculated through a heuristic editing distance algorithm, an alignment block and a difference list are obtained, in the dynamic alignment, compiling execution is carried out on source codes of two versions, test cases which do not pass operation are dynamically printed through gdb, and execution track information of the test cases is dynamically printed through gdb; error positioning is achieved, data flow backtracking analysis and scoring are carried out on error positions, and error feedback is generated. The method has the characteristics of high execution efficiency, high accuracy and easier understanding of generated feedback, automatic feedback of error codes can be realized, and the learning efficiency of students in programming tasks is improved.

This invention provides a Pinyin error correction method that simplifies the noise channel model used in most current Pinyin error correction algorithms by employing a real-time frequency counting method, effectively improving the efficiency and lightweighting of the algorithm. Furthermore, this invention uses a direct character substitution method instead of the traditional edit distance calculation method, avoiding frequent edit distance calculations in the Pinyin error correction algorithm. The dictionary database itself is localized to the individual user, resulting in a highly personalized, targeted, and small-scale dictionary that provides accurate candidate words. Compared to existing algorithms, this invention offers significant improvements in error checking rate, candidate word accuracy, and execution efficiency.

The invention provides a program error positioning and explanation generation method and system based on reinforcement learning, and belongs to the field of AI auxiliary programming education and natural language processing, and the method comprises the steps: S1, based on student submission records of a real teaching scene, screening out error codes and correct code pairing samples through an editing distance, and constructing a training set; s2, sampling and filtering the training set by using a pre-training language model, and constructing a sampling rejection data set to supervise and finely adjust the pre-training language model to obtain a base model; s3, constructing a multi-signal fusion reinforcement learning reward function for performing reinforcement learning training on the base model to obtain a trained error positioning and explanation generation model; and S4, inputting the question description and the error code into the trained error positioning and interpretation generation model, and generating structured feedback. According to the method, the error positioning accuracy is remarkably improved, the false drop rate is effectively reduced, and a feasible technical path is provided for intelligent programming education.

The embodiment of the specification provides a natural language query analysis and database field matching method based on a large model, which comprises the following steps: performing intent analysis and entity recognition on a user natural language query based on a large language model (LLM) to extract a potential keyword set; for each extracted keyword, binding an entity type through an STAM mechanism and constructing an entity-type mapping relationship; constructing a column description vector library, combining an approximate nearest neighbor (ANN) algorithm to realize efficient retrieval, and combining a multi-strategy fusion matching mechanism to combine the results of semantic vector matching, edit distance matching and traceability enhanced matching to form a final candidate column set; connecting a database metadata interface, analyzing information and constructing a basic structure to generate an enhanced semantic field description, and outputting in a four-tuple structure; reversing the corresponding table through column matching, judging the table structure value in combination with the information density in the table and the connectivity with other tables, eliminating redundant tables, and optimizing the database representation.

This disclosure provides a method, apparatus, program product, and related equipment for determining string similarity, relating to the field of artificial intelligence technology. The method includes: acquiring a first string and a second string; selecting a word from the first word in the first string that has not been compared before as the current word to be compared; performing semantic comparisons between the current word to be compared and the second word in the second string; if the second word has a semantic opposite to the current word to be compared, then determining the similarity between the first string and the second string to be zero; otherwise, continuing to select a first word that has not been compared before as the current word to be compared and continuing semantic comparisons; if, after traversing the first word, it is determined that there is no word in the second string with a semantic opposite to the first word, then using the edit distance similarity between the first string and the second string as the similarity between the first string and the second string. This method can improve the accuracy of string similarity determination.

The invention discloses an automatic process control system and method for a low-code platform, and relates to the technical field of computers.The method comprises the steps that source data of a user on the low-code platform are obtained according to the preset frequency, and snapshots are generated; the source data and predefined cue words are spliced to serve as large language model input; the large language model outputs structured suggestions, and the suggestions are classified and displayed for users to confirm; after the user confirms, recording related data as a training sample and adjusting the suggestion display frequency; constructing an event chain library, calculating a matching degree between an actual operation event chain and an operation event chain in the event chain library according to an editing distance algorithm, and recommending a component type; according to the method, the local network speed and computing power data of the user are obtained, the learning rate of the large language model is weighted and adjusted after normalization, training is completed in a local background, intelligentization and high efficiency of automatic process control of the low-code platform are achieved, and user operation experience and platform applicability are improved.

The invention relates to the technical field of natural language processing and information retrieval, and discloses a question and answer pair automatic generation method and device, computer equipment and a storage medium. The method comprises the following steps: firstly, extracting an initial question and answer pair from an input text through a large language model (LLM); slicing and vectorizing the original text, and constructing an original text vector database; vectorizing answers in the initial question and answer pair, matching the vectorized answers with an original text vector database, and positioning one or more original text fragments for each answer to form a traceable triple; performing quality verification on the triple by adopting a comprehensive verification mechanism based on integrity and accuracy evaluation and key information coverage calculation; and finally, de-duplication is carried out in combination with editing distance calculation and semantic similarity calculation, and high-quality question and answer pairs are output. According to the method, the problems of unstable generation quality, lack of answer basis and the like are solved, and automatic construction of the high-credibility knowledge base is realized.

Some implementations described herein relate to determining whether to modify segment(s) of responsive content, that is generated using a generative model (GM), based on a corresponding edit distance between the segment(s) of the responsive content and segment(s) of additional data. Processor(s) of a system can: receive user input associated with a client device, generate the responsive content using the GM and based on processing the user input, and determine whether to modify the segment(s) of the responsive content using the corresponding edit distance (e.g., fuzzy matching). Subsequent modification and / or processing of the responsive content can be dynamically adapted based on whether there is a match and, if there is a match, based on source(s) associated with the additional data that match. Further, the processor(s) can cause the responsive content, or modified responsive content, to be rendered at the client device.

This invention relates to the field of artificial intelligence technology, and discloses a text content comparison and matching method, apparatus, device, and medium. The method includes sequentially performing character adaptation and normalization, phonological encoding conversion, segmentation processing, fuzzy weighted edit distance calculation, peak analysis and path backtracking, and jump threshold traversal judgment on template text and speech recognition text to obtain text comparison and matching result data. In this invention, addressing the problem that existing text content comparison and matching schemes struggle to accurately identify the continuous inclusion of speech recognition text in template text, character adaptation and normalization and phonological encoding conversion can eliminate character form differences; segmentation processing and fuzzy weighted edit distance calculation can suppress accumulated errors; and peak analysis, path backtracking, and jump threshold traversal judgment are then applied. This method can be applied to the financial and medical fields, thus accurately identifying the continuous inclusion of speech recognition text in template text at the character granularity.

The invention relates to a multi-stage spelling error correction method and device for a medical image report and a medium, in a non-term error correction stage, for each text segment, a pinyin error correction knowledge base and a polyphone probability base are called, screening and replacement are carried out according to text similarity and editing distance, candidate correction for general spelling errors is generated, and the spelling error correction efficiency is improved. Obtaining a text fragment after non-term error correction; in the term error correction stage, a medical term package is called, candidate correction for term spelling errors is generated and screened through candidate interval positioning, set similarity screening and sequence similarity calculation, and text segments after term error correction are obtained; and enabling the text segments after non-term error correction and the text segments after term error correction to correspond to each other at the text position, correcting the candidates with overlapping or conflicting, and performing resolution and combination according to a preset priority rule to obtain a final error correction result. Compared with the prior art, the method has the advantages of wide range, high accuracy, high reliability and the like.

Anomalies are detected for a plurality of managed entities. Each configuration tree of a plurality of configuration trees represents configuration parameters at a corresponding managed entity, including annotations that identify a configuration that configures each parameter. For a particular configuration, a plurality of sub-trees is generated from the plurality of configuration trees. A set of weighted edit distances is computed from the plurality of sub-trees, each sub-tree representing a degree of difference between different pairs of sub-trees. The set of weighted edit distances is utilized to populate a distance matrix, and the distance matrix is used to identify an anomalous sub-tree(s) within the plurality of sub-trees for the particular configuration. In embodiments, a configuration corresponding to the anomalous sub-tree is deemed to have been abnormally applied by the particular configuration. Data of at least one managed entity and the particular configuration that is identified as being associated with the anomalous sub-tree is stored.

The invention discloses a cross-language editing distance and multi-model fused mixed word detection method and system, and belongs to the technical field of natural language processing. According to the method, firstly, preprocessing and candidate mixed word extraction are carried out on a text, a cross-language editing distance is calculated by utilizing dynamic programming, a large language model is called twice to carry out semantic scoring by adding a text window fragment without a context, then calibration weighted aggregation is carried out, and a final result is output through multi-model fusion and optimization. And high-precision recognition of Chinese-English mixed words is realized. By means of the scheme, effective combination of the structural features and the semantic features is achieved, the Chinese and English mixed words can be efficiently and accurately detected, interpretability and real-time performance are achieved, the accuracy and robustness of mixed word detection are improved, and the method can be widely applied to input methods, social platforms, network public opinion monitoring, news transmission analysis and other scenes.

This application provides a method and apparatus for translation quality classification. The method includes: acquiring source text and machine translation; inputting the source text and machine translation into a translation quality classification model, and acquiring the classification result output by the translation quality classification model; the translation quality classification model is optimized based on a first edit distance score, which is obtained based on the machine translation and the proofread translation, wherein the proofread translation is the machine translation after proofreading. The translation quality classification method and apparatus provided in this application can optimize a pre-trained language model using the first edit distance score between the machine translation and the proofread translation to obtain a translation quality classification model. This allows the translation quality classification model to output a classification result based on the input source text and machine translation, thereby improving the accuracy of translation quality classification.

The invention discloses a book name error correction method and system based on semantic embedding and editing distance fusion, and the method comprises the steps: carrying out the recall of two-stage candidate words from a candidate word bank for a to-be-corrected word, obtaining a target candidate word set, and carrying out the scoring and sorting of the candidate words in the target candidate word set; and based on the sorting result, selecting the target candidate word as an error correction result to be output. According to the technical scheme provided by the invention, firstly, the candidate words with similar fonts are ensured not to be omitted through the primary recall based on the editing distance, and then the similar candidate words are expanded through the secondary recall based on the semantic vector similarity, so that the coverage range and diversity of the candidate word set are remarkably improved; then multi-dimensional reordering is carried out on the candidate words through a comprehensive scoring model, decision making is carried out by fully combining font features, semantic relevance and context importance, and therefore the accuracy of error correction is greatly improved while the high recall rate is kept.

The invention relates to the technical field of knowledge graph construction, and discloses a railway disaster knowledge graph intelligent construction method based on a fine tuning Qwen model. The method comprises the steps of performing format unified conversion and deep cleaning on a multi-source railway disaster text to obtain sentence-level corpora, inputting the sentence-level corpora into a Qwen model, performing triple coarse extraction, performing fine tuning training by adopting an LoRA low-rank adaptation method, performing batch reasoning on full-amount corpora, and performing de-duplication standardization according to a minimum editing distance and Jaccard similarity fusion algorithm. And importing a Neo4j graph database to construct a railway disaster knowledge graph. According to the method, the automation degree of railway disaster knowledge graph construction, the knowledge extraction accuracy and the graph structure normalization are improved.

A method and a system for character-to-character modeling for next-word prediction and auto correction are provided. The method is based on character-to-character modeling. The method envisages the use of an edit distance and a hash table for creating valid candidates based on the cleaned input text, for next-word prediction and auto-correction. The next-word creation is performed when the input text ends with a space character and the auto-correction is performed when the input text ends with a letter. The edit distance and the hash table are employed to identify valid candidates and to remove invalid candidates. Furthermore, an encoder-decoder model is used to process the valid candidates, generate completed words, and subsequently create word suggestions based on the completed words. The word suggestions are subsequently optimized into highly probable candidates using a language model and a scoring and ranking algorithm.

This invention discloses a document conversion method, apparatus, device, and medium. This invention fully considers the specific context of words, and this context-aware error correction mechanism effectively avoids mechanical replacement without context, significantly reducing the false deletion rate and preserving the semantic integrity of the original document. Furthermore, this invention first recalls candidate words from a medical terminology database whose edit distance to the word to be verified is less than a distance threshold, and then calculates the semantic fit score. This two-stage strategy of "coarse screening + fine ranking" utilizes edit distance to quickly identify candidate words with similar shapes, and leverages the domain terminology database to ensure the effectiveness of candidate words within the medical semantic space. Compared to directly calculating semantic scores for all dictionary words, this method significantly reduces unnecessary semantic model inferences and improves processing efficiency; simultaneously, the domain specialization of the terminology database ensures that the recalled candidate words are highly relevant to the medical context.

This invention discloses a method and system for assessing the stability of long-term vocabulary memory based on real-time feedback, relating to the field of foreign language vocabulary learning technology. The method first collects user review and answer data, including time, answers, and results. Based on answer time thresholds, it categorizes ambiguous and regular vocabulary and records the number of reviews. The system dynamically calculates priorities based on vocabulary entry time and length, generating an initial push sequence. During the review process, it monitors cumulative answer time and error counts in real time, dynamically adjusts the sequence order, and triggers secondary review requests for frequently misspelled or ambiguous words. Based on edit distance algorithms and answer matching, it intelligently pushes confusing words for reinforcement practice to weak vocabulary. This invention generates personalized vocabulary review paths driven by real-time feedback data, significantly reducing confusion between similar words, promoting the solid formation of long-term memory, and improving learning efficiency.

This invention discloses a method and apparatus for generating negative samples based on lexical and semantic similarity in the Chinese domain. The method includes: extracting entities and all relations within one hop of each entity from a Chinese encyclopedia knowledge graph; concatenating the relations to form descriptive text for the entities; using Word2Vec trained on Chinese corpora to perform vector representation of the entities and descriptive text; performing similarity matching between the target matching entity and all entities, returning entities that meet the criteria as a set of semantic similarity candidate entities; calculating the edit distance between the target matching entity and all entities again, returning entities that meet the criteria as a set of lexical similarity candidate entities; and selecting high-confusion negative samples based on the semantic similarity candidate entity set and the lexical similarity candidate entity set. This invention combines lexical similarity and semantic similarity for negative sample selection, which can accelerate the convergence speed of the model and enable the model to train a good decision boundary.

The application discloses a similarity calculation model training method and device, equipment and medium, including: obtaining a first text and performing a text editing operation on the first text to obtain a second text; constructing corresponding training samples based on the feature vectors of the first text and the second text, and mapping the training samples to a preset text editing vector space through an initial neural network model to obtain corresponding first text editing vectors; calculating the similarity between different text editing vectors in the first text editing vectors, and constructing a loss function based on the similarity and a preset similarity label; training the initial neural network model using the loss function, and then jumping back to the step of obtaining the first text and performing the text editing operation on the first text until a preset iteration number is reached to obtain a similarity calculation model. The initial neural network is trained to obtain the similarity calculation model, so as to calculate the editing distance similarity of the text according to the text feature vector.

The invention relates to a voice recognition system fuzzy testing method based on uncertainty, and belongs to the technical field of testing. According to the method, a small amount of annotated speech is taken as a seed set, a large number of candidate samples are generated through multi-operator speech variation, and semantic offset samples are screened out by combining multi-angle semantic consistency verification of twin network similarity, BERT semantic similarity, editing distance and the like. And then comprehensively evaluating a sample risk by using multi-dimensional uncertainty indexes such as prediction entropy, variance, mutual information, sequence confidence and the like, and selecting high-uncertainty voice as key test data to trigger a potential defect. The system collects feedback information such as defect rate, average uncertainty and semantic offset in the testing process, and dynamically adjusts mutation operator weight, semantic consistency weight and uncertainty weight through an adaptive weight updating mechanism to realize closed-loop iteration. According to the invention, the effectiveness and coverage of the test sample can be continuously improved, so that the robustness of the speech recognition system can be systematically evaluated and enhanced.

The application relates to the field of text processing and discloses a text similarity detection method, device and system. The method comprises the following steps: obtaining a to-be-detected text, a comparison text and a preset comparison length value; dividing the characters in the to-be-detected text into a plurality of to-be-detected text fields; dividing the characters in the comparison text into a plurality of to-be-compared text fields based on the preset comparison length value and the to-be-detected text fields; comparing each to-be-detected text field with all to-be-compared text fields one by one; when the edit distance between the to-be-detected text field and the compared to-be-compared text field is within a preset edit distance, combining the to-be-detected text field and the compared to-be-compared text field to form a similar character comparison group and performing comparison; calculating the number of repeated characters and the number of continuous repeated characters in the to-be-detected text field, and then obtaining the text similarity of the two. The method provided by the application can efficiently and accurately locate the similar text position of the to-be-detected text in the comparison text and detect the text similarity of the two.

The application provides a building contour similarity measurement method, comprising the following steps: constructing a graph model of a building contour by converting the inflection points of the building contour into graph model vertices with area geometry and local direction feature labels; defining an edit cost of the vertices by using a geometric variation type variable, constructing an edit operation between the vertices of the graph model as an edit path, defining a sum of the edit costs as an edit distance of the edit path; and reflecting shape differences between building contours by the edit cost of the optimal edit path. The application has the beneficial effect that the optimal edit distance path problem is converted into a maximum weight clique solving problem of a vertex correlation graph based on graph theory, and a heuristic search method is designed to quickly and accurately search for the optimal edit path. The cognition result is highly consistent with the class cognition, the shape search effect is higher than that of the current mainstream measurement method, and the application can effectively support building matching, search and other application requirements.

The application provides a log anomaly detection model training method and device, the log anomaly detection model training method comprises the following steps: obtaining multiple sample log data generated by a signal system; analyzing the multiple sample log data based on an event template to obtain a first template sequence; sorting multiple log templates based on an edit distance between the multiple log templates and a word vector similarity between multiple parsed sample log data to obtain a second template sequence, adding statistical information of a first keyword and a target log template to the second template sequence to obtain a first count matrix; inputting the first count matrix into a clustering model for hierarchical clustering training, and obtaining a log anomaly detection model after the training is completed. The method can increase the dimension of the sample log data corresponding template sequence, thereby improving the training efficiency of the clustering model and the classification accuracy.

Diffing prior executions of an executable program. A computer system assigns a weighting to each entity type in a first ordered set of a first plurality of entities and a second ordered set of a second plurality of entities, including assigning different weightings to different entity types. The computer system calculates an edit distance between the first and second ordered sets based on calculating a minimal cost path through a diffing structure. Calculating the minimal cost path includes partitioning the diffing structure around intersecting entries that are identified based on (i) only permitting replacements between entities that have a common entity type, and (ii) utilizing the different weightings to identify single-entity replacements that have higher cost than adjoining multi-entity replacements. Calculating the minimal cost path also includes applying a differencing algorithm to these partitions, in order to identify corresponding segments of the minimal cost path.