15 results about "Noun" patented technology

The invention discloses a news standard term detection and correction method based on a large language model, which is characterized in that a generator-detector cooperative training architecture is adopted, a Qwen2.5-7B-Instruct model is used as a generator, a detector module is assisted to finely screen implicit vectors, the generator is guided to avoid excessive sensitivity and false change, and the accuracy of the news standard term detection and correction is improved. And the problem of news expression deformity caused by improper use of words, nonstandard language expression and noun errors in the news manuscript is effectively solved. Compared with the prior art, the method has the advantage that the problem of news expression deformity caused by improper use of words, nonstandard language expression and political noun errors in news manuscripts is solved. Experimental results show that the method obtains an ideal F0.5 index on the news term correction task, and shows that the large language model has a wide application prospect in the fields of text quality control and intelligent editing.

The application provides a search and recommendation method based on a tourism knowledge graph. The application comprises: in a label preparation stage, extracting nouns and corresponding characteristics from a tourism site introduction and comments according to rules and processing the nouns and the corresponding characteristics into labels; in a model training stage, transmitting the labeled data into a PEGASUS model pre-trained by LCSTS for training, and then transmitting unlabeled data and debugging; in a search construction stage, associating the graph with a source site and storing the graph in a database to construct a visual module and a query transmission module of a system; and in a recommendation construction stage, constructing a recommendation calculation module of the system. The application has the following advantages: 1. The labels retain the original meaning and the superior-inferior relationship, which facilitates future classification of the graph; 2. Synonyms are processed according to the principle of majority covering minority, so that the labels are generated as uniformly as possible, long-term work is reduced, and the growth of traditional inverted index vocabulary is avoided; 3. Site reputation feature retrieval is more in line with user needs than text retrieval; and 4. Compared with traditional recommendation algorithms, the application is more interpretable.

The invention discloses an intelligent inspection work order quality inspection method and system based on semantic features, and relates to the technical field of inspection work order quality inspection.The method comprises the steps that historical inspection work order data are read, and an identification noun candidate set is recognized; calculating a synonymous expression updating frequency set of the identification noun candidate set; dividing the identification noun candidate set into multiple types of identification noun candidate sets according to the size of the synonymous expression updating frequency set; and connecting an inspection work order quality inspection system, performing priority rule quality inspection on the identification noun set of the current inspection work order, and outputting an identification noun quality inspection return result. The technical problems that in the prior art, due to the fact that a single quality inspection rule is adopted for inspection work order identification nouns, expression changes of nouns of different active degrees cannot be adapted, the quality inspection efficiency is low, and the accuracy is insufficient are solved, and the technical effects that the nouns of different active degrees are accurately adapted, and the quality inspection efficiency and accuracy are improved are achieved.

The present disclosure provides a method and apparatus, electronic device, and medium for correcting terminology errors, relating to the field of artificial intelligence, and in particular, to the fields of natural language processing, data mining, and machine learning. The implementation scheme comprises: performing text segment identification on a text to be corrected to obtain a first text segment, wherein the first text segment is shorter than the text to be corrected and contains at least one terminology to be identified as having errors; obtaining a first terminology based on the first text segment, wherein the first terminology is a terminology identified as having errors; and determining a second terminology based on the first terminology to replace the first terminology.

A method, apparatus, and computer-readable medium for domain-specific execution of natural language queries, the method including receiving a natural language query comprising a plurality of query terms, identifying nouns in the plurality of query terms, querying an entity index based on the nouns to identify class entities corresponding to the nouns and class attributes corresponding to the class entities, generating structured query language (SQL) queries corresponding to the natural language query based on providing the class entities and the class attributes as context to a large language model (LLM), transforming the SQL queries into domain specific language (DSL) queries based on the SQL terms, additional terms semantically related to the SQL terms, and DSL mapping rules configured to map SQL to DSL, executing the DSL query on an enterprise data catalog to identify results corresponding to the natural language query, and transmitting the results in the user interface.

The application discloses a semantic feature-based intelligent inspection method and system for an inspection work order, and relates to the technical field of inspection work order inspection. The method comprises the following steps: reading historical inspection work order data and identifying a set of identified noun candidates; calculating a set of synonymous expression update frequencies of the set of identified noun candidates; dividing the set of identified noun candidates into multiple sets of identified noun candidates according to the sizes of the set of synonymous expression update frequencies; connecting an inspection work order inspection system; performing priority rule inspection on a set of identified nouns of a current inspection work order; and outputting an identified noun inspection return result. The application solves the technical problem that a single inspection rule is used for identified nouns of an inspection work order in the prior art, which cannot adapt to the expression changes of nouns with different activity levels, resulting in low inspection efficiency and insufficient accuracy. The application achieves precise adaptation to nouns with different activity levels, and improves the inspection efficiency and accuracy.

The invention discloses a method for solving mathematical problems through a machine. The method comprises the steps that sentences about noun objects in the mathematical problems are converted into direct corresponding formulas, contained formulas and / or formulas inferred from a knowledge graph; forming a formula association diagram according to the common variables in the plurality of formulas; and the value of each variable is calculated according to the formula association diagram, and self debugging can be carried out. And then interrogative sentences and corresponding answers are generated by utilizing the formulas and the variable values of the formulas to carry out interactive heuristic teaching with students.

The invention relates to a large language model Auto-CoT method based on sample iterative extraction. The method comprises the following steps: firstly, extracting a problem entity from a medical problem as an initial value; then, a large language model is used for generating noun entities with similar semantics with the entities, and semantic similarity scores between the noun entities and the problem entities are calculated; thirdly, searching entity nodes similar to the noun entities in the medical knowledge graph, and calculating entity similarity scores; when the entity similarity score is lower than 0.5, the similarity degree is low; when the value is higher than 0.5, the similarity degree is high. Therefore, entity nodes with similarity scores higher than 0.5 are used as candidate samples. And finally, organizing the node with the highest score and the related content into a sample and adding the sample into Auto-CoT so as to guide the generation of a large language model, thereby enhancing the accuracy and interpretability of the answer. A specific experiment verifies that the method shows a remarkable effect in the aspect of improving the medical query processing capability and interpretability of the large language model, and the performance of an intelligent question answering system in the medical field is remarkably improved.

The present invention relates to the field of artificial intelligence technology, and discloses a rapid annotation method based on an artificial intelligence large model, comprising: step 1, extracting entity sets and relationship sets in a specific field through a document set in a target field, wherein entity recognition technology is used to identify field-related nouns, terms and keywords in the document to form an entity set, and relationship extraction technology is used to analyze the logical and semantic associations between entities to generate a relationship set, and the entity set and the relationship set are combined to construct a knowledge triple, which contains the association relationship between entities. By introducing a domain knowledge graph, the entity and relationship information extracted from the target domain document is structured into a knowledge triple, and the semantic information of the annotation is matched and optimized with the data to be annotated, so as to achieve domain adaptability and semantic consistency of the annotation result, and obtain a higher accuracy and consistency effect of the annotated data in the professional field.

The invention relates to the technical field of natural language processing, and discloses a T5 and knowledge graph data enhancement-based small sample event extraction method, which comprises the following steps of: segmenting an original text according to a preset segmentation bound character, and screening a longest fragment as a covering object; performing prediction generation and backfilling on the covering object by utilizing the T5 model after fine adjustment to obtain a preliminary enhanced text; performing similar entity replacement on nouns or pronouns except the key information in the preliminarily enhanced text based on a Chinese concept knowledge graph to obtain a new sample; and combining the new sample with the original text to form an enhanced data set, and driving the event extraction model to train. According to the method, text distribution consistency is kept by adopting generative fine tuning, semantic diversity is increased by combining knowledge graph replacement, and meanwhile, tagging information is directly inherited by protecting trigger words and arguments, so that the accuracy and generalization of event extraction under a small sample condition are effectively improved.

The invention relates to the technical field of three-dimensional models, and discloses a three-dimensional model attribute binding method, system and device, a medium and a product, and the method comprises the steps: carrying out the depth extraction of each surface in a target three-dimensional model, extracting the image features of a reference picture through a pre-trained CLIP image encoder, and carrying out the recognition of the target three-dimensional model; inputting the attribute cues, the depth map and the image features into an IP-Adapter model, performing semantic binding on the modifiers and the nouns based on text cross attention, forcing the modifiers to be associated with the nouns, promoting the associated attributes to be associated with each other semantically, and improving the user experience. Therefore, the problem of attribute leakage occurring during attribute binding of different parts of the same object is avoided, and attribute binding of different parts of the same object is achieved. Meanwhile, loss calculation is not needed, and back propagation is not needed to update variables of a potential space, so that the calculation amount is remarkably reduced.

The present invention discloses a method, device, equipment and medium for extracting terms from multilingual texts, which relates to the technical field of natural language processing. The method obtains the original texts in different languages corresponding to the same text content, translates the text content of the original texts corresponding to each language into a unified language to obtain standard texts, then performs sentence splitting and word segmentation preprocessing on each standard text to obtain standard words of different standard texts, and then performs alignment recognition on the standard words in different standard texts to confirm whether the text contents in different languages are aligned. When the recognition result is aligned, the original texts corresponding to the aligned languages are used as term extraction texts, and part-of-speech analysis and word frequency statistics are performed on the words to be processed in the term extraction texts to determine high-frequency noun vocabularies, and then correlation relationship matching is performed on the high-frequency noun vocabularies to achieve the technical effect of extracting terms from the text contents in different languages.

A system and method for generating a list of tasks from an interaction recording may include a computing device; a memory; and a processor, the processor configured to: identify, for each sentence of one or more sentences of an interaction recording, whether the sentence comprises at least two nouns and at least one verb; generate a set of actionable items using a long short-term memory (LSTM) model by: when the sentence does not comprise at least two nouns and at least one verb, deleting the sentence; and when the sentence comprises at least two nouns and at least one verb, generating an actionable item; input the actionable items to a language model to generate a list of tasks. Computer systems may use a generated list of tasks to automatically allocate work to agents of a contact center or to automatically update a status of a completed task.

This invention discloses a method and system for integrating search and enhanced generation capabilities. The method includes: S1 constructing a basic SQL set, where each basic SQL corresponds to a preset business intent and includes placeholder condition fields; S2 establishing a mapping relationship between fields and noun types; S3 performing intent analysis using a large language model to match the basic SQL; S4 performing query condition reasoning, extracting specific nouns, and performing extended word generation and compound word processing; S5 generating complete SQL according to rules; S6 executing the SQL and dynamically correcting the query results to meet context length requirements; and S7 combining the question and result and inputting them into the large language model to generate an answer. This invention does not require a vector database, adapts to small parameter models, is low-cost, and has strong real-time performance, and can be widely applied to various traditional business systems that require integrated AI dialogue capabilities.

The present application relates to the field of place name and address translation technology, and specifically discloses a place name and address translation tool and method that integrates artificial intelligence proper noun and common noun splitting. The tool adopts deep learning-based artificial intelligence technology to perform semantic analysis and contextual semantic association enhancement on the place name and address to be translated, and facilitates the proper noun and common noun splitting and type annotation of the place name and address to be translated based on the Prompt-driven LLM model. Based on a preset geographic terminology knowledge base, an iterative minimum translation unit determination is performed on each place name and address segment, and a recursive decomposition is performed on non-minimum unit segments until all segments meet the minimum translation granularity requirement. Each translation unit is subjected to transliteration or standard translation conversion, and a standard place name and address translation is output through normalization processing. This can effectively solve the problems of inaccurate semantic splitting and insufficient standardization of translated names in traditional translation, and improve the quality of place name and address translation in scenarios such as cross-border map services and international logistics.