Patents
Literature
Patsnap Eureka AI that helps you search prior art, draft patents, and assess FTO risks, powered by patent and scientific literature data.

8 results about "Lexical semantics" patented technology

Lexical semantics (also known as lexicosemantics), is a subfield of linguistic semantics. The units of analysis in lexical semantics are lexical units which include not only words but also sub-words or sub-units such as affixes and even compound words and phrases. Lexical units make up the catalogue of words in a language, the lexicon. Lexical semantics looks at how the meaning of the lexical units correlates with the structure of the language or syntax. This is referred to as syntax-semantic interface.

Open vocabulary semantic segmentation method based on diffusion model

The invention discloses an open vocabulary semantic segmentation method based on a diffusion model, which comprises the following steps of: according to an input image, fusing global semantics and local semantics of the image, and automatically extracting a target category set and a text prompt in the image; for each target category, processing the target category and the text prompt based on an end filler embedding adjustment strategy and a target category embedding enhancement method to obtain optimized text embedding; for each target category, obtaining a target category activation diagram through hierarchical attention fusion according to the optimized text embedding and diffusion model; and performing post-processing on each target category activation graph, and performing fusion to generate a final semantic segmentation result. According to the method provided by the invention, the high-precision segmentation mask can be generated according to the input image; and secondly, due to the characteristics of open vocabularies, the model can extract any category in the image and generate a corresponding segmentation result.
Owner:BEIJING INFORMATION SCI & TECH UNIV

Court hearing verification method based on multi-mode Transform

The invention discloses a court trial verification method based on a multi-modal Transform, and the method comprises the following steps: collecting multi-modal information data, and carrying out the preprocessing and feature extraction; inputting the data to a modal encoder for dimension adjustment and standardization processing; inputting to a fusion module of the TriMERT model to realize semantic alignment and feature fusion; inputting the data to a reasoning module, and carrying out multi-layer feature coding and context modeling; comparing with the vocabulary semantic features, generating verification information, and integrating and judging; calculating a modal contribution score, and performing feature weighting and key frame labeling on each modal; and integrating the verification judgment result and the reasoning path indication, and displaying in a structured report form. According to the method, a multi-mode Transform modeling method is adopted, court hearing semantic verification and reasoning path generation are achieved, and the method has the advantages of being high in fusion degree, accurate in recognition and high in interpretability.
Owner:JINAN JIASHENG NETWORK TECHNOLOGY CO LTD

Multimodal alzheimer's disease early screening algorithm based on large language model

The application relates to the technical field of medical information, and specifically discloses a multi-modal Alzheimer's disease early screening algorithm based on a large language model. The algorithm first collects speech data and demographic information of participants. The speech data is processed through noise suppression, voice enhancement and speaker separation, and is transcribed into descriptive text. Then, the eGeMAPS feature set is extracted from the speech data and time-series spliced, and combined with the demographic information to form a multi-modal feature matrix. Key features are obtained using an attention pooling module, and feature classification is performed using a support vector machine to distinguish potential Alzheimer's disease patients from healthy individuals. The classification results, descriptive text and demographic information are input into the large language model, and zero-shot thinking chains are used to analyze four language barriers: reduced information content, lexical semantic defects, poor referential coherence and language loss of psychological state, and a comprehensive screening report is finally generated. The algorithm improves the accuracy of early diagnosis by combining audio and semantic features.
Owner:SICHUAN UNIV

Text vocabulary recommendation method, system and equipment and medium

The invention discloses a text vocabulary recommendation method, system and device and a medium, and the recommendation method comprises the steps: obtaining a first text input content of a user and a vocabulary semantic map of the user; performing text analysis on the first text input content to obtain a target keyword, text context information and text emotion information; according to the vocabulary semantic map, performing context matching and knowledge query processing on the target keyword and the text context information to obtain context expansion information corresponding to the first text input content; and inputting the text emotion information, the context expansion information and the first text input content into a pre-training language model for text generation to obtain a target vocabulary, and providing the target vocabulary to the user. According to the method, the individuation degree and accuracy of text vocabulary generation can be effectively improved, and the learning efficiency and learning experience of the user can be improved. The invention relates to the technical field of natural language processing.
Owner:SHENZHEN UNIV

Intelligent vocabulary learning method and system based on lexical semantic web and age grading

This invention provides an intelligent vocabulary learning method and system based on a semantic web and age-based classification. The method includes: constructing a dictionary containing semantic-level data and root-level data; semantic-level data including root association fields and difficulty quantification fields; and root-level data containing root-level shared metadata. It also includes establishing a mapping relationship between roots and grammatical variants based on the root association fields; recording learning behavior data by semantic unit; evaluating user mastery using a multi-dimensional mastery model; determining the English age representing the user's vocabulary learning level; dynamically matching vocabulary to be learned based on English age and difficulty quantification fields; calculating comprehensive mastery based on the multi-dimensional mastery model; calculating review intervals based on comprehensive mastery; dynamically adjusting the proportion of new words, due review, and reinforcement content in a single learning task based on the scale of activated vocabulary; and generating a learning conversation queue. This achieves structured reuse of vocabulary data and dynamic adaptive adaptation of learning paths.
Owner:BEIJING WENYAN INTERNATIONAL EDUCATION TECHNOLOGY CO LTD

An automatic error correction method and system based on a knowledge graph

The application relates to the technical field of text error correction, and discloses an automatic error correction method and system based on a knowledge graph. Based on the lexical semantic association and the syntax dependency rule of the knowledge graph, the surrounding fragments of potential errors of input text are extracted, and semantic similarity is calculated to obtain error association weight data; an influence ranking is generated in combination with the domain importance attribute of the error type, and a to-be-corrected error list is formed after context sentiment tendency analysis adjustment; global semantic representation vectors are constructed based on the knowledge graph, domain categories are recognized, and semantic constraint conditions are generated, a preliminary error correction path is screened, semantic fitness is calculated to determine an optimal path, text correction is completed in combination with a high-priority error correction rule, and a result is output. Through the method, semantic deviation can be effectively avoided, error correction accuracy and pertinence can be improved, the semantic integrity and expression coherence of the corrected text can be ensured, and the problem of insufficient adaptability of traditional error correction methods is solved.
Owner:SHANGHAI XIRUAN TECH CO LTD

An Open Lexical Semantic Segmentation Method Adaptive to Unlabeled Text Prototypes

This invention discloses an open-lexical semantic segmentation method based on unlabeled text prototypes. It comprises three core steps: reliability scoring throughout confidence pixel selection, text prototype updating, and prototype interval preservation. Without training a visual encoder, learning cue parameters, or relying on manual annotation, it mines reliable visual evidence from high-confidence pixels in the target domain image. Furthermore, instead of applying the same update magnitude to all categories, it calculates a reliability score based on the number of pixels in each category, the predicted interval strength, and the interval stability. This score is used to simultaneously control the text prototype update step size and the prototype orthogonalization priority. In this way, the invention can suppress the cumulative amplification of noise pseudo-labels while maintaining the prototype interval between similar categories. Thus, while keeping all network weights of the visual-language model frozen, this invention only utilizes unlabeled images of the target domain to iteratively correct category text prototypes, transforming text prototypes from prior anchor points of natural image language into discriminative anchor points that conform to the visual distribution of the target domain.
Owner:UNIV OF ELECTRONICS SCI & TECH OF CHINA

An Open Lexical Semantic Segmentation Method and Storage Medium Based on CLIP

This invention provides an open-vocabulary semantic segmentation method and storage medium based on CLIP, belonging to the fields of image processing and computer vision technology. First, the CLIP visual and text encoders are frozen. L A layer-embedded dynamic cue allocation module is used, setting a learnable parameter pool containing visual cue embeddings and indicator vectors. Cosine correlation is calculated layer by layer, and highly correlated cuees are selected to construct a binary cue allocation matrix. An attention bias is generated to execute restricted attention interactions, and cue information is fused to enhance local fine-grained representation. The SPE channel is enabled during training, utilizing frequency domain high-pass filtering and spatial random shuffling to achieve dual-domain perturbation. Weighted mixed features enhance structural semantic learning, and this channel can be disabled during inference. A candidate mask is output, and the category is predicted using cross-modal cosine similarity. Pixel backfilling yields the segmentation map. This invention retains the generalization capability of the CLIP open vocabulary, optimizes edge and small target segmentation, and improves zero-shot segmentation performance with a lightweight architecture.
Owner:TIANJIN POLYTECHNIC UNIV