11 results about "Concept vector" patented technology

The invention relates to the technical field of artificial intelligence, in particular to a concept activation vector-based medical image lesion analysis and clinical decision interpretation system, which comprises a feature analysis module, a semantic mapping construction module, a concept activation extraction module, a boundary response fusion module and an interpretation path generation module. According to the method, the gradient direction and the gray abrupt change position of continuous elements in a focus area are detected, texture density and edge consistency are jointly judged to generate structured alignment mapping, microscopic image features are accurately captured, cross comparison is carried out on co-occurrence probability of lesion nouns and modifiers in phrase combinations and activation frequency of the same area, and therefore the accuracy of the lesion nouns and the modifiers in the phrase combinations is improved. Constructing a deep correlation map of pathological semantics and image features, screening an activation group according to continuous response of phrase pairs and image regions under case input change, generating a concept vector trigger track to dynamically track pathological feature evolution, and performing clinical consistency judgment; and an interpretable decision basis with strict logic support is output while subjective interference is eliminated.

This application discloses a text label generation, model training, and text classification method and related equipment, which addresses the problem that category labels obtained in related technologies cannot accurately describe the category to which the sample corpus belongs, thus affecting the accuracy of the subsequently trained text classification model and the execution accuracy of text classification tasks based on the text classification model. The text label generation method includes: obtaining the meta-concept path corresponding to the sample text from a pre-constructed concept tree based on keywords in the sample text corresponding to the target classification task; the concept tree is used to represent the hierarchical relationship between multiple meta-concepts, and the meta-concept path is used to represent the hierarchical relationship between multiple target meta-concepts related to the sample text in the concept tree; searching for label words in a pre-set meta-concept table based on the concept vectors corresponding to the multiple target meta-concepts and the hierarchical relationship between the multiple target meta-concepts to determine the label words corresponding to the sample text and use them as the category labels corresponding to the sample text.

This invention discloses a knowledge tracking method and system decoupled from cognitive states, relating to the field of knowledge tracking technology. The decoupled cognitive state knowledge tracking system mainly includes: a problem and concept embedding representation module, used to construct a problem-concept heterogeneous relationship graph and its triple relationships, learn the embedding representations of problems and concepts based on the triple relationships to obtain problem vectors and concept vectors, and obtain basic interaction embeddings based on interactions; a cognitive state decoupling module, used to construct fluctuating cognitive states and stable cognitive states; and a cognitive state tracking module, used to fuse the fluctuating and stable cognitive states, perform decaying attention tracking, and obtain prediction probabilities. Implementing the decoupled cognitive state knowledge tracking method provided by this invention can improve the prediction performance, interpretability, robustness, and applicability of the knowledge tracking model.

The invention provides an interpretable image recognition system and a construction method thereof, the system comprises a feature encoder, a concept predictor, a concept locator and a category predictor, the feature encoder is used for encoding an image into a three-dimensional feature map; the concept predictor is used for encoding the three-dimensional feature map into a plurality of first concept activation maps and outputting a plurality of prediction concept tags of the image; the concept locator is used for covering other areas except a key feature area matched with a real concept label of each first concept activation graph to obtain a plurality of second concept activation graphs in one-to-one correspondence with the first concept activation graphs, performing spatial pooling on all the second concept activation graphs to obtain concept vectors of the image, and performing spatial pooling on all the second concept activation graphs to obtain the concept vectors of the image; outputting a second concept activation graph corresponding to each predicted concept label predicted by the concept predictor as a key feature region graph; and the category predictor is used for predicting and outputting a category label of the image based on the concept vector of the image.

The application provides a hypernym-hyponym relation extraction method based on concept definition and data enhancement, which comprises the following steps: extracting concept pairs from natural text by using a keyword extraction technology, constructing concept triples based on the concept pairs and the hypernym-hyponym relations corresponding to the concept pairs, and taking the set of concept triples as a training data set; obtaining concept vectors in each triple in the training data set, offset vectors between the concept vectors, and vectors of concept definitions; constructing a hypernym-hyponym relation prediction model with the training data set as the input and the fused vectors of the offset vectors between the concept vectors, the concept vectors, and the vectors of concept definitions as the output, training the hypernym-hyponym relation prediction model according to the training data set and the fused vectors; obtaining a to-be-predicted concept triple in a test text, inputting the to-be-predicted concept triple into the trained hypernym-hyponym relation prediction model, and predicting whether the to-be-predicted concept triple has a hypernym-hyponym relation according to the output components.

The invention discloses an interpretability method for a decision basis of a medical image classification model, and relates to an interpretability method for a classification model. A channel and a space attention mechanism are introduced in a prototype concept extraction stage, the model is guided to focus on a key semantic region related to a disease, background interference is reduced, and semantic discrimination and stability of a prototype concept vector are remarkably improved; a CARAFE up-sampling method based on content awareness is adopted to replace a traditional up-sampling mode, the problems of edge blur and structure distortion are effectively relieved in the prototype concept vector positioning and deconvolution visualization process, and important anatomical structures and focus contour information in medical images are better reserved; according to the method, semantic suppression of prototype concept vectors in a high-level semantic space is quantitatively analyzed in combination with an anti-fact intervention strategy, contribution of each prototype concept vector to model prediction can be objectively evaluated, and potential error correlation or redundant prototype concept vectors can be revealed, so that the fineness and credibility of model decision interpretation are improved.

The invention belongs to the technical field of patent classification method conversion, and particularly relates to a method for converting an international patent classification method and a Chinese library classification method. The method comprises the following steps of: 1, constructing an international patent classification semantic vector set I: fusing various target word vector sets Iw, concept vector sets Ic and phrase vector sets Ip of an international patent classification (IPC); 2, constructing a Chinese library classification semantic vector set C: fusing various target word vector sets Cw, concept vector sets Cc and phrase vector sets Cp of a Chinese library classification method CLC; step 3, performing IPC and CLC category mapping; and 4, verifying the mapping result by using an IPC and CLC category mapping sample set established by an expert. According to the method, the patent and the scientific and technical literature can be retrieved and analyzed uniformly in the precise subdivision field limited by the same classification number, switching operation in different databases is not needed, information separation is avoided, retrieval and analysis efficiency is improved, comprehensive and accurate intelligence support is provided, and integrity and continuity of intelligence information are guaranteed.

The invention discloses a precise question and answer method based on a knowledge graph and a large language model. The method comprises the following steps of 1, obtaining a dual-carbon field natural language question input by a user; step 2, executing TransE model training to obtain a global embedded vector set; step 3, generating node representation by adopting CompGCN; 4, semantic enhanced embedding of improved DeBERTa is constructed, and gating fusion is carried out on lexical elements of the question text and corresponding concept vectors in ConceptNet; 5, sensing a position code by adopting a syntax in the improved DeBERTa, generating the position code according to the dependency information and using the position code for decoupling self-attention calculation, and outputting context representation of a problem; and 6, executing a bidirectional cross-modal fusion mechanism, and determining an answer text. The method is suitable for scenes with high requirements on accuracy and context understanding in the fields of complex questions and answers, intelligent customer service, education question answering, dual carbon and the like.

The application relates to the technical field of coal mine safety risk construction reasoning, and discloses a coal mine safety risk multi-modal knowledge graph construction and intelligent reasoning method, which comprises the following steps: constructing an ontology and coding a concept vector by using a pre-training model to establish a benchmark semantic space; extracting a time-space feature and projecting the time-space feature to the benchmark space to calculate a semantic similarity with the concept vector; synthesizing conflict evidence by using evidence theory to instantiate a node, generating a real-time dynamic graph, eliminating a redundant node to construct a Bayesian network, mapping a weight to execute propagation and output a posterior probability, quantifying an entropy value to measure uncertainty, and feeding back a control device to trigger reasoning iteration when the value exceeds a threshold. The application projects heterogeneous sensor time sequence data and monitoring video data into a unified dimension feature vector by constructing a benchmark semantic space aligned with a static ontology, introduces evidence theory to process semantic conflicts among multi-source evidence, and effectively solves the problem that different modal physical signals cannot be directly fused at a feature level.

The invention discloses a conversation emotion recognition method and device based on task self-adaption and multi-level situation collaborative understanding and a storage medium, and belongs to the field of natural language understanding and emotion calculation. The method comprises the steps that firstly, a pre-training model is used for obtaining dialogue initial representation, emotion concept vectors are embedded into a unified space through task adaptive prompt and comparative learning, and high-discrimination utterance representation is generated; then, a bidirectional GRU and double-mask LSTM architecture is adopted, and coupling characteristics of global emotion flow and local personal inertia-instant interaction are captured respectively; and finally, directly parameterizing the classifier by using the emotion concept vector so as to realize whole-course task self-adaption from representation learning to final decision making. The method is obviously superior to the prior art in low-resource, cross-session and multi-modal conflict scenes, and can be widely applied to low-delay and high-robustness demand scenes such as intelligent customer service and social robots.

The invention discloses a knowledge tracking method and system for decoupling a cognitive state, and relates to the technical field of knowledge tracking, the knowledge tracking system for decoupling the cognitive state mainly comprises: an exercise and concept embedding expression module, which is used for constructing an exercise-concept heterogeneous relation graph and a triple relation thereof, learning the embedding representation of exercises and concepts according to the triple relationship to obtain an exercise vector and a concept vector, and obtaining basic interaction embedding according to interaction; the cognitive state decoupling module is used for constructing a fluctuation cognitive state and a stable cognitive state; and the cognitive state tracking module is used for fusing the fluctuation cognitive state and the stable cognitive state and performing attenuation attention tracking to obtain a prediction probability. By implementing the knowledge tracking method for decoupling the cognitive state provided by the invention, the prediction performance, interpretability, robustness and applicability of the knowledge tracking model can be improved.