6 results about "Evidence reasoning" patented technology

The invention discloses an intention recognition method based on multi-source evidence fusion reasoning, and belongs to the technical field of artificial intelligence, and the method comprises the following steps: S1, obtaining a multi-source heterogeneous evidence at least comprising a text, a behavior sequence and external knowledge data, converting the multi-source heterogeneous evidence into a standardized evidence unit comprising evidence content, a source reliability score and intention correlation degree, and storing the standardized evidence unit in a database; constructing an initial evidence set based on the evidence reasoning graph model; s2, inputting the initial evidence set into an inference engine based on a neural network, and executing multi-hop inference through multiple rounds of active query: analyzing evidence consistency in the first hop and generating a preliminary intention hypothesis and a question asking problem, and dynamically generating a new problem in the subsequent hop based on new evidence to directionally obtain supplementary information until a termination condition is met; s3, carrying out dynamic evidence fusion calculation; and S4, outputting a highest confidence degree intention conclusion according to a global fusion result, and generating an interpretable evidence chain report containing a complete multi-hop reasoning path and key evidence support.

The application discloses a video safety monitoring method and device based on evidence reasoning and a storage medium, and relates to the technical field of video reasoning. The method inputs video data and a monitoring question into a preset reasoning model, obtains a reasoning result output by the reasoning model, takes the reasoning result as an input parameter of a preset comprehensive evaluation function, determines a comprehensive evidence score of the reasoning result through the comprehensive evaluation function, determines a normalized answer corresponding to each candidate answer according to a preset mapping function, takes a key frame index set corresponding to the candidate answer as a reasoning branch index set of the normalized answer if the normalized answers corresponding to the candidate answers are the same, determines an evidence weighted total score corresponding to the normalized answer according to each reasoning branch index set corresponding to the normalized answer and each corresponding comprehensive evidence score, and finally determines a monitoring result in the normalized answer according to the evidence weighted total score, thereby improving the accuracy of the monitoring result.

The application discloses an intention recognition method based on multi-source evidence fusion reasoning, and belongs to the technical field of artificial intelligence, and comprises the following steps: step S1: acquiring multi-source heterogeneous evidence including at least text, behavior sequence and external knowledge data, converting into standardized evidence units containing evidence content, source reliability score and intention correlation degree, and constructing an initial evidence set based on an evidence reasoning graph model; step S2: inputting the initial evidence set into a reasoning engine based on a neural network, performing multi-hop reasoning through multiple rounds of active query: the first hop analyzes evidence consistency and generates a preliminary intention hypothesis and a follow-up question, subsequent hops dynamically generate new questions based on new evidence to acquire supplementary information in a directional manner, and the process is continued until a termination condition is met; step S3: dynamic evidence fusion calculation; and step S4: outputting an intention conclusion with the highest confidence according to a global fusion result, and generating an explainable evidence chain report containing a complete multi-hop reasoning path and key evidence support.

The invention relates to the technical field of natural language processing, in particular to a document-level relation extraction method based on collaborative evidence reasoning, which comprises the following steps of: preprocessing and uniformly coding an input document to obtain semantic representations of entities and sentences in the document; constructing a collaborative heterogeneous graph between the entity pairs and the sentences, updating node representation through a graph attention network, and calculating a prediction probability of taking each entity pair and each sentence as evidence; an evidence sentence set is retrieved from the document for each entity pair, a pseudo document is constructed based on the evidence sentence set, the pseudo document is a new document formed by splicing evidence sentences according to the sequence of the original document, entity marks are reserved, and the evidence sentences only reserve sentences with the evidence prediction probability larger than or equal to 0.5 in the collaborative evidence layer; and taking the original document and the pseudo document as double-source input, and reasoning through a self-adaptive fusion function to predict a final relationship between the entity pairs. The method has the advantages of being high in relation extraction precision, high in multi-hop reasoning capacity and good in generalization.

The present application relates to the field of fact checking, and provides a fact checking method and system based on debiased reasoning enhancement and joint relationship aggregation, which comprises the following steps: performing semantic analysis on a statement text to obtain a tendency label; obtaining an initial input text from the statement text and an original document; inputting the initial input text into a checking model to obtain evidence text; performing complex plane projection to obtain fusion bias of the evidence text, generating a category preference prompt word, and performing debiased reasoning enhancement on the evidence text to obtain debiased reasoning enhancement text; obtaining a debiased evidence vector and a debiased reasoning vector from the debiased reasoning enhancement text; expanding the debiased evidence vector and the debiased reasoning vector through a binary Taylor series to obtain evidence reasoning features; expanding the debiased reasoning vector through a binary Taylor series to obtain reasoning aggregation features; and inputting the evidence reasoning features and the reasoning aggregation features into a multilayer perception machine classifier to obtain fact checking output. The present application is used for completing fact checking.

This invention discloses a probabilistic anomaly early warning method for structural monitoring data fusion based on streaming over-thresholding and evidence-based reasoning. The method includes: establishing historical datasets for each sensor as sample data; separating the sample data by temperature effects and extracting the response components after removing the temperature effects as the initial dataset for anomaly detection; determining the initial threshold; constructing an over-threshold dataset; fitting the over-threshold dataset and obtaining the posterior probability distributions of shape and scale parameters through Bayesian inference; calculating the probability distribution of early warning trigger values; calculating the anomaly certainty index; dynamically updating the probability distribution of early warning trigger values ​​using the streaming over-thresholding method; and synthesizing the results using the evidence fusion method to obtain the fused anomaly certainty index. This invention can form a fused anomaly certainty index, thereby better handling conflicts between indicators, avoiding false alarms caused by single sensor failures, and providing data support for structural maintenance.