16 results about "Evidenced based" patented technology

A system makes evidence-based decision for actions associated with a user. The system receives information describing an action associated with the user. The system receives documents associated with the user and questions associated with the action. For each of the plurality of questions the system performs the following steps. The system evaluates sentences from the document in relation to the question using an evidence extraction model. The system classifies the evidence using an evidence classification model to determine whether the evidence refutes or supports a decision based on the question. The system makes a decision regarding the action based on the classifications of evidence sentences in relation to each of the plurality of questions. The evidence extraction model and the evidence classification model are trained neural networks.

The invention provides a generative knowledge graph completion method and system combined with dynamic narration, and relates to the technical field of knowledge graph completion, and the method comprises the steps: obtaining a to-be-completed knowledge graph triple; key information is intelligently selected from an internal structure and an external text corpus of the knowledge graph in combination with multi-dimensional embedding of an entity, and a section of natural language narrative which is rich in information and related to context is dynamically generated for a triple to be complemented; constructing a completion instruction based on the natural language narration, controlling the large language model subjected to multi-task instruction fine tuning to execute a completion task, and generating a preliminary completion result; whether the preliminary completion result needs to be corrected or not is judged through internal consistency verification and external selectivity verification, if the preliminary completion result needs to be corrected, an instruction is reconstructed through an evidence-based enhancement method, and a large language model is utilized to generate a completion result with additional interpretation; according to the method, the performance, the robustness and the interpretability of the large language model in the knowledge graph completion task are improved.

The invention provides a conclusion credibility analysis method based on evidence fusion and a related device, and relates to the field of large language models. The method comprises the following steps: acquiring a reference text set of a to-be-queried question by electronic equipment; wherein the reference text set comprises reference texts related to the to-be-queried question; calling a large language model to process the reference text set to obtain a to-be-evaluated answer of the to-be-queried question; determining an evidence text set of the answer to be evaluated from the reference text set; wherein the evidence text set comprises evidence texts supporting to-be-evaluated answers; and obtaining a credibility score of the to-be-evaluated answer according to the evidence text set. Therefore, after the to-be-evaluated answer is generated, the evidence text set supporting the answer is further identified from the reference text set, and the credibility score is calculated based on the evidence text set, so that the reliability evaluation of the answer is converted into a quantifiable evidence support degree analysis process, the problem of untrusted output caused by hallucination of the model is effectively relieved, and the reliability of the model is improved. And the objectivity and verifiability of the conclusion are improved.

The present invention discloses a multimodal student drowsiness monitoring method based on evidence theory. The drowsiness monitoring system as a whole uses visual acquisition equipment, electroencephalogram (EEG) acquisition equipment, and heart rate acquisition equipment to collect posture data, EEG data, and electrocardiogram (ECG) signal data generated by students during online learning. Based on the improved D-S evidence theory, these multimodal data are fused and calculated to improve monitoring accuracy, improve student learning efficiency, and assist teachers in their teaching work. An interface is used to open a data transmission channel between the teaching terminal and the drowsiness monitoring service system. Feedback on student drowsiness is provided to teachers through the teaching terminal, and sleepy students are reminded to pay attention and listen to the lecture. The present invention solves the paradoxical situation that violates common sense when traditional D-S evidence theory fuses high-conflict data. The advantage of the method proposed by the present invention is that even if high data conflict occurs during the fusion process, the fused result does not violate common sense.

The invention relates to a factual reasoning-based large model answer generation method and system. According to the method, after the answer draft is generated, an auditing process is carried out, firstly, a question is put forward for the answer draft, then fact checking is carried out based on the question, evidence is provided, finally, the answer draft is decided based on the evidence and the question, a deciding result is obtained, and if the deciding result comprises an accepting instruction, the current answer draft is used as a final draft; if the judgment result comprises a rejection instruction, modifying the current answer draft and re-executing the reviewing process; if the judgment result comprises a re-planning instruction, regenerating an answer draft, and performing an auditing process; and finally, optimizing the final draft, and obtaining and outputting an answer text. Compared with the prior art, the method has the advantages of remarkably improving the fact accuracy and logic preciseness of the output content and the like.

The application relates to the technical field of intelligent education and knowledge diagnosis, and provides a group knowledge diagnosis mastering state uncertainty objective quantification method, which comprises the following steps: obtaining historical answer data of a testee on test questions; based on the historical answer data, determining evidence parameters of a normal-inverse gamma distribution through a testee evidence encoder, wherein the evidence parameters comprise a mean value estimation of the testee ability, a mean value precision parameter, a variance shape parameter and a variance scale parameter; constructing a posterior evidence distribution of the testee mastering state based on the evidence parameters, and utilizing a variance decomposition property of the normal-inverse gamma distribution to explicitly decompose total uncertainty of the posterior evidence distribution into random uncertainty and cognitive uncertainty. The method of the application can explicitly separate and quantify uncertainty from different sources, improve the reliability and interpretability of a diagnosis result, and keep robust inference under the scene of data sparsity and response loss.

This invention presents a passive domain adaptive fundus segmentation method based on evidence representation and marginal screening. Under the condition that the source domain training images and their pixel-level annotations are inaccessible, a Student network and a Teacher network are constructed to learn segmentation of the target domain fundus image. The network outputs positive foreground evidence and negative background evidence for each segmentation channel, and constructs a pixel-level Beta distribution. Based on the Beta distribution, foreground probability, evidence margin, and evidence strength are calculated. Reliable positive pseudo-labels are obtained through joint screening, and a weighted segmentation loss is constructed by combining pseudo-label weights. Simultaneously, regularization loss and evidence strength constraint loss are introduced to optimize the Student network. Finally, the exponential moving average coefficient is dynamically adjusted based on the amount of evidence to update the Teacher network parameters. This method can effectively suppress the propagation of pseudo-label noise and improve the accuracy, stability, and cross-domain adaptability of optic cup and optic disc segmentation in the target domain fundus image.

The invention discloses a decision-making method and system based on multi-agent review. Three types of agents are established, and prior probability distribution is configured; carrying out multiple rounds of auditing interaction based on the evidence data, which comprises the following steps of: calling a conservative agent and a radical agent in parallel, and respectively generating structured arguments; performing Bayesian updating on respective prior probability distribution based on the structured discussion to obtain respective posterior probability distribution; calling a neutral agent to analyze the structured arguments, generating a conservative challenge list and a aggressive challenge list to drive the conservative agent and the aggressive agent to correct the respective structured arguments, and recalculating the respective posterior probability distribution; judging whether a preset convergence condition is met or not after each round of auditing interaction is finished; if yes, fusion processing is carried out, and fusion probability distribution used for representing the decision result is generated; if not, continuing the next round of auditing interaction; and outputting a decision result based on the fusion probability distribution or the total divergence.

The invention discloses an AI online education intelligent question and answer information processing method, and the method comprises the steps: carrying out the time alignment of a cold start interaction set, mapping the cold start interaction set into a knowledge point index, and completing the first-repair dependence binding in combination with a teaching structured constraint set to form a cold start sequence; based on the evidence group identifier, returning the evidence list, performing effective evidence screening, calculating a sample support degree index, comparing the sample support degree index with a support degree threshold value, and outputting a to-be-verified label set; obtaining a question discrimination level and a source reliability level, executing supervised learning regression to deduce and output a wrong marking risk upper bound value, comparing the upper bound value with a wrong marking threshold to determine a stable tag, and adding the stable tag into a recommendation set; an alternative label is generated for a solidification forbidding result, an interaction task set is retrieved and distinguished, a verification evidence sequence is collected, a consistency index is calculated and compared with a consistency threshold value, a pseudo-stable label mark is output, a label decision result is output, a recommendation set is updated, and the recommendation path locking risk is reduced.

The application belongs to the technical field of knowledge tracking, and particularly relates to a learning performance prediction method and system of a large language model and a multi-agent collaboration, a graph construction agent constructs a knowledge point relationship network; a statistical analysis agent combines a network operation knowledge tracking model to output statistical prediction probability and knowledge mastery degree, calculates a question passing rate, maps a time interval into a forgetting label, and quantifies a question related evidence score; a dynamic parameter adjustment agent adjusts a subsequent deduction temperature according to the evidence score; a logical deduction agent combines a last moment cognitive image, the knowledge mastery degree, the forgetting label and a question feature to perform recursive reasoning, outputs a logical prediction probability and updates the image; and a decision fusion agent adaptively weights and fuses double-track probabilities based on the evidence score, and outputs a final performance prediction result. The application dynamically tracks the knowledge mastery state of students by combining a knowledge tracking model and a large language model, and improves the accuracy of student learning performance prediction under data sparsity and cold start.

This invention provides a multi-view collective medical decision-making assessment method and system based on dual-concept learning, comprising: constructing a two-level decision interpretation mechanism to obtain decision concepts; modeling individual decisions of each view, inputting the decision concepts into an evidence-based neural network to obtain evidence quantity, and quantifying individual uncertainty; performing consensus judgment on individual decisions of each view, outputting a consistent collective opinion result, or obtaining prior importance and the importance of specific instances and modeling to obtain dynamic priorities, fusing each view according to the dynamic priorities to obtain a collective viewpoint and outputting it. This invention utilizes decision trajectory backtracking to achieve high transparency and interpretability in the decision-making process; it can proactively identify and warn of high-risk decisions, significantly improving the reliability of the system; thus, while retaining the differentiated information of the views, it achieves more logically reasonable and robust collective decision-making.

Described herein are methods, systems, and techniques for tailoring an Evidence based therapy (EBT) curriculum, said system comprising: one or more processors configured to receive demographic and psychological information of a user; a machine learning algorithm trained to analyze the received information and generate a series of tool kits forming a tailored curriculum based on the user's specific therapeutic needs and understanding of psychological principles; high-value items selected to reinforce self-worth and act as learning artifacts; and a series of educational materials configured to guide the user in understanding foundational principles of Evidence based therapy (EBT) and engage in therapeutic activities; wherein the generated series of tool kits are delivered over an extended period to provide an ongoing, adaptive, and sequential user support.

The application discloses a kind of knowledge conflict-oriented evidence constraint decoding type retrieval enhancement generation method, belong to artificial intelligence, natural language processing and information retrieval technical field.The method includes: generating structured fact representation set based on the question to be answered and constructing self-consistent fact chain;Based on the question to be answered, self-consistent fact chain and candidate answer hypothesis, perform multi-signal retrieval, and rearrange, conflict detection, counterfactual alignment and evidence coverage analysis, obtain evidence support information and conflict perception control signal;Based on evidence support condition distribution and control condition distribution, perform control type constraint decoding;According to evidence conflict degree, candidate divergence, evidence coverage and reference closure degree, construct risk control quantity, and dynamically adjust decoding parameter, output structured answer and evidence-based uncertainty report.The method can improve the fact consistency, robustness and explainability of the generated results in knowledge conflict and high noise retrieval scenarios.

The application discloses an AI online education intelligent question and answer information processing method, which aligns and maps a cold start interaction set time to a knowledge point index, combines a teaching structured constraint set to complete prerequisite dependency binding to form a cold start sequence; based on an evidence group identifier, retrieves an evidence list and performs effective evidence screening, calculates a sample support degree index and compares it with a support degree threshold to output a to-be-verified label set; obtains a question discrimination degree level and a source reliability level and performs supervised learning regression inference to output an upper bound value of a wrong label risk, compares the upper bound value with a wrong label threshold to determine a stable label and adds the stable label to a recommendation set; generates an alternative label for a forbidden solidification result, retrieves a differentiated interaction task set, collects a verification evidence sequence, calculates a consistency index and compares the consistency index with a consistency threshold to output a pseudo-stable label mark, outputs a label decision result and updates the recommendation set, and reduces a recommendation path locking risk.

The invention belongs to the technical field of evidence grouping, and particularly discloses a litigation case-oriented evidence grouping method and system, a terminal and a medium. Comprising the steps of obtaining multi-category evidences in a case, and performing content analysis on the evidences to generate uniform feature representation; performing fact element identification on the unified feature representation to obtain event features, time features and main body features, and constructing a graph structure for reflecting an association relationship between evidences; performing feature learning on the graph structure to generate a preliminary evidence grouping result; performing iterative optimization on the preliminary grouping result to obtain a stable optimized group; and reasoning and analyzing the time sequence, causal orientation and content correlation among the evidences based on the optimized grouping, and generating explanation information for representing evidence grouping logic. According to the method, multiple types of evidences in complex cases can be subjected to structured expression, association modeling and logically consistent evidence grouping, and technical support is provided for case fact carding and evidence chain construction.