190 results about "Quality Score" patented technology

The invention discloses a privacy protection data sharing and cooperative training method and system based on federated learning. The method comprises the steps of receiving software development log data, adaptively judging the sensitivity degree according to a data type, dynamically adjusting noise disturbance intensity according to the sensitivity degree to perform data desensitization, and generating a sensitivity index; selecting a feature extraction strategy, extracting time sequence correlation features from the desensitization data, constructing a dynamic graph structure with a weight, and obtaining a time sequence feature vector through iterative fusion; calculating the time sequence correlation of the time sequence feature vector to obtain a data quality score, and setting a contribution weight based on the quality score to perform parameter aggregation; combining sensitivity indexes with data quality scores to construct a security sharing domain, decoupling global training parameters into knowledge fragments in the domain, formulating a recombination rule, and selectively acquiring the required knowledge fragments by all parties for local training. According to the method, deep collaboration is realized on the premise of protecting data privacy, and the collaboration training effect is improved.

The invention discloses a knowledge graph-based search engine relation chain recommendation method, which comprises the following steps of: receiving a relation chain recommendation request submitted by a user through a search engine to obtain initial query information; performing semantic analysis and intention recognition on the initial query information, and determining a corresponding entity type and a relationship type; calling a dynamic knowledge selection and evaluation mechanism according to the entity type and the relationship type, and dynamically screening the original knowledge graph to obtain effective knowledge sub-graphs; performing diffusion type reasoning based on a knowledge graph diffusion model to generate a candidate relation chain path set; evaluating the quality score of each candidate path in real time; and performing dynamic screening according to the path quality score, and outputting a relation chain path recommendation result. According to the method, accurate mining and dynamic optimization of the knowledge graph relation chain path are realized, and the accuracy of a recommendation result and the flexibility of user interaction are improved.

Techniques are described for training and use of machine learning models to determine objective video quality scores. Video quality scores predict the quality of video content perceived by viewers. Quality scores have various uses, including the selection of encoding profiles and determination of encoding ladders. A core model and residual model may be used to determine quality scores.

Systems and methods are disclosed for generating an adaptive bitrate ladder for a content item based on the actual quality of the content item. A user uploaded content item is determined to have different quality scores in different time portions of the content item, where the quality score is based on an analysis of visual characteristics of frames of the content item. In response, an adaptive bitrate ladder (ABR) is generated for each time portion which contains a set of bitrates based on the quality score.

The invention discloses an RAG question and answer optimization method and system based on fusion retrieval, and relates to the technical field of large model knowledge question and answer. The method comprises the steps of obtaining user query, executing keyword retrieval and semantic vector retrieval in parallel, and obtaining a keyword retrieval result list and a semantic vector retrieval result list respectively; and calculating the quality score of each text unit in each retrieval result list based on multi-dimensional quality evaluation, and filtering out the text units with the quality scores lower than a preset quality threshold value. And sorting the text units in all the filtered retrieval result lists by adopting a quality weighted reciprocal sorting fusion algorithm. And based on the length of each text unit and the document format to which the text unit belongs, performing context content enhancement and combination on each text unit after sorting, and generating an enhanced context block list. And constructing structured prompt information based on the enhanced context block list and user query, inputting structured prompt words into the large language model, and generating and outputting a final answer.

The invention relates to the technical field of enterprise service digitalization and intelligent management, in particular to an enterprise management method and management platform based on dynamic portraits. The method comprises the following steps: S1, accessing and standardizing multi-source heterogeneous data, calculating a source-level quality score and generating an evidence chain; s2, constructing a portrait map and a time axis taking an enterprise as a main node; s3, the rule engine and the LLM-RAG output numbers in parallel; s4, consistent arbitration and belief propagation; s5, event-driven dependency subgraph increment recalculation is carried out; s6, carrying out portrait-strategy matching and task linkage; s7, carrying out closed-loop write-back and auditable playback; and S8, multi-tenant fine-grained permission. According to the scheme, the problems of data fragmentation and insufficient timeliness and traceability are solved, and the linkage of portrait credibility, interpretability and real-time decision is realized.

A large language model-based target sequence generation method, which belongs to the field of artificial intelligence technology, specifically to the fields of large language models, natural language processing, deep learning and other technologies are provided. The large language model-based target sequence generation method includes: determining quality scores of candidate paths corresponding to candidate sequence elements based on prediction probabilities of the candidate sequence elements obtained by a large language model; pruning the candidate paths based on the quality scores to obtain one or more pruned paths; determining a target search width based on the prediction probabilities, and determining one or more target sequence elements from one or more candidate sequence elements corresponding to the one or more pruned paths according to the target search width; and generating one or more target sequences based on the one or more target sequence elements.

The invention discloses a self-evolution emerging type data governance method and system, and relates to the technical field of data governance, the method comprises the following steps: inputting a multi-source heterogeneous initial data set into a large model, and executing the following processing by the large model: performing quality evaluation and defect identification on the initial data set to obtain a ternary data set; based on a preset quality score threshold and the data source knowledge graph, performing cyclic self-emerging data supplementation according to the ternary data set to obtain a complete data set; performing data governance on the complete data set by adopting a preset governance rule set to obtain a high-quality data set, performing self-evolution on governance rules, and updating parameters of a large model through closed-loop feedback so as to update the preset governance rule set; the updated preset governance rule set is used for participating in data governance of the next initial data set. According to the method, full-automatic data management can be realized in a self-adaptive, self-optimized and self-explored manner.

The present disclosure provides a method and an apparatus for evaluating the quality of model samples, a storage medium and a computer device. The method includes: inputting sample data into an AI-generated content detection model to obtain a hit probability of the sample data; matching a content evaluation system based on the attribute information of the sample data; processing the sample data based on the evaluation rule in the content evaluation system to determine a test value of the sample data relative to at least one preset evaluation index; and performing a weighted calculation on the hit probability and the test value based on a target weight corresponding to the hit probability and the preset evaluation criterion, to obtain a quality score of the sample data. This method is capable of filtering data that may mislead model training, and also realizing a high-precision evaluation of the training data.

The invention discloses a multi-modal data quality evaluation and optimization method and system for the Internet of Vehicles, and relates to the technical field of data quality evaluation and optimization, and the method comprises the steps: carrying out the calculation of the preprocessed multi-modal data through five dimension factors, and obtaining a corresponding quality score and a comprehensive score; obtaining multi-modal data needing to be optimized according to the comprehensive score; obtaining all quality scores and corresponding weights of the multi-modal data needing to be optimized, inputting the quality scores and the corresponding weights into a factor perception optimization network, outputting an optimization strategy suggestion value, and calculating corresponding quality scores of the optimized multi-modal data and the non-optimized multi-modal data through five dimension factors after optimization; and judging whether a convergence condition is met or not according to the comprehensive score variation and the marginal income. According to the method, efficient evaluation and closed-loop optimization of the Internet of Vehicles data set are realized through quality factor calculation, dynamic weighted scoring, and data completion and convergence control strategy of strategy perception optimization.

Systems and methods are disclosed for generating an adaptive bitrate ladder for a content item based on the actual quality of the content item. A broadcaster determines a quality score of a content item and multiplexes a message indicating the quality score with data packets of the content item to a headend. The headend uses the quality score to determine an ABR ladder with a set of bitrates for the content item. The headend then transcodes the content item into the set of bitrates. The headend then transmits a plurality of network addresses, each corresponding to a stream at a particular bitrate, to a user gateway to enable the user gateway to make the streams available.

The invention provides a chat question-answering method and device based on knowledge precipitation, equipment and a storage medium, relates to the technical field of data processing, in particular to the fields of artificial intelligence, domain knowledge modeling and the like, and can be applied to application scenes such as chat conversation intelligent question-answering and the like. The specific implementation scheme is as follows: in response to a current message event in a chat dialogue, determining context information of the current dialogue; performing intention recognition based on the context information to obtain an intention recognition result; retrieving related domain knowledge from a knowledge base based on the intention recognition result; using the retrieved domain knowledge and context information to generate candidate answers; and carrying out quality scoring on the candidate answers, screening the answers with scores reaching a preset threshold value as target answers, and sending the target answers to the chat conversation. According to the scheme, the context can be constructed in the chat conversation, the intention can be recognized, the precipitated knowledge can be retrieved, and the high-quality answer can be generated, so that real-time reliable reply is provided, efficiency and accuracy are improved, and knowledge reuse and dynamic update are promoted.

The invention provides an OCR (Optical Character Recognition) information dynamic verification method and system based on multi-modal fusion and a medium. The method comprises the following steps: firstly, acquiring OCR identification data and scene environment data, extracting a multi-modal feature set, then training a multi-modal quality evaluation model based on historical data, performing field correctness, format compliance and logic relevance scoring on each modal, calculating confidence to position a problem modal, and finally obtaining a multi-modal quality evaluation result. The method comprises the following steps of: obtaining an OCR (Optical Character Recognition) identification result, performing traceability analysis on a problem mode, implementing mode enhancement, recalculating an enhanced quality score and a confidence score, dynamically distributing each mode weight according to enhanced confidence, obtaining a global quality score after weight fusion, and finally judging whether the OCR identification result is qualified or not based on a global score threshold value. According to the method, the accuracy and the robustness of OCR information verification are remarkably improved through collaborative verification and dynamic weight fusion of the multi-modal information, and identification interference in a complex scene can be effectively coped with.

The invention discloses a news retouching method based on artificial intelligence and a related device. The method comprises the steps of obtaining a preset news database and a model training target corresponding to a target news mechanism; determining a target retouching AI model based on a preset news database and the model training target; obtaining a manuscript to be polished; determining a target report style and a target audience group corresponding to the target news institution; polishing the manuscript to be polished based on the target report style and the target audience group through the target polishing AI model to obtain a reference manuscript; determining a manuscript quality score corresponding to the reference manuscript; and when the manuscript quality score is greater than or equal to a preset quality score, determining the reference manuscript as a target manuscript. By adopting the embodiment of the invention, the manuscript polishing efficiency is improved.

The invention provides a business handling method, device and equipment for financial business, a medium and a product, and relates to the field of artificial intelligence. According to the method, when a customer carries out business handling, multi-modal features of a face image are extracted by responding to the face image of the customer, and a weight coefficient corresponding to each multi-modal feature is determined. And based on a core cluster reference library, determining a first quality score of each multi-modal feature, the core cluster reference library being obtained by performing clustering processing on historical face images. And determining a total quality score of the face image according to the plurality of first quality scores and the corresponding weight coefficients. And when the total quality score is greater than the quality score threshold, executing a corresponding business handling operation. According to the method, through multi-modal feature extraction and weighted evaluation, the face image quality can be comprehensively and accurately measured, the accuracy and efficiency of business handling are improved, and the business risk and handling error rate caused by the face image quality problem are reduced.

The invention relates to the technical field of course quality evaluation, in particular to an online financial and accounting course quality evaluation method and device and a storage medium, and the method comprises the steps: collecting target financial and accounting course information and user feedback information of a target financial and accounting course, setting a user subjective score according to the user feedback information of the target financial and accounting course, an accounting criterion text is collected in real time, a target course synchronization rate is extracted according to the accounting criterion text and a course accounting criterion text, then a course aging score is set based on the target course synchronization rate, and practical operation features of a target accounting course are extracted based on a practical operation item result of a user. And according to practical operation characteristics of the target financial and accounting course, course weak chapters are judged, course practical operation scores are set, quality scores of the target financial and accounting course are constructed, and an evaluation result of the target financial and accounting course is obtained. According to the invention, the evaluation efficiency of the online accounting course is effectively improved.

The invention relates to the technical field of multi-modal information processing, and discloses a multi-modal information generation and enhancement method based on an AI large model, comprising the following steps: step 1, acquiring multi-modal original data and task prompts, generating a unified token sequence and constructing a constraint graph; 2, setting a modal quota, a coverage threshold value and a consistency threshold value according to the importance degree of the constraint node; step 3, generating a visibility mask and establishing a quota account book, and controlling the visibility of dependent resources; 4, when the generation section meets a threshold value, determining a commitment section and generating an abstract fingerprint; step 5, weighting candidate output according to the quota account book, and adjusting mask regeneration when no feasible candidate exists; step 6, calculating a quality score, and performing backfill enhancement on the low-quality sub-segment; and 7, carrying out statistics on a modal conflict rate and a semantic drift rate, and adaptively adjusting a quota and a constraint weight. According to the method, semantic consistency maintenance, resource adaptive allocation and generation quality enhancement in the multi-modal information generation process are realized.

The invention provides a quality evaluation model, a model training method and a quality evaluation method and device, and belongs to the technical field of artificial intelligence, and the method comprises the steps: extracting a first feature and a second feature from a geometric part of a three-dimensional grid model through a geometric feature extractor; the appearance feature extractor extracts a third feature and a fourth feature from the projection image; the projection image is obtained by projecting the three-dimensional grid model; the text feature extractor extracts a fifth feature from the text cue word; the fusion comparison module generates a sixth feature according to the fifth feature, the third feature and the second feature; and the quality feature weighted fusion module performs weighted fusion on the first feature, the fourth feature and the sixth feature to obtain an overall quality score of the three-dimensional grid model. According to the method, the overall quality score of the three-dimensional grid model is obtained by extracting the geometric structure, appearance texture, semantics and other features of the three-dimensional grid model and performing weighted fusion, so that the accuracy of quality evaluation of the three-dimensional grid model obtained by a text 3D technology is improved.

The embodiment of the invention discloses a flow exit quality evaluation method, device and equipment in data transmission. According to the technical scheme provided by the embodiment of the invention, the initial score levels of the to-be-detected traffic exit are respectively added to the first score queue of the first manager and the second score queue of the second manager, and the first score level is determined according to the first number of initial score levels recorded in the first score queue; and determining a second score level according to a second number of initial score levels recorded in the second score queue, and determining a target score level of the to-be-detected traffic exit according to the first score level and the second score level. The technical problem is solved, the fluctuation of the outlet quality score is reduced, and the stability of the flow outlet scheduling is improved.

The invention provides a data asset quality evaluation method and system, and the method comprises the steps: obtaining an affiliated evaluation implementation object of a to-be-evaluated data asset and preset storage format data in the to-be-evaluated data asset, determining a quantification rule for the preset storage format data according to the affiliated evaluation implementation object, and carrying out the quantification of the preset storage format data. According to a quantification rule, carrying out quantitative evaluation on the preset storage format data to obtain a quality score of the to-be-evaluated data assets; according to the method, after the affiliated evaluation implementation object is determined, the quantification rule for the preset storage format data is determined according to the affiliated evaluation implementation object, so that the to-be-evaluated data assets are quantized and evaluated in multiple dimensions according to the quantification rule, an evaluation basis for quantitative calculation is set for the to-be-evaluated data assets, and the evaluation accuracy of the to-be-evaluated data assets is improved. And subjective evaluation factors are reduced, and the to-be-evaluated data assets are cooperatively evaluated through multiple dimensions, so that the accuracy of evaluating the to-be-evaluated data assets is improved.

Systems and methods are disclosed for generating an adaptive bitrate ladder for a content item based on the actual quality of the content item. A broadcaster determines a quality score of a content item and multiplexes a message indicating the quality score with data packets of the content item to an OTT provider. The OTT provider uses the quality score to determine an ABR ladder with a set of bitrates for the content item. The OTT provider then encodes a plurality of portions of the content item for each bitrate. The OTT provider then generates a manifest indicating a portion of a URL corresponding to each portion and transmits the manifest to a client device.

The invention discloses an intelligent fusion processing method, system and device for coal mine multi-source heterogeneous data and a storage medium, and relates to the technical field of data processing and distributed calculation scheduling, and the method comprises the steps: collecting multi-source heterogeneous data, constructing a feature signature, and generating a quality score; and generating a fusion result according to the feature signature and the quality score, collecting a node state through the fusion result, calculating a reward and punishment factor, and executing dynamic scheduling by using the reward and punishment factor. According to the method, format recognition is performed on collected multi-source heterogeneous data, and the multi-source heterogeneous data subjected to format recognition is output, so that unified input of the multi-source heterogeneous data in the same processing link is realized; constructing a feature signature object and generating a quality score based on the multi-source heterogeneous data subjected to format recognition, so that quality information is transmitted along with the data in a computable form; and the fusion processing has clear input, calculation paths and output.

The invention relates to the technical field of artificial intelligence, and particularly provides a model knowledge injection method and device and electronic equipment. The method comprises the following steps: if quality scores of two answers of the same question in a specified field meet the following preference requirements: the quality score of a first answer is greater than a first set threshold value, and the quality score of a second answer is lower than a second set threshold value, taking the two answers as a preference data pair corresponding to the question, constructing a preference data set based on the preference data pair corresponding to each question; in the alignment stage, outputting token prediction distribution of each question based on the input preference data set through the current knowledge answer model; calculating a first loss and a second loss based on the token prediction distribution and the answer of each question; and if the first loss and the second loss meet a preset iteration stop condition, taking the current knowledge answer model as a target knowledge answer model, otherwise, adjusting parameters of the knowledge answer model, and continuing iterative training.

Embodiments of the present application provide a prompt word generation method and device, equipment, storage medium and program product, relating to the technical field of financial technology and artificial intelligence. The method comprises: collecting first interaction behavior data between a user and a large language model and generating a first interaction feature vector based on the first interaction behavior data; inputting the first interaction feature vector into a first model to obtain a quality score of a first prompt word of the large language model; in the case that the quality score of the first prompt word is lower than a first threshold, generating a rewrite trigger signal and generating a new version of the first prompt word based on the rewrite trigger signal. The method of the present application can improve the efficiency of prompt word generation and the performance of the large language model.

The application belongs to the technical field of data asset management, and particularly relates to a data asset operation decision method and system based on artificial intelligence. Firstly, multi-dimensional characteristic data of target data assets is collected, and the collected multi-dimensional characteristic data is vectorized and fused to form a feature representation F. Then, the feature representation F is quality evaluated. Next, the feature representation F, a comprehensive quality score Q and historical data are used to predict and evaluate a future expected value Vp, a predicted risk value Ri and a predicted cost value Zm of the target data assets. Secondly, a multi-objective decision function is constructed with the future expected value Vp maximization, the predicted risk value Ri minimization and the predicted cost value Zm optimization as targets. Strategy optimization is performed through an optimization algorithm, and a recommended operation action Aopt is generated. The prediction result can be automatically converted into an executable operation decision, and the dependence of the decision on artificial experience can be greatly reduced.

The invention records a multi-target tracking method based on trajectory pipeline regression prediction and association, and the method comprises the following steps: obtaining and recognizing scene targets, including an existing target and a new target; matching and splicing the existing target and the track pipeline corresponding to the identity label of the existing target, and outputting the updated track pipeline of the existing target; constructing a track pipeline corresponding to the new target identity label, and outputting the track pipeline of the new target; performing position correction on the output track pipeline, and outputting a predicted position of the scene target in the current frame; two track pipelines are randomly selected, the space-time correlation degree, the position error and the shape error between the two track pipelines are calculated, quality scoring is carried out on the two track pipelines, and suppression operation is carried out on the track pipelines which have conflicts and are low in quality score; and executing the operation on all the track pipelines of the scene target, and outputting the identities of all the track pipelines which are not inhibited and the predicted positions of the current frames of the track pipelines to finish tracking of the scene target.

The present invention disclosure provides a system for performing root cause analysis, comprising a processor and a memory, the processor is configured to at least one problem associated with running of an application by at least one user, wherein identifying problem includes determining a priority level of one or more services being impacted based on service level agreements, identify at least one root cause associated with the problem, identify one or more root cause corrective actions (RCCA) based on root cause and implement at least one RCCA for troubleshooting the identified problem. Further, the system generates a RCCA form for use by a user group for identifying and validating an additional RCCA and update the database to include the additional RCCA based on a quality score for the identified root cause corrective action. The system is configured to create a RCCA form, auto-fill the RCCA form, identify one or more RCCA implemented for the identified problem, and determine a quality score of the RCCA.

A computerized-method for dynamically prioritizing social-media interactions in real-time, based on social-media posts in feeds, in a contact-center. The computerized-method includes: (i) monitoring by processors the social-media posts in the feeds of one or more social-media platforms which are integrated to the contact center. The social-media posts have been published during a preconfigured period, for each social-media post of a customer in each feed in the feeds: a. calculating a quality-score by operating an Artificial intelligence (AI) driven Content Quality Analysis (ACQA) module; and b. calculating a social-impact score based on the calculated quality score and one or more parameters; (ii) automatically prioritizing the social-media posts based on the calculated social-impact score to yield a priorities queue of social-media interactions. Each social-media post represents a social-media interaction, and (iii) automatically routing social-media interactions by a routine-engine to an available agent based on the yielded priorities queue of social-media interactions.

The invention discloses an AI psychological interview response quality management method and system, and relates to the technical field of artificial intelligence and digital psychological health. The method comprises the following steps: firstly, analyzing semantic features of user voice through a rule path of zero floating point operation to obtain a first quality score; when the score is in the section needing to be rechecked, starting a lightweight neural network model to analyze the voice energy envelope features to obtain a second quality score; and after fusing the two, outputting a single-bit quality flag bit. And the system decides whether to generate the gentle reminding verbal skill or not through a progressive gate according to the flag bit and the accumulated number of times, and automatically triggers zero-responsibility closed-loop feedback for the high-frequency low-quality response user when the session is ended, so as to generate a gentle closing declaration and management list. According to the invention, low-power-consumption and high-precision response quality real-time discrimination and low-invasion intervention on the edge terminal are realized, interview data quality and user experience are effectively considered, and a complete quality management closed loop is formed.

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating and updating project plans for software development projects. In one aspect, a method includes obtaining data indicative of initial project plan parameters associated with a software development project; identifying multiple program modules associated with the software development project; obtaining, for a first set of modules of the multiple modules, corresponding source code segments; executing a quality assessment process to compute, for each corresponding source code segment, a corresponding quality score; identifying, based on the quality scores, that at least a subset of the corresponding source code segments are integrable into a corresponding first subset of the first set of modules; and generating a revised project plan by generating updates to the initial project plan parameters by accounting for the corresponding source code segments being integrated in to corresponding ones of the first set of modules.