19 results about "Score fusion" patented technology

The invention provides a multi-attribute score fusion recommendation method based on user preferences and attention, and the method comprises a data collection and preparation module which is used for obtaining and cleaning user, project and score data; the attribute coding module is used for carrying out numerical coding on user and project attributes; the multi-attribute feature fusion module is used for obtaining user-project comprehensive representation by splicing attribute vectors and utilizing a deep neural network to extract interaction features; the fusion attention mechanism module is used for dynamically distributing each attribute weight and realizing weighted fusion of multi-attribute features; the comprehensive evaluation score calculation module is used for predicting the score of the user on the project according to the fusion feature and the attention distribution; the global preference modeling module is used for fusing historical behaviors of the user and improving the modeling capability of the overall preference; the recommendation result output module is used for generating a personalized recommendation list; and the method evaluating and updating module is used for evaluating recommendation performance and dynamically optimizing the model.

The invention provides an NL2SQL database table dynamic screening method and system based on multi-strategy fusion. The method comprises the steps that a preliminary candidate set is generated, a semantic encoder is used for encoding a user natural language problem and table and column information in a database into vectors, the semantic similarity is calculated, a preset number of tables are screened out from a whole database table based on the similarity, and a preliminary candidate table set is formed; performing deep analysis and reordering, performing multi-dimensional analysis on the tables in the preliminary candidate table set, and calculating a graph relationship compactness score and a content correlation score of each table; and performing fusion decision, inputting the semantic similarity score, the graph relationship compactness score and the content correlation score of each table into a fusion model to obtain a final comprehensive score of each table, and screening out a final correlation table set based on the final comprehensive score. The method has high accuracy and high recall rate, the generation speed and response efficiency of the NL2SQL system are improved, meanwhile, the robustness is high, and the practicability and generalization ability of the system are improved.

A hybrid recommendation system for generating personalized film recommendations, the system comprising the following: a processing unit for receiving metadata, configured to receive descriptive data related to films, the descriptive data including plot summaries, genre identifiers, cast lists and director names; a text preprocessing module that is communicatively coupled with the metadata ingestion processing unit and is configured to perform tokenization, stopword removal, stemming, and lemmatization on the descriptive data to obtain a processed corpus; a TF-IDF vectorization unit configured to encode the processed corpus into high-dimensional semantic feature vectors by calculating TF-IDF inverse document frequency values ​​over the entire film dataset, with the semantic feature vectors representing the contextual meaning of terms for individual films; a collaborative filter engine comprising a user-element interaction matrix, wherein the engine is configured to compute latent preference signals using one or more techniques selected from the group consisting of cosine similarity, k-nearest neighbor similarity, and matrix factorization; a score fusion controller coupled to both the TF-IDF vectorization unit and the collaborative filter engine, wherein the module is configured to normalize the semantic feature vectors and the collaborative preference scores and dynamically combine them according to an adaptive weighting coefficient, the coefficient being determined as a function of data density, interaction sparsity, and user history length; and a recommendation output module configured to evaluate candidate films for each user based on the combined score and generate a top-N recommendation list.

The invention relates to the technical field of big data, in particular to a personalized recommendation method and device based on structure and behavior layering and a medium, and the method comprises the steps: extracting global features and graph structure information of interaction data of a user and an article; modeling a personalized portrait of the user, and generating a filter coefficient; carrying out hierarchical coefficient fusion, and executing mixed spectral domain atlas filtering to obtain a personalized atlas filtering score; obtaining a basic recommendation score through multi-source score fusion, and executing dynamic score modulation to obtain a final recommendation score; and generating a recommendation list based on the final recommendation score, and returning the recommendation list to the user. According to the method, through layered and two-dimensional personalized modeling, a recommendation engine can dynamically adjust the signal response of each user in a low frequency-high frequency and global-personalized spectral domain, so that the diversity, novelty and interpretability of a recommendation result are remarkably improved while the recommendation precision is ensured.

The application relates to a supplier cooperation risk automatic identification method, system, device and medium. The method comprises the following steps: firstly, obtaining supplier decision maker historical text data and historical cooperation event logs, extracting decision maker personality characteristic keywords, sentiment tendency and dividing groups to obtain decision maker portrait labels; combining the portrait labels and the historical cooperation event logs, grouping and classifying delivery delays and quality dispute events to obtain correlation mode data; determining preliminary risk labels based on the correlation mode data, extracting quantitative indicators and obtaining a controlled state score through a weighted fusion algorithm; fusing the preliminary risk label quantitative value and the controlled state score to determine the overall risk level of the supplier; finally, obtaining supplier real-time emergency text, identifying keywords and evaluating the impact degree, and updating the overall risk level to obtain a dynamic risk perception result. The method can improve the risk identification accuracy and timeliness, avoid artificial subjective bias and reduce the control cost.

The invention discloses a table retrieval method and device for a database cold boot scene and electronic equipment. The method comprises the following steps: S1, preprocessing structure information; s2, semantic enhancement and vector representation generation; s3, constructing a reverse index; s4, performing semantic recall; s5, carrying out keyword constraint and filtering; and S6, performing multi-score fusion sorting. The method is suitable for performing accurate, stable and interpretable retrieval on new or semantic sparse data tables in an enterprise database under the condition of lack of historical query logs, manual annotation or statistical information access. According to the method, data table filtering, semantic enhancement description data table generation and table-level semantic vector generation, reverse index construction, semantic similarity score calculation and semantic similarity score calculation are carried out on the whole. The processes of keyword matching score calculation, coverage rate score query, accurate matching score calculation, filtered candidate data table generation, fusion sorting score calculation and sorting are executed in sequence, and all the steps form a complete and closed-loop retrieval processing link.

The invention provides an aviation field multi-modal retrieval enhancement generation optimization method based on a tree index, and relates to the technical field of aviation field large model retrieval enhancement, and the method comprises the following steps: S1, constructing a source data set of an aviation field multi-modal knowledge base; s2, constructing and retrieving a tree-shaped index based on an embedded model; s3, performing multi-dimensional score optimization on the retrieval candidate set; and S4, performing fusion sorting on the multi-dimensional scores to obtain a related context set for guiding generation of question and answer results. By improving index establishment and score calculation parts generated by traditional retrieval enhancement, the problems of low recall rate, poor multilayer semantic information understanding and the like when a retrieval enhancement generation system faces a multi-modal long document in the aviation field are solved, and increasingly complicated requirements of the aviation industry are met.

The invention discloses a multi-view outlier detection method fusing fuzzy particle information. The multi-view outlier detection method comprises the following steps: preprocessing input multi-view data; constructing a fuzzy information particle module, processing the mixed data, capturing the uncertainty of the data, and generating an outlier score; constructing a representation learning module, learning a potential representation of the view by using a variational auto-encoder, and generating an outlier score; carrying out outlier score fusion on the fuzzy information particle score and the representation learning score to generate a final score; and identifying outliers according to the final score. The fuzzy information particle module is used for processing mixed data and capturing uncertainty, deep features are mined in combination with the representation learning module, and the detection accuracy and robustness are remarkably improved; through separated expert product modeling and multi-granularity fuzzy rough set fusion, view sharing and effective separation and integration of specific information are realized, and the method is more suitable for complex scenes such as financial fraud detection and medical abnormality diagnosis.

This invention discloses a method and device for full hand vein biometric recognition to prevent spoofing attacks, relating to the fields of full hand vein recognition, biometric information recognition, and identity authentication. The method includes: Step 1, image acquisition and detection: acquiring a frontal hand image covering the palm and finger areas, ensuring the presence of a palm and a real living person in the frontal hand image; Step 2, ROI extraction: extracting the palm vein ROI and the multi-finger ROI; Step 3, feature extraction: extracting palm vein features and features of the four fingers, and generating corresponding feature hash codes; Step 4, feature matching: first matching the palm vein features, then matching the four finger features; similarity comparison is achieved by calculating the Hamming distance between the feature hash codes; Step 5, score fusion: calculating the matching scores for the palm vein features and the four finger features respectively, then performing Tanh normalization on the matching scores, and finally fusing the normalized matching scores to obtain the final recognition result.

The invention discloses a dialogue memory retrieval system based on multi-channel recall fusion. The dialogue memory retrieval system comprises a query preprocessing module, a multi-channel recall module, a score fusion module and a candidate output module. The query preprocessing module is used for receiving a user query Q, performing word segmentation, embedded vector generation, keyword extraction and metadata analysis, outputting a semantic vector, a keyword list and a structured condition, and providing input for each channel. According to the dialogue memory retrieval system based on multi-channel recall fusion, the recall coverage rate can be increased, and the Recall (at) 5 in a simulation experiment is increased by about 8-12%; the robustness is enhanced, and good performance is achieved on semantic rewriting, keyword accurate matching and structured query; weight adjustment and dynamic fusion enable the method to adapt to multi-task and multi-field requirements; the fusion step complexity is O (N), and the method is suitable for large-scale application.

The invention discloses a quality inspection method for literature and blog question and answer data and a related device, and belongs to the technical field of artificial intelligence, and the quality inspection method comprises the following steps: obtaining literature and blog question and answer data to be inspected; and obtaining a score set of a plurality of preset score indexes and the corresponding dynamic weight vectors, and determining an index value of each score index. And based on the target dynamic weight vector, fusing the index value of the formatting scoring index, the index value of the repetition scoring index and the index value of the content quality scoring index of the literature and blog question and answer data to obtain a quality inspection result of the literature and blog question and answer data. According to the scheme, score fusion in a single score index and fusion of a plurality of score indexes are carried out based on the dynamic weight of model training, expert experience driving of each score index and a quality inspection result is improved into data driving, and the objectivity and interpretability of quality detection are improved.

The invention relates to the technical field of anomaly detection, and discloses an anomaly detection system based on pluggable multi-model aggregation and context traceability, which comprises a detection and management module, a duration score fusion module, an association and traceability module and a relief and output module, through a pluggable multi-model architecture and a multi-behavior anomaly scoring technology, scene-based flexible adaptation of anomaly detection can be realized, and different detection scenes according to network traffic, host monitoring, application services and the like are supported; models such as seasonal trend detection (PROPHET technology), rare event detection (robust covariance method), multi-index combination detection (bidirectional generative adversarial network) and the like are flexibly added or deleted, and multi-dimensional behaviors such as seasonal trend, rare occurrence, multi-index association and the like of the indexes can be comprehensively covered on the basis of standardized abnormal scores output by the models, so that the accuracy of the indexes is improved. And a detection blind area of a single model is avoided.

The application relates to the technical field of information retrieval and natural language processing, and discloses a large model intelligent consulting report automatic generation method based on RAG technology. The method comprises the following steps: analyzing specifications and a basis library to construct a structured clause library, and analyzing a project data set to form a project portrait and a project evidence fragment set; then, based on the project portrait, a candidate specification set is screened and hierarchical constraint parameters are generated, a clause evidence candidate set is obtained through two-stage hybrid retrieval; after multi-dimensional score fusion reordering and same-theme clause conflict resolution, a conclusion-evidence pair bound by the specification evidence and the project evidence is constructed; finally, the pair is input into a large model to generate an evidence-constrained report paragraph text. The method effectively solves problems such as inappropriate clause application, conflict unexplainability and basis untraceability, improves report generation efficiency and professionalism, ensures that the report is reviewable and traceable, and is suitable for engineering field consulting report automatic generation scenes.

The invention discloses a doctor-patient communication co-situation language training scoring system and method based on multi-modal analysis, and relates to the field of artificial intelligence and medical informatization. Comprising the steps that a data collection module collects doctor-patient voice signals, and doctor voice text data and time sequence basic parameters are obtained through a data preprocessing and time alignment module; a common-situation language semantic feature extraction module analyzes the text to generate a common-situation language semantic feature vector; the emotional language semantic scoring module matches the feature vector with a template library to generate an emotional language semantic score value; an emotional language context correlation scoring module generates an emotional language correlation score value according to the patient voice emotion semantic features and the feature vectors; the speech speed analysis and speech speed scoring module generates a speech speed scoring value based on the time sequence basic parameters; the multi-dimensional score fusion module fuses the score values to generate a comprehensive common-situation communication score; and the feedback and training guide module generates an adjustment suggestion according to the comprehensive common situation communication score and each score value to form a training closed loop.

The application provides a text recognition method and related device. In the process of text recognition on a to-be-recognized text image, a first context feature and a first score vector are obtained. In the process of text recognition on a standard answer combined with the first context feature, a second score vector is obtained. Since the first context feature can enhance the attention to the target answer content which is the same as or similar to the shape of the to-be-recognized text in the process of text recognition on the standard answer, the second score vector can reflect whether the standard answer contains the target answer content. Therefore, whether to perform answer excitation on the first score vector can be determined according to the second score vector. If yes, score fusion is performed on the first score vector and the second score vector to obtain an answer excitation recognition result. As can be seen, when the standard answer contains the target answer content, the application can perform answer excitation on the recognition result of the to-be-recognized text itself, thereby improving the recognition accuracy.

The invention relates to a spoken language expression ability AI evaluation method and device, equipment and a medium. The method comprises the following steps: performing multi-granularity feature decoupling processing on an original voice signal to obtain acoustic, linguistic and sub-linguistic feature sets; generating a specific feature representation of each scoring task based on a shared underlying encoder and a task specific expert network; constructing a differential scoring rule graph to quantify a relationship between tasks; identifying a gradient conflict task pair through forward gradient prediction and conflict detection; implementing a dynamic parameter freezing strategy according to the conflict index to generate an optimized gradient direction; model parameters are updated in combination with a multi-target loss function of dynamic weight calibration; and finally, generating a scoring result through multi-task score fusion. By adopting the method, the problems of gradient conflicts, inconsistent scoring standard semantics and calibration failure in multi-task learning can be effectively solved through task relation modeling and dynamic gradient reconciliation, and the evaluation accuracy and stability are improved.

The invention relates to the technical field of generative search engines, and discloses a generative search engine-oriented content acquisition probability evaluation and structure optimization system. The system comprises a multi-dimensional probability quantitative evaluation module which decomposes answer content into a fact statement unit, calculates source credibility, semantic association degree and multi-source consistency three-dimensional sub-scores in parallel, fuses a generation unit credibility acquisition probability, aggregates an output unit credibility acquisition probability, and identifies a low-confidence evaluation factor; the credibility-driven structured optimization module is used for carrying out semantic segmentation on answer contents according to the unit credit collection probability, carrying out enhanced presentation on high-confidence facts and carrying out risk prompt and evidence association display on low-confidence statements; and the closed-loop feedback optimization module is used for triggering supplementary retrieval according to the evaluation factors, inputting new evidences into re-evaluation and driving the structured optimization module to perform real-time adjustment. According to the method, credibility quantification, structured presentation and retrieval-optimization closed loop are realized, and the credibility and interpretability of the generated answer are improved.

The invention relates to an electric power field Text2SQL (Structured Query Language) generation method based on structure control and feedback enhancement, which focuses on the problems of weak structure adaptability, low semantic matching degree, poor execution stability and the like in an electric power field Text2SQL task. The method comprises the following steps: firstly, constructing a multi-view driven Prompt construction module, and fusing electric power query semantic coding, an electric power database Schema structure and historical execution feedback to generate a Prompt example set with semantic consistency and structure controllability; secondly, designing a structure-driven multi-SQL generation module, generating a plurality of SQL candidates with consistent semantics and various structures in combination with a temperature regulation and structure template guide mechanism, and meanwhile, ensuring the effectiveness of a candidate set in cooperation with structure legality detection and a conflict suppression strategy; and finally, providing a structure-semantic fusion driven SQL reordering method, performing feature extraction and score fusion from three dimensions of structure similarity, semantic consistency and execution feasibility, and selecting an optimal SQL as a final query instruction. An execution result feedback chain mechanism is constructed, an example pool and a Prompt construction strategy are continuously optimized, and a self-adaptive closed-loop generation system is formed.

The application relates to the technical field of big data, in particular to a personalized recommendation method and device based on structure and behavior stratification and a medium, the method comprising the following steps: extracting global features and graph structure information of user and item interaction data; modeling a user personalized portrait to generate filter coefficients; stratified coefficient fusion is performed to execute hybrid spectral domain graph filtering to obtain a personalized graph filtering score; a basic recommendation score is obtained through multi-source score fusion, dynamic score modulation is performed to obtain a final recommendation score; and a recommendation list is generated based on the final recommendation score and returned to the user. The method can dynamically adjust the signal response of the recommendation engine in the low-frequency-high-frequency and global-personalized spectral domain for each user through stratified and two-dimensional personalized modeling, so that the recommendation accuracy is ensured, and the diversity, novelty and interpretability of the recommendation result are significantly improved.