17 results about "Query Rewriting" patented technology

The present disclosure provides a training method and device of a query rewriting large model, equipment and medium. The method comprises: performing preliminary optimization on the query rewriting large model based on a fine-tuning dataset to obtain a reference model; and performing re-optimization on the reference model based on a reinforcement learning method to obtain a target model, wherein a reward signal in the reinforcement learning process is generated based on a deviation between a question and answer result of the reference model and a standard answer. Thus, the model can be quickly learned to the basic query rewriting mode by using the fine-tuning dataset for preliminary optimization. Then, the reinforcement learning is introduced, and the deviation between the question and answer result of the reference model and the standard answer is used as the reward signal for fine adjustment, which can effectively correct the generation deviation that may exist in the preliminary model, so that the final target model can generate a rewriting result closer to the real intention of the user in a complex query scenario, thereby improving the overall performance of the downstream question and answer system.

This application relates to the field of human-computer interaction technologies, and provides a dialog interaction method, an electronic device, and a computer-readable storage medium. In this application, a historical dialog record is screened for selecting an associated historical dialog record to perform query rewriting on query information, to reduce impact of a historical dialog record with an unrelated topic on the query information, improve accuracy of the query information, improve accuracy of reply data of the query information, and improve interaction accuracy. In addition, the associated historical dialog record is selected to perform query rewriting on the query information, to support a user in simplifying input of the query information in a multi-round interaction process.

Embodiments of the present application provide a natural language-based query processing method and device and electronic equipment, and relate to the technical field of industrial internet. The method extracts multi-dimensional features from an obtained query sentence to obtain a feature vector representing the complexity of processing the query sentence, the feature vector being constructed based on at least one of the dimensions of semantic entropy, entity size, multi-modal dependency, and governance demand correlation. The complexity value of the query sentence is calculated based on the feature vector, and the corresponding target processing path is matched for the query sentence according to the complexity value, the target processing path including at least one of a direct generation path, a query rewriting optimization path, a query decomposition execution path, and a multi-modal coordination path. The query sentence is processed according to the target processing path to generate an execution result. The method guarantees the query processing efficiency while improving the cross-modal query accuracy of multi-source heterogeneous data.

This invention discloses a system and method for dynamic semantic mapping and secure pass-through of big data supporting AI context protocols. The system includes a data acquisition module, a semantic knowledge graph construction module, a context generation module, an identity authentication module, an SQL processing module, and a feedback optimization module. The method includes: Step S1: Multi-source metadata fusion and semantic graph construction; Step S2: Intent-based dynamic context resource generation; Step S3: Dynamic pass-through of user identity and session binding; Step S4: Context-based query rewriting and security interception; Step S5: Feedback closed-loop optimization. This invention achieves "what you see is what you get" secure data querying, ensuring data security while improving query efficiency.

Provided are a retrieval optimization method based on a multi-level knowledge base, a medium, and a system. The method includes: hierarchically managing knowledge data to construct a multi-level knowledge base; retrieving the multi-level knowledge base according to a user query to generate an initial vector chunk set; then performing first query rewriting on the query and retrieving the multi-level knowledge base again to generate an updated vector chunk set. The strong and weak direction positioning of this retrieval can generate query questions that are both professional and comprehensive.

The present disclosure provides an agent-based query rewriting model training method and a search method, and relates to the technical field of artificial intelligence such as deep learning, agents, intelligent search, and intent understanding. The agent-based query rewriting model training process comprises: processing sample scene labels by using a first scene understanding layer of a teacher rewriting model to generate sample intent information and sample parameter constraint information; processing sample initial query words, sample intent information, and sample parameter constraint information by using a first query word rewriting layer of the teacher rewriting model to generate sample rewriting results of the sample initial query words; training a second scene understanding layer of a student rewriting model, which is smaller in size than the first scene understanding layer, by using sample scene labels, sample intent information, and sample parameter constraint information; and training a second query word rewriting layer of the student rewriting model, which is smaller in size than the first query word rewriting layer, by using sample initial query words, sample intent information, sample parameter constraint information, and sample rewriting results to obtain a target rewriting model.

The embodiment of the application provides a training method and device of a query rewriting model, a storage medium and a program product, and belongs to the technical field of artificial intelligence. The method comprises the following steps: generating a first training sample through a sample generation model based on a first to-be-rewritten query, the first training sample comprising the first to-be-rewritten query, a first rewriting thought chain and a first rewritten query; then adding the first training sample to a fine-tuning sample set, using the fine-tuning sample set to fine-tune a first query rewriting model, and obtaining a second query rewriting model. The application can make the first query rewriting model rewrite the query description defects in the case of keeping the query intention unchanged, so that the query rewriting of the second query rewriting model obtained by training is more accurate, and the accuracy of the search result is improved.

The application belongs to the field of knowledge retrieval enhancement generation, and discloses a multi-path mixed knowledge retrieval enhancement generation method and system applied to a vertical field. Through collaborative optimization of query understanding, multi-path recall, result reordering and knowledge fusion generation, the whole process from query to content generation is enhanced. A query rewriting strategy based on a large language model is introduced to improve semantic analysis and intent recognition capability. A multi-path recall mechanism is adopted, combined with text feature retrieval, vector semantic retrieval and multi-hop retrieval mechanism based on brain inspiration, to realize high-coverage recall of deep semantic information. Through multi-stage reordering optimization by a multi-stage reordering model, the accuracy and robustness of the recall result are improved. Finally, in the knowledge fusion and generation stage, weighted average or graph attention mechanism is used to realize dynamic fusion of multi-source knowledge and context-aware generation, thereby significantly improving the knowledge accuracy, generation consistency and semantic interpretability in professional field tasks.

The present disclosure relates to a method and apparatus for recommending a harmonized system code for goods on the basis of a large language model. The method comprises: receiving goods information that is input by a user; inputting the goods information into a harmonized system code recommendation large language model to obtain an initial recommended harmonized system code for goods; inputting query rewriting information based on the goods information into a query rewriting large language model to obtain a goods name and goods specifications; inputting the goods name and the goods specifications into a hybrid retrieval module to obtain hybrid retrieval results of harmonized system codes; and inputting the initial recommended harmonized system code and the hybrid retrieval results of the harmonized system codes into a decision fusion module to obtain a final recommended harmonized system code for the goods.

The application relates to a judgment document auxiliary writing method and system based on retrieval enhancement generation, which comprises the following steps: step 1, case fact structured block and law article library construction; step 2, generation of a judgment-oriented judicial abstract; step 3, hierarchical query rewriting to generate a retrieval expression; formation of a multi-granularity query expression from coarse to fine; step 4, vectorization law article retrieval and reordering; multi-layer query parallel retrieval is carried out in a pre-constructed law article vector retrieval library, and a relevant law article set is output by weighting and reordering according to semantic similarity and law article effectiveness level; step 5, construction of a structured generation input; step 6, segmented controlled generation of a judgment reasoning text; a staged generation strategy is adopted to sequentially complete case nature identification, dispute focus analysis, legal application demonstration and judgment conclusion expression, and finally splice to form a complete reasoning text. The application improves the consistency and integrity of the reasoning text and the case facts.

A SQL statement rewriting method includes: obtaining a first SQL statement to be rewritten; selecting at least one first rewriting case from multiple rewriting cases based on the similarity between the first SQL statement and multiple rewriting cases, wherein a rewriting case includes an example SQL statement and a rewritten SQL statement after rewriting the example SQL statement; evaluating multiple rewriting rules based on the similarity between the first SQL statement and the at least one first rewriting case, and the association between the at least one first rewriting case and multiple rewriting rules, to obtain the evaluation results of each of the multiple rewriting rules, wherein the rewriting rules are rules used to rewrite the SQL statement query; and rewriting the first SQL statement based on at least one first rewriting rule from the multiple rewriting rules to obtain a second SQL statement, wherein the at least one first rewriting rule is a rule whose evaluation result among the multiple rewriting rules meets the requirements. This method can improve rewriting performance.

This invention discloses a large-model-enhanced query rewriting synthesis method for detecting logical errors in database engines. The method includes: constructing a rule tree and systematically traversing all applicable rewriting rules; implementing a large language model-enhanced Monte Carlo tree search, combining Monte Carlo tree search with the large language model; and performing root cause analysis, utilizing a historical error pattern repository to precisely locate the abstract syntax tree nodes leading to the logical errors. This invention systematically explores the query rewriting space by combining rule tree search and large language model-enhanced Monte Carlo tree search techniques, improving the efficiency and coverage of logical error detection.

The application relates to the technical field of computer data processing, and provides a storage-computing separation database query optimization method and system for massive data analysis, which solves the technical problems of low system resource utilization and high query response delay. The method comprises the following steps: collecting business query logs and metadata information of all basic tables; constructing a syntax tree based on a query text, and calculating fluctuation indexes of each basic table; constructing a query topology graph based on a fingerprint sequence to obtain a candidate substructure set; taking the remaining available space as a space constraint, determining a target substructure combination through a multi-dimensional knapsack problem, and generating a query rewriting rule; matching the target fingerprint sequence with the query rewriting rule to determine a target substructure; replacing the target substructure with a read operation of pre-computed data corresponding to the target substructure combination; and generating an execution plan for optimizing the query, so that transparent rewriting of the query based on the pre-computed result is realized, and the query performance is maximally improved under the condition of limited maintenance cost.

A user-generated search query is received. A first machine learning model is used to retrieve a plurality of embeddings representing a plurality of query rewrite candidates. A plurality of query pairs is generated between the user-generated search query and each query rewrite candidate. A second machine learning model is used to generate a plurality of similarity scores based on the plurality of query pairs. A plurality of similarity scores is ranked to identify a query rewrite candidate used for generating query results. The query results are caused to be displayed.

The application discloses a kind of real-time processing and fast retrieval optimization method for big data, it is related to big data storage and retrieval technical field, first, the method is collected multi-source heterogeneous big data to complete standardization pretreatment, constructs hierarchical distributed storage architecture to realize data hierarchical storage, adopts event-driven stream processing framework to execute real-time data calculation, constructs multidimensional hybrid index system, and optimizes retrieval by intelligent query rewriting and parallel execution, deploy three-level cache mechanism, comprehensive evaluation system performance and realize closed-loop iterative optimization.The application improves the utilization efficiency of storage resources and the load balancing capability of cluster, enhances the efficiency of real-time data processing and multidimensional retrieval, optimizes query execution and cache hit effect, continuously improves the overall performance and adaptability of system through closed-loop iteration.

Artificial intelligence techniques for query management are described. A method comprises generating, by a context detection module, context information for a first query comprising natural language information to request a result from one of a plurality of machine learning models, modifying, by a query modification module, the first query based the context information to form a first modified query, determining, by an intent module, an intent type for the first modified query, selecting, by a routing module, a machine learning model from the plurality of machine learning models based on the intent type, and routing, by the routing module, the first modified query to the selected machine learning model. Other embodiments are described and claimed.