Query Tagging Model Using Stratified Lexicons and Graph Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current query tagging models require substantial human effort for maintenance and annotation, and existing semantic class lexicon acquisition methods are inadequate for accurately understanding user intent in search queries, leading to inefficient retrieval of relevant information.
Innovation Solution
A technique for updating query tagging models using stratified lexicons, which involves creating a list data structure, filtering phrases, assigning lexicon probability distributions, and applying graph learning to enhance query tagging performance and reduce reliance on training data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If traditional query tagging models are used, then query tagging can be performed, but substantial human effort is required for maintenance and annotation
Solution Approach 1:
The system performs self-updating by automatically acquiring semantic class lexicons from web resources, filtering phrases, and updating lexicon probability distributions without requiring manual annotation. The graph learning algorithm enables the system to maintain and update its own knowledge base autonomously, eliminating the need for continuous human intervention in maintenance and annotation activities.
Solution Approach 2:
The system pre-acquires and pre-processes semantic class lexicons from web resources before they are needed for query tagging. By continuously updating the lexicon database in advance and maintaining ready-to-use semantic class mappings, the system eliminates the need for manual annotation at the time of query processing, significantly reducing operational time requirements.
2Measurement precision
If existing semantic class lexicon acquisition methods are used, then lexicons can be obtained, but they are inadequate for accurately understanding user intent
Solution Approach 1:
The system applies local quality by filtering and selecting phrases based on specific criteria (maximum phrase length, minimum phrase occurrence count) to create high-quality semantic class lexicons. This targeted approach ensures that only phrases meeting defined quality thresholds are included, improving the accuracy of user intent understanding compared to generic lexicon acquisition methods.
Solution Approach 2:
The system uses feedback mechanisms where the graph learning algorithm continuously refines lexicon probability distributions based on the relationships between phrases and semantic classes. This iterative feedback process improves the accuracy of user intent understanding by adjusting lexicon assignments based on observed patterns in the data, thereby enhancing the reliability of relevant information retrieval.
3Productivity
If manual annotation is used for training data, then query tagging models can be trained, but the process is inefficient and time-consuming
Solution Approach 1:
The system automatically generates training data by acquiring semantic class lexicons from web resources and processing them through the graph learning algorithm. This self-service approach eliminates the need for manual annotation of training data, significantly improving productivity and reducing the time required for model training while maintaining data quality through automated filtering and validation processes.
Data Source
AI summary
A user's search experience may be enhanced by providing additional content based upon an understanding of the user's intent. Query tagging, the assigning of semantic labels to terms within a query, is one technique that may be utilized to determine the context of a user's search query. Accordingly, as provided herein, a query tagging model may be updated using one or more stratified lexicons. A list data structure (e.g., lists of phrases obtained from web pages) and seed distribution data (e.g., pre-labeled probability data) may be used by a graph learning technique to obtain an expanded set of phrases and their respective probabilities of corresponding with particular lexicons (e.g., semantic class lexicons). The expanded set of phrases may be used to group phrases into stratified lexicons. The stratified lexicons may be used as features for updating and/or executing the query tagging model.