Machine Learning Query Enhancement for Relevant Document Retrieval
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing search engines in documentation repositories often return irrelevant results due to mismatches between user queries and document metadata, even when users are unfamiliar with the tagging conventions.
Innovation Solution
Utilizing machine learning models to predict a target label from keyword search queries by generating search value vectors and applying algorithms like KNN to identify the most relevant label, enhancing the search query with the predicted label.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If keyword search is used to match documents in documentation repositories, then search speed is improved, but search result relevance deteriorates
Solution Approach 1:
The patent introduces machine learning models as an intermediary between the keyword search query and the document metadata. The ML model predicts the appropriate label from the query keywords, which then serves as a mediator to bridge the gap between user intent and document classification, improving relevance without sacrificing search speed
Solution Approach 2:
The patent replaces the mechanical string-level matching system with a machine learning-based prediction system. Instead of relying on exact keyword matches against metadata, the ML model learns semantic relationships and predicts relevant labels, substituting the rigid mechanical matching approach with a more flexible intelligent system
2Measurement precision
If users are familiar with tagging conventions, then search result relevance is improved, but ease of operation deteriorates
Solution Approach 1:
The patent enables the system to automatically predict and apply the appropriate labels based on the query keywords, eliminating the need for users to manually understand or specify tagging conventions. The ML model performs the labeling task autonomously, making the search process easier while maintaining high relevance
Solution Approach 2:
The patent performs label prediction in advance as part of the search query processing pipeline. The ML model pre-computes the relevant labels from the query keywords before the actual search is executed, so users benefit from enhanced relevance without any additional effort or knowledge of tagging conventions
3Measurement precision
If machine learning models are used to predict target labels, then search result relevance is improved, but device complexity increases
Solution Approach 1:
The patent designs the machine learning model to serve multiple functions: it not only predicts labels for improving search relevance but also learns from feedback to continuously improve performance. The same ML infrastructure supports both the prediction task and the feedback learning process, reducing overall system complexity through multi-functionality
4Measurement precision
If feedback is collected to retrain machine learning models, then measurement precision is improved, but loss of time increases
Solution Approach 1:
The patent implements periodic retraining of the machine learning model using collected feedback data. Instead of continuous retraining that would cause significant delays, the system periodically updates the model at intervals, balancing the need for improved accuracy with the constraint of time loss. This periodic approach allows the system to maintain good performance while minimizing disruption to normal operations
Data Source
AI summary
Techniques for generating terms to replace an initial set of search terms for a query are disclosed. A system generates a training data set for training a machine learning model. Generating the training data set includes generating search value vectors for each of a set of labels based on sets of search values associated respectively with the labels in the set of labels. The system trains a machine learning model to predict a target label for a target search vector based on the set of labels and the respectively associated search value vectors. The system generates a target search value vector based on an initial set of search values. The system then applies the trained machine learning model to the target search value vector to predict the target label. The target label is used as a search term, that replaces the initial set of search values, for executing the query.