Hybrid Speech Search Combining Supervised and Unsupervised Retrieval
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing content repository search systems face limitations in effectively classifying new queries due to the need for large volumes of training data and the inability to identify content added after classifier training, especially when using supervised search methods, which restricts the scope of search results and user convenience.
Innovation Solution
A hybrid approach combining supervised and unsupervised search techniques, where a classifier is trained on a selected portion of content and unsupervised methods are used to identify additional relevant content, allowing for reduced training data requirements and expanded search results, including the use of speech recognition and ontology supplementation to enhance search queries.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If supervised search methods are used with trained classifiers, then search accuracy for previously classified content is improved, but the ability to identify newly added content and reduce training data requirements deteriorates
Solution Approach 1:
The patent combines supervised search (using trained classifiers) and unsupervised search (using word search algorithms) into a hybrid search system. The supervised component maintains high accuracy for previously classified content, while the unsupervised component enables identification of newly added content without requiring retraining of classifiers. This merging resolves the contradiction by allowing both precise classification and adaptability to new content simultaneously.
Solution Approach 2:
The system dynamically selects and weights search results from both supervised and unsupervised approaches based on query characteristics and content availability. This dynamic adjustment allows the system to optimize between using trained classifiers for accuracy and using word search for discovering new content, adapting to different search scenarios and resolving the static limitation of purely supervised systems.
2Measurement precision
If large volumes of training data are used for classifier training, then classification accuracy is improved, but training time and computational resources increase
Solution Approach 1:
The patent applies partial action by using unsupervised word search to handle portions of the content repository that do not require classifier training. This allows the system to achieve comprehensive search coverage without investing computational resources in training classifiers on all content, thereby reducing training time while maintaining search effectiveness through the complementary unsupervised component.
3Ease of operation
If speech recognition is used to process user queries, then user convenience and accessibility are improved, but speech recognition errors and ambiguity increase
Solution Approach 1:
The hybrid search system acts as an intermediary that processes speech recognition results through both supervised and unsupervised search pathways. When speech recognition produces ambiguous or erroneous results, the unsupervised word search component can still identify relevant content by matching recognized words against the content repository, thereby compensating for speech recognition errors and maintaining reliability while preserving the convenience of voice-based querying.
Data Source
AI summary
According to some aspects, a method of searching for content in response to a user voice query is provided. The method may comprise receiving the user voice query, performing speech recognition to generate N best speech recognition results comprising a first speech recognition result, performing a supervised search of at least one content repository to identify one or more supervised search results using one or more classifiers that classify the first speech recognition result into at least one class that identifies previously classified content in the at least one content repository, performing an unsupervised search of the at least one content repository to identify one or more unsupervised search results, wherein performing the unsupervised search comprises performing a word search of the at least one content repository, and generating combined results from among the one or more supervised search results and the one or more unsupervised search results.