Search String Construction Using Item Action Actor Tuples
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image searching technologies fail to effectively convey the essence of an item, particularly for non-reading users who struggle to create appropriate search strings and often receive images of product packaging instead of usage illustrations, leading to inadequate information retrieval.
Innovation Solution
A computer-implemented method for constructing, evaluating, and improving search strings by using tuples of item classes, actions, and actors, with techniques such as ontology siblings, n-grams, and word embeddings to generate alternative search strings that better indicate item use, thereby enhancing image retrieval results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional image searching is used with simple item names, then search speed is fast, but the retrieved images do not accurately represent item usage (relevance is low)
Solution Approach 1:
The search string is segmented into structured tuples containing item class, action, and actor components. This segmentation allows the system to systematically construct comprehensive search queries that capture different aspects of item usage, improving retrieval relevance while maintaining manageable complexity through modular construction.
Solution Approach 2:
An intermediary processing system is introduced that automatically generates and evaluates multiple search string alternatives using ontology siblings, n-grams, and word embeddings. This intermediary layer handles the complexity of search string construction, allowing users to benefit from improved relevance without directly managing the complexity themselves.
2Measurement precision
If multiple search string alternatives are generated and evaluated, then image retrieval accuracy improves, but processing time increases
Solution Approach 1:
The system generates multiple search string alternatives beyond what a single query would provide, evaluating each against image similarity thresholds. By performing partial evaluations and stopping when sufficient results are found, the system balances thoroughness with efficiency, improving accuracy without excessive time consumption.
Solution Approach 2:
The system performs preliminary generation of search string alternatives using ontology siblings, n-grams, and word embeddings before the actual image search. This preliminary action prepares multiple candidate queries in advance, allowing the system to quickly evaluate and execute the most promising search strings, reducing overall processing time.
3Measurement precision
If image similarity comparison is performed to evaluate search strings, then search quality improves, but computational complexity increases
Solution Approach 1:
The system substitutes manual or simple image comparison with automated image similarity computation using algorithms that calculate similarity scores between retrieved images. This substitution handles the computational complexity automatically, providing accurate evaluation of search string effectiveness without requiring manual intervention or overly complex processing systems.
Data Source
AI summary
Examples of techniques for constructing, evaluating, and improving a search string for retrieving images are disclosed. In one example implementation according to aspects of the present disclosure, a computer-implemented method includes receiving, by a processing device, a plurality of images as search results returned based at least in part on a search string for an item in the form of a tuple including an item class, an action and an actor. The method further includes determining, by the processing device, whether the search string is effective at indicating a common item use based on image similarity. The method further includes, based at least in part on determining that the search string is ineffective at indicating the item use, generating, by the processing device, an alternative search string.