Introduction:
Item Similarity Finder is an AI-enabled service from Excelsoft. It allows item authors to look for items similar to a given item in a repository. The service can be used while authoring a new item or periodically scanning the repository to identify similar items. Authors can then weed out these similar items or modify them to reduce the similarity.
What's under the hood?
- The Item Similarity Finder service leverages both semantic and elastic technologies to identify and group similar items.
- Items are encoded into semantic vectors, and then cosine similarity is used to identify and group similar items. The cosine similarity metric provides a measure of semantic closeness between items.
- This process is a form of semantic search, as it goes beyond exact keyword matching and considers the semantic content and meaning of the items.
- A slider in the user interface allows users to dynamically set a similarity threshold. Adjusting the threshold controls the sensitivity of the grouping, providing elastic control over the analysis.
- The application forms groups of similar items based on the calculated semantic similarity. A set is used to efficiently track and avoid redundant groupings, contributing to the elastic nature of the analysis.