Google provides impressive search capabilities across domains. This chrome extension attempts to mimic that behaviour within a tab so that the user can search content in their current page.
There are three features to this extension: locating image/video and searching text.
The omnibox (the area where you enter the url in Chrome) is where you would enter your query and it will attempt to find the 10 closest matches. Initially, the text matching algorithm consisted of a combination of Levenshtein distance and n-gram. However, it was extremely slow and produced inconsistent results such as matching words with similar spelling instead of meaning (eg "cat" with "hat").
After experimenting with several thesaurus APIs, Big Huge Thesaurus was chosen as it had a larger maximum daily requests limit and it returned JSON objects from REST calls. Using synonyms of search query had several advantages as seen in the video where the query "beaver" matched "rodent" on the page.
What's next for Delphi
The image and video locator could use have more fine tuning as currently it is traversing the DOM and searching for specific node types. A more accurate matching algorithm should be used based on well known data analysis techniques.