A search bar for testing different visual cues of data matching. Try it out! http://kchovercam.com/search-bar/
This experiment's goal is to gauge user response to finding similar words in the search results. User’s responses could be used to inform a good UI design that made it clear what was happening with their search query.
The search bar differentiates between complete words and incomplete words. For incomplete words, a word is autosuggested.
For complete words, exact matches are bolded, and similar matches are in italics. The results dropdown and similar words on the side of the page are updated dynamically.
User testing can be found at the end of the Summary Report
- Grab the user’s query.
- Split it into words.
- Discard less useful words according to a list from http://www.stopwords.org
- Categorize the words.
- Every word but the last is trusted as a “complete” word. (Unsafe, to be sure.)
- The last is either “incomplete” or “complete”, as determined by the results from https://api.datamuse.com/sug?s=word
- Generate related words.
- “similar” words are generated from sending “complete” words to https://api.datamuse.com//words?ml=word
- A “suggested” word is generated from sending “incomplete” words to https://api.datamuse.com/sug?s=word
- Assign scores to the words, to simulate relevancy.
- “complete” words are worth the most.
- “similar” words are worth almost as much as complete words.
- “incomplete” words are not worth very much.
- “suggested” words are worth even less.
- Each word found in a data entry adds that word’s score to the entry’s cumulative score according to a few rules:
- The first word found is worth more than the next words.
- If the first character matches, the entry receives a lot more points. (To reduce the ‘tree is in street’ problem.)
- Title matches are worth more than summary matches.
- Sort the data entries by their score and display them to the user with matching words highlighted.