How does Dynamic Signal sort search results?
Search results are sorted by relevance. In the web experience, members can choose to receive the results in chronological order (Oldest to Newest or Newest to Oldest).
How is search relevance determined?
Relevant search results are based a modified TF/IDF (or "term frequency") score for each search term.
How does term frequency work?
Term frequency calculates how often a term occurs in a document. The search algorithm is weighted by (i) how rarely a term occurs in all documents in the search index, and (ii) by field length. For example, in a search for "Bill Gates," if the post title is 10 words, the author name is 2 words, and the content is 600 words, relevant results are sorted accordingly:
- A post written by Bill Gates
- A post that mentions Bill Gates in the title
- A post that mentions Bill Gates only in the body.
A post's category is not included in weighting.
What fields are included in the search algorithm?
Several fields are searching, including title, author name, description, and content. The highest field score is used as the post score. This prevents a post that has the same text in its title, document, and content from getting an excessively high score.
What about variations of a word?
We have an edge-ngram analyzer on the fields we search, meaning it will index the first 1-10 characters of each word as a search term. For example, `onboards` in the post will be indexed as `o`, `on`, `onb`, `onbo`, `onboar` , `onboard`, and `onboards`. The more a variation occurs in the search results, the less weight its given for relevance.