Hacker News Reader

Hacker News

Comment by sandGorgon | Hacker News Reader

Parent Full thread

sandGorgon

Does anyone know how to play with relevance and spell corrections in FTS ?

In elasticsearch, relevance is Tf-IDF or BM25. I don't see anything similar in FTS.

Not sure how I can results around ranking as well as spell correction/suggestions simultaneously.

View on HN

As far as I am aware, you cannot get ES/Solr quality ranking + fuzzy search out of the box.

I think it depends on your specific situation and your project requirements. In my project for example I had exactly the search quality I need with Django and PostgreSQL without installing another software (Elastic), using a different query syntax, synchronizing the data and installing another layer (haystack).

Maybe in other project ES will be a better solutions.

My comment was in response to a question whether PG can do TF/IDF or BM25 and fuzzy search (handle misspelling) simultaneously. And as far as I am aware, PG cannot do it. My comment had nothing to do with your blog post.

I used Elastic in past project as I reported in my article, but I never used relevance, can you add a link to the elastic official documentation about it ?

Depending on the ES version, it either does TF/IDF (older versions) or BM25 (newer versions) by default. You have to explicitly disable it if you don't want it.

https://www.elastic.co/guide/en/elasticsearch/reference/curr...

I might shamelessly plug my own book, Relevant Search (http://Manning.com/books/relevant-search) which introduces relevance on Elasticsearch.

Ping me (http://o19s.com/doug) if you’d like a discount code!

here - https://www.elastic.co/guide/en/elasticsearch/reference/curr...

similarity/relevance is the mathematical difference between search and database query. Its an inexact match.

Thanks. I'll read it.

If you want to reason about these things from first principles, then I recommend the IR book, which is free online: https://nlp.stanford.edu/IR-book/

Reply on news.ycombinator.com