Hacker News new | comments | show | ask | jobs | submit login
Ask HN: What are search tools that works for the average user?
4 points by ChrisCinelli 7 days ago | hide | past | web | favorite | 7 comments
When it comes to search on internal company documents or any no completely open document, the standard search feature in these days is based on Elastic Search. Most of the search APIs out there are based on Lucene or Solr and, at best, they do synonym matching and edit distance to deal with misspelled words. But they still just look at how many words in the query are matched for the ranking.

These engines work if the user know how to search and who write content knows how to be found. If you look at the searches of a person that was not using the Internet until 6-10 years ago, they are not trained to search on those engines. You would be amazed looking at what they type in a search box.. Unfortunately, Google “delivers” on those queries so they do not have to learn.

And by the way, we all became spoiled by Google a little bit. For example I rarely end up changing my query on Google to find what I want while I used to have to try at least 4 or 5 similar phrases before.

Confluence wiki's I think is based on Lucene and the search is terrible even if you know how to search.

Try to put "tool to fix a clog" in Amazon and contextualwebsearch.com (I think they use Lucene) to laugh out loud. Compare them with Google. Bing does pretty well too on this query.

Is there anything that goes even close? I prefer something opensource but I am open to close source solutions.






Somebody brought up also Swiftype. Their brochure site is good to sell the product. It seems to have a bunch of good features.

I usually test a few alternative and see how they work and pick one base on some criteria. The main reason I did not did that this time is that I could not see anything that seem going a lot further than the usual ES.

I am curious to hear from somebody that had the same problem and found a satisfactory solution. It would also be interesting to hear from those that tried with modest results and hear what they end up setting on.


Tried Algolia?, HN use them. They offer some nice features.

For finding stuff:, sometimes moderation is needed. Check Information Architecture articles to see how to index and catalog information. Some solutions are: site maps, and an alphabetical index, metadata, etc. And as mentioned, sometimes the best solution relies in: Mods and curators.


Thank you for the input. I know them but I never used Algolia except hn.algolia.com - It seems ok but I usually need to be creative and go through a bunch of pages when I am looking for something. I think Algolia is also using elastic search. Do you use them? How much are they better than an ES cluster?

(algolia employee) Hi Chris, Algolia is actually not using Elasticsearch. We've been building our own search engine from the ground up to target the very (end-)user-facing search use-case (big focus on UX, speed & typo-tolerance) while one of Elasticsearch's sweet spot is more billion of line of logs indexing, aggregations & filtering.

You can read more about the differences on Quora, our CTO answered there a few years ago: https://www.quora.com/How-does-Elasticsearch-relate-and-or-c...


Not sure why, but you need to remove the Ad-Blocker to see our CTO (Julien)'s answer /o\

Just sign-in in Qura or add ?share=1 in the URL...

I read his answers and a few others. Thanks! It sounds it is definitely worth to test it.



Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: