We use Elasticsearch for the search in Community Analytics; for developers it is extremely nice and quick to set up. It takes some tweaking to get proper human-friendly results (this is an endeavour we haven’t yet undertaken at CA); but once you have it going it is very customisable.
Why do you need the search functionality? What should it provide? Is this for the end-users? Why do you want them to be able to search this well, what do you expect them to be able to find that they can’t find right now?
We have a very old search now. it’s based on Fulltext / Sphinx. if a member posts a thread "Prince announced dead’ (RIP) a new member searching for this info won’t find it until after approx 15 mins of it being posted. Thats is of course terribly slow and would result in multiple postings. We have members upset that they can’t search for their content directly after posting it. (totally understandable)
Elasticsearch will be instant.
But what we are very keen on is its ability to be configured to provide “search facets” (they are called aggregates in ES) relating to a product database we maintain.
Learn more about facets in the links below
We are also interesting in relevancy tuning of Elasticsearch
We have a mixture of old and new content
We could segment search results like this
Latest - relevant - display most recent at top - click to see more Recent past - relevant but with different weighting - perhaps more on key words - click to see more "From the archives" - with different relevancy weighting - favoring perhaps most read / most liked / most popular - click to see more
I feel this might go some way to solve the issues my theory raises, namely that a long running discussion community forum is like a giant planet of dust and gas. How can a casual visitor retrieve useful information from the centre of it?
If users are presented with results from the surface of the planet alongside results from the middle and the deep core of the planet - they might then get an at a glance overview of the huge potential of both recent and legacy information available to them.
But results brought up from the core of the ‘planet’ had better be good! It would be a shame to present legacy discussions that were trivial and non relevant!
I am excited about the potential of adaptive search parameters that change the older the content is.
Yes, a delay in search isn’t acceptable anymore in this day and age
Faceted search is really powerful, and it’s something that ES does extremely well. (do make sure to disallow the links in your robots.txt, it can easily blow up since it’s an infinite path)
Sounds like you thought it through pretty thoroughly!
Are you planning to show relevant posts as well? I think stack overflow and discourse do something like this to prevent double posts; could be interesting in a large forum such as yours.
I have learned that fancy search costs lots of money.
Quotes of £40k seemed average
I have entreprise tastes on a small business budget.
Did it take long to get Coveo to do what you wanted?
How was that journey?
Did you have to hire Coveo expert / search programmer to get it configured?
But I have been recommended ElasticSearch and to not host it myself but to pay for a hosted version. I bought my developer the definitive handbook for ES and have a few emails of developers that I could hire to get us up and running with it.
ElasticSearch is free but you need developer to install and configure it.
The Xenforo platform has a basic ElasticSearch available (as an add on for $100 I think it’s called “Enhanced search”) its a nice, fast forum search but it doesn’t offer the ‘search facet’ functionality I was after - so my set up is custom configured. The developers 'learned" Elasticsearch (and I bought them a big fat manual for it)
In this screenshot you can see 3 custom facets and one automatic facet (the monitors between $1,000 - $2,000 link)
I will be writing guides and “how to” articles for the top search querys
With facets I can always link to them.
Lest you think this is just a keyword trigger system (the above example is) But it’s much more, the system has been primarily configured to expand user journey 'discovery" in relation to our 35,000+ product and product category database (but thats all too boring to write out here and my fingers would drop off! )
Here is a brand name search, for Neumann Microphones, lets say the user has misspelled it too
Search term: Neuman (missing an end at the end)
As well as displaying threads discussing Neumann products it has made suggestions to visit 5 product pages we have for the products that act as ‘hubs’ to display links to spec, reviews, videos, discussions, purchase links and more.