Search terms should map to nearest term in the database

[anselmbradford]

Related search terms should map to a synonym of the term in the database. For instance, searches for "food," "meal," "dinner," etc. might all resolve to "Food" or "Food Program"

This was issue codeforamerica/ohana-web-search#18 in HSF and has been moved here.

[monfresh]

This doesn't come for free with elastic search. You have to manually configure synonyms: http://www.elasticsearch.org/guide/reference/index-modules/analysis/synonym-tokenfilter/

Since Elastic Search is not trivial to understand and implement, the easiest solution — which would yield the same benefits — would be to define additional keywords that should map to each of the categories. Once that list is defined, it would take a few seconds to add them to the DB.

[monfresh]

Also, this list doesn't have to be defined in advance. It should be data-driven. We can use Google Analytics to see which terms people are searching for, and if they don't already exist in the DB, we can add them. This can probably be automated using the GA API and a script that updates the DB.

[monfresh]

This doesn't sound like something the API should be responsible for. It seems more like a data quality issue to me. If a search for a certain term should return a particular location, then the location's description, the service's keywords and/or categories field should be updated to include those terms.

[anselmbradford]

Could this be the responsibility of a script in /scripts that goes through each entry and looks for specific keywords in the description and adds synonyms of those keywords in the keywords field if they are absent? Seems like that could be a straightforward way to handle this. Also seems setting that up may be out of scope :), but maybe as a hack request?

[monfresh]

Nyet. Such a script wouldn't help if the description itself doesn't contain the right terms. This is a data issue that should be dealt with by maintainers of the data. I'm closing this.