Open jlepird opened 7 years ago
I wasn't exactly sure what you meant when you referred to elastic search, but I think I get the idea now. Besides what we showed for 61A billets, what additional info would you show that are unstructured?
It'd probably just be the same unstructured description. Even giving them the ability to do a keyword search would be helpful-- a really cool solution would be to have a D3-driven wordcloud that people could use to filter down on, dc.js style
Should we do the preprocessing to have the structured data? I think most jobs would have similar attributes
It might make sense to have a parallel data structure that contains preprocessed text-- probably stemmed, punctuation + stopwords removed, etc. You could also build a table that lists (billet, word, count) triplets. This table would make it super easy to query the most popular words of the remaining billets
I'm building a python-based backend (https://github.com/jlepird/tm-backend) that we can use to run tasks like this. It'll have access to the same SQL database as the front-end and I'm configuring it now to be REST-ful as well
As a user, I need to be able to query the unstructured text fields so that I can find information I need to prioritize the officers/billets I'm interested in.