Thanks! The keywords are searched in the story title. The current (somewhat naive) approach simply splits the title by non letter/number characters, essentially getting rid of punctuation chars, symbols, etc
I see. Trying to detect keywords in an article could be nice improvement.
Another question, why did you limit the lower bound for the score w/ 200 ?
I assume you start indexing the HN submissions. In the long run you can create a context aware version of HN, e.g. 'Bring HN submissions with keyword X submitted during the last week.' This could be a brand new way to browse HN.
Quick question : Do you search for the specified keywords in the title or in the link content ?