i guess the question is simply, how to predict searches based on a history?
or maybe it's a question of doing ai with elasticsearch?
i'm not sure. i just know i've got tons of historical data and i figure it
would be a decent idea to see if i could use this data to generate searches
instead of doing it myself. since this data was generated with a human, i'm
guessing there will have to be some fudge factor built in if i can do this.
i guess, what i want is a netflix recommendation like solution - i've
looked at certain things in the past, i'm likely to like certain things. if
someone could explain to me (or reference) what the theory is behind this
process, that might help.
On Sunday, August 26, 2012 4:44:13 PM UTC-4, Otis Gospodnetic wrote:
I don't follow 100% what you are after. Are you after a search analytics
tool? If so, see http://sematext.com/search-analytics/index.html . If
not, please clarify.
On Sunday, August 26, 2012 1:33:11 PM UTC-4, shawn wilson wrote:
for over two years, data from a web site was mined by hand. i created
a scraper to do it and can search pretty well. however, since i have
so much historical data about what i'm interested in, is there
anything i can use to generate what search should find data like it?
in the reports generated by hand, they put the url the data was
grabbed from, so i can use that field to refer to data i want.
i'm thinking that fully automating this might be impossible (or more
time consuming than figuring out good searches on my own), so maybe
some hints as to how i might use the analytics to look into this more