TF-IDF query engine in context of terms weight

I'm looking for public algorithm which gives the engine these abilities:

• Query by ranked terms
• Limit outcome by date/time range

Basically, i'd like to concentrate articles (generally title|text|timestamp) identify the source and make N-N correlation to terms (is term for datasource same marking as term for dataentry?)

Given the database of such information

entry_data_type:[type_id|title|description]
entry_data:[entry_id|data_type_id|data_content]
entry:[id|entry_type(data,source)|parent_entry_id|created|updated]
terms(keywords):[id|keyword]
entry2term:[entry_id|term_id|term_weight]


Where keywords are both automatically defined (text frequency analysis) and manually assigned (probably abstract terms in context to entry contents)

I should be able to query by keywords like this: kw1:3 kw2:10 kw3:-2 [range:-7 days]
and output shall be entries sorted by given keyword weights (pattern keyword:weight)

I thought about something similar to EdgeRank, but that is social-graph-oriented, and I'm looking for more straight-forward solution (more selfish, meaning input filter is given by personal preferences, not social-graph-near preferences or social-score ranking)

Also TF-IDF would have to be limited by time, so the document base to calculate the entry score is inserted in given date/time range only. Is there any possible break-down of TF-IDF ranking, eg. to pre-calculate raw-data for each day and then, based on query, merge them for given date-range?

This question is independent of any particular programming language, platform, etc. I'm generally looking for keywords to look for, papers to read or ready implementations to study, but accepted are only answers not using paid or closed-source software parts or non-public-domain patents.

• What does "technology agile" mean? – David Richerby Apr 7 '14 at 16:21
• @DavidRicherby that means, the question is not specific to used programming language, platform, etc.. – Marek Sebera Apr 7 '14 at 16:38