Overview:
There is so much content in the Internet that its availability becomes its own problem, we still need to manually extract information from each content. Sure there's a lot of ways this is being addressed like use of meta data, tags, web semantics, micro-format but it only yields data. As example, a news article has headline and title; its tagged and categorized. We used web semantic analysis and find Persons, places, events and etc. Still we dont know anything really about it, how does the mentioned persons related to the places and events. Where they on it or referring to it? How about answering "Is the article a good or bad news?" To whom is it good to and by what degree?
Proposal:
I propose a project that would allow Drupal extract gist INFORMATION from nodes using Semantic Web (like http://www.opencalais.com/about) assisted Natural Language Processing (maybe using http://web.media.mit.edu/~hugo/montylingua/).
It will basically be available as an API and pre-integrated with Views. This makes it possible to extract known web semantic terms from a node and is relationship; example:
Subjects by relevancy:
Health_care(+10), Obama(+9), Biden(-2), etc
this says the current node is related to the above subject, its say +10 good about Health care, scores +9 good to Obama and scores -2 bad for Biden.
If you drill down on Health_care(+10) to find the data supporting "facts" it will reveal something like:
Supporting facts to Health_care(+10) by score:
1. "Its a big_fucking deal" (+7)
2. "Biden to Obama at Health_care" (+5)
3. ....something..something.. (-2)
So why the need do all these? Blame it to SEO where they keep polluting the net with content, using search keywords, tagging it but when you read it it actually doesn't make sense, just snippets of articles jumbled to create new content -- but that's my personal rant :) But who wouldn't want it, it extract the gist of the article. No need to figure out whats its all about or be duped by catchy titles that lies.
Output:
1. API
2. Views integration, samples
3. Pivot browsing?
Overall difficulty: Hard Core
Fun rating: "non-non-non heinous!" - by Bill & Ted :D
- Feel free to add or revise the idea esp on how output is presented.
Comments
The need for this exists, but
The need for this exists, but everything will hinge on the implementation and ease of use. I know Calais already does this, but organizations that handle sensitive information are skeptical about sending it outside their own firewalls. As such, being able to analyze this information using their own software would prove to be immensely valuable.