Hello.
In the past I made a software to recognize keywords, text, secuences of words, to determine if a website was or wasn´t in the interest of a client. The clients were for example politicians, celebrities, McDonnalls.
So if someone in a lost blog in internet talk about McDonnall´s this software give the page´s website a score of interest. It work like a internet spider, browsing very very fast (around 8k pages per minute I think)
I still have the program, source code and everything
So, this project sounds very similar don´t you think?