In Stackoverflow's Podcast 32 Jeff Atwood discusses how "related questions" feature works. I would suggest listening to it for more detail but basically he says that it is a combination of comparing question tags, question title and the question text.
However, he takes all the question's text and removes the top 10,000 most common English words. What is left are basically unique words related to that question with all the filler-english removed.
So while I don't have linq-to-sql information for you, hopefully this can get you on the right track.
与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…