Relation extraction

I was thinking about what steps are required for building a algorithm for accomplishing the following task:

Suppose you were given a text, and you wanted to extract named entities (NER) and further you want to get the text passages that are related to this named entities (coreference?). Now, what step would be required to build such a system?

Afterwards, one had names (of persons, companies or even products) and the corresponding references to these words like "is friendly" or "went bankrupt" etc.

What NLP techniques are necessary for this and how should they be combined?

Edit: I see my question is not specific enough. I try to give an example. Suppose, I had given the following text (copied from a tech review of a tech site):

But the Xoom, which sported a 10.1-inch screen, was a bit too heavy (1.6 pounds) and much 
too expensive, and the 3G and 4G models were available 
only through Verizon.

This sentence describes some features of the Xoom tablet. These features are

(Xoom is) a bit too heavy, too expensive, available only through Verizon

As I understand coreference resolution this would be a correct result for coreference resolution? If so, you could apply a coreference resolution algorithm on this sentence which would (ideally) extract these features of the Xoom. Then, I could use those features like "a bit too heavy" to predict wether this is positive or negative. So the result would look something like that (no ofense to Xoom users ;) ):

a bit too heavy (negative), 
too expensive (negative),  
available only through Verizon (neutral)

At the end I could say: Okay this review of the Xoom is negative.

Returning to my question: Is this technically (on the basis of currently existing algorithms) realistic? And if so, what type of algorithms would solve such a problem?