L~R, if I'm following you, your suggestion may be the closest yet to what I want. I'll think about that.

Not sure about the mechanics....

Is this what you mean? Take first two words, count the frequency of each such word pair against the headlines....

(BTW, this would be done after removing "stopwords" (and, the, a, and the rest of a long list)