Yes, "dumb" as in simplistic algorithm insufficiently informed by NLP AI not as in insulting either the designer or the coder. When selecting a limited sample of articles from a source, an article with a title that precisely matches the title of the Factbook entry, ought to be given highest priority for selection. Instead I am given articles that brief mention the topic rather than discussing the topic. I would expect a routine that tosses around terms like TF-IDF weighting, BM25 transformation,, pivoted length normalization ... i.e. there are standard algorithms designed to avoid this "dumb problem". And yes, I am unlikely to know the most current solutions - I'll plead being a bit dumb on the topic. [;)]
