I used to work there back in the early 90s when it was still the Public Affairs Video Archives. Not long before I left, I wrote software to parse closed-captioning and generate metadata for the program. It would collect things like what part of the session congress was in, the topics being discussed, who was talking, vote outcomes, etc.
The biggest problem by far was there because it is a live program, there were a lot of misspellings that had to be accounted for, as the people doing the closed captions didn't really pay any attention to what was being said, just what sounds they heard. A lot of times, a word or part of a word, would be spelled like a similarly sounding word.
I'm not sure what ever happened to it after I left.....