Wka writeup of Brin 1999

This is a review of brin_1999_extracting_patterns_and_relations_from_the_world_wide_web by user:wka.

Prefer precision over recall. Each pattern needs only small coverage; defines "specificty" of pattern.

• Ofcourse a more involved template of patterns can be used (syntactic and semantic features)
• Avoids computer-generated listings like Amazon because patterns would be cueless; SML methods solve this problem.
• I wonder (just for the fun of it) whether it would have been even more influential if the author actually took the time to answer all the questions he poses and attempt all the improvements he suggests!
  • Divergence
  • Handling bogus extractions
  • Relation to SVDs
  • Low recall of specific category (science fiction)

[In-Class note: compare to SEAL?]