Hacker News new | ask | show | jobs
by teraflop 4088 days ago
To be a little bit pedantic: in NLP terms, this is a sequence labeling problem, not parsing. "Parsing" is normally only used for methods that can handle CFGs or similar nested structures. This tool just seems to be breaking up a string into a linear sequence of labeled segments.

I don't mean that as a criticism of the project itself, though. The demos alone are pretty cool, and the framework looks incredibly useful.

1 comments

you're incorrect. To be more pedantic, sequence labeling is a special case of parsing. Parsing need not be hierarchical. Chunking, aka shallow parsing, is a sequence labeling task.

Parsing is commonly understood to infer an hierarchical structure, but that's not required.

Indeed, the Latin root leads us to understand that parsing is about breaking an object into parts. That these parts nest according to a particular grammar is something of an important implementation detail.