- [[direct quotation model]]
- Similar to [[attention mechanism]], but attention itself needs to be fed back to the human
- [[Viterbi algorithm]] used in [[hidden Markov model]]
- [[Sentence granularity]]
- I need something a little longer than keyphrase extraction.
- I want it to be natural as "[[speech]]."
- With key phrases, it's too short.
- It’s a little tricky when it’s ppoi-based.
- Multiple discriminators are trained from a single data set.
- It’s a big deal to create the data naively, because it’s an identification task for each character in a sentence.
- 280,000 lines
- Interactive mode is slow.
- Downsampling implemented.
- I took the 5 letters before and after each point and put them in unknowns, but most of them are negative examples, and it’s hard to make a teacher out of just 5 letters.
- Is this a pattern that requires making some teachers in advance?
This page is auto-translated from /nishio/注目部分を切り出す using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I’m very happy to spread my thought to non-Japanese readers.