NLPIR/ICTCLA2018 ACADEMIC SEMINAR 8th ISSUE COMPLETED
Last week, Shen Li gave a presentation about the paper, Design Challenges and Misconceptions in Neural Sequence Labeling and shared some opinion on it.
After the presentation, several questions were asked. The Q&As are listed as following:
Q: What is oov, oobv and ooev?
A: oov is Out-of-vocabulary, oobv is out-of-both-vocabulary, ootv is out-of-training-vocabulary, ooev is out-of-embedding-vocabulary.
Q: How CNN and LSTM work?
A: The authors constructed a neural sequence labeling framework with three layers , a character sequence representation layer, a word sequence representation layer and an inference layer. LSTM and CNN are used on the character and word level, to caputure sequence representation information without hand-engineered features. More detailed information can be seen in figure 2 and 3
The powerpoint of the presentation can be download.
