- Predicting Discourse Structure using Distant Supervision from Sentiment [EMNLP 2019] Patrick Huber, Giuseppe Carenini.
- In DP Task, the corpus is so small that NN method not work in this task.
- But the annotated is very expensive, so need a method to generate the dataSet.
- And the Discourse Parsing Task is a fundamental task in NLP which can be use by some downstream task like sentiment analysis.
- So having a mind whether can use the sentiment info to generate the dataSet.
-
- Using a big sentiment to train a SOTA Sentiment Analyzer Model.
-
- Using the SOTA Hidden -> Generation CKY-style tree.
- Evaluation in inter-domain corpus to prove the transfer info is useful.
Repository navigation
DiscourseParsing
Directory actions
More options
Directory actions
More options
聽
聽