Hierarchical document representation for summarization
Most extractive summarization models usually employ a hierarchical encoder for document summarization. However, these extractive models are solely using document-level information to classify and select sentences which may not be the most effective way. In addition, most state-of-the-art (SOTA) mode...
Saved in:
Main Author: | Tey, Rui Jie |
---|---|
Other Authors: | Lihui Chen |
Format: | Final Year Project |
Language: | English |
Published: |
Nanyang Technological University
2022
|
Subjects: | |
Online Access: | https://hdl.handle.net/10356/157571 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Nanyang Technological University |
Language: | English |
Similar Items
-
Deep metric based feature engineering to Improve document-level representation for document clustering
by: Xu, Liwen
Published: (2022) -
Deep learning-based automatic document categorization and organization
by: Foo, Shawn Nicholas Say Yan
Published: (2021) -
TFIDF meets deep document representation : a re-visit of co-training for text classification
by: Chen, Zhiwei
Published: (2020) -
Automated source code summarization via transformer
by: Viswen Kumar Mariammalle
Published: (2021) -
Reddit post summarization based subreddit analysis
by: Ng, Jesline
Published: (2023)