Section Extraction from Texts
summary
Presents document sectioning as a sequence labeling task and compares different labeling methods. Evaluates Hidden Markov Models and Conditional Random Fields enhanced with word embeddings, and compares those to various RNN-based models to show the state of the art for extracting text sections.
AI-written summary · about
chapters (17)
- 0:00Welcome
- 3:02Client document features
- 7:56Why extract sections
- 9:23Sequence models
- 12:11Plain text processing
- 14:44Word embeddings
- 15:47HMM for tagging
- 21:56CRF vs HMM
- 22:18Neural sequence models
- 25:27LSTM + CRF model
- 26:54Evaluation metrics
- 28:43Results visualization
- 29:47Unknown words handling
- 30:49Characters for rare words
- 32:14Conclusions
- 33:37Literature challenges
- 35:02Q&A
related talks
39mRNN sequence labeling for document parsing in Tensorflow
44mExtracting Structured Data from Legal Documents
58mAutomatic text summarization
38mThe Evolution of Supervised Machine Learning Approaches for Text Analysis
29mDetecting anomalous sequences using text processing methods
41mAutomatic Text Categorization
30mStarting deep learning projects without enough labelled data
01h 31mNatural Language Understanding: Foundations and State-of-the-Art