Explore scientific text mining and knowledge graphs in this comprehensive conference talk from KDD 2020. Delve into phrase mining techniques, including quality estimation and phrasal segmentation, with a focus on the SegPhrase and AutoPhrase algorithms. Learn about named entity recognition (NER), covering supervised methods, neural language models, and BERT. Discover distantly supervised NER approaches like SwellShark and AutoNER, with comparisons in the biomedical domain. Examine meta-pattern mining for information extraction, including pattern grouping and type adjustment. Investigate the PENNER framework for pattern-enhanced nested NER in biomedical literature, featuring weakly-supervised pattern expansion. Gain insights into cutting-edge techniques for mining and analyzing scientific text, presented by experts from the University of Notre Dame and the University of California, San Diego.
Scientific Text Mining and Knowledge Graphs - Part 2-1
Association for Computing Machinery (ACM) via YouTube
Overview
Syllabus
Why Phrase Mining?
Phrase Mining: A Keystone
Quality Phrase Mining from Massive Domain-Specific Corpora
Quality Estimation using Expert Labels
Phrasal Segmentation using Viterbi Algo
SegPhrase (SIGMOD'15): Quality Estimation Phrasal Segmentation
SegPhrase (SIGMOD'15): Reliance on Expert-Provided Labels
AutoPhrase (TKDE'18): Negative Sampling from Noisy Negative Pool
Phrase Mining: Empirical Evaluation - Precision Recall Curve
AutoPhrase (TKDE'18): Results of Chinese Phrases from Wiki Articles
What's Named Entity Recognition?
Supervised Methods: Training Data
Supervised Methods: Neural Models
"Data-Driven" Philosophy
What's (Neural) Language Model?
Neural LM: Example Generations
BERT: Introduce Transformer
Questions
Distantly Supervised NER Methods
SwellShark: Distantly Supervised Typin
AutoNER: Dual Dictionaries
AutoNER: Tailored Neural Model
Comparison - Biomedical Domain
Summary & Q&A
Meta-Pattern Mining for Information Extraction
Our Meta-Pattern Methodology
Grouping Synonymous Patterns
Adjusting Types in Meta Patterns for Appropriate Granularity
PENNER: Pattern-Enhanced Nested Name Entity Recognition in Biomedical Literature
Framework Overview
Weakly-supervised Pattern Expansion
Comparison with Pub Tator
Taught by
Association for Computing Machinery (ACM)
