What is X up to? - NER and Relationship Extraction for Information Extraction
summary
A concrete pipeline extracts named entities and classifies pairwise relations into a graph, using transformer-based NER and relation classifiers (tokenizers to vectors, pairwise attention), plus practical tooling for scraping with BeautifulSoup, indexing with Elasticsearch, and a Dash app for visualization. Demo processes Reddit, Wall Street Journal and bioarchive text, deduplicates mentions but still mislinks cases like GME vs graduate medical education, and reports about 42 visiting relations, showing the system works end-to-end yet needs entity linking and domain fine-tuning.
AI-written summary · about
chapters (10)
- 0:00Information extraction intro
- 1:11Why extraction matters
- 4:46Named entity recognition
- 7:36Entity linking and graphs
- 10:24Implementation pipeline
- 13:54Data acquisition
- 16:29Storing extracted data
- 17:57Applying NER and relations
- 20:03Building knowledge graphs
- 26:39Summary
related talks
58mRelation Extraction from Online PH News Source
01h 30mRelation extraction from texts
34mExposing Dark Data in the enterprise with custom NLP
49mNamed-Entity Recognition and Name Resolution Using Similarity Queries
33mXAI meets Natural Language Processing
37mMachine Learning models utilization
35mUnderstanding the News around the World with Web Scraping and NLP at Scale
39mHerding Entities- Information Search and Synthesis in the Context of Transaction Data