Polyglot AI: The Role of Natural Language Processing (NLP)
summary
Argues that practical polyglot enterprise AI must master both human and company vocabularies and that companies operate in 170+ languages, so scale matters. Shows a working pipeline using multilingual BERT plus top-layer fine-tuning for domain and company customization, reports zero-shot 57% sentiment accuracy in an untrained language versus 54% baseline and 64% with some labels. Presents a rule path using Annotation Query Language for low‑data, privacy‑sensitive domains and a human‑cocreation loop that refines machine‑generated rules.
AI-written summary · about
chapters (12)
- 0:00Intro
- 1:12Key Takeaways
- 2:37Complexities of understanding human language
- 3:462019 was a defining moment for NLU – computers are passing humans on some basic tasks
- 5:25Enterprise Al: From Process Automation, Virtual Assistants, Recommender Systems, to Decision Support Systems
- 6:33Enterprise data is complex and needs domain understanding
- 7:51How about narrowing the problem to make progress? As the domain narrows, learning accelerates!
- 9:19General purpose Vs. Customized Speech
- 11:20Conventional Approach towards Language Enablement
- 17:02Semantic Role Labeling (SRL)
- 17:12Cross lingual Information Extraction
- 17:23Polyglot Information Extraction using Rules
related talks
17mNatural Language Processing in Automation
55mDemystifying Natural Language Processing across several industry verticals
45mHow multilingual is your NLP model?
43mBreaking the language barrier: how do we quickly add multilanguage support in our AI application?
26mNatural Language Processing (NLP) going mainstream
06mAdvancements in Natural Language Processing (NLP) and Text Analytics
22mEnterprise NLP with BERT | Optimized Text Analysis Framework
01h 16mGeneralization in natural language processing