Towards Contextual intelligence in Natural Language Processing
Rinaldo Rex (~rinaldo-rex) |
Starting from Eliza, to today’s Google/Amazon/Microsoft/Apple virtual assistants, we’ve come a long way in Natural Language Processing. However, it’s still not about the extreme level of processing that makes a system intelligent. It’s Context and Common sense.
The talk is going to explore how to bring about the contextual intelligence and emulate common sense in such a way that it will further the development of NLP based products/systems in the future.
NLP is one of the frequently discussed topics when it comes to AI. Of course, artificial intelligence is at par when you can 'literally' communicate your intentions to a system. We've been dabbling with NLP for half a century now, and it's important to understand that we're closing in on understanding how language works and how we can replicate it, artificially. Just like how 'customer is king' in any business, 'context is king' in any NLP task. And we need to work towards contextual awareness.
Outline for the talk:
- Gist of what NLP (NLU and NLG) is. (~3 mins)
- How modern trends started and evolved in NLP (~7 mins)
- Starting from Eliza towards modern assistants(Google, Microsoft, Apple, Amazon)
- Discussions of what the cutting edge evolution is. (BERT, ELMo, GPT-2, etc)
- Why and how knowledge graphs help in NLP? (~10 mins)
- Why context matters.
- How we can possibly achieve contextual awareness in future.
- Why knowledge graphs are critical to AI systems.
- Beginner to intermediate level python (Intermediate level preferred)
- Understanding of basic NLP terminologies.
- Understanding of Knowledge graphs.
Slides are here. (Still evolving. ;) ) A short video shall be added later.
I work as Backend Engineer at PositiveNaick Analytics, an energetic startup focused on everything AI. I love coming up with smart(as in AI) solutions to common and tricky problems, and hardcore lover on Open source. We work exclusively on NLP, and one such product out of it is the core built with advanced NLP engine on top of which we have Yekaliva, a really smart platform providing chatbot solution!
P.S: I have already given a lightning talk on PyCon 2014. (On Naive Bayes Classification)