Recent Advances in NLP
Anuj Gupta
Lead Data Scientist, FreshWorks
@anujgupta82
anujgupta82@gmail.com
About Me
I am a Lead Data Scientist at Freshworks; working in the areas of NLP, Machine
Learning, Deep learning. Earlier he was heading ML efforts at Airwoot (now
acquired by Freshworks) and Droom.
Speaker at forums like Ainthill Indside, PyData, Fifth Elephant, ICDCN, PODC,
IIT Delhi, IIIT Hyderabad and special interest groups like DLBLR.
@anujgupta82
anujgupta82@gmail.com
What is NLP
• Concerned with programming computers to fruitfully process large natural
language.
• It is at the intersection of computer science, artificial intelligence and
computational linguistics
Analogies
Demo
https://blog.openai.com/unsupervised-sentiment-neuron/
Sentiment Neuron
Sentiment analysis
Resurrect your dead friend as an AI
Luka - Eugenia lost her friend Roman in an accident. Determined not to lose his memory, she gathered
all the texts Roman sent over his short life and made a chatbot – a program that responds automatically
to text messages. Now whenever she is missing Roman, Eugenia sends the chatbot a message and
Roman’s words respond.
Google smart reply
Spell checking
Retail Search
• Co-reference (e.g. What does "he" or "it" refer to given a document?)
• Machine Translation (e.g. translate Chinese text to English)
http://www.spndev.com/neww2v.html
Word2Vec model on Fox News broadcasts
Artworks mapped using Machine Learning.
Art work Mapped using t-SNE
https://artsexperiments.withgoogle.com/tsnemap/#47.68,1025.98,361.43,51.29,0.00,271.67
t-SNE Applications
https://aiexperiments.withgoogle.com/bird-sounds/view/
Story generation from images
Story generation from images
https://www.youtube.com/watch?v=cersRTtjFcU
Document Vectors
Visualization of wikipedia paragraph vectors
https://code.facebook.com/posts/1686672014972296/deal-or-no-deal-training-ai-bots-to-negotiate/
● Shakespeare’s work ● Linux Source Code
● Algebraic Geometry ● NSF Research Awards abstracts
Writing with machine
“Toward Controlled Generation of Text” - Zhiting Hu et.al 2017
Dr QA
NLP is not easy !
NLP is not easy !
• The examples we have seen are cherry picked success stories.
• We are still far from building general AI to understand NLP – there
are a lot of fundamental pieces that need to be solved as a precursor.
• None the less in last 5 yrs lot of exiting work that has happened.
• A key quest has been building systems/models that are not black box.
• There are no shortcuts.
Thanks
Questions ?

Recent Advances in NLP