Anthill Inside 2017

One of the main problems in NLP is the Named Entity Recognition(NER).The NER problems are addressed using traditional Machine Learning techniques, It mainly involves feature representation(Common step in all NLP problems), and then make use of a ML classifier to train and predict the correct Named Entity. The evolution of better feature representation methods and RNN based neural networks really helped to improve the accuracy much better than the traditional methods.

Outline

This talk covers following points,

1. How to approach the NER problems.
2. Need for feature representation.
3. Type of feature representation methods (Just passing over it, as there is a worshop entirely based on this topic.)
4. Window based classifier (Kind of 1-d convolution) applied on features. (Detailed analysis)
5. How some hand tuned features helps the standard classifiers.
6. RNN based models to improve more context awareness and gives better accuracy than other models. (Detailed Analysis)
7. Other tweaks and suggestions.
8. QA

Requirements

ML basics, Some idea about NLP and Deep learning.

Speaker bio

Senior Development Engineer part of Imaginea Labs team in Pramati Technologies. Involved with solving different types of Machine learnings problems and integrate ML based services to existing products to enhance the buiness value of the product. Have worked in end-to-end Machine Learning pipeline. Right now more focused with the Deep learning techniques to solve Image classifications, NLP problems like Sentence classification, NER, seq2seq, etc. I would like to use this opportunity to share my experience and learn further.

Links

• https://docs.google.com/presentation/d/1MflzUKCBwj1Ic0CAr1hmqz9zNXjvjZAGdcDaXcFfjsE/edit?usp=sharing
• https://docs.google.com/presentation/d/1iZ8TszkxdOHw6ZrJZrGz5v1rTR4wezCpNXjJqY-5Yts/edit?usp=sharing
• http://haridas.in
• http://github.com/haridas