Speaker "Pranjal Daga" Details Back



Deep Learning for Speech Recognition


In the field of Automatic Speech Recognition (ASR), the state of the art for generic conversations have reached super human levels. However, things are not nearly as good in specialized knowledge domains: attempting to transcribe vendor-customer or intra-vendor conversations often results in high double-digit error rates. Considering the low performance of ASR on real data, it becomes imperative to customize the end-to-end probabilistic model instead of analyzing Language Model and Acoustic Model separately. This session will focus on discussing grapheme based end-to-end Recurrent Neural Network architectures which can transcribe audios directly. We will also have a reality check to reduce latency by tweaking the model during inference time.


Pranjal Daga is a Data Scientist focused on strategizing and developing Deep Learning Proof of Concepts at Cisco Services Machine Learning R&D. He graduated from Purdue University with a Masters in Computer Science, specializing in Deep Learning and NLP. In the past, Pranjal has worked with researchers at Adobe Research, University of Alberta, Northwestern University, IBM Research and MIT. He has spent the past few years exploring new technologies, hacking on the quirky side projects and bringing together his research and engineering experiences. To understand the practical aspects of identifying business ideas and moving them forward, Pranjal recently joined Stanford Graduate School of Business' Ignite program.
On a side for fun, Pranjal loves building new things from scratch, which is why he regularly goes for hackathons.