Nitro hosts first NLP meetup in Dublin
2 weeks ago Nitro hosted the first NLP meetup in Dublin … EVER!!!
THAT is kind of surprising. Or at least it was for me.
First, because we here at Nitro believe that ML (Machine Learning) and NLP (Natural Language Processing) based software engineering approaches will revolutionize how we provide value to customers in the next 5 to 10 years.
Second, because Dublin is without a doubt one of the most vibrant locations in Europe to do software engineering and here at Nitro we have been hosting a nice mix of meetups (e.g. Spark Meetup, Reactive Systems Meetup, Scala Meetup, Machine Learning Meetup), but I never realized that there is no NLP meetup.
In any case: The problem is fixed now. Sebastian has created the meetup! For the first one we had +50 attendees and 3 speakers.
Chris Hokamp talked about “Task Based Learning for NLP”. Excellent talk! My main take-away was training the model “blindly” using well-known, standard approaches will not give you the best results. Training with an objective in mind can make a big difference.
Next Sebastian delivered a talk on “Softmax Approximations for Learning Word Embeddings and Language Modeling”. This is based on the work that he is doing with Aylien and the Insight Centre for Data Analytics. He presented on very new research (CNN-Softmax; Jozefowicz et al., 2016) on how to improve on traditional softmax approaches.
Last but not least, Jing Su from CeADAR presented on “Topic Listener — Observing Key Topics from Multi-Channel Speech Audio Streams”. There are a lot of call centers in Ireland. The research is focused on identifing topics in audio streams (e.g. the audio that is recorded when you call a call center). The visualization of the topics was amazing.
Originally published at nitro.github.io on August 22, 2016.