[w18/19] Modern Approaches in Natural Language Processing (Heumann)

In the last few years, there have been several breakthroughs concerning the methodologies used in Text Mining and Natural Language Processing (NLP). These breakthroughs originate from both new modeling frameworks as well as from improvements in computational resources.
In this seminar, we’re planning to review these frameworks starting with the neural probabilistic language model (Bengio et al, 2003) and continuing with discussing techniques like Word2Vec (Mikolov et al., 2013), Doc2Vec (Mikolov and Le, 2014) and GloVe (Pennington et al, 2014) as well as various other variants.
We will also put a focus on the computational aspects of NLP, discussing implementations in R (e.g. “text2vec”) and in Python (e.g. “gensim”, “tensorflow”).
As most of these modern approaches rely heavily on machine learning, one part of the seminar will deal with some fundamentals of the theoretical background of machine learning, like the Backpropagation algorithm, the Vanishing gradient problem, LSTMs, etc.
The third part of the seminar is about various applications of neural nets or other models in combination with NLP, e.g. Parsing, Speech Recognition, Chat Bots, Image-text representations, Machine Translation, Information Retrieval, Latent Dirichlet Allocation, etc.

Preliminary meeting: 

Thursday, 18.10.18, 16.00 - 17.00
(Seminarraum 144, Ludwigstr. 33)

Talks:

Thursday, 17.01.19, 09.00 - 12.00 & 14.00 - 16.00
Friday, 18.01.19, 09.00 - 12.00 & 13.00 - 15.00
(Seminarraum 144, Ludwigstr. 33)

Advisors:

Christian Heumann, Daniel Schalk, Matthias Aßenmacher