Deep text mining

For decades, humans have dreamed of computers that understand natural language
in the form of text or speech. The interaction between human and machine using
natural languages is achieved through Natural language processing (NLP) with the help
of smart assistants. Natural Language Processing and Text Mining or Text Analytics
are Articial Intelligence (AI) technologies that endow users in transforming the key
content in text documents into quantiable and actionable insights.
Text in the documents is a rampant form of communication. The analyzing and
understanding of these text includes multiple tasks which needs to instruct the computer
to understand word-sense disambiguation. It addresses toilsome scaling and language
challenges where traditional NLP techniques are less eective. Deep Text Mining, a
deep learning-based text understanding helps to acquire this task to some extent. Deep
text utilizes several deep neural network architectures like convolutional and recurrent
neural networks and is able to perform word-level and character-level based learning.
Deep learning uses a mathematical concept called word embeddings that preserves the
semantic relationship among words. So, when represented properly, the word embedding
allows capturing the in-depth semantic meaning of words. Word embedding techiques
like GloVe, Word2Vec are euclidean approaches which fails to represent for hierarchical
structures. So a better approach for representation of word embeddings is a requisite
in the eld of text mining. This work attempts to builds a deep learning model with
poincare word embeddings for the task of intent classication.

MPhil CS 2018-2019 INT Dr Asharaf S