Improving Convolutional Neural Networks for Text Coherence Modelling using Class Balancing…

Lately, Deep Learning is gaining huge popularity due to its supremacy in terms of accuracy when it comes to very complex problems. It proved efficiency in NLP and was widely adopted by many problems to address them opening new doors for more meaningful and accurate modeling approaches. While many problems in NLP involve text syntheses such as text generation and multi-document summarization, text quality measures became a core requirement, and modeling them is an active problem. Of these measures, the problem of text coherence is key and needs special handling. Text coherence, which means the degree of the logical consistency of text, is a problem that dates back to the 1980s, where several models were suggested.