Text to Code Conversion Using Deep Learning for NLP
Youssef Mellah, Zakaria Kaddari, Jamal Berrich, Toumi Bouchentouf, Mohammed Ghaouth Belkasmi · 2020
Semantic parsing is the problem of mapping natural language sentences to a formal representation like lambda calculus expressions. We work on a specific sub-problem in semantic parsing: producing python code snippets from natural language descriptions. We address this sub-problem using the CoNaLa Dataset in order to improve precision and to help developers in their development phases. For this task, we use Deep Learning (DL) for Natural Language Processing (NLP), based mainly on an Artificial Neuron Networks (ANN). Unlike the simple ANN, which learns in one direction, we use a Recurrent Neural Networks with LSTM cells since it performs better in Machine Translation (MT) task. We get a BLEU score of 15.2 beating the previously reported baseline of the CoNaLa challenge.