Evaluation of Python Based NLP Frameworks for Template Based Automatic Question Generation

Abstract

Automatic question generation techniques emerged as a solution to the challenges facing test developers in the development of smart e-tutoring systems. The current challenge in selecting the available developer tools is depend on several aspects, including the kind and source of text, where the level, formal or informal, may influence the performance of such tools. This tool, popular packages for NLP: NLTK, spaCy, TextBlob, and CoreNLP. Our experiences show that spaCy is several times faster than others in tokenization, tagging and parsing. It has also the best feature set of neural network models and of entity recognition methods. Based on our test results spaCy would be an optimal choice for the implementation of template based automatic question generation. The downside of spaCy is the limited number of supported languages. The choice which NLP package to choose depends on the specific problem you have to solve.