Master's Thesis Iman Jundi

Using Multitask Deep Learning for Question Answering: A Use Case on Insurance Question Answering Dataset

Question Answering is one of the most challenging tasks in Natural Language Processing. Like with many other research fields in computer science, Deep Learning has been heavily used in recent years to solve NLP problems including its use for Question Answering. To have a good performance in deep learning, a huge amount of data is required which is not always available in certain tasks and domains. Multi-tasking is one of the answers to this problem where it's possible to train an architecture on multiple tasks. The performance in tasks with smaller datasets could benefit as a result from training the model on related tasks with more data. Related tasks could also benefit from discovering important features in the data that might not be easy to see from a single task. Our goal is to investigate this by implementing a model using supervised learning to perform multiple tasks in the area of question answering, namely reading comprehension and answer selection in the insurance domain.