Interpreting natural language processing (NLP) models and lifting their limitations
Abstract
There have been many advances in the artificial intelligence field due to the emergence of deep learning and big data. In almost all sub-fields, artificial neural networks have reached or exceeded human-level performance. However, most of the models are not interpretable and they perform like a black box. As a result, it is hard to trust their decisions, especially in life and death scenarios. In recent years, there has been a movement toward creating explainable artificial intelligence, but most work to date has concentrated on image processing models, as it is easier for humans to perceive visual patterns. There has been little work in other fields like natural language processing. By making our machine learning models more explainable and interpretable, we can learn about their logic, optimize them by removing bias, overcome their limitations, and make them resistant against adversarial attacks. This research dissertation is concentrated on making deep learning models that handle textual data, more understandable, and also use these insights in order to boost their performance by overcoming some of the common limitations. In addition to that, we use this knowledge to target words for designing efficient and effective textual adversarial attacks.
Collections
- OSU Dissertations [11222]