Architecture of machine reading comprehension models
Abstract
Early machine reading comprehension (MRC) models are mostly based on information retrieval technology, that is, search and return statements in the article related to the question. However, information retrieval is heavily dependent on keyword matching, which often gives answers irrelevant to the question. With the development of deep learning, MRC has entered the era of neural networks. Advances in related technologies have considerably improved the efficiency and quality of models. The accuracy of MRC models have been steadily increasing, even reaching or exceeding human performance in some datasets. In this chapter, we introduce the architecture of these MRC models and explore ...
Get Machine Reading Comprehension now with the O’Reilly learning platform.
O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.
