The Reader

In the literature of academic research, the reader is usually of readers, ie does not pay little attention to the subjective individual to read the specific text. A reading of the paper David S. Miall and Don Kuiken: is the empirical research historical view of Bun読国 house.

The first concept we have to understand is a text model reader. Model readers are not experienced readers (neither you nor me). It is a reader living in the author's mind, and the author builds it when writing a story. This model leader can help the reader understand as the author will make the author judge the detailed information necessary for his work. Let's explain this with an example. When I wrote an article saying "I am sitting in front of my computer and starting coding", I have made several choices about the amount of information that needs to be communicated based on the model leader. Explain that you enter the program using the keyboard. In most cases, this is how it is done. In that sentence, my model leader is a person who is familiar with how the computer works and how to enter programs on the computer today.

I started implementing two multi-choice reading comprehension models, AS leader and Gauge attention leader (GA leader). Our library deep_qa is used for detailed learning of various questions and task answers. In the model reading comprehension task, paragraphs and questions about them are displayed. You need to select the correct choice (if multiple selections) or select the paragraph substring (in this case) to answer directly to the question. answer). As a person who lacks application experience when building a deep learning model, I learned a lot when I reimplemented these models in code base. There are lots of hints and tricks that seems not important for implementation and training details, but it is important to correctly reproduce the results of other users.