Paper Reading #16

Comments:
Reference:
Title: Personalized Reading Support for Second-Language Web Documents by Collective Intelligence
Author: Yo Ehara, Nobuyuki Shimizu, Takashi Ninomiya, Hiroshi Nakagawa
Venue: IUI 2010
Summary:
This paper discusses a potential adaptive system to help users browsing the web who are not particularly fluent in English. There are current systems that allow a user to click on a word, and the definition of that word is displayed. This is referred to as a word glossing system. However, these systems do not try to keep accounts of the words that the user clicks, so they cannot predict the words most likely to be clicked in the future. That is the aim for this proposed system: keep records of the words a user has to look up so they can predict which words the user will not be able to know in the future. The system uses the Item Response Theory, which is "a statistical method for analyzing the results of tests of human abilities including language tests." The group used 16 different subjects to test their system. The library they used contained 12,000 words, which were selected from the Standard Vocabulary List as the 12,000 fundamental words that a person should learn to speak English. Logistic Regression produced the most accurate results of the five algorithms they evaluated. An example of a word glossing system, pop jisyo is given below:

Discussion:
I found it incredibly funny that a paper about a system to help users who aren't adept in English to be so hard to read. This paper had tons of technical details that I felt were not fully explained or were explained in too technical of details. I had to skip whole paragraphs because the words looked like pure nonsense. Perhaps my attention was not fully devoted to reading the paper, but I definitely felt that they could have done a much better job of explaining what exactly they were doing. As for the system itself, I think it is a good tool for those who are struggling with a language. I can see this being a very widely used algorithm. However, the bottom line for me is that for a paper about helping others read, this paper should have been written much more clearly.