iCQA project -- ideas

[Yu Wang]

Hello,

I am Yu Wang, fourth year Ph.D. student at Emory University. I am currently working with Nikita and Qiaoling on our CQA project. Below is a brief summary of what we are trying to do for this project.

Any comments/feedback will be appreciated.

Thanks,

Yu

  

Predicting askers’/answerers’ level of expertise:

The questions and answers provided by the user in Question-Answering websites reflect the knowledge that he has and the one he is trying to gain. So the content generated by the user in the question answering community (e.g. Stack Overflow) indicates his level of expertise on certain topics. In Stack Overflow, a question-answering thread is manually tagged by its corresponding topics, such as “java”, “python” etc. We propose to estimate and predict the user’s level of expertise by examining and modeling the questions and answers he generated in the question answering website.

Definition of expertise level:

We have several options for defining user’s level of expertise:

1.       The expertise score on Stack Overflow.

2.       Basic measurements of the usefulness of the content. For example, the vote of his questions and answers.

3.       Advanced measurements, such as user’s authority.

Tasks:

1.       Estimating user’s level of expertise.

2.       Modeling the evolution of expertise over time and therefore predicting user’s level of expertise in the future.

Organizing questions into syllabus-like structures. 

Question answering websites usually organize questions and answers by tagging them with proper topics or put them into a hierarchical topic tree. However, few efforts have been made to index the questions for learners. Instead of searching in a question pool, we argue that a syllabus-like indexing structure is more helpful and effective for users to build their knowledge. Stack Overflow is manually selecting representative questions according to the topics in a pre-defined syllabus, and make it more like learning materials other than random questions. We propose to build this syllabus automatically by grouping and sorting questions according to common learning procedures, such as showed in textbook.

[Nikita Zhiltsov]

Thanks Yu,

let me also provide an example syllabus for the second task. 

[Ramis Yamilov]

HI Yu,

Very interesting tasks. You have mentioned some texbook in description of the 2nd task. Could you specify its name?

понедельник, 22 октября 2012 г., 20:30:19 UTC+4 пользователь Yu Wang написал:

[Nikita Zhiltsov]

I believe Yu meant a hypothetical textbook. For example, we would borrow the outline from any textbook on Java or existing Java course.

[Qiaoling Liu]

Nice summary, Yu!   I just want to add a little more about the first task that we've discussed:

As an application, it is interesting to use the system in a class teaching scenario. Based on the question answering, browsing, and searching behaviour of a student, the system could be used to predict the student's final performance on the class.

--
You received this message because you are subscribed to the Google Groups "iCQA" group.
To post to this group, send email to ic...@googlegroups.com.
To unsubscribe from this group, send email to icqa+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msg/icqa/-/mU2Az5Mu0igJ.
For more options, visit https://groups.google.com/groups/opt_out.