DiscourseDB

[oliver....@gmail.com]

As part of our DANCE efforts, we are creating a common analysis structure for web discussions called DiscourseDB. DiscourseDB is supposed to represent online discussions from different sources (e.g. forums, chats, instant messaging, etc.) in a unified format that allows researchers to perform discourse analyses across sources without having to take the specific properties of each particular source into account.

We are still at an early design stage and welcome any feedback.

The links below lead to (1) a brief overview document summarizing the basic ideas that govern the current design of DiscourseDB and (2) to an Entity-Relationship diagram that lays out the current structure of DiscourseDB.

DiscourseDB overview document

https://docs.google.com/document/d/1OoT3lA-JiL5NKtZ_iuXxR7SD6MjQGX2rgzxTXQKDVlU

DiscourseDB ER diagram draft v0.1

https://drive.google.com/file/d/0BwCBBnex0H4AWWZCRlBBODEteWM

Looking forward to your feedback.

[DANCEcollab]

Hi Oliver,

Recently I have been looking at Wikipedia discussion pages, I noticed that was one of your use cases.  What I have noticed is that there is different types of content that might be interleaved in a single contribution, including discussion pertaining to content, discussion pertaining to wikipedia practices, and discussion pertaining to distribution of work.  Let's say someone posts something involving all three and someone else responds.  The parts of the contributions might be marked out in the UIMA part of your framework.  Let's say that marks one part as assigning work to someone.  Now that person inserts a contribution saying he will do that and then includes some discussion about other wikipedia practices, not specifically related to ones mentioned in the first post.  Part of that is a response to the part of the original post assigning work, but the whole turn is not a reply to the whole contribution of the original poster.  So then the reply links seem to be overly broad.  How would you handle that?  

Carolyn

[Diyi Yang]

Could I ask a question? What is the connection and comparison of DiscourseDB and MoocDB? I thought MoocDB does similar things as DiscourseDB in the "Forum Thread" case. 

[Carolyn Rose]

As I understand it, MOOCdb does store discourse information, but only about discourse that occurs within the MOOC platform.  It doesn't store anything from related discussions like facebook studygroups, twitter, or other environments, including interventions linked to the platform through the LTI protocol.  So then if one wanted to analyze discussion in MOOCs more broadly, MOOCdb as it stands would not support that.  DiscourseDB fills that gap.  It only focuses on discussion, but it does so more broadly.  And it is meant to connect ultimately to DiscourseDB and also the LearnLab Datashop under the Umbrella of the LearnSphere project http://learnsphere.org/about.html.

[Oliver Ferschke]

Hi all,

the most recent version of DiscourseDB (v.0.2) can be found on github and is openly accessible.

https://github.com/DiscourseDB/discoursedb-schema

We managed to integrate many of the great suggestions we received from the DANCE community and from researchers working on scripted collaboration, knowledge building and discourse analysis in education.

We are now proceeding with the next steps in the iterative design process that will allow us to load the first real-world datasets into DiscourseDB. This will give us a better understanding of the current limitations and allow us to update the scheme accordingly.

An updated documentation for the revised database scheme will follow as soon as possible.

Best,

Oliver