QUESTIONS / COMMENTS

Adrian Rossi

DERA

Visualization of Massive Retrieved Newsfeeds in Interactive 3D

 

Helen Braiter - Does the output give you an indication of the best document, the one with the most hits? Does it rank the documents? How are the key words determined?

Margaret Varga - The key words used are generated by the system, not by the user.

X - How scaleable is it?

Adrian Rossi - That is a good question. I used a cityscape because even with many documents, you can easily see which bar is higher.

Helen Braiter - What about colour?

Adrian Rossi - You could do it.

Milan Kuchta - Right now you have it on a flat grid. Could you provide for alternatives?

Adrian Rossi - Yes.

Justin Hollands - The problem with displaying bars in this way is that the bars in the front block the ones in the back. You could order them, or rank the documents, or rank the words, or use 3D.

Adrian Rossi - You can zoom in and fly through this.

Milan Kuchta - What is the file format?

Adrian Rossi - I am taking it real-time from a text file. It hasn't actually been linked yet.

Milan Kuchta - So you don't have at input format yet.

Adrian Rossi - No, but you could use anything.

Margaret Varga - We have an Oracle database.

Zack Jacobson - What research would you have to do next?

Adrian Rossi - I think that the first thing to do is to see which 3D techniques are useful to the user. You need to know what the user wants and what is useful. You should start with 3D, not go from 2D to 3D.

Zack Jacobson - I see that you work from the task.

Adrian Rossi - I work from both ends and go to the middle.

Milan Kuchta - Have you catalogued what 3D methods are useful?

Adran Rossi - I have found papers that list them, but there are not many methods that are fundamentally different.

Jim Wise - The taxonomy's are all different.

X - Have you considered getting an idea of how a human would perform using this?

Adrian Rossi We have considered it but have not performed any tests.

Helen Braiter - How does the system choose the keywords for you?

Margaret Varga - The system determines the key words based on your input. Documents are retrieved based on the words the user inputs, and then you pick the documents you like. The system will then re-rank the documents. You go through the process until you are happy with the key elements, and then the system does a final re-ranking.

Martin Taylor (to Jim Wise) - How would you handle this problem of key words and concept vectors?

Jim Wise - First I would like to ask Adrian, how do you see this being used? The question is what is the conceptual concept of a database? In this case, it seems to me that I would not use this. However, I suppose that if I wanted to go sort documents and select a few, then I would use this. It is not for looking at conceptual content.

Adrian Rossi - It would be difficult to sort through thousands of documents. It comes back to scalability.

Zack Jacobson - We found that when visualization clicks, it starts to incorporate and replace the search engine. We use visualization to extend the search capacity.

Margaret Varga - This would allow you to have more detail with regards to the structure, etc.

Jim Wise - What goes on in front of this is what I call the authoring stage.

Joe Collins - Have you developed a filter?

Margaret Varga - Yes. On this project we are focusing more on retrieval than on routing. People grasp retrieval better than routing.

Jim Wise Do you have something to display updates on the information? For instance, a Starfield Display with motion, where you can see that something is developing (i.e., a story coming in from Reuters).

Adrian Rossi We have run into problems with motion. How can you determine what speed means what? How do you quantify how different one type of movement is from another?

Margaret Varga They do it in the visual stock market monitoring. They are able to show motion to demonstrate stocks increasing and decreasing in value.

Milan Kuchta The interesting thing about motion is that for humans, the best place to put motion is in their peripheral vision. That is where we are most sensitive to motion.

Martin Taylor It depends if you are using it to track something or to alert the user to something.

Milan Kuchta Have you considered oral cues?

Adrian Rossi No.

Jim Wise There is an international society for aural displays, and they have combined text visualization with aural cues to represent more information. Most of their work is available through Iowa State University. They have a website called "The Big Picture".