yalibian / crowdspire

CrowdSPIRE: Crowdsourced based StarSPIRE

• 👋 Hi, I’m @yalibian
• 👀 I’m interested in human-in-the-loop machine learning, interpretable deep learning, human-ai collaboration, visual analytics.
• 🌱 I’m currently working at Intel Labs.
• 💞️ I’m looking to collaborate on ...
• 📫 How to reach me ([email protected])...

crowdspire's Issues

Next Step

How to show whether crowdsourcing could be used on investigative analysis/intelligence analysis:

The advantage of Crowdsourcing is "Could make high-level concepts, when coworkers read and conclude the documents", lead to finding more related documents based on their semantic meanings.

What's the new view for express this strength, should take this advantage of crowdsourcing into consideration, also, what we should provide the analysts. (Does the users need this kind of views.)

Implement the background update model of semantic interaction

Implement the underlying analytic models (right now, the basic text analysis method: only help change the weighting vector)

Maybe a more complex ML model after the basic one.

Add background algorithms when semantic interaction executed

Test if OPEN_DOCUMENT interaction changed the links

Update the value of d.x, d.y and label of Document Rectangles when it be resized.

Rect collision detections don't work well based of the changing size of documents rectangles

From StarSPIRE to CrowdSPIRE

Right now, finished the StartSPIRE part, then use the crowd.json to update the layout.

Make interaction detection

When two document level node are overlaid each other, CrowdSPIRE could detect it.

Add Controller Components

The Controller Components have at least two parts:

• Entity-Level:
  Searching part

• Document Level:
  Show one document content and you can edit it.

• Cluster Level:
  Compare part(When documents overlap each other, this panel need to show the most similar contents, and show the user why)

Make document node appropriate to contents

Define semantic interactions

When adding a search tag, the whole view will change based on new nodes: "doc nodes" and "keyword nodes"

Design contents of Document Level Rectangle

Instead of a rectangle, it's a complex sub visualization view or dialog, to help represent contents of document