Ushine from the Data Science Fellows

Ushine is a tool created by the Data Science for Social Good Fellows. Nathan Leiby, Kayla Jacobs, Kwang-Sun Jim and Elena Evena joined Emmanuel Kala, our community and network to dive into data cleaning and data analysis to assist Ushahidians and others on their data missions. The created tool is possible to use with Ushahidi or, with some code mashing, other software. Emmanuel Kala also created an Ushahidi plugin building on the collective braining.

Today we held an 1-hour Google + Hangout to show the work, answer questions and talk about the next steps:

(You can see the comments on the G+ site)

We all want to see this work continue. Emmanuel and I will work with the fellows and our community network to build out the next steps on the wiki. To see the full presentation:

Thank you

Thank you to the DSSG fellows (Nathan, Kayla, Kwang and Elena), the Data Science for Social Good program, the sponsors, fellow DSSG project participants, and the Ushahidi community for all their input, testing and guidance during this expedition.

Additional Resources

More DSSG Blog posts:

Machine Learning for Human Rights
Data Fellows in the Community Weekly Reports

4 Responses to “Ushine from the Data Science Fellows”

  1. Excellent overview of the new technology. Thanks for posting it!

    If I can attempt a summary: Ushine helps the Ushahidi report verification process by making suggestions related to content, locations, links and categories.

    The slide at 8:25 in the video does a good job showing this. http://bit.ly/1dVJDyf

    But just to clarify, Ushine flags reports with helpful tips such as to location and categories (http://bit.ly/18pvaWG). It does NOT automate the process. (http://bit.ly/IXakVH). The number of reports requiring human interaction is the same with or without Ushine.

    Just a suggestion: it would be great to have the option to turn the automation on or off. The video made a good argument for keeping automation turned off, namely the need to have high quality reports and that machine automation is not 100% accurate.

    But there are cases where machine automation is better than none at all. Consider a situation where there are more reports coming in than people to review them. Automation, even if not perfect, could help improve situational awareness.

    And if a report were flagged as such, i.e. “This report has been machine reviewed and may include mistakes”, the viewer could take appropriate care with the info.

    My 2 cents:)

    Location
    ========
    Can you clarify are locations PLACE specific or POINT specific?
    (PLACE meaning bounding box of a given location, such as the polygon boundary of an entire city. POINT meaning exact lat/lon coordinates)

    Some reports may contain general PLACE information (i.e. the name of a city) while other reports may contain POINT information (a specific landmark in the city). How does Ushine handle this?

    Ushahidi Integration
    ====================

    Is Ushine included with Ushahidi automatically? Are there extra configuration steps?

    Is Ushine used in the Chambua NLP project (https://github.com/ushahidi/Chambua) or vice versa? Are there plans to integrate this functionality into SwiftRiver? If not, can UShine be used as a stand-alone product, for example, to do post processing analysis on droplets already in a SwiftRiver River or Bucket?

    In the 3 months since this video has been posted are there any real-life situations where Ushine has been used? If not would it be possible to stage a test incident to see the workflow in action?

    Thanks to Emmanuel, Rob, David and all who contributed to the project (http://bit.ly/1gdD8a7) and keep up the good work.

    Dan King
    http://www.viewpoint.pro