Organisation: Newscorp & IBM (Australia)
Publication Date: 03/04/2016
DescriptionComment Sweeper a tool for Journalists and readers Some journos and bloggers engage and form relationships with readers but most don’t have time to engage with hundreds or thousands of comments posted on their articles in many social channels. But there are experts, witnesses and people with experience of issues in the comments. How do we find them? Some readers don’t engage because the discussion is hard to follow, goes off topic, can be full of white noise, or is just rude. The Fix: A hybrid app to search comments for keywords or a topic and harvest relevant comments. Sort and filter users, or by sentiment to find insights and people who can help move the story along. For readers it's a tool to listen to ideas, opinions and thoughts of thousands of people online that can be searched, filtered and analysed. For example explore the community sentiment at all the news outlets (left and right wing) to new policy or events. Compare attitudes of right/left wing readers with others. Look for patterns and find trolls who pop up on more than one news outlet saying the same thing. For Journalists the tool adds value by finding experts, witnesses and people with experience on an issue. Once an expert is identified they can be contacted and appear as “verified” for future searches. With eyewitnesses the tool can help find networks of people as well as people who say “I’m there!” Insights into commenters can be built with a credibility rating both for trolls, people who comment one sided arguments and experts who can be used as sources in future. Recent examples (attached pics): 1. A house fire covered by nine news was posted onto a local newspaper Facebook page which produced residents comments with photos of the house lit up by Xmas lights and their memories of it, we followed up with a rich story based on these. 2. A newspaper story was posted to social media about a photographer accused of abusing kids at a kindergarten. The comments revealed parents who had kids at the school were still paying the photographer for class pictures. This app provides journalists a way to catalog people of interest, with available contact info and monitor topics across platforms to start investigating stories. We also wanted to: 1> Create a connections diagram of possible relationships between users posting comments, like friends/colleagues 2> Once an expert is identified they can be contacted before the next story and will appear “verified” in next comments. 3> Create analysis of overall sentiment to issues at each outlet and across all outlets, a virtual poll on reactions to news 4> Filter out trolls or paid persuaders that appear in comments across channels with the same or similar comment by analysing similar phrases/users
Technologies used for this project:We scraped comments from Facebook, The Guardian and ABC on a topic to use as dummy data for the prototype using a Kimono API and a Dataminer recipe. Tech: Treemap with D3, Angular.js, IBM Watson -Alchmey API / Tone Analyzer, Tomcat. The idea was to gain targeted insights from the comments that could help journalists and readers. Insights that without technology would go unnoticed given the sheer volume of comments. We had to make a decision to either gather data or do analysis of the data. We choose to tackle the analysis. However,we did have some cool ideas around gathering data : > Creating a browser plugin to capture all the websites a journalist visits as part of their background research to capture comments from those sites. > Or traditional API calls that exist like NYT, HackerNews, NPR, Alchemy News For the analysis, we scoped the target insights to 4 main areas per comment and at an overall cross platform level: 1> Entities mentioned in comments like people names, organisations, locations, white papers, etc < completed> 2> Keywords and concepts generation
You have to be connected to contribute
You have to be connected to follow
Leave this project and no longer be informed about this project
By joining this project, you will be informed by email when an update or a new contribution is posted on the website.
Thank you for your active participation !
The GEN Community Team