— Philip Mai (@PhMai) January 26, 2017
It seems that even an old dog can learn new tricks. I was experimenting recently with the list view settings in DiscoverText and was pleased to find a new way to review Twitter data. This 2-minute video outlines the steps you can take to get an optimal list view display for Twitter data in DiscoverText.
We have been very busy creating a set of unique Twitter data samples for the fake news student detection challenge. There are new collections of bot-mentioning Tweets, also by users with >100,000 followers, other influencer segments, top RT segments, and a 1.1 million item sample representing the seed of every duplicate group in the original 20 million Tweet collection. To learn about the metadata enhancements that make this and the influencer segments valuable, we have prepared a short video. Watch the video below to learn more about the free student data challenge and to get tips on why the best data samples are in fact great tricks for doing other kinds of important social media research. New teams are still getting organized. Write to firstname.lastname@example.org for more information. The deadline for submissions is April 1, 2017.
We often get field reports from graduate students who are deep into the exploration of Twitter data using DiscoverText. This excerpt below from Stanford’s Anita Tseng further illustrates why so many academics are using DiscoverText to collect and analyze Twitter data to better understand public sphere discourse.
“I am currently conducting a large-scale content analysis of Tweets on controversial science issues. My dissertation focuses on misinformation about science in new user-generated media, and one of my projects deals with illustrating the scope and nature of misinformation about controversial science on social media. For seven months, I collected Tweets on “vaccination safety” and “GMO safety” using multiple variations on each search term, collected at various times each day on a weekly basis. I aim to analyze these Tweets for sentiment, as well as the presence of errors in scientific reasoning, based on an existing framework in recent research on philosophy of science. After trying a number of data collection and analysis tools, I came across DiscoverText during a workshop on campus last year and was thoroughly impressed by the functionality and most importantly for me, user experience. After dealing with a number of other badly programmed analysis tools that were both slow and unintuitive, DiscoverText was fast for me to pick up, Web-based and did the grunt work of collecting onwards to 56,000 Tweets for me over the course of several months in 2016. I’m now in the analysis portion of this project and excited for the findings to develop — I am qualitatively coding a subset of my data and training the built-in machine learning algorithm to code the remainder so I can have a broader picture of the data. This spring, I will be presenting the project as a Computational Social Science Fellow at Stanford University, and at the National Association for Research in Science Teaching as part of the Informal Education division, which includes research on social media and its impact on public understanding of science.”
Anita is a Doctoral Candidate at Stanford University’s Graduate School of Education. Even though she is now an experienced user, we look forward to seeing her at the upcoming DiscoverText workshops January 17, 2017 on the Stanford campus.
– Build a model of fake news on Twitter
– Submit the model using a short video (=<60 seconds)
– Teams must use our dataset and online collaborative tools, which we will provide for free.
– Data export is prohibited.
– The model must focus on the nature and scope of fake news itself, not external analyses of it.
– Qualitative, quantitative, and mixed methods models are all welcome.
– Collaborative teams must include 2 or 3 students at any educational institution.
– Faculty supervisors may join one or more teams.
– The challenge is open to students in any country.
– The final report simply needs to be in English.
– Entry Deadline is now extended to April 1, 2017
– See Video Update #1 about the data samples
– Links to videos presenting the model in 60 seconds or less on YouTube must be Tweeted with the hashtag #fakenewsdetection.
1st Prize: $100 for each student.
2nd Prize: $50 for each student.
3rd Prize: $25 for each student.
Every team that submits a video will retain an academic DiscoverText license for the remainder of 2017.
There are many ways to explore the metadata, Tweet text data, images, news links (both fake and real), to test and refine student hunches about the scope and nature of fake news disseminated on Twitter. Our goal is to share these models with academic research community and to support a variety of methodologies for human or automated fake news detection. There is no a priori labeled data or ground truth. Everyone gets the same unlabeled data and the same tools to build models. Everyone builds their own models. Anyone can collaborate with anyone else on the system (not restricted to your team). How you define and deliver the model is entirely up to you. We have no idea what fake news is…yet. It seems to mean a lot of different things to different people. We are looking for models that make sense to most people.
– Have each member of the team sign up for a free trial DiscoverText account.
– Send an email to email@example.com with your team name and a list of team member names and emails.
– Identify a student team leader who can manage the project and serve as a point of contact.
– Schedule a web meeting to go over some of the e-discovery, human coding, and machine-learning techniques.
– Review the DiscoverText help guides and FAQs: https://texifter.zendesk.com/hc/en-us
– Check out some use cases and methods in previous scholarly mentions of DiscoverText.
For more information, contact us at firstname.lastname@example.org.
Better late than never, thanks to the holidays, we are pleased to have drawn the final two winners for our Tools and Data Give-A-Way. Today’s lucky winners are @socmednew and @JoAnnLivingston for these Tweets. Thanks to everyone who entered. We hope to see more publications on this list as a result in 2016.
Woohoo! Did an estimate on https://t.co/M8frNto0LV. So cool: unlimited free estimates; able to customize what I need for my paper!
— JoAnn Livingston (@JoAnnLivingston) December 18, 2015
— Your Social Media (@socmednew) October 22, 2015
We did it again. Nine Fridays in a row we have drawn two winners for our Tools and Data Give-A-Way. Today’s lucky winners are @yelenamejova & @jaganadhg for these Tweets. We are doing one final give-away next friday. Who will get prize #19 & 20?
Check out https://t.co/2GELWhAnE6 for highly customizable comprehensive historical Twitter datasets.
— Yelena Mejova (@yelenamejova) October 20, 2015
— Jaganadh G (@jaganadhg) November 3, 2015
— Nigel Williams (@Org_PM) December 9, 2015
— Moa Eriksson (@MoaLEriksson) October 20, 2015
There are still two drawings left before we close out the year. There is no time like the present to try out Sifter and Tweet your review to be entered. The entire process can be completed in about 5 minutes.
We are giving away software and custom historical Twitter prizes today, so it must be Friday. Congratulations to @karawhytas and @Bpowder87 who won based on these fine Tweet reviews of Sifter. There are still three drawings left before we close out the year. There is no time like the present to try out Sifter and write your review.
— Maximilian Franke (@Bpowder87) November 8, 2015
Sifter is the most efficient way to gather historical Twitter data. https://t.co/u9PdYN3odg
— Kara Whytas (@karawhytas) November 20, 2015