Tool for journalists: Reveal Project's tweet verification assistant

Reveal Project's tweet verification assistant is a free tool that analyses the elements of a tweet, including multimedia content, language and punctuation, to determine its veracity.

How is it of use to journalists?

The use of eyewitness media has become ubiquitous in most newsrooms and both news organisations and social networks have started working towards establishing best practices to improve the process of sourcing and verifying material online.

Part of this process consists of developing guides and tools for journalists reaching out to eyewitnesses in breaking news scenarios, to help them find relevant images and determine their origin or verify video footage.

Reveal Project, an initiative co-funded by the European Commission, has released a tool for determining the accuracy of tweets that contain images and videos by analysing elements such as punctuation, number of hashtags and mentions, and whether or not it includes external links.

The verification assistant can be a useful starting point in the verification process, as it can show potential red flags, although reporters should consider using it alongside other resources, such as TinEye, YouTube Data Viewer, Montage and First Draft's Visual Verification Guide for images and videos.

There have been instances where news organisations have only determined an image was fake or wrongly attributed after publishing it online, so this tool can help journalists avoid risking a correction, retraction or loss of trust from their audience by providing some insight into whether the material posted by a Twitter used is likely to be true.

How it works:

Copy and paste the URL or ID of the tweet you wish to analyse into the verification assistant's search box, and click 'verify'. Bear in mind that the tool is likely to provide more accurate results for tweets containing multimedia, rather than just text.

The tool can provide an analysis based on the tweet or on the user's profile, with different elements taken into consideration according to which option you prefer.

You can toggle between the two by clicking the 'tweet-based' and 'user-based' buttons displayed above your tweet, on the left side of the page.

For the tweet-based analysis, the verification assistant will look at how many question and exclamation marks the person has used if any, and if the sentence contains pronouns, slang words or words that highlight a positive or negative sentiment.

It will also take into account the use of external links, sad or happy emoticons, how many times it has been retweeted and how the tweet is ranked in online searches.

Screenshot from Reveal's Tweet Verification Assistant

By comparison, the user-based analysis checks for the total number of tweets a user has posted, their follower ratio and the information included in their profile, such as profile image, location and the date the account was created.

For each of these categories, the tool will provide data and charts comparing the current tweet to previous tweets analysed and included in the database.

A tweet used as an example by Reveal, containing an image of two children hugging and a description connecting them to the Nepal earthquake was determined as being fake by the verification assistant. The tweet did not include any URLs, so the tool highlighted that 51 per cent of overall tweets with a similar value are usually fake. However, the tool also found that 59 per cent of tweets that include one hashtag, similarly to the tweet analysed, are real.

Reveal provides a few more examples of both real and fake tweets analysed using the tool, although the platform does not provide an explanation for how these metrics are weighed to determine a tweet's veracity.

Symeon Papadopoulos, post-doctoral research fellow at the Centre for Research and Technology (CERTH) in Greece, told Journalism.co.uk via email there is no explicit weighting of the different parameters considered by the tool, and this decision is based on a classification algorithm developed by the team.

He also pointed out the tool provides a verdict on whether a tweet is fake or not, by placing a red or green border respectively around it once the analysis has been completed, although the way in which this is conveyed is subject to change to make it more clear to users.

The tweet verification assistant is still undergoing development, and journalists can contact Papadopoulos with feedback.

Update 29/09: This article has been updated with comment from Symeon Papadopoulos.

Free daily newsletter

If you like our news and feature articles, you can sign up to receive our free daily (Mon-Fri) email newsletter (mobile friendly).