Table of contents
- Criterion: Massive
- Criterion: Harmful
- Criterion: Artificial
Some actors still consider creating their own detection tools. While this can lead to innovation, it is most certainly not the fastest way to detect disinformation.
Moreover, these tools entirely depend on access to social networks data - which is not easily provided - and need regular maintenance to be kept functional.
Crowdtangle was created in 2011 and acquired by Facebook in 2016. Although it is a Facebook partner, relying mostly on Facebook data, it is also available accross Twitter, Instagram and Reddit.
It has both an online social analytic platform with an API, and a Chrome extension.
The extension can track accounts who shared a specific link, offers interaction data on them, including commentaries, and monitors “over two million influential social accounts”.
Out of over a million social media accounts, the Crowdtangle platform displays news to which much attention is drawn, “whether it is an internal viral hit, breaking international news, a heartwarming hyperlocal story, or a competitor’s biggest hit of the day”, on a real-time dashboard that is customizable by the user.
It offers the possibility to set notifications for weekly digests, viral or referral alerts, or even automated reports - among others. Users can also track trending topics.
Crowdtangle also provides users with a set of customizable visualizations, case studies and other resources.
In addition, the platform allows users to analyze their own acc ounts as well as “discover other influential accounts across all the major social platforms”, and facilitates social referrals.
This is done through the help of an automated spreadsheet “that monitors performance metrics” for the user’s brand(s) on all tracked social platforms.
Crowdtangle is used mostly by journalists and professionals. As described on the website, users comprehend “newspapers, television stations, digital media outlets, investigative journalists, entertainment companies, sports teams and nonprofits” worldwide.
Ezyinsights offers services to professionals through two main online platforms available in 4 languages (English, French, German and Italian) and several tools.
The Ezyinsights Realtime online platform allows subscribers to gather news in a faster and more efficient way than traditional methods. Indeed, the platform monitors millions of stories daily, all of which can be filtered according to criteria such as source, time table, keywords, topics, categories, etc.
It is also possible to see how a story is covered and the number engagements it receives using Ezyinsights. Both websites and social platforms are monitored “to deliver real-time metrics on engagement”. The Ezyinsights Intelligence platform can track subscribers overall social performance as well as the performances of competitors.
Moreover, timetable of best posting times (heatmaps) are available and a “Traﬃc Light” functionality relying on engagement acceleration warns subscribers to post or hold off a post to “reach maximum potential engagement”.
Ezyinsights also provides services such as consulting services, training sessions, performance review, advice, management meetings, analyses and reports.
Ezyinsights functions are also partially available through a Chrome extension (which allows users to see how selected articles are performing on different social media) and a mobile app (which sends notifications when filtered news emerge).
Buzzsumo is a social analytics tool destined to professionals.
It provides insight on content shared accross major social networks such as Twitter, Facebook, LinkedIn, Google Plus and Pinterest, including activity around said content.
Its search engine allows users to find “the most shared content for a topic, author or a domain” on these networks. The search is customizable and searchable data goes back up to 12 months. Trending topics and news, sortable by domains and types, are displayed on Buzzsumo’s website.
Its Facebook analyzer can track and process posts performance. In addition, a Question Analyzer is open to subscribers to search “millions of real questions asked on forums and e-commerce sites, including Amazon Q&A, Reddit, and Quora”.
Buzzsumo users can run detailed analysis reports for specific topics (based on data such as the number and network of shares, content length and format, date, etc.) including domain comparison reports, find influencers, identify and track backlinks, and set content alerts.
One of several monitoring tools available through Buzzsumo, alerts are sent by e-mail and can be daily digests or instant notifications, and rely on a filter system with at least two criteria, i.e. the minimum number of shares and keywords.
This detection bot uses results provided by Buzzsumo to notify users of suspicious content.
It is accessible via the “Detection” channel of our collaboration chat. Results are based on chosen sources.
TrendsMap’s name is completely transparent. Although access requires a social media account (Twitter, Facebook or Instagram), TrendsMap focuses exclusively on Twitter activity, showing the latest Twitter trending hashtags and topics on a world map.
More information is available when clicking on the keywords displayed on the map. Then, trend graphs show users the activity related to the keyword both locally and globally, based on the number of tweets in the last day, and the interface directly displays these tweets starting with the most popular ones.
It is also possible to choose a geographical area on the map and zoom in to find what is trending there. Search can go backwards, allowing users to find out where specific top tweets come from.
Analyses can be launched based on hashtags, keywords or phrases, among others. Search results can be filtered by region or language, for instance.
TrendsMap also provides subscribers with several analytics and visualization tools, such as interactive heatmaps showing areas where a topic is trending (most tweets about it, within or without a conversation) as well as sentiment (positive or negative) towards a topic.
TrendsMap can produce detailed statistics on chosen topics (such as the number of tweets and engagements regarding said topic, the gender repartition of the Twitter users who engaged, their bot activity score, etc.).
It offers exhaustive account analytics, among which are the number of followers and tweets over time, the location of followers and followees (through an interactive map), the accounts with whom interactions are frequent and common, top hashtags and words used in tweets, detailed and time-framed tweet activity, etc.
Subcribers can set detailed alerts for any topic, “anywhere in the world, and in any language”.
Trendolizer’s main functionality is the display in real time of the highest trending topics on the web.
Its dashboard shows respectively top trending articles, videos, images, tweets and rising topics across online social platforms. It supports various source types (“SS feeds, HTML scraping, Facebook pages, Twitter searches and subreddits”).
Trendolizer works with an API. It automatically indexes approximately 300 000 new links daily, from thousands of online sources on the internet. It constantly assesses their engagements (number of likes, tweets, pins, video views, shares, etc.) and then calculates their current rate of increase per hour.
The so-tracked links are then stored and open to filtered search. Filters comprehend keywords, source, author, domain, date, language and type.
Users can receive by e-mail alerts of the “biggest Internet trends” or customized alerts according to criteria like the increase rate, keywords, sources, and the number of engagements. They can also pick sites “to be used as sources and track all new links that appear in them”.
Trendolizer also provides visualizations (graphs) of trends and their rates.
Trendolizer’s API allows subscribers to extract all information into JSON-formatted data “for easy use with a variety of programming/scripting languages”.
The creator of Trendolizer, Maarten Schenk, is also the co-founder of Lead Stories, a online fact-checking and debunking media outlet which uses the Trendolizer engine to detect and further analyze trending news. A sample of their work can be found here. It shows how Trendolizer can actually be used in fighting disinformation.
Automatic video transcription and annotation
For video content, the first difficulty is to make content discoverable without relying on the viewing of all streams in their entirety. Two initiatives supported by the EU Media Hub transcribe audio streams and annotate the video streams with object and personalities recognition.
AlgoTransparency, co-created by ex-YouTube engineer Guillaume Chaslot and Soline Ledésert from the ICIJ, detects and exposes the most recommended videos on YouTube. Such “recommended” videos appear as the next video to be played, are autoplayed by default, and their titles and thumbnails are displayed next to the currently played video on a user’s screen.
The aim of the service is to show which content has its visibility increased by the YouTube algorithm.
AlgoTransparency’s program gathers daily all recommended videos after playing the last videos from over 1,000 USA channels and 1,400 French channels. Then, the channels with the most recommended videos are computed. The process is repeated until recommendations from 2,000 channels are gathered. The most recommended videos, as well as the the number of channels recommending each video, are made public.
Users can search for the most recommended videos depending on themes or search terms. The code of the program is open source.