Part
01
of one
Part
01
What companies are pursing automated fact-checking? How sophisticated is the best software in this space?
Hi! Thank you for your question about what companies are pursuing automated fact-checking. The short answer is that Facebook, Full Fact, The Washington Post, Le Monde, Factmata, RumorLens, FactMinder, Watson Angles, Trooclick, and Claimbuster are key players in the automated fact-checking industry. Given that many technologies are new or in development, there is little data available to gauge their level of sophistication. Challenges to the industry include cost, the need for human judgment, and language barriers. Below you will find the deep dive of my research.
METHODOLOGY
Using corporate websites, industry reports, and trusted media sites, I found a rich variety of sources with which to answer your question. I began by broadly researching companies that have fact-checking projects underway. I then searched industry reports for specific information on those companies, with regard to the level of sophistication, as well as the pros and cons of their fact-checking technology. It is important to note that fact-checking technology is relatively new. Therefore, most companies have not had a general roll out of their products, which means that their products have not been fully vetted or reviewed as to levels of sophistication. I have included information as the level of sophistication, pros, and cons as it was available.
COMPANIES WITH FACT-CHECKING PRODUCTS
1. Facebook. The social media platform has teamed up with the fact checkers at ABC News, the Associated Press, FactCheck.org, Snopes, and Politifact. All of these organizations will use a tool created by Facebook to check the truthfulness of stories that have been noted as fake. If the story is found to be fake, it will be demoted in the news feed.
2. Full Fact. A British charity, Full Fact has raised $52,000 from Google's Digital News Initiative to invest in new technologies for automated fact-checking. Their investment includes a tool for news rooms and a mobile app that journalists can use during live events.
3. The Washington Post. The news outlet recently released a Chrome plug-in specifically designed to fact check posts from US President Donald Trump's Twitter account. This plug-in adds a gray box to all of Trump's tweets that adds relevant context and flags misinformation.
4. Le Monde. This French newspaper has Les Decodeurs, a 13-person fact-checking division that has been working on a "hoax busting" database that is designed to identify false information. It also has a search engine that allows readers to fact check the statements of politicians. Le Monde has received funding from Google's Digitial News Initiative to create a Google Chrome Extension that will allow people to see which stories have been verified using a red, yellow or green colored flagging system.
5. Factmata. Factmata is a company offering a "state-of-the-art" fact checking system using artificial intelligence. The Factmata Project has been awarded official backing from the Google Digital News Initiative as a reward for their many years of work in the Natural Language Processing (NLP) problem of statistical fact checking. Their work thus far has only allowed the capability to check claims against a statistical database, but they seek to allow their research to let anyone check any of the facts they come across on the web. Factmata has indicated that they want to "make the fact-checking experience fun, engaging and powerful, and build an environment where readers love identifying false claims, correcting news articles, and ensuring more accurate content for the rest of the world." According to Factmata, automated fact-checking is cheaper than human fact-checking, less prone to error, unbiased, and fast enough to keep up with the Internet.
6. RumorLens. This is a system that can be fed information about rumors and will then search for tweets that are either spreading them or correcting them. The system was developed at the University of Michigan.
7. FactMinder. A team of French scientists created FactMinder as a browser extension that allows users to get context about individuals by removing data about those from a website.
8. Watson Angles. IBM's Watson group will release a beta version of the Watson Angles app that will check news stories against 55 million previously published news articles. Additionally, it collects basic facts that are relevant to the topic and offers context, timelines, and key quotes.
9. Trooclick. This company has developed an engine available as an application programming interface (API) that can extract speech from text and attribute it to the correct speaker. The engine can also crawl news and social media sites to find instances of dialogue, which it then delivers to the fact-checkers along with other potentially checkable material. This allows the fact-checkers to circumvent the laborious search process.
10. Claimbuster. This is an automated live fact-checking system that uses a mixture of human coding and Neuro-Linguistic Programming (NLP) algorithms to rank statements according to whether the claim is "objectively verifiable" and whether there is public interest in its truthfulness.
CHALLENGES TO AUTOMATED FACT-CHECKING
There are several challenges to automated fact-checking. These challenges are affecting the current levels of sophistication of the available services.
The first is the costly nature of collecting and organizing the massive amount of data available. Checking new information against existing facts requires having databases in place. While there are many huge repositories of information, none are complete nor hold information on all imaginable topics. Additionally, media outlets and organizations do not have developers on hand to create and maintain these Content Management Systems (CMS). Thereby, important elements for the development of automated fact-checking are funding, international collaboration, and common standards.
Next, while some claims are easy for machines to check, some will always require human judgment. Even if all the possible information on a topic exists, it is a difficult task to teach computers to compare claims against past statements. Nuance is a very difficult thing to automate.
Another large challenge to automated fact-checking is speech recognition and language barriers. Solutions in this industry will need to determine effective and efficient ways to retrieve information from television and radio. Most automated solutions are currently English language only and as a result could create further misinformation with poorly translated results.
CONCLUSION
To wrap it up, Facebook, Full Fact, The Washington Post, Le Monde, Factmata, RumorLens, FactMinder, Watson Angles, Trooclick, and Claimbuster are key players in the automated fact-checking industry. Challenges to the industry include cost, the need for human judgment, and language barriers. Given that many technologies are new or in development, there is little data available to gauge their level of sophistication. Important elements for the development of automated fact-checking are funding, international collaboration, and common standards.
Thanks for using Wonder! Please let us know if we can help with anything else!