Behind the Breaking News Sourced from Social Media: Using AI to Quickly Detect and Report News-Worthy Accidents and Events

The FASTALERT service introduced in this article is intended for use in Japan and utilizes information posted in Japanese on social networking sites.

In an age where anyone and everyone is an information provider, the "accuracy" of information is a major issue

According to a 2017 survey of consumer communication trends released by the Ministry of Internal Affairs and Communications, 60.9% of the people in Japan in 2017 had smartphones. And, over half of these smartphone users used their phones to access social networking sites (SNSs). (Just so that everything is clear, a smartphone is a telephone equipped with a camera, recorder and text typing capabilities.) In other words, current-day Japan can be looked at as having about 100 million cameramen who, if happening upon an accident or caught up in a natural disaster, could potentially provide first-hand information "from the site" via SNSs. In fact, one of the triggers to this use of the smartphone was the Great East Japan Earthquake that struck eight years ago. Alongside the explosive diffusion of the smartphone that followed, SNSs have become an indispensable source of first-hand information during disasters that have occurred since then.

In order to get the word out as quickly as possible, even news organizations in Japan gather first-hand information from SNSs when an event occurs. However, ensuring the reliability of that information is a major issue because it is often fragmented and contains noise unrelated to the content. And, the longer it takes to pick apart and organize the information, the greater the chance there is of delays releasing information when there is not a moment to lose.

FASTALERT, a service provided by news venture JX Press Corp., screens emergency information and delivers highly accurate news in real-time. Restricted to just Japan, this service carefully extracts "when, where and what kind of event occurred" from information posted on SNSs and sends it as a news bulletin to news organizations. The news organizations then sift through the bulletins to compose and release their breaking news.

In just the first six months after the beta version was made available in September 2016, NHK and all major private broadcasting stations joined the big telecom and newspaper companies in subscribing to FASTALERT. Today, FASTALERT is an indispensable tool that Japan's news organizations cannot do without, as it is being used to find images and collect information on events. As a result, TV news is giving increasingly more credit to its viewers for providing content.

Technology-ripped AI for automating breaking news

Under a mission banner of "mechanizing news reporting," JX Press aims to solve the news industry's reliance on antiquated methods of acquiring information. This purpose alone puts JX Press in a unique position. Though listed as a communication company, JX Press does not have a single reporter on their payroll and engineers make up two-thirds of their workforce.

Yoshiko Nagami of JX Press explained how FASTALERT came to be: "New organizations have for some time recognized the need to gather information from SNSs, but labor-intensive practices such as having reporters do manual searches that take several minutes per search or enlisting a large number of part-time workers to sift through hordes of information are the norm. We developed FASTALERT to automate this tedious, time-consuming work and pool open resources for reporting that only humans can do, thus enabling a new way to go about work in the process."

What should be pointed out is that FASTALERT properly composes news stories for delivery as breaking news without requiring human intervention. Moreover, it automatically assigns titles such as "Earthquake Damage in ** City, ** Prefecture" and handles all editing without need for human input. Packed with technology from head to toe, FASTALERT is all about making news reporting more accurate and easier to do.

FASTALERT does the job via a news engine for analyzing news articles and the text and other content that they are composed of and a system for making editing work efficient, which JX Press developed after their founding in 2008. Together with the huge amount of data that was amassed in the process, the service is the lifeblood of the company. Known as "XWire," the hi-tech news engine is the basis of FASTALERT and has been adopted by the Sankei Shimbun Group as the backbone for their flagship "Sankei Plus" news app.

The ace-up-the-sleeve is AI for analyzing text, still images and movies. It holds the key to efficiently sorting out noise from still images and movies contained in information from the site of a disaster or accident, as well as ensuring a high level of natural language processing. Ms. Nagami described how it works: "Training data is set first and incoming data is matched to it. For example, with an image of a fire, an analysis is done to determine whether the fire was caused by a real fire or whether the image is of meat on the grill at a barbeque restaurant. A wide variety of patterns are archived and used for comparisons."

The news organizations that introduced the service have highly praised FASTALERT for its speed and comprehensiveness. In just these few years, it has been used for distributing news bulletins to media organizations across Japan on too many occasions to count, such as the train bridge on the Kyudai Line that was washed away because heavy rainfall in Kyushu, the Eastern Iburi Earthquake and the fires that ravaged Itoigawa City. As for its being comprehensive, there were cases where news organizations could not have possibly covered stories by conventional reporting methods like the child that was born on a train on JR's Joban Line and the herd of pigs that escaped from a truck on the Hanshin Expressway.

In any case, the more quickly information about fires, events and so forth can be detected, the more quickly action can be taken and secondary damage contained. As Ms. Nagami pointed out, "News outlets can dispatch reporting crews and helicopters to sites based on information obtained from FASTALERT, so it is an indispensable tool at the forefront of news reporting." This can be considered a good example of technology being effectively used to make up for the shortfalls of limited human resources.

A helpful tool for enhancing safety and security in disaster-prone Japan

News organizations are not the only ones needing fast accurate information on emergencies. FASTALERT is also drawing strong interest from public organizations, social infrastructure administrators and even private businesses who are seeking better and more effective approaches to disaster prevention and risk management. In disaster-prone Japan, FASTALERT might be the help society is looking for.

Catching sight of this potential, Panasonic System Solutions Japan (hereinafter "Panasonic") teamed up with JX Press as a distributor. In the B2B market, they were already supplying safety and security solutions like wireless systems to public organizations and social infrastructure administrators all across Japan, so, when Panasonic got involved, they were well positioned to create synergies by combining FASTALERT into packagable solutions.

Toshihiro Toh from Panasonic explained, "Panasonic is at an advantage because we are well versed in the operations of the public and social infrastructure sectors. In the public sector, you have governments, police, fire departments, administrative offices and so forth, while in the social infrastructure sector, you have communications, traffic, energy, broadcasting and other service providers. We're confident that we can offer both of these sectors high value-added solutions by bundling in FASTALERT."

Panasonic's encounter with JX Press was at a tradeshow related to disasters. Seeing for himself how breaking news was produced on the system screen practically in real-time, Mr. Toh recounted, "The speed and accuracy were very compatible with our solutions. FASTALERT provides valuable information for making decisions and launching investigations in a disaster." Persons in charge of crisis management at public organizations have been wanting to use SNSs, but because of the risk it poses to residents, information sourced from SNSs cannot be released to the public unless it can be trusted.

Mr. Toh added, "Their biggest concern was how reliable it was. For that reason, JX Press and Panasonic conducted a month-long trial with it. Many of our customers who tried it said it was 'simply wonderful' and we continue to receive lots of inquiries not only from public organizations but private businesses as well."

While all of that was going on, Panasonic was feeling out customers for new demands and feeding them back to JX Press. Adding the ideas from places other than news organizations would only improve FASTALERT more. Mr. Toh highlighted that "What makes JX Press so good is the speed with which they work. In the past, we had them quickly develop features on more than one occasion." In response to that, Ms. Nagami noted, "JX Press is very open to trying new development languages and technologies. Asking more of us only motivates our engineers more." This collaborative relationship has had and continues to have a visible synergistic effect.

Sharing his hopes for the future, Mr. Toh commented, "I'd like to see AI analyses of various information sources other than SNSs. By fusing FASTALERT into our solutions, I think we can grow it into an indispensable service during disasters in Japan." JX Press feels similar, as Ms. Nagami explained, "By gathering a wide range of information from the general public, we are aiming at tapping into 'collective intelligence.' We want to use that to create a future where damage and losses across society or an entire region are minimized."

- Reproduced from the website "Mirai-kotohajime", by courtesy of Nikkei Business Publications, Inc.

# # #