You make the news: How Spectee mines social media
Spectee, a Tokyo-based information analysis company and online news agency, is looking for partnership in Southeast Asia where it is seeing huge growth in demand for video content.
The company specialises in curating newsworthy videos for mainstream media enterprises and other customers based on social media feeds of breaking events. Its artificial intelligence (AI) technology gives broadcasters and publishers real-time access to user-generated content (UGC) from all over the world.
Spectee also provides alerts via its simple notification service (SNS) in Japan for media companies, government bodies and transport companies.
The company's clients are mostly news agencies with over 100 media partners including NHK, CNN, The Associated Press and Reuters. It also serves clients from other sectors such as transport companies including All Nippon Airways.
"We utilise our AI technology to curate the latest news from the content on social media, which we mainly get from Facebook and Instagram, and we provide all this information for news agencies," Spectee marketing manager Edel Sng told Asia Focus.
"Over 90% of the news agencies in Japan are our clients and right now we are trying to expand overseas," he said. "Transport companies also want to know what is happening with their planes and trains because the fastest way to get first-hand information nowadays is from people through social media."
The Spectee AI application is enhanced by natural language processing, image recognition and frame-to-frame analysis. But despite all the high-tech input, all of the content is fact-checked and licence-cleared by human moderators to ensure accuracy.
"Whatever information is made public on any social media platform, we are able to use it, which is already agreed in the statement before you join any such platform," said Mr Sng.
"Nevertheless, if we are going to publish a video on Spectee.com, we do contact the user to get their permission to share and redistribute those videos."
The inspiration to start the company came during the disastrous earthquake and tsunami of March 2011 in eastern Japan. The company's founder, Ken Murakami, volunteered in the disaster area and witnessed how SNS worked as an effective communication tool among victims, while traditional mass media could not deliver accurate information that quickly.
"During the disaster in 2011, there was a lot of fake news being spread online," Mr Sng recalled. "Not only disinformation, but also misinformation where people deliberately posted pictures that had already been posted in the past, but they shared it as something that happened after the earthquake."
With the help of AI, Spectee can detect when content was posted. By using multiple sources from different users, it can confirm where and when the content was created with geolocation.
"A person cannot just sit down at a computer while searching for videos 24 hours a day," he noted. "They can search for videos by typing in different hashtags on social media but most of the time, they will end up with content that they don't need, since basing things on words only is insufficient."
For example, AI is being taught to detect fire by looking for different images such as the presence of a fire engine, bellowing smoke and other characteristics of the fire itself. Back in July, when a man set Kyoto Animation on fire and the blaze killed 33 people, the fire happened shortly after 10.30 on a Thursday morning. The company's AI was able to detect the fire on social media by 10.39am.
"Within minutes, we were able to let our clients know what was happening, and within 16 minutes after it happened, we delivered 10 different articles with information such as the exact location of the animation studio," Mr Sng said. "Most large media corporations reported the news at least one hour afterward. By then, we already had 20 different articles."
Spectee's AI is now specialised in fire incidents, natural disasters and weather-related videos, road and other transport accidents, along with protests around the world. The AI-human collaboration is also at work to help identify each incident to confirm that it is authentic.
"We have a team of specialised people who understand the area and the culture to help identify the situation, such as the difference between an ongoing protest and commuters in different parts of the world," Mr Sng noted.
Apart from avoiding unwanted content and detecting when and where the content it needs was created and uploaded in real-time, AI can also screen and block hate comments along with the people who post them.
According to HubSpot, a US-based developer of marketing software, people dramatically increased content consumption on the three most popular social networks between 2015 and 2017, with a 57% increase for Facebook, followed by Twitter (25%) and LinkedIn (21%). It also found that 45% of the users it surveyed watch more than an hour of Facebook or YouTube videos a week.
The findings were based on a 2017 survey with a sample size of 1,091 global internet users. When asked where they go to catch up on news, business and lifestyle stories online, 79% said they searched for relevant content in search engines and 76% cited their Facebook feed.
When asked what they tend to skim and what they read or watch thoroughly, video was the type of content people pay close attention to at 55%, followed by social media posts (52%), and news articles (49%).
According to Mr Sng, one of the challenges Spectee is facing as it considers regional expansion is language. "We want local partners such as in Thailand since most of the social media here are not in English and we need them to help us teach the AI about the Thai language," he said.
Spectee employs 15 engineers and over 30 local experts in Asia as it is branching out into the US and Europe right now.
"Some of the current limitations for the AI are in language because we still have to identify the fastest way to filter all the key words to add on to the image recognition," he explained.
Also, Spectee is limited to content from accounts that are open to the public. China is a huge market but most people there use WeChat and if you are not their friends, you cannot access their accounts the way you can with public accounts on Facebook, Twitter and Instagram.
"The good thing about Thailand is that the world average for social media users is about 42% [of the population], but in Thailand it is around 67%, which is massive where every two out of three people are using social media here," Mr Sng added.