Babel Street: The Cacophony of Social Media is the New Newsroom
Natural language processing and sentiment analysis are enabling computers to understand our social media posts. What are the implications for public safety and user privacy?
You may think that your friends and family are the only audience liking and sharing your social media posts, but data analytics company Babel Street has made it their business to leverage open source data for event detection. Using powerful algorithms, Babel Street鈥檚 software is parsing through our day-to-day social media conversations and extracting relevant insights.
A Computer That Really Gets Me
Babel Street is a data analytics company that provides software as a service (SaaS) products to users in intelligence and law enforcement. Their technology is able to consolidate and interpret data from millions of publicly available sources including social media and news.[1] Babel Street鈥檚 customer promise is to transform this unstructured, multilingual data into actionable insights in real-time. The company is using machine learning to organize the terabytes of noisy data published on the internet each day and extract signals that can be used by public safety teams to react quickly to 鈥� or even preempt 鈥� security threats.[2]
In order for their software to make sense of all this data, Babel Street invests heavily in natural language processing (NLP), a form of artificial intelligence that helps computers understand human language.[3] The sophistication required by Babel Street鈥檚 natural language processing algorithms is much greater than the industry standard for two reasons. First, Babel Street is using data sourced from communication platforms like social networking websites and forums, which are characterized by complex, intertwined conversations. Second, while the field of natural language processing has made significant strides in analyzing the semantic dimension of human discourse, there are still gaps in the ability to evaluate contextual nuances.[4]聽 Babel Street aims to close that gap and boasts to “possess the most sophisticated sentiment analysis tool on the market.”[5] Sentiment analysis enables computers to interpret the meaning behind text and understand the user鈥檚 attitude, emotions, and intentions.[6]
Machine learning is important to Babel Street because artificial intelligence drives the core event detection capability of the company’s software. Humans are not able to process data with the same speed and volume that computers can handle, and for Babel Street鈥檚 customers, speed and accuracy are critical. For example, Babel Street was the first source to detect the publication of the thirteenth issue of Dabiq, ISIS鈥檚 official magazine, when it was posted by propaganda accounts on Twitter.[7]
The results of machine learning algorithms also improve over time as they are given more training data.[8] Some of Babel Street鈥檚 customers have even started using the software to predict events based on trends and anomalies in the data. Counterterrorism units in the Pentagon, for example, use Babel Street to track terrorist conversations on social media to predict attacks.[9]
Next Stop, Emojis
In the short-term, Babel Street is adapting its natural language processing capabilities in response to evolving trends in internet communication. For example, criminals are increasingly using emojis to communicate in order to circumvent text analysis.[10] In the medium-term, the company is planning to expand its market to other industries. Reputational risk has become increasingly more difficult for companies to manage as customers air their grievances publicly on Twitter and Facebook. Babel Street plans to expand its enterprise solution and offer software products for brand management.[11]聽With their sentiment analysis capabilities, Babel Street can offer corporate clients the ability to gauge positive or negative customer sentiment toward brands.
In order for Babel Street to remain competitive in the evolving internet communication landscape, they will need to invest in image analysis as a complement to natural language processing. With the increased use of image sharing platforms like Instagram, events will be chronicled through photos and videos more frequently. Babel Street will need to use image analysis to identify significant events in media posts when there is no text present.
Helpful Or Harmful?
While the public good delivered by Babel Street鈥檚 software is undeniable, the company operates in an industry where privacy concerns around user data have made recent headlines. An important question to raise is, where do we draw the line when making the tradeoff between public safety and user privacy? In 2016, Babel Street鈥檚 competitor, Geofeedia, was heavily criticized by the ACLU for providing a tool that allowed Chicago police to surveil protests by searching for public social media posts by location[12]. The ACLU argued that Geofeedia failed to protect free speech rights, especially for activists of color. However, a counterargument could be made that public social media posts are fair game and that the onus should be on users to be educated about the implications of sharing information publicly. As machine learning becomes more advanced, it becomes more important for us as savvy internet users to be aware of how we share our data.
(Word Count: 776)
[1] Babelstreet.com. (2018). [online] Available at: https://www.babelstreet.com/ [Accessed 13 Nov. 2018].
[2] Gregg, A. (2017).聽For this company, online surveillance leads to profit in Washington鈥檚 suburbs. [online] The Washington Post. Available at: https://www.washingtonpost.com/business/economy/for-this-company-online-surveillance-leads-to-profit-in-washingtons-suburbs/2017/09/08/6067c924-9409-11e7-89fa-bb822a46da5b_story.html?noredirect=on&utm_term=.13d83f405bc3 [Accessed 13 Nov. 2018].
[3] Mills, T. (2018).聽What Is Natural Language Processing And What Is It Used For?. [online] Forbes. Available at: https://www.forbes.com/sites/forbestechcouncil/2018/07/02/what-is-natural-language-processing-and-what-is-it-used-for/#6aee142a5d71 [Accessed 13 Nov. 2018].
[4]聽Abbas, Ahmed, Yilu Zhou, Shasha Deng, and Pengzhu Zhang. 2018. “Text Analytics To Support Sense-Making In Social Media: A Language-Action Perspective”.聽MIS Quarterly42 (2): 427-464. doi:10.25300/misq/2018/13239.
[5] “Meet Babel Street, The Powerful Social Media Surveillance Used By Police, Secret Service, And Sports Stadiums”. 2018.聽Motherboard. https://motherboard.vice.com/en_us/article/gv7g3m/meet-babel-street-the-powerful-social-media-surveillance-used-by-police-secret-service-and-sports-stadiums.
[6] Socher, Richard. 鈥淎I鈥檚 Next Great Challenge: Understanding the Nuances of Language.鈥澛�性视界 Business Review Digital Articles, July 2018, pp. 2鈥�4.聽EBSCOhost, ezp-prod1.hul.harvard.edu/login?url=http://search.ebscohost.com/login.aspx?direct=true&db=bth&AN=130975191&site=ehost-live&scope=site.
[7] “Meet Babel Street, The Powerful Social Media Surveillance Used By Police, Secret Service, And Sports Stadiums”. 2018.聽Motherboard. https://motherboard.vice.com/en_us/article/gv7g3m/meet-babel-street-the-powerful-social-media-surveillance-used-by-police-secret-service-and-sports-stadiums.
摆8闭听Brynjolfsson, Erik, and Andrew McAfee. 鈥淲HAT鈥橲 DRIVING THE MACHINE LEARNING EXPLOSION? Three Factors Make This AI鈥檚 Moment.鈥澛�性视界 Business Review Digital Articles, July 2017, pp. 12鈥�13.聽EBSCOhost, ezp-prod1.hul.harvard.edu/login?url=http://search.ebscohost.com/login.aspx?direct=true&db=bth&AN=124641872&site=ehost-live&scope=site.
[9] Gregg, A. (2017).聽For this company, online surveillance leads to profit in Washington鈥檚 suburbs. [online] The Washington Post. Available at: https://www.washingtonpost.com/business/economy/for-this-company-online-surveillance-leads-to-profit-in-washingtons-suburbs/2017/09/08/6067c924-9409-11e7-89fa-bb822a46da5b_story.html?noredirect=on&utm_term=.13d83f405bc3 [Accessed 13 Nov. 2018].
[10] Ibid.
[11] Ibid.
[12] “Police Use Surveillance Tool To Scan Social Media, A.C.L.U. Says”. 2018.聽Nytimes.Com. https://www.nytimes.com/2016/10/12/technology/aclu-facebook-twitter-instagram-geofeedia.html.
I thought the content in this piece was accurate and thoughtful. I have used Babel Street technology before and can attest to the speed, efficiency, and general common good that it provides. So, thank you for the holistic view point. The last question you raised is critical and has been a central debate in policy, dating back to 2003 and the release of the Patriot Act. U.S. policy has simply not caught up with technology yet. As AI advances, the federal government鈥檚 response lags behind. Regardless, I believe it won鈥檛 be until some type of standard (law/policy) is set that this fundamental, and arguably philosophical, question can be answered.
What I found particularly compelling about your piece was your underlying point regarding speed at the tradeoff of human judgement. While at first glance this may seem enticing, I am hesitant to believe it will be albeit of profiling. I wonder in what ways you can restructure the learning to only raise topics of a certain confidence level 鈥� much like Watson. To that point, I wonder at what threshold of crime, crisis, or general situation, utilizing this technology makes sense and does not infringe on citizen鈥檚 rights.