Language is beautiful, unique, nuanced, complex, ubiquitous and powerful all at the same time. Communications through language and text surround and power our personal interactions. This power in language extends into all our interactions, including within the enterprise.
In the enterprise, language data is how your customers communicate with you, it’s how you learn about the risks and the environment where you operate, it’s what you gather to make decisions. Today, enterprises are capturing more data than ever before and among the many ways we can use this information—create efficiencies, drive decision making, etc. — it offers a tremendous opportunity to effect positive change in our businesses and the world at large.
This was the idea behind our third hackathon, which officially wrapped up last month. “Turn Language into Action: A Natural Language Hackathon for Good” challenged developers to use the expert.ai API in a web application project that tackled some of the most difficult challenges that language presents us—and turn it into a tool for good.
We announced the project winners on November 17, and now we’d like to share a closer look at the challenge itself and winning projects.
About the Challenge
Language is the vehicle for how we communicate and share ideas, and the online world of the internet and social networks has transformed how information spreads, for the better and for the worse. While disinformation, misinformation, fake news and cyber bullying all happen through language via written text, technologies that work with language—like natural language processing (NLP) and natural language understanding (NLU)— offer us a tremendous opportunity—and very powerful tools— to combat them.
The challenge to developers was this: Build or enhance an existing app by embedding natural language capabilities, word meaning and sentiment analysis with the purpose of highlighting or enabling ways to improve the world using deep natural language processing. For example: analyzing reactions on social media to detect cyber bullying, for identifying fake news, for monitoring for ESG risks on social media.
Competitors were invited to leverage our Natural Language APIs and “smart from the start” knowledge models in specific domains, such as Hate Speech Detection, ESG (Environmental, Social, Governance) and Emotional Traits & Sentiment Analysis. In evaluating the projects, we were looking for creative ways of leveraging our API in the application, for overall creativity in applying the tools at hand, and for ideas that had real impact.
The result? Out of 530 registrants, 46 projects were submitted, and we selected three category winners and three finalists. We were absolutely blown away by the creativity and the high caliber of submissions that showcase the variety of use cases where our AI-based NL capabilities can be a game changer for building innovative applications that leverage language data.
Read on to learn all about the winners!
The Winning Projects
1st honorable mention: Hate speech detection: citizen5
Citizen5 is a secure tool that supports open-source digital investigations into war crimes that is designed to resist disinformation and attempts to discredit investigations by promoting transparency in the investigative process. Citizen5 uses the expert.ai API to analyze text and identify hate speech, detect and anonymize personally identifiable information (PII) and to extract relations it contains. This could be any kind of communications (articles, transcripts, interviews) but also social media postings and other media.
2nd honorable mention: ESG: SustainaMeter
SustainaMeter provides insights to companies and the public to increase transparency about companies’ attitudes towards ESG issues and to facilitate comparison within their industry and with public perception. It uses expert.ai to classify companies’ LinkedIn posts into ESG categories like sustainability, ethical labor sourcing and corporate board diversity with sentiment scores for each post.
3rd honorable mention: Sentiment & Emotions: NewsReport
NewsReport is an app that helps readers know if a website is untrustworthy or if it is generating negative emotions, such as fear and outrage. NewsReport uses expert.ai to find instances of hate speech and negative emotions in articles to determine the density of these elements—right down to the most hateful and emotionally charged sentence—and calculate a “grade” for each website. Users, anyone,? can even run the analysis on their own articles.
Third place: MeMo
MeMo is a cloud-based Metaverse Moderator solution that uses AI to create a safer metaverse. MeMo uses the expert.ai NL API to analyze metaverse content creation before it goes live to identify hate speech, the use of PII and overall sentiment. It also provides end users with information about the sentiment, emotional traits or content warnings in a metaverse story to protect them from inappropriate content and it blocks PII to protect the identity of users.
Second place: atlat
atlat’s Understand Human Rights Complaint Analyzer is a digital global grievance system that provides workers with a channel to safely and anonymously submit human rights violations, whistleblower complaints or other workplace concerns directly to the brands they work for. atlat uses the expert.ai API to analyze and classify complaints up and down a company’s supply chain according to ESG risk areas, to understand the sentiment of the complaint and to identify any PII so that the anonymity of the worker is preserved.
First place: PagePal
PagePal is an assistant tool that helps children with autism spectrum disorder (ASD) to better recognize emotions in written texts, such as books or other forms of storytelling. PagePal uses the expert.ai NL API to label emotional traits in text so that young readers are able to identify and follow the emotional elements in a text, helping them to identify and understand these emotional cues and ultimately to more easily navigate the world. PagePal is available for anyone to start using: they have already tagged a few stories, and users can add their own text to generate labels for emotional traits.
So again, congratulations to all of the winners and to all of the teams who participated in this exciting challenge. We were really impressed. Hopefully everyone had fun building their projects and learning about others. Watch our livestream announcing the winners for a look at the winning projects.