Meltwater: A Global Disruptor in Media Intelligence and Outside Insight Powered by AI

September 10, 2018

Aditya JamiMeltwater, a pioneer of media intelligence and now Outside Insight, gives businesses the information advantage they need to stay ahead. The world’s online data can unlock key indicators about a market or competitor. Making sense of this data creates an information advantage; something the company calls Outside Insight. But harnessing the volume and variety of global online data is a time-intensive and expensive task, both for businesses and data scientists.

Meltwater creates outside insight solutions from PR and marketing to finance and executives to give businesses the information advantage they need to stay ahead. Its solutions are used by more than 30,000 clients across the globe.

The company was founded by Jorn Lyseggen in 2001 in Oslo, Norway with just $15,000. Meltwater’s mission at inception which still holds true till date is to give businesses and its decision-makers the information advantage from publicly available sources on the internet. At first, Meltwater focused on the nascent field of digital news and became the leader in media monitoring. Then, with the advent of social media, the company was one of the first to integrate social media into its platform creating Media Intelligence. More than 30,000 companies in 121 countries have used Meltwater’s media intelligence to stay on top of billions of online conversations and extract relevant insights to strategically manage their brands.

Now, Meltwater is pioneering again with the introduction of, a comprehensive insights platform for developers and data scientists to create analytics on global online data sources with configurable AI models. The company has grown to include 2,000 employees in 60 offices across six continents. Meltwater is also committed to fostering the data science ecosystem through MEST, a pan-African entrepreneurial program and incubator, and Shack15, a global data science community.


The Influential Leaders

Jorn Lyseggen, a serial entrepreneur and philanthropist, is the Founder & CEO of Meltwater. Launched in his home country of Norway in 2001 with just a $15,000 investment, Jorn and his team have grown the company to become a global leader in media intelligence. Today, Meltwater has more than 2,000 employees in 60+ offices across six continents, with more than 30,000 clients worldwide.

Jorn released his first book, Outside Insight: Navigating a World Drowning in Data, in October 2017. The book describes the emergence of a new software category that will leverage insights gleaned from the vast amount of external information available on the web to transform corporate decision-making.

Given his experience as an entrepreneur, Jorn is committed to supporting the next generation of tech leaders. Based on the notion that talent is evenly distributed, though opportunity is not, Jorn founded the Meltwater Entrepreneurial School of Technology (MEST) in 2008. Headquartered in Accra, Ghana–with a presence in South Africa, Kenya, Nigeria and Ivory Coast–MEST is a pan-African training program, seed fund and incubator for African tech entrepreneurs. In 2016, Jorn launched the world’s first data science hub, SHACK15, as a place where tomorrow’s leading data science startups can come together to meet, collaborate and innovate.

Aditya Jami is the Chief Technology Officer of Meltwater, responsible for the overall technical strategy and drives the company’s AI & Data Science initiatives across all product lines. He works very closely with Jorn on acquisitions and leads their technology integration.

Aditya has been responsible for leading the transformation of underlying media intelligence technology into an expanded, data science platform – – which powers all Meltwater’s products across multiple business lines. He is motivated by solving problems in the areas of large-scale multi-modal data management, machine learning on the cloud, information network analysis and knowledge graphs. Aditya was a founding member of the Cloud Solutions team at Netflix, where he built the first Chaos Monkey (later extended to Simian Army) that ran on the production cloud, Cassandra/Priam as well as several other cloud migration initiatives. At Yahoo, he worked on a real-time data platform that collected and analyzed 300 billion web events (10TB) a day to power its news feed recommendation and behavioral targeting advertising models. He has served as chief architect of the cross-university joint project called RoboBrain, a large-scale computational system that learns from publicly available internet resources, computer simulations, and real-life robot trials. RoboBrain was listed among the MIT Technology Review’s Top 10 Breakthrough Technologies in 2016.


Delivering Competitive Intelligence with Deep Learning

Meltwater uses AI and deep learning at various stages of the underlying platform and in its products.

•  Data Acquisition: Meltwater’s crawling engine is a fully automated unsupervised full website extraction system. Based on a few samples, it figures out how to navigate web content, what that content is about, and then how to extract structured attributes from the data and sends it to further downstream processing. In total, its crawlers bring in data from tens of millions of sources across online news, broadcast, social media, reviews, forums and corporate websites.

•  Natural Language Processing: Meltwater analyses billions of data points every day and in order to extract meaningful insights, all its data is constantly augmented with very rich linguistic annotations tasks including Semantic (Named entity recognition, sentiment analysis, event & relationship extraction, topic modeling), Syntactic (POS tagging, brown clusters, and lemmatization), Discourse (Automatic Summarisation, Coreference resolution) across 20 different languages.

•  Knowledge Graph Management: Knowledge graph is at the heart of everything Meltwater does. All the data the company acquires and the insights it derives are linked to the graph. The company extracts factual information about entities (companies, products, brands, and influencers), associated events, and relations from these data sources to continuously update the knowledge graph. Rule mining and link prediction techniques are applied to ensure high quality of the graph. The knowledge graph is used for improving its search interface, mining user intent, recommendation engine, provide contextual information and reasoning capabilities to the Insights Meltwater delivers.

•  Insights Creation: Data science has been held back by the need to curate and wrangle data into shape and extract meaning from unstructured sources such as news. Through Meltwater’s knowledge graph, those tasks are largely taken care of and data scientists can focus on creating higher-level insights — say, that a company is entering an expansion phase — quickly and learn which models provide the most value. Meltwater provides a library of such higher-level insights together with easy ways to create new ones through the combination of information from the knowledge graph, from news documents, or from the web.


Artificial Intelligence to Power the Future of Media

Aditya believes artificial intelligence (AI) is going to have a profound impact on media intelligence; essentially business intelligence on outside data. Lots of companies in this space to date have largely focused on elegant solutions that deliver measurements and manage content (owned, earned, paid) but less on providing insights that will help businesses understand trends, drive communications, and business strategy. Most of these tasks require a context builder that can systematically connect the dots across different data sources (instead of looking at them in isolation), understand causalities and surface insights.

“The key is to figure out a way to programatially get customer feedback through product hooks and personalize their experience. All of these require in-depth use of AI and ML methods to get it right,” he added.


At the Forefront of Disruptive Innovation

The expertise behind Meltwater’s technology reflects years of research and development from an experienced team of in-house engineers and data scientists, many who arrived through a series of AI acquisitions or have strong ties with academia. Meltwater’s platform is also supported by collaborations with top universities and research groups, such as Carnegie Mellon University and the University of Oxford. In addition, the platform is informed by a Scientific Advisory Board. Its founding members – Regina Barzilay, Georg Gottlob, Jure Leskovec and Eric Nyberg are considered pioneers in the fields of natural language processing, information retrieval, web data extraction and information network analysis.


The Remarkable Accomplishments

Recently, Meltwater opened up its in-house data science platform that’s currently powering all its product lines– to give data scientists and developers access to 1.2 trillion documents from 10 million sources, organized by 12 million entities, growing by 600 million documents per day and library of pre-trained and configurable AI models to accelerate creation of novel insights. also makes it easy to integrate its knowledge graph with organizations’ internal data and existing models. For example, companies can integrate into enterprise risk management to navigate complex supply chains and use the powerful knowledge graph to trace events through global supply chains, connecting suppliers, products, and their dependencies.

Advancements in data science rely on easy access to large-scale data, but collecting and processing a large volume of unstructured data frequently requires a cost-prohibitive investment in engineering. That’s why Meltwater has also given researchers at top universities such as Oxford and Carnegie Mellon an access to, to advance key research projects that will address real-world problems like identifying fake news detection.


Challenges & Hurdles

Aditya feels that large labeled training datasets are critical building blocks of supervised learning methods and key enablers of deep learning techniques as well. Creating these labeled training sets still remains to be the most time-consuming phase for Metawater’s data scientists to ship new models and hasn’t changed over time. The company started experimenting with domain adaptation and distant supervision techniques and while the results are promising, they are only applicable for a subset of tasks at hand.


Future Through the Lens of Media Intelligence

The internet has changed the way decisions are made. Today, companies are still focused on mining internal data. But there are valuable insights hiding within the vast amount of external information available online that can drive more informed, forward-looking business decisions. “Those that embrace this new digital reality will have an information advantage over those that don’t – that’s what we mean by outside insight.  We believe outside insight will be the new decision-making paradigm for mastering this new digital reality, and that’s why we’re bringing new tools to give businesses access to that information advantage,” Aditya said.