Director of data science Gregor Lämmel on “AI in action”

Market Logic Team

Gregor Lämmel, Director of Data Science at Market Logic Software, joined host Anthony Kelly on the AI in Action podcast to discuss his foray into the world of AI-driven insights technology and the role Market Logic plays in the exciting SaaS landscape. Powered by Alldus International, premium AI Workforce Solutions, the AI in Action podcast breaks down the hype around AI and explores the impact that Data Science, Machine Learning and Artificial Intelligence have on our daily lives. It shares insights from technologists and data science enthusiasts and showcases top AI developments from around the world.

From transportation systems to consumer insights

Gregor Lämmel joined Market Logic in 2018 after a career in research and academia, where he worked on large-scale transport issues and in aerospace. His transition from theory to practice was gradual but not particularly difficult – the major difference is, he says, the faster pace of change and shorter iteration cycles in the private sector, which means there is more room for fast innovation.

Fast innovation for insights

At Market Logic, Gregor and the rest of the technology teams work on providing B2B SaaS solutions to help clients run truly insights-driven businesses. Major firms like Coca-Cola, Unilever and Procter & Gamble spend millions on market research to truly understand their consumers, but that data is often processed manually and then forgotten in siloes. Market Logic provides a centralized market insights platform for all market research data, comprising thousands of presentations, reports and table data, where insights professionals can organize their data, find new insights from their data and can easily collaborate with their peers. A Market Logic platform is similar to a proprietary Google, LinkedIn and Twitter for each client’s market research.

The technology driving data-driven insights

Crunching the data involves many state-of-the-art AI and Big Data technologies, like NLP, Deep Learning. Technologies like named entity recognition and multi-word expression extraction are processed with market taxonomies, and the resulting data is stored in the knowledge graph – which helps make sense of it all, create new insights, and draw conclusions from it. (Read all about knowledge graphs here.)

Symbolic AI vs Deep Learning

Gregor weighed in on symbolic AI vs Deep Learning and which one has a better practical implementation. Although symbolic AI was famous in the 1980s and 1990s, nowadays Deep Learning and Neural Networks have become more in fashion. Market Logic technology uses the best of both worlds. The knowledge graph is a symbolic representation (you look at it and understand what the relations are about – almost like natural language). But if you want to understand the similarity or difference between concepts, you need a numerical representation, which is where Deep Learning comes into play. Gregor’s team tries to use symbolic representation wherever possible, especially because it’s easier to debug.

Finding insights quickly with the Auto-Summarizer

Market Logic’s flagship AI product is called the Auto-Summarizer, which helps users get through large amounts of documents in a short amount of time. Instead of users scanning hundreds of documents to find the insights they need, the Auto-Summarizer builds upon clean data that is already extracted in the data ingestion layer, then automatically evaluates the data to determine if it contains “Findings”: insights within artefacts that are used directly for learning and decision making. For example, if a user wants to understand how well a certain shampoo is selling in the UK, they would see 500 results. Within the documents, there are common entities and synonyms expressing a sentiment, like “consumers don’t like olive oil scented shampoo.” The Auto-Summarizer extracts this information and displays it as a Finding.

Communicating the benefits of AI to business stakeholders

In summary, Gregor explained how to prove the ROI of AI to business stakeholders. He thinks there needs to be a real exchange of ideas: data science teams, in particular, need to be cross-functional and include a business intelligence expert. It is important to provide continuous education and exchange of information within a company so both the technology and business sides understand the value the other is creating.