Today, more than one-third of American adults (i.e., 78.6 million) are obese, and nearly 70% are overweight, leading to an estimated medical cost of $147 billion in 2008. Although food journaling is a useful tool for weight loss, existing diet tracking applications are typically too time-consuming for many users, requiring manually entering each eaten food by hand and selecting the correct item from a long list of entries in the nutrient database. In collaboration with Dr. Susan Roberts and her team of nutritionists at Tufts University, our goal is to provide a simpler diet tracking option for anyone who wants to track their food intake by applying speech and language understanding technology to automatically detect food entities and find the corresponding nutrition facts in a database.
Our nutrition analyzer uses cutting-edge speech and language understanding technology to allow people to track their dietary intake more easily and efficiently than existing applications. A user simply speaks or writes a sentence describing their meal in natural language (e.g., “This morning I had a bowl of Kellogg’s corn flakes”), and our system automatically determines the corresponding nutrient database entries (e.g., “Cereals ready-to-eat, KELLOG, KELLOGG’s Corn Flakes”) and related quantities (e.g., “a bowl”). We are currently exploring dialogue mechanisms for asking the user followup clarification questions (e.g., "Was that 2% or nonfat milk?") and providing personalized nutrition advice.