Recipes lined up on a table with someone holding and iPad

Bringing NLP expertise to Cookpad

Liverpool computer scientist, Dr Danushka Bollegala, is bringing his expertise in natural language processing (NLP) to the world’s largest recipe sharing platform, Cookpad, to work on a range of recipe and ingredient recognition projects.

NLP is a branch of artificial intelligence that helps computers understand, interpret and manipulate human language and Dr Bollegala’s research focuses on developing algorithms that allow computers to represent the meaning of texts written by humans.

Dr Bollegala is working with Cookpad’s R&D team based in Bristol.

Challenge

Cookpad is one of the most popular cooking platforms in the world, currently operating in 22 languages across 70 countries worldwide. It allows visitors to upload and search through original, user-created recipes. Described as `recipes by home cooks for home cooks’, it has 40 million monthly unique users globally.

Dr Bollegala describes the current problem, “At the moment it is very difficult to automatically evaluate the difficulty of the recipe so we’re building machine learning models that can grade the difficulty of a recipe. We are developing the NLP capabilities to enable people to search more easily for recipes with certain ingredients or find out what can replace ingredients they may not have."

Solution

The project aims to develop algorithms using NLP that can recognise and analyse key terms in the recipes across the platform to enable users to search for recipes with certain food types, by certain skill levels and to recommend recipes based on users likes, skills and previous posts.

Dr Bollegala said, "The first stage is manually assessing and grading existing recipes, so the machine learning algorithms can start using this information to apply it elsewhere. There are certain key pieces of information we can use, like if a recipe has been cooked by lots of people it’s likely to be easier, people write comments on recipes so we can look at these, to decide the difficulty level.”

Impact

For Dr Bollegala, all of this work is a practical application of his academic learning and offers opportunities to carry out further research in a real-world setting.

Tomo Yasuda, managing director of Cookpad, said: “We’re really keen to build closer partnerships with leading academics in fields such as machine learning and feed into their research as well as benefitting from their industry leading findings.

“Dr Bollegala’s extensive expertise is in exactly the areas where we want to develop functionality so it’s proving to be an exciting and productive partnership.”