Finding the perfect product with technology
Finding the perfect product with technology
NAVER I Tech I Kim Do-hui
NAVER search is now an inseparable part of our everyday life, but it continues to develop in order to change with the times. In addition to providing the right answer to users’ questions in a timely manner, NAVER’s search service now also recommends content that users would like based on their search history. Kim Do-hui has been working as a developer for 11 years now, and is in charge of recommendation modeling development for personalized search. Do-hui started off her career as a backend developer at NAVER, but her interest in data and users motivated her to study machine learning on her own. She changed jobs when her current team came to know about her passion for machine learning. Do-hui’s goal is to provide users with a practical and valuable experience using recommendation. Just like how the technology development for NAVER’s search service is ongoing, Do-hui’s aspirations to help users find the perfect content is ongoing.
Tell us about the work you are doing at NAVER?
I am in charge of ‘MY Subscription’, a service that recommends content created by the users themselves to other customers according to their personal preferences. MY Subscription was just a simple platform that displayed new posts from the channels the users’ have subscribed to, but I wanted to recommend the right content to the right user by understanding their personal interest and taste to make the experience more meaningful.
What is the difference between recommendation and search?
You might not notice the difference between the two because both recommendation and search are services that help users find information that they didn’t know before. But they are different in that search only shows information with user input, whereas recommendation provides information to users without search queries. In short, recommendation is ‘a search service offered to users without entering queries’. In order to offer recommendations, we need to understand the users by analyzing the user logs that are stored in existing services. The queries that find the answers for us are extremely important, but there are users who don’t know what they are looking for even when they are entering search queries. So we have developed AiRSearch, a service that offers a great search experience to users by showing them what they have been looking for all along.
Tell us more about AiRSearch.
One of the projects related to AiRSearch is a project called Smart Block which is a service that goes one step further from providing only the correct answer when a query is entered based on a slightly larger category such as interior, fishing, and camping. I came to think that there would be different documents or collections that users might be interested in. That is mainly the reason why we are working on this project that makes a variety of recommendations to suit individual tastes. For example, when users search for ‘home interior’, those who have recently bought a house may want more information that will help them change the look of the entire house beyond simple home accessories.
I heard you changed your field of expertise afterwards. Was there a specific occasion?
I first started off as a backend developer and then I switched my career to machine learning. It was around the time when Lee Se-dol competed against AlphaGo in a Go match. I was interested in machine learning even before the match, so I started studying on my own. It happened to be the time when I was working for news development and gained experience in modeling and data engineering. Above all else, I was more interested in doing data-related work because NAVER has copious user logs and is one of the companies with a lot of data generated from those logs. I wanted to take part in analyzing the user logs to extract meaningful data and gain insights in order to take the service to the next level to offer services that add value to the everyday lives of our users.
What does it mean to be good at your job?
There could be a long list of skills and characteristics of a good employee, but in my opinion, I think problem-solving ability, from observing certain phenomena, to defining problems and finding solutions, is the most important quality. In addition, you must have the skills to fix the problem. For developers, it seems like an endless loop. You raise the problem yourself, solve it and find another problem.
What should we bear in mind when it comes to machine learning?
I think there are a lot of people who have recently started taking interest in AI and machine learning. There are abundant materials and data to study and learn related technologies, and libraries that you can actually try implementing yourself. Learning these skills is of course important, but I think imagining how these skills are going to be used in services is equally important. We should think about how to add more value into users’ lives and to offer more meaningful user experiences. No matter how great the technology is, it becomes worthless unless it is in use.
