SK Telecom Wins Excellence Award for Recommendation Model Algorithm Research in the US

SIGIR 2024 Conference on Information Retrieval
'One Model Version 2.0' Best Paper Award
Up to 3 Times Higher Response Rate Compared to Previous Versions↑

SK Telecom announced on the 5th that its self-developed recommendation model algorithm research won the Best Paper Award at a globally prestigious conference in the field of information retrieval.

At the SIGIR 2024 conference held in Washington DC, USA, on July 18, SKT representatives received the Best Paper Award. From the left in the photo are SKT Personalization Modeling Team managers Hyungjun Yoon, Jung Park, and Taesan Kim.
[Photo by SKT]

SKT received the Best Paper Award at 'SIGIR 2024,' held last month in Washington DC, USA, for research on 'One Model version 2.0.' This study proposed an algorithm that enhances recommendation prediction performance by creating synergy among data from various service domains.

The algorithm was highly evaluated for its novelty, commercial deployment feasibility, and reliability of results through extensive experiments, earning the Best Paper Award given to only the top 0.6% of submitted papers.

SKT's self-developed recommendation model, One Model, was commercially deployed last year with version 1.0. Version 2.0 improved recommendation performance compared to version 1.0 while also increasing learning efficiency.

SKT integrates and refines various types of individual behavior logs in chronological order and predicts customers' next actions through the One Model algorithm, performing personalized recommendations that consider customers' multidimensional characteristics.

For example, it comprehensively analyzes behavioral data from various customer service domains such as subscription plan history, T Deal shopping history, and membership usage history, recommending service benefits or products that match the customer's needs and interests at the most recent point in time.

At the SIGIR 2024 conference, Park Jung, Manager of the Personalization Modeling Team at SKT, is presenting on the algorithm of 'One Model Version 2.0'.

This approach is called 'multi-domain sequential recommendation,' and the One Model actually learns more than 10 different data domains simultaneously, providing integrated recommendations across various channels within SKT through a single model.

Applying this model in practice resulted in up to three times higher customer response rates compared to existing recommendation methods. Currently, the model is applied to SKT’s AI personal assistant service A. Dot’s recommendation system, T Membership, and subscription plan recommendations. Within this year, it is expected to be expanded to recommend various products such as the subscription service T Universe and AI curation commerce T Deal.

Jung Do-hee, head of AI Data at SKT’s AI Service Business Division, stated, "We will continue to apply advanced personalization technologies across our services to further increase customer satisfaction and accelerate our evolution into a global AI company."

This content was produced with the assistance of AI translation services.