May 18, 2024

In the rapidly evolving realm of online marketplaces, the efficacy of product searches is crucial for user satisfaction and business prosperity. Acknowledging this critical aspect, Rakuten Group has embarked on a transformative journey to enhance search and recommendation outcomes through the formidable power of deep learning AI technology.

At the forefront of this ambitious initiative is Lee Xiong, a trailblazer renowned for his pioneering work in applying deep learning technology at scale, particularly in semantic search. Semantic search represents a paradigm shift in search technology, transcending traditional keyword matching to grasp the essence and context of queries and thereby delivering markedly more relevant results.

Xiong elucidates on this innovative approach, detailing how semantic search harnesses deep learning models to transform text into vectors. This enables a deeper understanding of user queries and yields more precise search outcomes. It involves training neural networks with vast amounts of query and item pairs to create multi-dimensional vector representations, thereby enhancing the accuracy and relevance of search results.

With an unwavering focus on achieving the pinnacle of e-commerce search and recommendation quality, Rakuten Group leverages its unparalleled wealth of user search history and purchase behavior data to refine and optimize these models. The culmination of this exhaustive process is developing and deploying state-of-the-art semantic search models poised to redefine the e-commerce search experience.

The journey toward implementing this groundbreaking search technology was marked by collaboration and innovation, with a diverse team of engineers and researchers working tirelessly across multiple time zones. Through a meticulous process spanning seven months, the semantic search models were meticulously trained and fine-tuned, culminating in their successful integration into Rakuten’s flagship services.

The impact of these efforts has been nothing short of transformative, with initial tests showcasing remarkable outcomes. Across Rakuten Fashion and Ichiba services, no-match searches have plummeted, while searches per session have surged, underscoring the enhanced user experience and increased engagement facilitated by semantic search.

As Rakuten Group continues to innovate in e-commerce, the widespread adoption of semantic search technology promises to unlock new opportunities and drive sustained growth across the Rakuten ecosystem. With its steadfast commitment to excellence and relentless pursuit of technological advancement, Rakuten Group is poised to shape the future of e-commerce in Japan and beyond.