LANGUAGE PROGRAM

Create structured product data using AI

The language program is another vertical of Rakuten Institute of Technology. This group comprises of AI researchers who are specialized in natural language processing techniques. The goal of this group is to convert the vast unstructured data of Rakuten into structured one. The focus lies with catalog data primarily. Benefits of structured catalog data are numerous, ranging from better product search in Rakuten Ichiba, effective filtering for products, quicker understanding of demand of products, to superior product matching and recommendation. Key techniques used for converting unstructured data to structured ones include attribute extraction for must-have attributes, attribute normalization for critical attributes to ensure consistency, key-word extraction, and others to name a few. We also have a dedicated Machine Translation team on a mission to help Rakuten strengthen its global presence and lower language barriers by developing language translation solutions for Rakuten’s portfolio businesses.

Fully automated AI catalog

Rakuten is one of the largest e-commerce platforms in Japan with more than 300 million products. The clean, structured item data that captures the key features is one of the foundational technologies to innovate the shopping experience. It leads to the purchase decision-making for the customers, such as size and color information for fashion and spec information for electronics. Once we have the data, we believe it is possible to make a much better discovery and search experience based on AI. However, the challenging with creating such structured data is that it is very costly and requires manual processing.

In the Language program, we strive to develop a fully automated AI catalog where all the necessary technologies for structures such as item categorization, feature extraction, query, and customer behavior understanding and normalizations are conducted by automated AI with small inputs from annotations.

In addition, Rakuten aims to become one of the largest communication platforms with Rakuten Mobile and Viber. Rakuten's dedicated Machine Translation team is supporting the vision by focusing on eliminating language barriers and facilitating communications for Rakuten customers worldwide.

Machine Translation

Rakuten's dedicated Machine Translation team is on a mission to help Rakuten strengthen its global presence and lower language barriers by developing language translation solutions for Rakuten’s portfolio businesses. Rakuten Translate helps translate content of Rakuten’s global services, such as video subtitles, e-commerce product descriptions, and travel booking information, into multiple languages using deep learning technology. The Rakuten Translate team was awarded a special prize at the Nikkei Deep Learning Business Awards in October 2019 in recognition of its excellence in generating new businesses and social impact using deep learning.

PUBLICATIONS

VIEW ALL

Sakura’s Participation in WAT 2021: Effectiveness of Pretrained Models for Multilingual and Multimodal Machine Translation

Conference: WAT 2021

Author: Raymond Susanto, Dongzhe Wang, Sunil Kumar Yadav, Mausam Jain, Ohnmar Htun

Publication Year: 2021

CONFERENCES InternationalMultimodal Item Classification fully based on Transformers

Conference: ECNLP4

Author: Lei Chen, Houwei Chou, Yandi Xia, Hirokazu Miyake

Publication Year: 2021

Label-Guided Learning for Item Categorization in e-Commerce

Conference: North American Chapter of the Association for Computational Linguistics (NAACL)

Author: Lei Chen, Hirokazu Miyake

Publication Year: 2021

Goku’s Participation in WAT 2020

Conference: Workshop on Asian Translation (WAT)

Author: Dongzhe Wang, Ohnmar Htun

Publication Year: 2020

  • Sakura’s Participation in WAT 2021: Effectiveness of Pretrained Models for Multilingual and Multimodal Machine Translation

    Conference: WAT 2021

    Author: Raymond Susanto, Dongzhe Wang, Sunil Kumar Yadav, Mausam Jain, Ohnmar Htun

    Publication Year: 2021

Item 1 of 4

CAREERS

We are always looking for great talent and researchers to lead us and work with us.

LEARN MORE