Research and Development

The LY Corporation Group (the “Group”) aims to contribute to the development of information technology society through the promotion of R&D and open innovation including the development of open source software (OSS).

R&D in the LY Corporation Group Companies

LY Corporation

LY Corporation (the "Company")  aspires to make Japan more convenient with cutting-edge technologies under its mission statement, " Create an amazing life platform that brings WOW! to our users."  Supporting the Company with advanced technology, LY Corporation Research is engaged in the research and development of next-generation internet-related technologies.
Today, various devices are used in different everyday life situations, and a variety of data is collected from them. By further deepening its understanding of individuals and trends in the world, the Company aims to create value that can help people overcome new challenges. The research and development of LY Corporation is conducted by broadly collaborating with service divisions ofthe Company, and numerous universities and research institutes, in order to achieve maximum results from the valuable environment in which enormous amount of data and feedback from the customers using the Company's services are collected every day. LY Corporation Research also hosts an internship program for students in the hope of providing them with firsthand experience that will broaden and further their future research and development.
In order to contribute to the development of the information technology society, some of the outcomes of these research and development efforts are made available to the public in the form of open source software (OSS), etc. and are utilized in various situations.

Related Links

ZOZO Research - ZOZO Group's R&D wing

ZOZO Research, the ZOZO Group's R&D organization, conducts research and development under the mission: "quantify fashion." Leveraging the information asset possessed by the ZOZO Group, it is engaged in research and development to scientifically unravel fashion.

Related Links

Cases in Which Research Resulted in Services

LY Corporation

Released “Tappy,” a UX tool that visualizes tap accuracy

In January 2024, the Company released Tappy, a tool that creates a model to estimate the distribution of tap coordinates when users select UI elements such as a publish button or link. When the users enter a link, the tool calculates the probability of successfully tapping those buttons or links. Tappy automatically detects tappable UI elements on smartphone webpages and estimates the tap success rate based on the size of these elements. Different device models have different values of scale and pixel resolution (pixels per inch, PPI), and the size of physical UI element that is actually seen and tapped varies. These differences are also automatically calculated to determine the percentage of successful taps. Tappy is used to improve the services provided by the Company and is also open to the public. Creators can benefit from this tool when they make analyses based on quantitative figures and make comparative discussions.

Related Links

Scientific validation of character and line spacing to enhance readability of text content and application to Yahoo! JAPAN News

The selection of font properties is largely dependent on the designer's sensibilities, and since the fonts provided by operating systems and browsers are developed based on the English language, there was room for optimization of Japanese fonts. By conducting a large-scale experiment and statistically analyzing the data obtained, an attempt was made to identify Japanese font properties that many would find easy to read.

In a large-scale preliminary experiment using crowdsourcing, 6,480 data samples collected from 1,080 participants were statistically analyzed together with a university laboratory specializing in interaction, to determine how the combination of the size of characters and line spacing affects the way people perceive the text. As a result, combinations with good visual balance and low psychological load were identified.

Using this analysis results, LY Corporation Research conducted a demonstration experiment with Yahoo! JAPAN News and analyzed the impact of readability on service KPIs. The experiment confirmed that visually balanced characters and line spacing with low psychological load contribute to service KPIs, such as increasing time spent on the website and page views per visit, as well as significantly increasing page views. These factors led to the application of these results to Yahoo! JAPAN News, and this is an example of how the design methodology of improving small experiences can improve larger experiences and KPIs of services.

Related Links

Certified with "FIDO 2," a standard that enables safe login through use of biometrics authentication devices, etc.

Former Yahoo Japan Corporation (“Yahoo Japan”) is certified with "FIDO2," a new standard established by an industry association called FIDO Alliance, whose mission is to promote the standardization of simple and robust authentications. The company acquired this certification in the world's first FIDO2 certification test held in August 2018, and it is the only FIDO2-certified company in Japan. From October 2018, users of Android devices can log into Yahoo! JAPAN services on the Google Chrome browser by using biometric authentication, such as fingerprint authentication, and iOS device users on the Safari browser from December 2020. Both for Android and iOS, this became the world's first ever case of applying FIDO2-certified authentication to commercial services for consumers.

Related Links

Congestion prediction feature (Yahoo! JAPAN MAP, Yahoo! JAPAN Search, Yahoo! JAPAN Car Navigation Apps)

The results of research that uses location history data to "predict" population anomalies several days in advance have been implemented as a function to seek "usual congestion" in various applications.

Related Links

Spelling correction for Yahoo! JAPAN search queries

Using a large amount of search history data, Yahoo Japan developed a machine learning model that auto-corrects spelling errors in search queries.

Related Links

Judgment of positive/negative tweets for Yahoo! JAPAN Realtime Search

In the renewal of the positive/negative judgment of tweets in Yahoo Japan Realtime Search, Yahoo! JAPAN introduced an efficient machine learning approach to achieve a highly accurate judgment feature.

Related Links

Applying tags to stickers for sticker suggestions feature

In sticker suggestion feature, which recommends stickers appropriate to the users’ conversational content, it is important that stickers are tagged correctly. However, manual input of tags results in variations depending on the sensitivity of the individual entering the tags. For this reason, a video language model has been developed for automatic tagging.

Fuzzy search

Former LINE Corporation has continuously developed a library of fuzzy search to enable the extraction of information that matches the user's intent, even when the users make errors in their input. The result of this research is used in the search in Demae-can and suggestion system for LINE stickers (sftrie). The ability to present results that match user needs contributes to a better user experience.

Related Links

OCR (Optical Character Recognition)

Former LINE Corporation has promoted the development of high-precision OCR, and in July 2019, launched "CLOVA OCR" that utilizes this engine for automatic deciphering of business documents. (From April 1, 2023, CLOVA OCR is provided by WORKS MOBILE Japan Corp., an affiliated company to which the business was succeeded.) In 2021, CLOVA OCR was adopted in the project to digitally convert documents possessed by the National Diet Library of Japan.

Related Links

Automation of call centers

The results of speech recognition/synthesis technology development are being used to automate call centers. The technology is used not only to listen to user inquiries with high accuracy, but also to process them by assigning them to appropriate actions and returning the results in an easy-to-understand, friendly voice. These outcomes have been reflected in an AI telephone response service, "LINE AiCAll," which was released in July 2019. (From April 1, 2023, CLOVA OCR is provided by WORKS MOBILE Japan Corp., an affiliated company to which the business was succeeded.)

Related Links

ZOZO, Inc. ("ZOZO")

Application to workforce management in customer support

ZOZO Research has been working on how to achieve more efficiency in the back-end operations with various technologies. One specific example is workforce management (WFM) implemented in the ZOZOTOWN customer support.* The task allocation problem required in WFM is formulated as a mixed integer optimization problem, a kind of mathematical optimization problem, and the optimal task allocation is calculated.

* A term used to describe overall efforts to achieve more efficient and higher performance through the appropriate allocation of human resources.

Related Links

Development of technology that enables AI to automatically interpret ambiguous fashion expressions by focusing on the characteristics of individual clothing items

ZOZO has developed a machine learning model to realize a new functionality by improving the accuracy of the Fashion Intelligence System, which automatically learns and interprets images of fashion to obtain answers to ambiguous questions from users. The newly developed model can obtain answers to more detailed questions from users such as “How casual is the upper body (lower body, shoes, etc.) of this outfit?” and “What would the outfit look like if the upper body of this outfit were more business casual?” The results are expected to support users' every choice and action regarding fashion, including the clothes they wear and the items they purchase.

Related Links

Ultimate personalized-styling service, "niaulab by ZOZO" which explores "your style," and gives styling suggestions based on AI-generated coordination images

The ultimate personalized styling service "niaulab by ZOZO," launched in December 2022, uses ZOZO's original "your style AI." The AI is built on massive coordination data possessed by WEAR, ZOZO's fashion coordination app. Using the three patterns of coordination images proposed by "your style AI" as reference, professional stylists and hair and makeup artists interview customers in detail to find the best-suited style for the customers.

Related Links

Active Communication of Research Results

LY Corporation

Supporting the Company with advanced technology, LY Corporation Research is engaged in the research and development of next-generation internet-related technologies.

The research results are presented at conferences both in Japan and overseas, and the number of presentations in 2023 doubled from the previous year to 192 papers (101 overseas and 91 in Japan), partly due to the merger of former LINE and former Yahoo Japan Corporation.

Research areas include mainly natural language processing, speech processing, image processing, information search, machine learning, data science, semantic web, security and privacy, interaction, crowdsourcing, multimodal, and generative AI.

2023 Major Examples of Presentations

Related Links

ZOZO

ZOZO Research presented six papers in Japan and overseas in FY2020, 16 in FY2021, and 12 in FY2022. Of these, five were published in top conferences and top journals in the fields of image recognition and machine learning, including ECCV 2020 and NeurIPS 2021.

Related Links

Research Papers by LY Corporation Research That Led to Patent Applications

Numerous patents have been obtained from research papers published by LY Corporation.

 Research. Please refer to the following;

Paper before 2022

2022	ReflecTouch: Detecting Grasp Posture of Smartphone Using Corneal Reflection Images	INTERNATIONAL
2022	FIDO Microservices: Decoupling Public-key Based Authentication	DOMESTIC
2022	ReflecTouch: Detecting Grasp Posture of Smartphone Using Corneal Reflection Images	INTERNATIONAL
2022	Verifiable Certificate System to Prove Diverse Behaviors	DOMESTIC
2022	Robust Adaptive Noise Canceller Algorithm with SNR-Based Stepsize Control and Noise-Path Gain Compensation	INTERNATIONAL
2022	ReflecTouch: Detecting Grasp Posture of Smartphone Using Corneal Reflection Images	INTERNATIONAL
2022	Investigation of Private Information Unintentionally Included in Screen Sharing and Proposal of a System for Its Visual Concealment	DOMESTIC
2022	ShiftTouch: Extending Touchscreens with Passive Interfaces using Small Occluded Area for Discrete Touch Input	INTERNATIONAL
2021	Evaluation of Grasp Posture Detection Method using Corneal Reflection Images through a Crowdsourced Experiment	INTERNATIONAL
2021	Context-Free TextSpotter for Real-Time and Mobile End-to-End Text Detection and Recognition	INTERNATIONAL
2021	Detecting Grasp Posture of Smartphone Using Corneal Reflection Image	DOMESTIC
2021	ShiftTouch: Sheet-type Interface Extending Capacitive Touch Inputs with Minimal Screen Occlusion	INTERNATIONAL
2021	LightTouch Gadgets: Extending Touchscreen Interactions by Converting Light Emission into Touch Inputs	DOMESTIC
2021	ShiftTouch: Sheet-type Interface Extending Capacitive Touch Inputs with Minimal Screen Occlusion	DOMESTIC
2021	Proposal of a Query Generation Method for Refined Search on Multi-Intent Entity Queries	DOMESTIC
2021	Developing Personal Knowledge Questions Using Location Trajectory	DOMESTIC
2021	Investigation of the Effects of Curve Radius and Road Width on Driving in a Driving Simulator during Curve Driving	DOMESTIC
2020	ScraTouch: Extending Interaction Technique Using Fingernail on Unmodified Capacitive Touch Surfaces	INTERNATIONAL
2020	LightTouch: Passive Gadgets for Extending Interactions on Capacitive Touchscreens by Automating Touch Inputs	INTERNATIONAL
2020	A Simple Text-based Relevant Location Prediction Method using Knowledge Base	INTERNATIONAL
2020	Diverse and Non-redundant Answer Set Extraction on Community QA based on DPPs	INTERNATIONAL
2020	Context-Free TextSpotter for Real-Time and Mobile End-to-End Text Detection and Recognition (external site)	INTERNATIONAL
2020	PredicTaps: Latency Reduction Technique for Single-taps Based on Recognition for Single-tap or Double-tap	INTERNATIONAL
2020	ScraTouch: Extending Touch Interaction Technique Using Fingernail on Capacitive Touch Surfaces	INTERNATIONAL
2020	Extracting interpersonal relationships from Wi-Fi Connection Data (Japanese only)	DOMESTIC
2020	Association Analysis of Pet Foods with Diseases in dogs Using E-Commerce Data	DOMESTIC
2020	Estimation of Relevant Locations from News Articles using Knowledge Base	DOMESTIC
2020	News Headline Generation Based on Reinforcement Learning Considering Human Evaluation (external site)	DOMESTIC
2020	Click-Through Rate Prediction with Confidence for News Headlines Using Natural Gradient Boosting (external site) (Japanese only)	DOMESTIC
2020	Parasitic Location Logging: Estimating Users' Location from Context of Passersby	INTERNATIONAL
2020	SheetKey: Generating Touch Events by a Pattern Printed with Conductive Ink for User Authentication	INTERNATIONAL
2020	Fast Convergence Algorithm for Adaptive Noise Cancellers with SNR-Based Stepsize Control	INTERNATIONAL
2020	Fast Start-Up Algorithm for Adaptive Noise Cancellers with Novel SNR Estimation and Stepsize Control	INTERNATIONAL
2020	Yosegi: Columnar format for efficient nested data processing by schema conversion	INTERNATIONAL

Promoting Open Innovation

LY Corporation

In order to contribute to the development of the information technology society, LY Corporation's development results and intellectual properties created in ternally are made available to the public in the form of OSS, etc. Many of them are being utilized in various situations. The Company believes that encouraging the creation of a healthy market will also lead to the future growth of LY Corporation.

OSS developer certification program

This is a program that certifies selected employees as developers for OSS which is strategically adopted by LY Corporation. Time used for developing specific OSS is regarded as work hours and a maximum of JPY 1 million budget per year is granted for the development of specific OSS and related activities, providing an environment in which the certified employees can work as OSS committers.

Related Links

Similar product image search using Neighborhood Graph and Tree for Indexing (NGT)

Similar product image search which uses NGT, researched and developed by LY Corporation Research, is applied in the apps of Yahoo! JAPAN and Yahoo! JAPAN Shopping and in Yahoo! JAPAN Browser.

In this similar product image search, areas that are thought to be products are detected from product images on Yahoo! JAPAN Shopping and used as search targets. Vector data representing the characteristics of the products are extracted from these areas. To search the vector data at high speed, an index is generated for the search using NGT. Applying this index, similar products can be searched at high speed using product images in Yahoo! JAPAN Shopping or images captured by cameras as queries.

NGT is software that enables high-speed retrieval of features (vector data) extracted from various data, such as text, images, and product/user data, and is available as OSS.

Related Links

LINE services uses OSS in many of its services. The Company also encourages in-house developers to contribute to the open source community, and has released some of its in-house systems as OSS. As a member of the open source community, the Company also supports the Apache Software Foundation, a prominent non-profit organization, in addition to contributing to and releasing OSS to live in harmony with this huge ecosystem.

Released Resembla, a library for similar text string search, as OSS

Former LINE released Resembla, a library for similar text string search (approximate text string matching), to the public as OSS. Resembla was developed with a focus on improving quality and speed: improving quality by selecting and combining scales that match the properties of the text strings; and improving speed by narrowing down candidates and preprocessing the corpus.

Related Links

Released LINE Distil BERT, a light-weight Japanese language model, as OSS

Language models have been in the limelight in recent years, but from a practical standpoint, the scale of the model is an issue. Meanwhile, model contribution is important from the perspective of generating smooth dialogues, and "distillation" of large models has been proposed to resolve this tradeoff. The Company makes developments in these areas and releases some of its outcomes as OSS.

Related Links

ZOZO

The ZOZO Group makes active use of its vast amount of data and technologies. In particular, ZOZO Research is engaged in research and development to build a more convenient website and improve services for customers, by adopting AI-powered recommendation and search technologies into its products.

Released ZOZOTOWN fashion recommendation data and algorithm R&D platform, as OSS

ZOZO Research released the large-scale fashion recommendation data and R&D platform "Open Bandit Data & Pipeline" to the public as OSS. The project, named "Open source development and implementation infrastructure for social decision-making algorithms," was conducted by a joint research team consisting of Hanjuku-kaso, Inc. and ZOZO Research researchers, and received the Prime Minister's Award at Japan Open Innovation Prize, an award program held by the Cabinet Office.

Related Links

Released large-scale coordination data and experimental infrastructure for verifying fashion trends, as OSS

As the second version of an open data project, ZOZO Research released a large-scale data set, "Shift15M" composed of coordination data posted to the fashion app "IQON" and its experimental infrastructure. This dataset includes approximately 2.55 million fashion coordination data posted from 2010 to 2020, the period when IQON's service was provided, as well as approximately 15 million items that make up these fashion coordination, and related data such as the amount of features related to the items, data on item categories, and the number of "likes" on the coordination posts. Through this initiative, ZOZO aims to contribute to distribution shift research, which is the mathematical basis for changes in fashion trends.

Related Links

Collaborations with External Partners

The LY Corporation Group collaborates with external partners from diverse fields, including enterprises, universities, and research institutions, to make the lives of users more convenient and more enriching. It also aims to solve various social problems through these collaborative initiatives.

In collaboration with universities, former Yahoo Japan develops a method using cameras for estimating how smartphones are held

Former Yahoo Japan Corporation, Keio University, and Tokyo Institute of Technology developed a new method for estimating the posture of users grasping smartphones. The method uses machine learning to classify corneal reflection images in the facial photos taken by front-facing cameras of smartphones.

Related Links

Former LINE Corporation and Osaka University collaborate to establish “LINE Virtual Human Joint Research Course”

On March 1, 2023, former LINE Corporation and Osaka University entered into an agreement on the establishment of a joint research course, and established “LINE Virtual Human Joint Research Course,” facilitating the research and development of virtual characters. This research course is a project to promote research and development of virtual characters, which are attracting attention in the virtual space (the Metaverse, etc.), and will conduct activities aimed at a world in which any user can freely handle virtual characters.

Related Links

Former LINE Corporation and NAVER jointly develop a super-giant language model specialized in the Japanese Language — Enabling the generation of various Japanese-language AI, such as for dialogues and translations without new development work

In 2020, former LINE Corporation and NAVER jointly made an announcement on the development of a super-giant language model specialized in the Japanese language and the construction of the infrastructure necessary for its processing. Using AI, the super-giant language model (general-purpose language model generated from a huge amount of data) enables a more natural language processing and language expression.
As of 2022, the two companies have succeeded in developing a foundation model specialized in Japanese, in the size of 82 billion parameters. As training data, the model uses more than 500 billion tokens mainly in Japanese. This is the largest amount of parameters and training data for a Japanese-based language model in the world. The companies plan to continue improving performance by increasing the amount of training data and the number of parameters.

Related Links

ZOZO Technologies collaborates with the University of Tokyo and HOSOO Co., Ltd. in a joint research on expanding design potential for functional textiles

The joint research brings together the strengths of the three partners: knowledge in fashion, cutting-edge materials, and technologies from ZOZO Technologies; expertise in interaction technologies and experience design derived from the use of materials and technologies from the research team led by Associate Professor Yasuaki Kakehi, the Graduate School of Interfaculty Initiative in Information Studies, The University of Tokyo; and innovative textile design capabilities backed by long-established traditional Nishijin textile technology and artistry from HOSOO Co., Ltd.

Related Links

ZOZO NEXT, The University of Tokyo and HOSOO host “Ambient Weaving Collection ─Environment and Textiles” exhibiting the results from joint studies

Ambient Weaving is a concept created in the course of a joint research project among ZOZO Next, the University of Tokyo and Hosoo Co., Ltd. Since the launch of the project, the joint team has attempted to combine the structure and design of Nishijin textiles, a traditional craft, with cutting-edge materials, devices, and interaction technology to achieve a balance of various functions and expressions mediated by the textiles and the surrounding environmental information.

This exhibition proposed the possibility of new smart textiles that achieve both functionality and beauty, realized by combining the structure and design of Nishijin textiles with advanced materials and devices. Many visitors were able to enjoy the exhibition of innovative textile works while imagining the future of fashion and textiles, and the reactions from visitors obtained through the exhibition will be utilized for future joint research.

Related Links

ZOZO Inc. conducts joint research with Yale University to apply economics to understanding of customer incentives and behavioral psychology and optimization in customer communication

The objective of this joint research is to understand customer incentives and psychology behind their purchasing behavior and to optimize communications tailored for each user, utilizing both the expertise of Dr. Kosuke Uetake, Associate Professor of Marketing, Yale School of Management, who specializes in quantitative marketing and empirical industrial organization, and ZOZOTOWN's large-scale and diverse data. Algorithms developed based on the vast amount of past purchase data and other data accumulated in ZOZOTOWN have been implemented in ZOZOTOWN, and are still yielding significant results. Through this initiative, ZOZO aims to further enhance its services in ZOZOTOWN, and to maximize the long-term value for ZOZOTOWN users.

Related Links

List of Awards

FY2023

FY2022

Best Poster Honorable Mention	Annual Conference on Intelligent User Interfaces	LINE Corporation
2022 IPSJ Industrial Achievement Award	Information Processing Society of Japan (IPSJ)	Yahoo Japan Corporation
Young Incentive Award at NLP2023	The Association for Natural Language Processing	LINE Corporation
Special Committee Award at NLP2023	The Association for Natural Language Processing	Yahoo Japan Corporation
Kiyoshi Awaya Academic Incentive Award	Acoustical Society of Japan	LINE Corporation
DBSJ Kambayashi Young Researcher Award	The Database Society of Japan	LINE Corporation
Best Paper Runner-up at DEIM 2023	The Database Society of Japan	Yahoo Japan Corporation
Best Paper Award at 76 th IPSJ-SIGUBI	Information Processing Society of Japan (IPSJ)	Yahoo Japan Corporation
Award of the Journal of the Society for Art and Science Volume. 20	The Society for Art and Science	Yahoo Japan Corporation
42th Case Study Award	The Operations Research Society of Japan	ZOZO NEXT, Inc.
IPSJ Yamashita SIG Research Award	Information Processing Society of Japan (IPSJ)	Yahoo Japan Corporation
IEICE Distinguished Achievement and Contribution Award	The Institute of Electronics, Information and Communication Engineers (IEICE)	Yahoo Japan Corporation

FY2021

2021 IPSJ Industrial Achievement Award	Information Processing Society of Japan (IPSJ)	Yahoo Japan Corporation
Language Resource Award at NLP2022	The Association for Natural Language Processing	Yahoo Japan Corporation
IPSJ Yamashita SIG Research Award	Information Processing Society of Japan (IPSJ)	Yahoo Japan Corporation
Best Presentation Award at WISS2021	Japan Society for Software Science and Technology	Yahoo Japan Corporation
69th Electrical Science and Engineering Promotion Award	The Promotion Foundation for Electrical Science and Engineering	Yahoo Japan Corporation
Best Paper Award at CSS	Information Processing Society of Japan (IPSJ)	Yahoo Japan Corporation
Best Paper Award at DEXA 2021	DEXA Conference	Yahoo Japan Corporation

FY2020

Student Presentation Awards at DEIM 2021	The Database Society of Japan	Yahoo Japan Corporation
JSAI Annual Conference Award	The Japanese Society for Artificial Intelligence	Yahoo Japan Corporation

*Company names, titles and divisions are as of the time when the awards were received.