Research and development in Japanese language analysis, deep neural networks (DNNs), emotional speech, and related areas
First written in FY2018. 13 of 23 entries say "Restated". Latest entry: FY2026.
No other company in this theme has enough history to compare yet.
- FY2026AdvancedAnnual report
「As part of its R&D into high-quality speech synthesis incorporating the latest neural-vocoder technology, the company improved the neural network that generates acoustic features, enabling it to produce more natural speech. These achievements are used in AITalk®6 products and services and the A.I.VOICE®2 speech-synthesis software for consumers.」
The improvements to the neural vocoder and acoustic-feature generation network were implemented in AITalk6 and A.I.VOICE2. - 2026-05-14AdvancedEarnings
「Through measures including the absorption-type merger of ATR-Trek Co., Ltd. effective 1 September 2025, we are optimizing research and development resources for AI voice-related technologies and establishing a flexible development structure.」
The absorption-type merger of ATR-Trek advanced the optimization of R&D resources and flexibilization of the development structure, including speech recognition and translation-related technologies. - FY2026-H1AdvancedAnnual report
「The company decided to implement the merger to accelerate decision-making, unify its business strategy and R&D organization, optimize R&D resources for AI voice-related technologies, and build a rapid and flexible R&D structure.」
With the absorption-type merger of ATR-Trek, the company unified its R&D organization, including speech-recognition and translation-related technologies, and optimized its resources. - FY2025ReframedAnnual report
「During the consolidated fiscal year, the company developed speech-synthesis technology, speech-recognition technology, sound-based AI inspection technology, and Visionary Cloud, among other technologies.」
The scope of R&D expanded from a focus on conventional Japanese-language analysis, DNN speech synthesis and emotional speech to multiple fields, including speech recognition, sound-based AI inspection, and Visionary Cloud for CRM. - FY2025-H1RestatedAnnual report
「The amount spent on R&D during the interim accounting period was ¥37.321 million. There were no material changes in the status of the company's R&D activities during the interim accounting period.」
The company continued its R&D activities and disclosed R&D expenses of ¥37.321 million. It stated that there were no material changes in the status of its activities. - FY2024AdvancedAnnual report
「During the fiscal year, the company conducted basic research to improve the quality of the Japanese speech-synthesis engine AITalk®. In product development, it focused on developing products and services using the next-generation DNN speech-synthesis engine AITalk®6, as well as the A.I.VOICE®2 speech-synthesis software for consumers.」
The company applied its R&D achievements in Japanese-language analysis and DNN speech-synthesis quality improvements to the development of AITalk6 and A.I.VOICE2 products. R&D expenses were ¥65.593 million. - FY2024-Q1RestatedAnnual report
「The amount spent on R&D during the cumulative first-quarter period was ¥18.562 million. There were no material changes in the status of the company's R&D activities during the cumulative first-quarter period.」
The company continued its R&D activities and stated that there were no material changes in the status of its R&D activities. - FY2023-Q2RestatedAnnual report
「The amount spent on R&D during the cumulative second-quarter period was ¥55.141 million. There were no material changes in the status of the company's R&D activities during the cumulative second-quarter period.」
The company continued its R&D activities and disclosed R&D expenses of ¥55.141 million. It stated that there were no material changes in the status of its activities. - FY2022AdvancedAnnual report
「In addition to continuously improving the language dictionary that forms the basis of Japanese-language analysis quality, the company conducted R&D on technology using DNNs (deep neural networks) to estimate pause positions and sentence-ending expression commands in synthesized speech, with the aim of achieving more natural linguistic expression.」
The company pursued multiple R&D activities, including Japanese-language analysis, DNN speech synthesis, neural vocoders, joint research with universities, and development and expansion of A.I.VOICE characters. R&D expenses were ¥144.040 million. - FY2022-Q2RestatedAnnual report
「The amount spent on R&D during the cumulative second-quarter period was ¥73.154 million. There were no material changes in the status of the company's R&D activities during the cumulative second-quarter period.」
The company continued its R&D activities and disclosed R&D expenses of ¥73.154 million. It stated that there were no material changes in the status of its activities. - FY2022-Q1RestatedAnnual report
「To expand sales in fields where demand is growing, the company is strengthening its sales and R&D organizations while developing the A.I.VOICE™ original brand for individual users and pursuing joint research with the Toda Laboratory at Nagoya University and the Tokuda, Nangaku and Hashimoto Laboratories at Nagoya Institute of Technology.」
The company continued its R&D activities and pursued joint research with Nagoya University and Nagoya Institute of Technology. R&D expenses were ¥37.162 million, and it stated that there were no material changes in the status of its activities. - FY2021AdvancedAnnual report
「The company developed a speech-synthesis method capable of controlling emotion in DNN parametric speech synthesis. Experiments confirmed that the method could maintain the previous level of quality while providing sufficient controllability.」
The company disclosed multiple R&D achievements, including emotion control, spoken-language analysis, dictionary creation using limited recordings, neural vocoders, and functional improvements for Cerence. The R&D team comprises 11 people, and R&D expenses were ¥130.359 million. - FY2021-Q3RestatedAnnual report
「The amount spent on R&D during the cumulative third-quarter period was ¥93.037 million. There were no material changes in the status of the company's R&D activities during the cumulative third-quarter period.」
The company stated that it continued its research and development activities and that there were no material changes in their status. - FY2021-Q2RestatedAnnual report
「The amount spent on R&D during the cumulative second-quarter period was ¥62.653 million. There were no material changes in the status of the company's R&D activities during the cumulative second-quarter period.」
The company continued its R&D activities and stated that there were no material changes in the status of its R&D activities. - FY2021-Q1RestatedAnnual report
「The amount spent on R&D during the cumulative first-quarter period was ¥30.597 million. There were no material changes in the status of the company's R&D activities during the cumulative first-quarter period.」
The company continued its R&D activities and disclosed R&D expenses of ¥30.597 million for the cumulative first-quarter period. There were no material changes in the status of its activities. - FY2020AdvancedAnnual report
「For DNN parametric speech synthesis, the company developed a training method that takes into account the statistical characteristics of speech and confirmed a substantial quality improvement compared with conventional DNN speech synthesis. It also completed basic validation of a real-time neural vocoder and carried out a technology transfer.」
The company disclosed multiple specific R&D achievements, including improvements in DNN speech-synthesis quality, recordings from 18 new speakers, completion of basic verification of a real-time neural vocoder, and development of AITalk®5. The R&D team comprises 11 people, and R&D expenses were ¥115.133 million. - FY2020-Q3RestatedAnnual report
「The amount spent on R&D activities during the cumulative third-quarter period was ¥82.911 million. There were no material changes in the status of the company's R&D activities during the cumulative third-quarter period.」
The company continued its R&D activities and disclosed R&D expenses of ¥82.911 million for the cumulative third-quarter period. There were no material changes in the status of the activities. - FY2020-Q1RestatedAnnual report
「The amount spent on R&D activities during the cumulative first-quarter period was ¥29.445 million. There were no material changes in the status of the company's R&D activities during the cumulative first-quarter period.」
The company continued its R&D activities and disclosed R&D expenses of ¥29.445 million for the quarter. There were no material changes in the status of the activities. - FY2019AdvancedAnnual report
「The company developed a new normalization method to address degradation in sound quality when information outside the training data is input in DNN-based speech synthesis, and confirmed that sound quality improved.」
The company disclosed specific results, including development, quality improvements, and completion of basic validation for Japanese-language analysis, DNN speech processing, a new prosody model, a neural vocoder, and DNN voice-quality conversion. The R&D team comprised 10 people, and R&D expenses were ¥101.256 million. - FY2019-Q3RestatedAnnual report
「The amount spent on R&D activities during the cumulative third-quarter period was ¥70.922 million. There were no material changes in the status of the company's R&D activities during the cumulative third-quarter period.」
The company continued its R&D activities and disclosed R&D expenses of ¥70.922 million. There were no material changes in the status of the activities. - FY2019-Q2RestatedAnnual report
「The amount spent on R&D activities during the cumulative second-quarter period was ¥44.047 million. There were no material changes in the status of the company's R&D activities during the cumulative second-quarter period.」
The company continued its R&D activities and recorded R&D expenses of ¥44.047 million. There were no material changes in the status of the activities. - FY2019-Q1RestatedAnnual report
「The amount spent on R&D activities during the cumulative first-quarter period was ¥21.854 million. There were no material changes in the status of the company's R&D activities during the cumulative first-quarter period.」
The company continued its R&D activities and disclosed R&D expenses of ¥21.854 million for the quarter. There were no material changes in the status of the activities. - FY2018First statedAnnual report
「To improve the quality of synthesized speech, the company created a prototype DNN speech-synthesis engine, which is attracting attention as a next-generation speech-synthesis technology. The company plans to further improve quality and increase processing speed toward practical implementation.」
The company also implemented an update to MeCab, improvements to pauses and accents, a framework for registering new words in the dictionary, a database for dialogue, and voice-quality conversion technology for emotional speech. The R&D team comprised 7 people, and R&D expenses were ¥64.360 million.