Kakao Recognized for 'AI Research Achievements' at 12 Global Conferences

[Asia Economy Reporter Buaeri] Kakao announced on the 9th that its AI subsidiaries Kakao Enterprise and Kakao Brain have published 26 papers in 12 global academic conferences.

This month, Kakao and Kakao Enterprise co-authored a paper with Korea University and Hanshin University that was accepted at the international conference representing natural language processing and linguistics, the International Conference on Computational Linguistics (COLING). The paper studied the effectiveness of a new evaluation metric called 'RDASS' instead of the traditional 'ROUGE' used to measure the performance of text summarization models. The joint research team plans to conduct further studies incorporating factors that determine the truthfulness of machine-generated summaries.

At the annual Neural Information Processing Systems conference (NeurIPS) held in Vancouver, Canada, one paper from Kakao Brain and two papers from Kakao Enterprise were accepted. NeurIPS is a top-tier conference in AI, machine learning, and neuroscience, selecting only the top 30% of submissions from leading research groups in each field.

Kakao Brain, in collaboration with Professor Younghoon Kim from Sungshin Women's University, proposed a new data augmentation technique in their academic paper, demonstrating effective adaptation to diverse data transformations. Kakao Enterprise published papers presenting a speech synthesis model and research on the quality of TTS (text-to-speech) models. Based on the technology and experience gained from this research, they plan to conduct audio synthesis studies performing various tasks such as E2E (end-to-end) TTS, speech compression, sound quality enhancement, and generating speech in various styles depending on context.

Additionally, Kakao Brain and Scatter Lab's joint research team published a paper at the Association for Computational Linguistics and Natural Language Processing conference 'AACL-IJCNLP 2020'.

Last month, Kakao Enterprise's Context team published three papers at 'EMNLP 2020'. Kakao Brain also published one paper at the same conference, releasing datasets for Korean language understanding and evaluating the performance of Korean language understanding models.

In October, at 'Interspeech', the world's largest conference for speech processing technology papers, Kakao Brain published a paper on the construction of a new dataset converting Chinese characters to Pinyin. A joint paper by Kakao and Kakao Enterprise AI Lab's speech processing team was also accepted. They proposed 'JDI-T', an architecture that trains speech synthesis and phoneme-audio alignment models simultaneously, receiving recognition for excellence in speech synthesis technology research.

In the same month, Kakao Enterprise published a paper on end-to-end speech recognition technology in 'Signal Processing Letters', and Kakao, Kakao Enterprise, and Pohang University of Science and Technology jointly published a paper at the international image processing conference 'ICIP 2020'.

Furthermore, six papers were accepted at Europe's leading computer vision conference 'ECCV 2020' and the world-renowned conference for computer vision and pattern recognition 'CVPR 2020'. The topics included face recognition and fashion image search.

A Kakao representative stated, "We will continue to dedicate ourselves to AI research and share more advanced technological cases."

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