Developing conversational ai requires a large amount of speech data for ASR, TTS, and DM, etc. We provide high quality voice corpus to ensure the best results.
The trickiest part of NLP research was collecting and refining raw corpus. Now, leave it to us. As we are NLP people, we know which training corpus is good.
Vision matters in NLP. You will need video corpus for speaker recognition, multimodal discourse analysis, and speech & lip movements synchronization, etc.
Send us guidelines for data collection or annotation. Then we start a new corpus generation project.
The human + AI collaboration proceeds through deep learning models, heuristic, and human manipulation.
During the process, we perform multi-level automated and manual quality assurance over all annotations.
Train and validate your models by exporting results in a variety of formats, such as JSON, CSV, and XML.
"Tacotron is an end-to-end generative text-to-speech model that synthesizes speech directly from characters."
Human and AI cooperatively processed data while considering the purpose of corpus construction.
High-quality training corpus increases your AI research performance. Let the best team do the data processing.
On January 2, Park Won-soon, Mayor of Seoul, visited Yangjae R&CD Innovation Hub. He saw our TTS demo mimicking his own voice and asked us to continue working on artificial intelligence. He promised that he and the city would be a testbed for startups developing AI technology… Read More
The Yangjae R&CD Innovation Hub, an AI startup incubator operated by KAIST and ModuLabs, supported by Seoul Metropolitan Government, held the '2018 AICON' and 'Startup Demo Day & Talk'. We participated in the demo day and introduced our 'Human + AI collaboration for AI' business plan with demo videos… Read More