To produce high quality training data for deep learning, it requires a lot of worker's manual labor. DeepNatural AI improves quality and efficiency with "Human + AI Collaboration for AI" processes where people and AI collaboratively generate training data.
AI Assist for Voice
Case A - Dialogue Corpus Construction
For developing Voice User Interface, AI Contact Center, and Dialogue System, etc., it is essential to build high-quality voice conversation corpus. You need to segment the audio to chunks and transcribe each of them into text to construct a speech corpus. DeepNatural AI improves corpus building efficiency by performing tasks quickly and then verifying by human.
AI Assist for Text
Case B - NLU Corpus Construction
NLU is an essential part of conversational AI, and NER is one of the important NLP tasks to make computers understand the human languages. To construct a corpus for NLU, people need to read the text and tag entities one by one manually. DeepNatural's "Human & AI Collaboration Workflow" makes it efficient in terms of time saving and more accurate annotation results.
"As the data production batch progresses, we periodically train AI with the data created. This will increase AI performance as the operation progresses."