This notion comes to mind via speech recognition, but it applies to any platform where multiple inclusive functions correlate. I notice on my android phone that I am constantly correcting the voice to text output. The same mistakes occur over and over. My idea is for the speech software to notice the type of text edits I am retyping, and utilize that data to amend it's own errors for future voice scanning. If it could learn from my corrections, and thereby tweak itself to be more in tune with me, by anticipating my intent -next time it wouldn't keep making those same errors. Subtagged: Machine Learning, Natural Language Processing Technologies, Speech segmentation, Speech recognition
No comments:
Post a Comment