Users can use voice dictation capabilities of their devices to provide text input (or commands) and/or to edit text by using their voice. Speech recognition mechanisms are imperfect, leading to personalized names and phrases not being correctly transcribed, and requiring cumbersome manual corrections. This disclosure describes personalized automatic corrections to the transcribed text generated from voice dictation using a user-editable personal dictionary of correction pairs. With user permission, the personal dictionary can be generated automatically based on manual corrections to the transcribed text and shared across devices and applications associated with the user. Entries are added to the dictionary only after the user performs the corresponding corrections at least a threshold number of times. With user permission, user-specific and/or contextually relevant terms can be inferred from data sources such as the user's contacts, calendar, interaction data, locations, etc. The techniques reduce the effort to provide manual corrections and enhance the user experience of voice dictation.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Chao, Pu-sen, "Personalized Phrase Dictionary for Voice Dictation", Technical Disclosure Commons, (November 14, 2023)