SHACER: a Speech and Handwriting Recognizer

摘要: the task domain of a multi-party, multimodal meeting focused on creation whiteboard schedule chart, we have designed and implemented general method aligning handwriting speech for capturing out-of-vocabulary terms, dynamically enrolling them in system's recognition modules, then using to improve subsequent tracking recognition. Our approach involves use an ensemble syllable phoneme recognizers whose output is integrated with redundantly delivered We refer our conceptual framework as Multimodal Out-Of- Vocabulary Recognition (MOOVR — pronounced mover). Within that this paper describes Speech HAndwriting reCognizER module (SHACER shaker), which observes human-to-human spoken handwritten interactions, analyzes off-line contributes improved recognitions record form project schedule. examine example show how technique corrects four five label errors including implicitly discovering semantics abbreviation.