The mapping of a raw phonetic transcription to an orthographic word sequence is carried out in three steps: First, a syllable segmentation of the transcription is bootstrapped, based on unsupervised subtractive learning. Then, the syllables are grouped to word entities guided by non-linguistic distributional properties. Finally, the phonetic word segmentations are mapped onto entries of a canonic pronunciation dictionary by means of a co-occurrence based aligner. For syllable segmentation accuracies between 89 and 96% are obtained, and for word segmentation accuracies between 92 and 98%. The transcription to word conversion performance amounts 77%.
Bibliographic reference. Reichel, Uwe D. (2013): "From segmentation bootstrapping to transcription-to-word conversion", In INTERSPEECH-2013, 1443-1447.