I tried to create model for speech recognition, based on several guides.
will use the LJSpeech dataset to train model.
test loss in final version of model โ 102; word error rate โ 0.76
prediction: end quote afterd the fhr shot te quote braved that o mersyta lap end quote
real: end quote after the third shot he quote revved that motorcycle up end quote
prediction: colls fland wase taov hav worf fof hisol nonally occupided by a preturrtraing as mmolt te an compity
real: coles plan was to have a wharf of his own nominally occupied by a creature trading as maltby and co
prediction: whed beconsisten wuith the periid win the oswalds reling mun nele stret sincs the paprtment wais rentin on marchh thre nineteen sixtythtree
real: would be consistent with the period when the oswalds were living on neely street since the apartment was rented on march three nineteen sixtythree