Inventor · New York, NY, US

Hasim Sak

39Patents

8h-index

25Co-inventors

71Inventor score

Filing activity: Mar 14, 2013 → Mar 28, 2024

Most-cited inventions

Patent	Title	Area	Cited by	Status
US9292489B1	Sub-lexical language models with word level pronunciation lexicons	Physics	241	Active
US9786270B2	Generating acoustic models	Physics	45	Active
US10229672B1	Training acoustic models using connectionist temporal classification	Physics	30	Active
US9460088B1	Written-domain language modeling with decomposition	Physics	30	Active
US9728185B2	Recognizing speech using neural networks	Physics	21	Active
US9620108B2	Processing acoustic sequences using long short-term memory (LSTM) neural networks that include recurrent projection layers	Physics	20	Active
US10706840B2	Encoder-decoder models for sequence to sequence mapping	Physics	9	Active
US9818409B2	Context-dependent modeling of phonemes	Physics	8	Active
US9818410B2	Speech recognition with acoustic models	Physics	7	Active
US10127904B2	Learning pronunciations from acoustic sequences	Physics	6	Active
US10325602B2	Neural networks for speaker verification	Physics	6	Active
US9721562B2	Generating representations of acoustic sequences	Physics	6	Active
US9208779B2	Mixture of n-gram language models	Physics	5	Active
US10803855B1	Training acoustic models using connectionist temporal classification	Physics	4	Active
US10134393B2	Generating representations of acoustic sequences	Physics	3	Active
US10783900B2	Convolutional, long short-term memory, fully connected deep neural networks	Physics	2	Active
US10733979B2	Latency constraints for acoustic modeling	Physics	2	Active
US9594744B2	Speech transcription including written text	Physics	2	Active
US11741947B2	Transformer transducer: one model unifying streaming and non-streaming speech recognition	Physics	2	Active
US10535338B2	Generating representations of acoustic sequences	Physics	2	Active
US10431206B2	Multi-accent speech recognition	Physics	1	Active
US11769493B2	Training acoustic models using connectionist temporal classification	Physics	1	Active
US10923112B2	Generating representations of acoustic sequences	Physics	1	Active
US11341958B2	Training acoustic models using connectionist temporal classification	Physics	1	Active
US11961515B2	Contrastive Siamese network for semi-supervised speech recognition	Physics	1	Active

Source: USPTO / EPO open patent data. Inventor disambiguation is heuristic; counts are objective bibliographic measures.