Abstract: A method includes receiving context biasing data that includes a set of unspoken textual utterances corresponding to a particular context. The method also includes obtaining a list of carrier phrases associated with the particular context. For each respective unspoken textual utterance, the method includes generating a corresponding training data pair that includes the respective unspoken textual utterance and a carrier phrase. For each respective training data pair, the method includes tokenizing the respective training data pair into a sequence of sub-word units, generating a first higher order textual feature representation for a corresponding sub-word unit, receiving the first higher order textual feature representation, and generating a first probability distribution over possible text units. The method also includes training a speech recognition model based on the first probability distribution over possible text units.

Abstract: An ASR model includes a first encoder configured to receive a sequence of acoustic frames and generate a first higher order feature representation for a corresponding acoustic frame in the sequence of acoustic frames. The ASR model also includes a second encoder configured to receive the first higher order feature representation generated by the first encoder at each of the plurality of output steps and generate a second higher order feature representation for a corresponding first higher order feature frame. The ASR model also includes a decoder configured to receive the second higher order feature representation generated by the second encoder at each of the plurality of output steps and generate a first probability distribution over possible speech recognition hypothesis. The ASR model also includes a language model configured to receive the first probability distribution over possible speech hypothesis and generate a rescored probability distribution.

Abstract: A joint segmenting and ASR model includes an encoder and decoder. The encoder configured to: receive a sequence of acoustic frames characterizing one or more utterances; and generate, at each output step, a higher order feature representation for a corresponding acoustic frame. The decoder configured to: receive the higher order feature representation and generate, at each output step: a probability distribution over possible speech recognition hypotheses, and an indication of whether the corresponding output step corresponds to an end of speech segment. The j oint segmenting and ASR model trained on a set of training samples, each training sample including: audio data characterizing a spoken utterance; and a corresponding transcription of the spoken utterance, the corresponding transcription having an end of speech segment ground truth token inserted into the corresponding transcription automatically based on a set of heuristic-based rules and exceptions applied to the training sample.

Abstract: An ASR model includes a first encoder configured to receive a sequence of acoustic frames and generate a first higher order feature representation for a corresponding acoustic frame in the sequence of acoustic frames. The ASR model also includes a second encoder configured to receive the first higher order feature representation generated by the first encoder at each of the plurality of output steps and generate a second higher order feature representation for a corresponding first higher order feature frame. The ASR model also includes a decoder configured to receive the second higher order feature representation generated by the second encoder at each of the plurality of output steps and generate a first probability distribution over possible speech recognition hypothesis. The ASR model also includes a language model configured to receive the first probability distribution over possible speech hypothesis and generate a rescored probability distribution.

Abstract: A hypothesis space of a search graph may be determined. The hypothesis space may include n hypothesis-space transcriptions of an utterance, each selected from a search graph that includes t>n transcriptions of the utterance. An evidence space of the search graph may also be determined. The evidence space may include m evidence-space transcriptions of the utterance that are randomly selected from the search graph, where t>m. For each particular hypothesis-space transcription in the hypothesis space, an expected word error rate may be calculated by comparing the particular hypothesis-space transcription to each of the evidence-space transcriptions. Based on the expected word error rates, a lowest expected word error rate may be obtained, and the particular hypothesis-space transcription that is associated with the lowest expected word error rate may be provided.

Abstract: Finite-state transducers and weighted finite-state automata may not be determinizable. The twins property can be used to characterize the determinizability of such devices. For a weighted finite-state automaton or transducer, that weighted finite-state automaton or transducer and its inverse are intersected or composed, respectively. The resulting device is checked to determine if it has the cycle-identity property. If not, the original weighted finite-state automaton or transducer is not determinizable. For a weighted or unweighted finite-state transducer, that device is checked to determine if it is functional. If not, that device is not determinizable. That device is then composed with its inverse. The composed device is checked to determine if every edge in the composed device having a cycle-accessible end state meets at least one of a number of conditions. If so, the original device has the twins property. If the original device has the twins property, then it is determinizable.

Abstract: Finite-state transducers and weighted finite-state automata may not be determinizable. The twins property can be used to characterize the determinizability of such devices. For a weighted finite-state automaton or transducer, that weighted finite-state automaton or transducer and its inverse are intersected or composed, respectively. The resulting device is checked to determine if it has the cycle-identity property. If not, the original weighted finite-state automaton or transducer is not determinizable. For a weighted or unweighted finite-state transducer, that device is checked to determine if it is functional. If not, that device is not determinizable. That device is then composed with its inverse. The composed device is checked to determine if every edge in the composed device having a cycle-accessible end state meets at least one of a number of conditions. If so, the original device has the twins property. If the original device has the twins property, then it is determinizable.

Abstract: Finite-state transducers and weighted finite-state automata may not be determinizable. The twins property can be used to characterize the determinizability of such devices. For a weighted finite-state automaton or transducer, that weighted finite-state automaton or transducer and its inverse are intersected or composed, respectively. The resulting device is checked to determine if it has the cycle-identity property. If not, the original weighted finite-state automaton or transducer is not determinizable. For a weighted or unweighted finite-state transducer, that device is checked to determine if it is functional. If not, that device is not determinizable. That device is then composed with its inverse. The composed device is checked to determine if every edge in the composed device having a cycle-accessible end state meets at least one of a number of conditions. If so, the original device has the twins property. If the original device has the twins property, then it is determinizable.