Method and system for automatic speech recognition
First Claim
1. A method of recognizing speech, comprising:
- generating a decoding network for decoding speech input, the decoding network comprising a primary sub-network and one or more classification sub-networks, wherein;
the primary sub-network includes a plurality of classification nodes, each classification node corresponding to a respective classification sub-network of the one or more classification sub-networks, wherein each respective classification sub-network is distinct from the primary sub-network; and
each classification sub-network of the one or more classification sub-networks corresponds to a group of uncommon words;
receiving a speech input; and
decoding the speech input by;
instantiating a token corresponding to the speech input in the primary sub-network;
passing the token through the primary sub-network;
when the token reaches a respective classification node of the plurality of classification nodes, transferring the token to the corresponding classification sub-network;
passing the token through the corresponding classification sub-network;
when the token reaches an accept node of the classification sub-network, returning a result of the token passing through the classification sub-network to the primary sub-network, wherein the result includes one or more words in the group of uncommon words corresponding to the classification sub-network;
outputting a string corresponding to the speech input that includes the one or more words.
1 Assignment
0 Petitions
Accused Products
Abstract
A method of recognizing speech is provided that includes generating a decoding network that includes a primary sub-network and a classification sub-network. The primary sub-network includes a classification node corresponding to the classification sub-network. The classification sub-network corresponds to a group of uncommon words. A speech input is received and decoded by instantiating a token in the primary sub-network and passing the token through the primary network. When the token reaches the classification node, the method includes transferring the token to the classification sub-network and passing the token through the classification sub-network. When the token reaches an accept node of the classification sub-network, the method includes returning a result of the token passing through the classification sub-network to the primary sub-network. The result includes one or more words in the group of uncommon words. A string corresponding to the speech input is output that includes the one or more words.
25 Citations
15 Claims
-
1. A method of recognizing speech, comprising:
-
generating a decoding network for decoding speech input, the decoding network comprising a primary sub-network and one or more classification sub-networks, wherein; the primary sub-network includes a plurality of classification nodes, each classification node corresponding to a respective classification sub-network of the one or more classification sub-networks, wherein each respective classification sub-network is distinct from the primary sub-network; and each classification sub-network of the one or more classification sub-networks corresponds to a group of uncommon words; receiving a speech input; and decoding the speech input by; instantiating a token corresponding to the speech input in the primary sub-network; passing the token through the primary sub-network; when the token reaches a respective classification node of the plurality of classification nodes, transferring the token to the corresponding classification sub-network; passing the token through the corresponding classification sub-network; when the token reaches an accept node of the classification sub-network, returning a result of the token passing through the classification sub-network to the primary sub-network, wherein the result includes one or more words in the group of uncommon words corresponding to the classification sub-network; outputting a string corresponding to the speech input that includes the one or more words.
-
-
2. The method of claim 1, wherein the returned result is a respective result in a plurality of possible token-passing results through the classification sub-network, the returned result having a higher rollback probability than any other result in the plurality of possible token passing results through the classification sub-network.
-
3. The method of claim 1, wherein:
-
transferring the token to the corresponding classification sub-network further includes preserving one or more phones obtained prior to the token reaching the classification node as a starting index for the classification sub-network; and returning the result of the token passing through the classification sub-network to the primary sub-network includes preserving one or more phones obtained prior to the token reaching the accept node of the classification sub-network as a returning index for the primary decoding sub-network.
-
-
4. The method of claim 1, wherein the decoding network is a weighted finite state transducer.
-
5. The method of claim 1, wherein the one or more classification sub-networks include a medical terminology sub-network, a personal names sub-network, a place names sub-network, and a computer terminology sub-network.
-
6. An electronic device, comprising:
-
one or more processors; memory; and one or more programs, wherein the one or more programs are stored in memory and configured to be executed by the one or more processors, the one or more programs including an operating system and instructions that when executed by the one or more processors cause the electronic device to; generate a decoding network for decoding speech input, the decoding network comprising a primary sub-network and one or more classification sub-networks, wherein; the primary sub-network includes a plurality of classification nodes, each classification node corresponding to a respective classification sub-network of the one or more classification sub-networks, wherein each respective classification sub-network is distinct from the primary sub-network; and each classification sub-network of the one or more classification sub-networks corresponds to a group of uncommon words; receive a speech input; and decode the speech input by; instantiating a token corresponding to the speech input in the primary sub-network; passing the token through the primary sub-network; when the token reaches a respective classification node of the plurality of classification nodes, transferring the token to the corresponding classification sub-network; passing the token through the corresponding classification sub-network; when the token reaches an accept node of the classification sub-network, returning a result of the token passing through the classification sub-network to the primary sub-network, wherein the result includes one or more words in the group of uncommon words corresponding to the classification sub-networks; output a string corresponding to the speech input that includes the one or more words.
-
-
7. The electronic device of claim 6, wherein the returned result is a respective result in a plurality of possible token-passing results through the classification sub-network, the returned result having a higher rollback probability than any other result in the plurality of possible token passing results through the classification sub-network.
-
8. The electronic device of claim 6, wherein:
-
transferring the token to the corresponding classification sub-network further includes preserving one or more phones obtained prior to the token reaching the classification node as a starting index for the classification sub-network; and returning the result of the token passing through the classification sub-network to the primary sub-network includes preserving one or more phones obtained prior to the token reaching the accept node of the classification sub-network as a returning index for the primary decoding sub-network.
-
-
9. The electronic device of claim 6, wherein the decoding network is a weighted finite state transducer.
-
10. The electronic device of claim 6, wherein the one or more classification sub-networks include a medical terminology sub-network, a personal names sub-network, a place names sub-network, and a computer terminology sub-network.
-
11. A non-transitory computer readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by an electronic device with one or more processors and memory, cause the electronic device to:
-
generate a decoding network for decoding speech input, the decoding network comprising a primary sub-network and one or more classification sub-networks, wherein; the primary sub-network includes a plurality of classification nodes, each classification node corresponding to a respective classification sub-network of the one or more classification sub-networks, wherein each respective classification sub-network is distinct from the primary sub-network; and each classification sub-network of the one or more classification sub-networks corresponds to a group of uncommon words; receive a speech input; and decode the speech input by; instantiating a token corresponding to the speech input in the primary sub-network; passing the token through the primary sub-network; when the token reaches a respective classification node of the plurality of classification nodes, transferring the token to the corresponding classification sub-network; passing the token through the corresponding classification sub-network; when the token reaches an accept node of the classification sub-network, returning a result of the token passing through the classification sub-network to the primary sub-network, wherein the result includes one or more words in the group of uncommon words corresponding to the classification sub-network; output a string corresponding to the speech input that includes the one or more words.
-
-
12. The non-transitory computer readable storage medium of claim 11, wherein the returned result is a respective result in a plurality of possible token-passing results through the classification sub-network, the returned result having a higher rollback probability than any other result in the plurality of possible token passing results through the classification sub-network.
-
13. The non-transitory computer readable storage medium of claim 11, wherein:
-
transferring the token to the corresponding classification sub-network further includes preserving one or more phones obtained prior to the token reaching the classification node as a starting index for the classification sub-network; and returning the result of the token passing through the classification sub-network to the primary sub-network includes preserving one or more phones obtained prior to the token reaching the accept node of the classification sub-network as a returning index for the primary decoding sub-network.
-
-
14. The non-transitory computer readable storage medium of claim 11, wherein the decoding network is a weighted finite state transducer.
-
15. The non-transitory computer readable storage medium of claim 11, wherein the one or more classification sub-networks include a medical terminology sub-network, a personal names sub-network, a place names sub-network, and a computer terminology sub-network.
Specification