SPEECH RECOGNITION FOR USER SPECIFIC LANGUAGE
A system, method and computer program is provided for generating customized text representations of audio commands. A first speech recognition module may be used for generating a first text representation of an audio command based on a general language grammar. A second speech recognition module may be used for generating a second text representation of the audio command, the second module including a custom language grammar that may include contacts for a particular user. Entity extraction is applied to the second text representation and the entities are checked against a file containing personal language. If the entities are found in the user-specific language, the two text representations may be fused into a combined text representation and named entity recognition may be performed again to extract further entities.
1 . A computer implemented method of converting an audio file into a text representation, the method comprising:
receiving, at a general automatic speech recognition module, the audio file as a general input, wherein the audio file comprises personal language and wherein the general automatic speech recognition module comprise a general language model;
generating a general text representation of the audio file using the general automatic speech recognition module;
receiving, at a personal automatic speech recognition module, the audio file as a personal input, wherein the personal automatic speech recognition module comprises a personal language model;
generating a personal text representation of the audio file using the personal automatic speech recognition module;
generating a first phoneme sequence representing the general text representation and a second phoneme sequence representing the personal text representation;
aligning the first phoneme sequence and the second phoneme sequence;
determining whether the second phoneme sequence is more likely to represent a portion of the audio file than the first phoneme sequence; and
merging the general text representation and the personal text representation to create a fused text representation when the second phoneme sequence is more likely to represent the portion of the audio file than the first phoneme sequence.
2 . The method of claim 1 , wherein aligning is performed using dynamic programming.
3 . The method of claim 1 , wherein the personal language model comprises contact information from a mobile device.
4 . The method of claim 1 , wherein determining comprises performing named entity recognition on the personal text representation to extract entities, and cross-referencing the entities with an electronic repository.
5 . The method of claim 4 , further comprising performing named entity recognition on the fused text representation to extract additional entities.
6 . The method of claim 5 , wherein the named entity recognition includes semantic role labeling.
7 . The method of claim 1 , wherein the personal language model is updated upon receiving the personal language via an application on a mobile device.
8 . The method of claim 7 , wherein the application updates the personal language model when the application synchronizes the personal language with an external service.
9 . The method of claim 8 , wherein the external service comprises a social media network.
10 . The method of claim 1 , wherein aligning comprises using an operation selected form a group comprising: dynamic programming, A* star search alogirthm, and a Viterbi algorithm.
11 . A computer system for converting an audio file into a text representation, the system comprising:
a communications device for receiving the audio file as input, wherein the audio file comprise personal language;
a memory for storing the audio file during processing and for storing programming instructions; and
a processor in communication with the memory and the communications device, the processor configured to generate a general text representation of the audio file using a general automatic speech recognition module, wherein the general automatic speech recognition module comprises a general language model, the processor further configured to generate a personal text representation of the audio file using a personal automatic speech recognition module, wherein the personal automatic speech recognition module comprises a personal language model, the processor further configured to generate a first phoneme sequence representing the general text representation and a second phoneme sequence representing the personal text representation, align the first phoneme sequence and the second phoneme sequence, determine whether the second phoneme sequence is more likely to represent a portion of the audio file than the first phoneme sequence, and merge the general text representation and the personal text representation to create a fused text representation when the second phoneme sequence is more likely to represent the portion of the audio file than the first phoneme sequence.
12 . The system of claim 11 , wherein the processor is further configured to align by using dynamic programming.
13 . The system of claim 11 , wherein the personal language model comprises contact information from a mobile device.
14 . The system of claim 11 , wherein the processor is further configured to perform named entity recognition on the personal text representation to extract entities, and cross-referencing the entities with an electronic repository.
15 . The system of claim 14 , wherein the processor is further configured to perform named entity recognition on the fused text representation to extract additional entities.
16 . The system of claim 15 , wherein the named entity recognition includes semantic role labeling.
17 . The system of claim 11 , wherein the personal language model is updated upon receiving personal language via an application on a mobile device.
18 . The system of claim 17 , wherein the application updates the personal language model when the application synchronizes the personal language with an external service.
19 . The system of claim 18 , wherein the external service comprises a social media network.
20 . The system of claim 11 , wherein the processor is further configured to use an operation selected form a group comprising: dynamic programming, A* star search alogirthm, and a Viterbi algorithm.