Abbreviation classification for speech synthesis and recognition
A method, computer system, and computer program product are provided for synthesizing and/or recognizing speech. Training data is provided to a machine learning model, wherein the training data comprises a plurality of labeled examples of acronyms and initialisms. The machine learning model is trained to classify strings into an acronym class or an initialism class. An input string is classified with the machine learning model into the acronym class or the initialism class. Based on the classifying, a pronunciation is generated for the input string.
1 . A computer-implemented method comprising:
training a machine learning model to classify strings into an acronym class or an initialism class;
classifying, with the machine learning model, an input string of an input text provided by a user into the acronym class or the initialism class, wherein the input text comprises the input string and one or more additional strings;
providing a confidence score indicating a probability of the input string being correctly classified into the acronym class or the initialism class;
converting, based on the classifying, a plurality of graphemes of the input string into a plurality of phonemes; and
combining one or more audio samples of the plurality of phonemes with a pronunciation for the one or more additional strings to generate a synthesized pronunciation for the input text.
2 . The computer-implemented method of claim 1 , further comprising:
updating a lexicon of an automatic speech recognition model using the synthesized pronunciation; and
employing the automatic speech recognition model to perform speech-to-text conversion processing.
3 . The computer-implemented method of claim 1 , wherein the synthesized pronunciation comprises an audio pronunciation generated using a text-to-speech conversion model.
4 . The computer-implemented method of claim 1 , further comprising:
adding the synthesized pronunciation to a dictionary that includes other pronunciations of words; and
using the dictionary to perform text-to-speech conversion processing.
5 . The computer-implemented method of claim 1 , further comprising:
collecting user feedback based on results of classifying the input string; and
updating the machine learning model based on the user feedback.
6 . The computer-implemented method of claim 1 , wherein converting, based on the classifying, the plurality of graphemes of the input string into the plurality of phonemes includes using a natural language processing model to perform grapheme-to-phoneme conversion for the input string that is classified into the acronym class.
7 . The computer-implemented method of claim 1 , wherein converting, based on the classifying, the plurality of graphemes of the input string into the plurality of phonemes includes converting a grapheme of each letter of the input string that is classified into the initialism class into one or more phonemes.
8 . The computer-implemented method of claim 1 , wherein the machine learning model is a long short-term memory model or a transformer model.
9 . The computer-implemented method of claim 1 , wherein training the machine learning model to classify the strings into the acronym class or the initialism class includes training the machine learning model based on a plurality of labeled examples.
10 . A computer system comprising:
one or more computer processors;
one or more computer readable storage media; and
program instructions stored on the one or more computer readable storage media for execution by at least one of the one or more computer processors, the program instructions comprising instructions to:
train a machine learning model to classify strings into an acronym class or an initialism class;
classify, with the machine learning model, an input string of an input text provided by a user into the acronym class or the initialism class, wherein the input text comprises the input string and one or more additional strings;
provide a confidence score indicating a probability of the input string being correctly classified into the acronym class or the initialism class;
convert, based on the classifying, a plurality of graphemes of the input string into a plurality of phonemes; and
combine one or more audio samples of the plurality of phonemes with a pronunciation for the one or more additional strings to generate a synthesized pronunciation for the input text.
11 . The computer system of claim 10 , wherein the program instructions further comprise instructions to:
update a lexicon of an automatic speech recognition model using the synthesized pronunciation; and
employ the automatic speech recognition model to perform speech-to-text conversion processing.
12 . The computer system of claim 10 , wherein the synthesized pronunciation comprises an audio pronunciation generated using a text-to-speech conversion model.
13 . The computer system of claim 10 , wherein the program instructions further comprise instructions to:
add the synthesized pronunciation to a dictionary that includes other pronunciations of words; and
use the dictionary to perform text-to-speech conversion processing.
14 . The computer system of claim 10 , wherein the program instructions further comprise instructions to:
collect user feedback based on results of classifying the input string; and
update the machine learning model based on the user feedback.
15 . The computer system of claim 10 , wherein converting, based on the classifying, the plurality of graphemes of the input string into the plurality of phonemes includes using a natural language processing model to perform grapheme-to-phoneme conversion for the input string that is classified into the acronym class.
16 . The computer system of claim 10 , wherein converting, based on the classifying, the plurality of graphemes of the input string into the plurality of phonemes includes converting a grapheme of each letter of the input string that is classified into the initialism class into one or more phonemes.
17 . A non-transitory computer program product comprising one or more computer readable storage media collectively having program instructions embodied therewith, the program instructions executable by a computer to cause the computer to perform operations including:
training a machine learning model to classify strings into an acronym class or an initialism class;
classifying, with the machine learning model, an input string of an input text provided by a user into the acronym class or the initialism class, wherein the input text comprises the input string and one or more additional strings;
providing a confidence score indicating a probability of the input string being correctly classified into the acronym class or the initialism class;
converting, based on the classifying, a plurality of graphemes of the input string into a plurality of phonemes; and
combining one or more audio samples of the plurality of phonemes with a pronunciation for the one or more additional strings to generate a synthesized pronunciation for the input text.
18 . The non-transitory computer program product of claim 17 , wherein the program instructions further cause the computer to:
update a lexicon of an automatic speech recognition model using the synthesized pronunciation; and
employ the automatic speech recognition model to perform speech-to-text conversion processing.
19 . The non-transitory computer program product of claim 17 , wherein converting, based on the classifying, the plurality of graphemes of the input string into the plurality of phonemes includes using a natural language processing model to perform grapheme-to-phoneme conversion for the input string that is classified into the acronym class.
20 . The non-transitory computer program product of claim 17 , wherein converting, based on the classifying, the plurality of graphemes of the input string into the plurality of phonemes includes converting a grapheme of each letter of the input string that is classified into the initialism class into one or more phonemes.