Method for speech synthesis
View Patent ↗A method, an arrangement and a computer program synthesize speech by grapheme/phoneme conversion. In this case, a search is made for subwords of a given word in a database which contains phonetic transcriptions of words. If at least one subword of the given word is found in the database, a phonetic transcription registered in the database is selected for the subword found. In addition to the subword found, the given word has at least one further constituent, which is not registered in the database. This further constituent is phonetically transcribed with the aid of an OOV treatment, and the phonetic transcription of the subword found and the phonetic transcription of the further constituent are combined.
1. A method for speech synthesis by a grapheme/phoneme conversion, comprising:
searching for subwords of a given word in a database which contains phonetic transcriptions of words, the given word having a subword registered in the database, and a further constituent which is not registered in the database;
selecting a phonetic transcription from the database for the subword;
phonetically transcribing the further constituent of the given word with the aid of an out-of-vocabulary (OOV) treatment, the out-of-vocabulary (OOV) treatment of the further constituent being performed based on phonetic context, as a function of the phonetic transcription of the subword; and
combining the phonetic transcription of the subword and the phonetic transcription of the further constituent, wherein
the out-of-vocabulary (OOV) treatment for phonetic transcription of the further constituent is performed by a neuron network,
the given word has at least first and second subwords registered in the database,
a search is made for both the first and second subwords in the database,
a phonetic transcription is selected from the database for both the first and second subwords,
the phonetic transcription of the first and second subwords and the phonetic transcription of the further constituent are combined,
the further constituent in the given word is arranged between the first subword and the second subword, and
the out-of-vocabulary (OOV) treatment for phonetic transcription of the further constituent is performed as a function of the phonetic transcription of the first subword and the phonetic transcription of the second subword.
2. The method for speech synthesis as claimed in claim 1 , wherein
the searching for subwords in the database is performed by searching for subwords which have a prescribed minimum length.
3. The method for speech synthesis as claimed in claim 1 , wherein
if a plurality of subwords are found for the same word part, the longest subword is selected therefrom.
4. The method for speech synthesis as claimed in claim 1 , wherein
the out-of-vocabulary (OOV) treatment for phonetic transcription of the further constituent is performed by a rule-based method.
5. The method for speech synthesis as claimed in claim 1 , wherein
the first and second subwords are found in a first database, and
the out-of-vocabulary (OOV) treatment for phonetic transcription of the further constituent is performed by a second database which contains the phonetic transcription of filling particles normally used in the case of composite words.
6. A method for speech synthesis by a grapheme/phoneme conversion, comprising:
searching for subwords of a given word in a database which contains phonetic transcriptions of words, the given word having a subword registered in the database, and a further constituent which is not registered in the database;
selecting a phonetic transcription from the database for the subword;
phonetically transcribing the further constituent of the given word with the aid of an out-of-vocabulary (OOV) treatment, the out-of-vocabulary (OOV) treatment of the further constituent being performed based on phonetic context, as a function of the phonetic transcription of the subword; and
combining the phonetic transcription of the subword and the phonetic transcription of the further constituent wherein
the searching for subwords in the database is performed by searching for subwords which have a prescribed minimum length,
if a plurality of subwords are found for the same word part, the longest subword is selected therefrom,
the out-of-vocabulary (OOV) treatment for phonetic transcription of the further constituent is performed by a neuron network,
the given word has at least first and second subwords registered in the database,
a search is made for both the first and second subwords in the database,
a phonetic transcription is selected from the database for both the first and second subwords,
the phonetic transcription of the first and second subwords and the phonetic transcription of the further constituent are combined,
the further constituent in the given word is arranged between the first subword and the second subword, and
the out-of-vocabulary (OOV) treatment for phonetic transcription of the further constituent is performed as a function of the phonetic transcription of the first subword and the phonetic transcription of the second subword.
7. The method for speech synthesis as claimed in claim 6 , wherein
the out-of-vocabulary (OOV) treatment for phonetic transcription of the further constituent is performed by a rule-based method.
8. The method for speech synthesis as claimed in claim 7 , wherein
the subwords are found in a first database, and
the out-of-vocabulary (OOV) treatment for phonetic transcription of the further constituent is performed by a second database which contains the phonetic transcription of filling particles normally used in the case of composite words.