Speech translation apparatus, speech translation method, and non-transitory computer readable medium thereof

• US 9,471,568 B2
• Filed: 09/12/2014
• Issued: 10/18/2016
• Est. Priority Date: 09/19/2013
• Status: Active Grant

• Current Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Original Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Inventors: Kamatani, Satoshi, Sumita, Kazuo, Kawamura, Akinori
• Primary Examiner(s): Han, Qi
• Application Number:
  US14/484,483
• Publication Number:
  US 20150081270A1
• Time in Patent Office: 767 Days
• Field of Search: 704/2, 704/4, 704/5, 704/7, 704/8, 704/9, 704/10, 704/231, 704/251, 704/257, 704/277
• US Class Current: 1/1
• CPC Class Codes: G06F 40/58 Use of machine translation, e.g. for multi-lingual retrieval, for server-side translation for client devices or for real-time translation
• CPC Class Codes: G10L 15/005 Language recognition
• CPC Class Codes: G10L 15/065 Adaptation

Abstract

According to one embodiment, a speech of a first language is recognized using a speech recognition dictionary to recognize the first language and a second language, and a source sentence of the first language is generated. The source sentence is translated into a second language, and a translation sentence of the second language is generated. An unknown word included in the translation sentence is detected. The unknown word is not stored in the speech recognition dictionary. A first pronunciation candidate of the unknown word is estimated, from a representation of the unknown word. A second pronunciation candidate of the unknown word is estimated from a pronunciation of an original word included in the source sentence corresponding to the unknown word. The unknown word, the first pronunciation candidate and the second pronunciation candidate, are registered into the speech recognition dictionary correspondingly.

Claims

1. A speech translation apparatus comprising:

a speech recognition unit configured to convert a speech of a first language to a source sentence of the first language by recognizing the speech using a speech recognition dictionary,

the speech recognition dictionary storing words of the first language and pronunciation candidates corresponding to the words, and words of the second language and pronunciation candidates corresponding to the words;

a translation unit configured to convert the source sentence to a translation sentence of the second language by using a translation dictionary,

the translation dictionary storing words of the first language and translated words of the second language corresponding to the words;

an unknown word detection unit configured to detect an unknown word of the second language from the translation sentence by using the speech recognition dictionary, the unknown word being unregistered word not stored in the speech recognition dictionary;

a pronunciation estimation unit configured to estimate a first pronunciation candidate of the unknown word from a character string of the unknown word included in the translation sentence by using Text-To-Speech technique, and to estimate a second pronunciation candidate of the unknown word from a pronunciation of an original word included in the source sentence, the original word corresponding to the unknown word; and

a dictionary update unit configured to register the unknown word, the first pronunciation candidate and the second pronunciation candidate, into the speech recognition dictionary correspondingly,

wherein the speech recognition unit recognizes a next speech by using the speech recognition dictionary updated by the dictionary update unit.

2. The apparatus according to claim 1, wherein

the pronunciation estimation unit estimates a third pronunciation candidate of the unknown word from a speech sound included in the speech, the speech sound corresponding to the original word.

3. The apparatus according to claim 1, wherein

the dictionary update unit registers the unknown word so as to be preferentially selected than other words already registered into the speech recognition dictionary, the other words corresponding to the first pronunciation candidate or the second pronunciation candidate in the speech recognition dictionary.

4. A speech translation method comprising:

converting a speech of a first language to a source sentence of the first language by recognizing the speech using a speech recognition dictionary,

the speech recognition dictionary storing words of the first language and pronunciation candidates corresponding to the words, and words of a second language and pronunciation candidates corresponding to the words;

converting the source sentence to a translation sentence of the second language by using a translation dictionary,

the translation dictionary storing words of the first language and translated words of the second language corresponding to the words;

detecting an unknown word of the second language from the translation sentence by using the speech recognition dictionary, the unknown word being unregistered word not stored in the speech recognition dictionary;

estimating a first pronunciation candidate of the unknown word from a character string representation of the unknown word included in the translation sentence by using Text-To-Speech technique;

estimating a second pronunciation candidate of the unknown word from a pronunciation of an original word included in the source sentence, the original word corresponding to the unknown word;

registering the unknown word, the first pronunciation candidate and the second pronunciation candidate, into the speech recognition dictionary correspondingly; and

recognizing a next speech by using the speech recognition dictionary updated by the registering.

5. The method according to claim 4, wherein

the estimating a second pronunciation candidate comprises

estimating a third pronunciation candidate of the unknown word from a speech sound included in the speech, the speech sound corresponding to the original word.

6. The method according to claim 5, wherein

the registering comprises

registering the unknown word, the first pronunciation candidate, the second pronunciation candidate and the third pronunciation candidate, into the speech recognition dictionary correspondingly.

7. A non-transitory computer readable medium for causing a computer to perform operations for translating speech, the operations comprising:

converting a speech of a first language to a source sentence of the first language by recognizing the speech using a speech recognition dictionary,

the speech recognition dictionary storing words of the first language and pronunciation candidates corresponding to the words, and words of a second language and pronunciation candidates corresponding to the words;

converting the source sentence to a translation sentence of the second language by using a translation dictionary,

the translation dictionary storing words of the first language and translated words of the second language corresponding to the words;

detecting an unknown word of the second language from the translation sentence by using the speech recognition dictionary, the unknown word being unregistered word not stored in the speech recognition dictionary;

estimating a first pronunciation candidate of the unknown word from a character string of the unknown word included in the translation sentence by using Text-To-Speech technique;

estimating a second pronunciation candidate of the unknown word from a pronunciation of an original word included in the source sentence, the original word corresponding to the unknown word;

registering the unknown word, the first pronunciation candidate and the second pronunciation candidate, into the speech recognition dictionary correspondingly; and

recognizing a next speech by using the speech recognition dictionary updated by the registering.

8. The non-transitory computer readable medium according to claim 7, wherein

the estimating a second pronunciation candidate comprises

estimating a third pronunciation candidate of the unknown word from a speech sound included in the speech, the speech sound corresponding to the original word.

9. The non-transitory computer readable medium according to claim 8, wherein

the registering comprises

registering the unknown word, the first pronunciation candidate, the second pronunciation candidate and the third pronunciation candidate, into the speech recognition dictionary correspondingly.