Method and apparatus for inputting text

• US 5,526,259 A
• Filed: 04/22/1994
• Issued: 06/11/1996
• Est. Priority Date: 01/30/1990
• Status: Expired due to Fees

• Current Assignee: Hitachi, Ltd.
• Original Assignee: Hitachi, Ltd.
• Inventors: Kaji, Hiroyuki
• Primary Examiner(s): Weinhardt, Robert A.
• Assistant Examiner(s): NOT, DEFINED
• Application Number:
  US08/231,901
• Time in Patent Office: 781 Days
• Field of Search: 381/36, 381/41, 381/43, 381/44, 364/419.01, 364/419.02, 364/419.03, 364/419.09, 364/419.16, 364/419.07, 434/156, 434/157, 434/167, 434/169, 395/2.86, 395/2.63
• US Class Current: 704/3
• CPC Class Codes: G06F 3/16 Sound input; Sound output speech processing G10L
• CPC Class Codes: G06F 40/268 Morphological analysis
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G06F 40/55 Rule-based translation
• 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 15/183 using context dependencies, e.g. language models
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence

Abstract

By inputting an input text in voice or phonogram and also the source information which is a source of the input text, extracting candidate words composing the input text by analyzing the source information, and using the candidate words, ambiguity in converting the voice or phonogram to characters is eliminated and highly accurate conversion is realized. For example, highly reliable translation is effected by extracting candidate words comprising the translation text using a bilingual dictionary containing words in the source language, translations in the target language, and extracting associated words in the target language which correspond to each other. When ambiguity occurs in converting the voice or phonogram input of the translation text to characters, by selecting the above candidate words preferentially, a voice typewriter for translation or a word processor for translation for converting the inputted translation text to characters a highly accurate translation is realized.

Claims

1. A translation text inputting apparatus for reading a source text in character string form of expression and a translation text in voice form of expression and converting the translation text to a character string, the apparatus comprising:

a source text analysis and bilingual dictionary retrieval module for predicting at least a translation text word composing the translation text, by identifying at least a source text word which composes the source text and retrieving a bilingual dictionary by use of the identified source text word as a key;

a phoneme recognition module for converting the translation text in voice form to a lattice of phonemes with a recognition probability exceeding a predetermined value; and

,

a phoneme to character conversion module for converting the lattice of phonemes outputted by said phoneme recognition module to the character string, wherein a path in the lattice is searched, said path matching the phonemic expression of one of the translation text words predicted by said source text analysis and bilingual dictionary retrieval module and said path being converted to the character string of the matched translation text word.

2. A translation text inputting apparatus according to claim 1, further comprising:

a source text display module for displaying at least a part of the source text on a display apparatus;

means for analyzing the part of the displayed source text contained in the source text analysis and bilingual dictionary retrieval module; and

,

means for converting the translation text from the voice form of expression to the character string using words predicted by analyzing the part contained in the displayed part of the phoneme source text.

3. A translation text inputting apparatus according to claim 2, further comprising:

a module for judging whether the conversion process for the translation text corresponding to the displayed part on said display apparatus is finished; and

,

means for displaying a subsequent part of the translation text on said display apparatus when said processing is finished.

4. A translation text inputting apparatus according to claim 1, further comprising:

means for displaying a first part of the source text contained in the source text display module in a format different from that of the other remaining parts;

means for analyzing only the first part of the source text contained in the source text analysis module, which is displayed in a format different from that of the other remaining parts; and

,

means for converting the translation text from the voice form of expression to the character string using candidate words which are words predicted by analyzing the first part contained in the source text displayed in a format different from that of the other remaining parts.

5. A translation text inputting apparatus according to claim 4 further comprising:

a module for judging whether the conversing process for the translation text corresponding to the displayed part on said display apparatus in a format different from that of the other remaining parts is finished; and

,

means for displaying a subsequent part of the translation text in a format different from that of the other remaining parts when said processing is finished.

6. A translation text inputting apparatus according to claim 1 further comprising:

a candidate word display module for displaying predicted words as candidate words which are predicted by the source text displayed on a display apparatus for the word forming the translation text.

7. A translation text inputting apparatus according to claim 1, further comprising:

a translation text display module for displaying the translation text by the character string which is an output of said phoneme to character conversion module on a display apparatus.

8. A translation text inputting apparatus for inputting phonograms of at least one of a kana and romaji form of expression, and converting the at least one kana and romaji phonogram to a kanji and kana character string in a second form of expression, said phonograms being in said translation text which includes at least a word translated from a source text, the apparatus comprising:

means for inputting translation text phonograms of at least one of kana and romaji;

a source text analysis and bilingual dictionary retrieval module for predicting at least a translation text word composing the translation text, by identifying at least a source text word which composes the source text and retrieving a bilingual dictionary by use of the identified source text word as a key;

a phoneme recognition module for converting the inputted translation text phonograms to a lattice of phonemes with recognition probability exceeding a predetermined value; and

,

a phoneme to character conversion module for converting the lattice of phonemes outputted by said phoneme recognition module to a character string, wherein a path in the lattice is searched, said path matching the phonemic expression of one of the translation text words predicted by said source text analysis and bilingual dictionary retrieval module and said path being converted to the second form of expression of the matched translation text word.

9. A translation text inputting apparatus according to claim 8, further comprising:

a source text display module for displaying at least a part of the source text on a display apparatus;

means for analyzing the part of the displayed source text contained in the source text analysis and bilingual dictionary retrieval module, which is displayed on the display apparatus; and

,

means for converting the translation text from the phonograms to the kanji and kana character string, using words predicted by analyzing the displayed part of the source text.

10. A translation text inputting apparatus according to claim 9 further comprising:

a module for judging whether the conversion process for the translation text corresponding to the displayed part on said display apparatus is finished; and

,

means for displaying a subsequent part of the translation text on said display apparatus when said processing is finished.

11. A translation text inputting apparatus according to claim 8, further comprising:

means for displaying a first part of the source text contained in the source text display module in a format different from that of the other remaining parts;

means for analyzing only the first part of the source text contained in the source text analysis module, which is displayed in a format different from that of the other remaining parts; and

,

means for converting the translation text from the phonograms to the kanji and kana character string, using candidate words which are predicted by analyzing the first part contained in the source text displayed in a format different from that of the other remaining parts.

12. A translation text inputting apparatus according to claim 11 further comprising:

a module for judging whether the conversion process for the translation text corresponding to the displayed part on said display apparatus in a format different from that of the other remaining parts is finished; and

,

means for displaying a subsequent part of the translation text in a format different from that of the other remaining parts when said processing is finished.

13. A translation text inputting apparatus according to claim 8 further comprising:

a candidate word display module for displaying predicted words as candidate words which are predicted by the source text displayed on a display apparatus, for the word forming the translation text.

14. A translation text inputting apparatus according to claim 8, further comprising:

a translation text display module for displaying the translation text by the kanji and kana character string outputted from the phoneme to character conversion module on a display apparatus.

15. A translation text inputting method in an apparatus for reading translation text in voice which includes at least a word translated from a source text, and for converting it to a character string, the method comprising:

a source text analysis step of predicting at least a translation text word composing the translation text, by identifying at least a source text word which composes the source text and retrieving a bilingual dictionary by use of the identified source text word as a key;

a phoneme recognition step of converting the translation text in voice to a lattice of phonemes with a recognition probability exceeding a predetermined value; and

,

a phoneme to character conversion step of converting the lattice of phonemes outputted by said phoneme recognition module to the character string, wherein a path in the lattice is searched, said path matching the phonemic expression of one of the translation text words predicted by said source text analysis and bilingual dictionary retrieval module and said path being converted to the character string of the matched translation text word.

16. A translation text inputting method in an apparatus for inputting phonograms of at least one of kana and romaji forms of expression and converting the inputted phonograms to a kanji and kana character string in a second form of expression, said phonograms including at least one word translated from a source text, the method comprising:

a source text analysis step of predicting at least a translation text word composing the translation text, by identifying at least a source text word which composes the source text and retrieving a bilingual dictionary by use of the identified source text word as a key;

a phoneme recognition step of converting the inputted translation text phonograms to a lattice of phonemes with recognition probability exceeding a predetermined value; and

,

a phoneme to character conversion step of converting the lattice of phonemes outputted by said phoneme recognition module to a character string, wherein a path in the lattice is searched, said path matching the phonemic expression of one of the translation text words predicted by said source text analysis and bilingual dictionary retrieval module and said path being converted to the character string of the matched translation text word.

17. A translation method for converting source language words into target language words in a voice typewriter apparatus including a central processing unit, a display unit, a voice input device and a storage unit storing a source text file, a bilingual dictionary and a translation text file, the method comprising steps of:

retrieving a source language word from the source text file of the voice typewriter apparatus for predicting at least a target language word which corresponds to said first source language word;

displaying the retrieved source language word on the display unit of the voice typewriter apparatus for reference by a human operator of the voice typewriter apparatus;

retrieving, from the bilingual dictionary of the voice typewriter apparatus, a set of target language candidate words corresponding in meaning to the retrieved source language word;

with the voice input device, generating a voice input signal of the human operator uttering a target language word corresponding in meaning to said displayed source language word;

converting the voice input signal to a plurality of target language phonetic strings; and

,

when at least one of the plurality of target language phonetic strings matches at least one target language word among said set of target language candidate words, writing said at least one target language word to the translation text file as a translation of the source language word.

18. The translation method according to claim 17 wherein the step of retrieving the set of target language candidate words from the bilingual dictionary includes retrieving a plurality of target language words corresponding to a range of meanings of said source language word as said set of target language candidate words.

19. The translation method according to claim 17 wherein the step of retrieving the source language word from the source text file includes:

morphologically analyzing a portion of the source text file to recognize at least one source language word comprising the source text file with reference to a dictionary in the source language

20. The translation method according to claim 17 wherein the step of converting the voice input signal includes:

segmenting and extracting a plurality of phonetic features from said voice input signal; and

,

associating each of the extracted phonetic features with a phonogram.

21. The translation method according to claim 20 wherein the step of converting includes:

constructing a lattice of phonograms which includes a plurality of phonogram strings of the associated phonogram; and

,

converting one of the phonogram strings into a character expression of the target language word among said set of target language candidate words.

22. The translation method according to claim 21 further comprising:

converting a second one of the phonogram strings into a second character expression of a second target language word among the entire vocabulary of the target language by referring to a target language dictionary of the voice typewriter apparatus.

23. The translation method according to claim 17, further comprising:

displaying the source language word on a first area of the display unit of the voice typewriter apparatus; and

,

displaying said at least one target language word on a second area of the display unit of the voice typewriter apparatus.

24. A translation method for converting source language words into target language words in a word processor apparatus including a central processing unit, a display unit, an input device and a storage unit storing a source text file, a bilingual dictionary and a translation text file, the method comprising the steps of:

retrieving a source language word from the source text file of the word processor apparatus;

displaying the retrieved source language word on the display unit of the word processor apparatus for reference by a human operator of the word processor apparatus;

retrieving, from the bilingual dictionary of the word processor apparatus, a set of target language candidate words corresponding in the meaning to the source language word;

inputting a kana translation text through the input device of the word processor apparatus from the human operator, the kana translation text corresponding in meaning to said displayed source language word;

converting the inputted kana translation text to a plurality of kana-kanji mixed translation text; and

,

when the at least one of the plurality of kana-kanji mixed translation texts matches at least one target language word among said set of target language candidate words, writing the kana and kanji translation text to the translation text file as a translation of the source language word.