# US 2015/0178274 A1 · Speech translation apparatus and speech translation method

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2015/0178274 A1
- Document ID: US20150178274A1
- Status: Published application
- Kind: A1
- Application number: 14/581,944
- Filed: 2014-12-23
- Published: 2015-06-25
- Earliest claimed: 2013-12-25
- Assignees: Kabushiki Kaisha Toshiba
- Applicants: Kabushiki Kaisha Toshiba
- Inventors: Hiroyuki TANAKA
- Web page: https://patentgrep.com/patent/US20150178274A1

## CPC

- 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
- G10L 15/26: Speech to text systems (G10L15/08 takes precedence)
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING

## Abstract

According to an embodiment, a speech translation apparatus includes an allocator, a searcher and a sorter. The allocator allocates, to each of the phrases in the set of phrases, a weight dependent on a difference between a current dialog status and a dialog status associated with an original speech sound that corresponds to a text in which each of the phrases appears. The searcher searches the plurality of examples in the first language for an example including one or more phrases included in the set of phrases to obtain a hit example set. The sorter calculates a score of each of hit examples included in the hit example set based on the weight and the degree of similarity to sort the hit examples based on the score.

## Claims

### Claim 1

A speech translation apparatus comprising:
a speech recognizer that performs a speech recognition process on a current speech sound to generate a current speech recognition result;
a machine translator that performs machine translation from a first language to a second language on the current speech recognition result to generate a current machine translation result;
a first storage that stores a dialog history for each of one or more speeches constituting a current dialog;
an extractor that extracts a phrase from a text group to obtain a set of phrases, the text group including the current speech recognition result, a past speech recognition result and a past machine translation result both included in the dialog history;
an allocator that allocates, to each of the phrases in the set of phrases, a weight dependent on a difference between a current dialog status and a dialog status associated with an original speech sound that corresponds to a text in which each of the phrases appears;
a second storage that stores a plurality of examples in the first language and translations in the second language, each of the translations corresponding to each of the examples in the first language;
a searcher that searches the plurality of examples in the first language for an example including one or more phrases included in the set of phrases to obtain a hit example set;
a calculator that calculates, for each of hit examples included in the hit example set, a degree of similarity between the hit example and the current speech recognition result; and
a sorter that calculates a score of each of the hit examples included in the hit example set based on the weight and the degree of similarity to sort the hit examples based on the score.

### Claim 2 (depends on 1)

The apparatus according to claim 1, wherein the weight allocated to a given phrase is dependent on a difference between a speaker of the current speech sound and a speaker of an original speech sound that corresponds to a text in which the given phrase appears.

### Claim 3 (depends on 1)

The apparatus according to claim 1, wherein the weight allocated to a given phrase is dependent on a difference between a chronological order of an original speech sound in the current dialog, the original speech sound corresponding to a text in which the given phrase appears, and a chronological order of the current speech sound in the current dialog.

### Claim 4 (depends on 1)

The apparatus according to claim 1, wherein, the allocator, if a given phrase appears in a plurality of texts, calculates a weight for the given phrase by summing weights dependent on a difference between a dialog status associated with an original speech sound that corresponds to each of the plurality of texts and the current dialog status.

### Claim 5 (depends on 1)

The apparatus according to claim 1, wherein the text group includes at least one of a second or subsequent candidate text of the current speech recognition result, a second or subsequent candidate text of the past speech recognition result, and a second or subsequent candidate text of the past machine translation result.

### Claim 6 (depends on 5)

The apparatus according to claim 5, wherein the weight allocated to a given phrase is further dependent on a candidate order of a text in which the given phrase appears if the text is any one of the second or subsequent candidate texts of the current speech recognition result, the second or subsequent candidate texts of the past speech recognition result, and the second or subsequent candidate texts of the past machine translation result.

### Claim 7

A speech translation method comprising:
performing a speech recognition process on a current speech sound to generate a current speech recognition result;
performing machine translation from a first language to a second language on the current speech recognition result to generate a current machine translation result;
storing a dialog history for each of one or more speeches constituting a current dialog;
extracting a phrase from a text group to obtain a set of phrases, the text group including the current speech recognition result, a past speech recognition result and a past machine translation result both included in the dialog history;
allocating, to each of the phrases in the set of phrases, a weight dependent on a difference between a current dialog status and a dialog status associated with an original speech sound that corresponds to a text in which each of the phrases appears;
storing a plurality of examples in the first language and translations in the second language, each of the translations corresponding to each of the examples in the first language;
searching the plurality of examples in the first language for an example including one or more phrases included in the set of phrases to obtain a hit example set;
calculating, for each of hit examples included in the hit example set, a degree of similarity between the hit example and the current speech recognition result; and
calculating a score of each of the hit examples included in the hit example set based on the weight and the degree of similarity to sort the hit examples based on the score.

### Claim 8

A non-transitory computer readable storage medium storing instructions of a computer program which when executed by a computer results in performance of steps comprising:
performing a speech recognition process on a current speech sound to generate a current speech recognition result;
performing machine translation from a first language to a second language on the current speech recognition result to generate a current machine translation result;
storing a dialog history for each of one or more speeches constituting a current dialog;
extracting a phrase from a text group to obtain a set of phrases, the text group including the current speech recognition result, a past speech recognition result and a past machine translation result both included in the dialog history;
allocating, to each of the phrases in the set of phrases, a weight dependent on a difference between a current dialog status and a dialog status associated with an original speech sound that corresponds to a text in which each of the phrases appears;
storing a plurality of examples in the first language and translations in the second language, each of the translations corresponding to each of the examples in the first language;
searching the plurality of examples in the first language for an example including one or more phrases included in the set of phrases to obtain a hit example set;
calculating, for each of hit examples included in the hit example set, a degree of similarity between the hit example and the current speech recognition result; and
calculating a score of each of the hit examples included in the hit example set based on the weight and the degree of similarity to sort the hit examples based on the score.
