JP2009146043A - Unit and method for voice translation, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a voice translation unit translating a filler also. <P>SOLUTION: The voice translation unit includes: a voice information reception section 11 for receiving original language voice information; a voice recognition section 13 for voice recognizing the original language voice information; a machine translation section 15 for machine translating voice recognition result information; a voice generation section 17 for generating target language voice information corresponding to translation result information; a filler time position identification section 20 for identifying a filler position in the original language voice information; a filler information extraction section 21 for extracting original language filler information including a paralanguage of the filler; a filler text position identification section 22 for identifying the filler position in the voice recognition result information; a filler insertion position identification section 23 for identifying a filler insertion position, that is, a filler position in the target language voice information; a filler information insertion section 25 for inserting the target language filler information, corresponding to the original language filler information, into the filler insertion position; and a voice information output section 19 for outputting the target language voice information including the target language filler information. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a speech translation apparatus that inserts information related to fillers into translated target language speech information.

In a conventional translation device, a technique for reflecting information other than text itself in a source language in a target language has been developed. For example, a machine translation device has been developed that reflects character modification information such as italics in the source language text in the target language text (see, for example, Patent Document 1).
JP 2000-123012 A

  As described above, it has been required to transmit information other than the text itself by reflecting information other than the text itself in the source language in the target language. It is. That is, in speech translation, it has been demanded to realize translation of spoken words with a more realistic feeling by transmitting information other than the spoken language itself.

  The present invention has been made in view of the above circumstances, and an object thereof is to provide a speech translation apparatus and the like that can transmit information other than the language itself spoken in the original language.

  In order to achieve the above object, a speech translation apparatus according to the present invention receives a speech information receiving unit that receives source language speech information that is information obtained by collecting spoken source language speech with a microphone, and the speech information receiving unit receives the speech information. A speech recognition unit that recognizes the original language speech information and obtains speech recognition result information that is text information corresponding to the source language speech information; and machine translates the speech recognition result information obtained by the speech recognition unit. A machine translation unit that acquires translation result information that is text information of a target language corresponding to the speech recognition result information, and a target language that is speech information of a target language corresponding to the translation result information acquired by the machine translation unit In the source language speech information received by the speech generation unit that generates speech information and the speech information reception unit, the temporal of filler that is an utterance inserted between significant utterances Information including at least a para language that is non-linguistic information in speech information, with respect to a filler time position specifying unit that specifies a position and a filler in which the temporal position is specified by the filler time position specifying unit in the source language speech information. A filler information extracting unit for extracting certain source language filler information, and a filler text for specifying a position in the speech recognition result information corresponding to a temporal position of the filler specified by the filler time position specifying unit in the source language speech information A position specifying unit, a filler insertion position specifying unit for specifying a filler insertion position that is a temporal position in the target language speech information corresponding to a position in the speech recognition result information specified by the filler text position specifying unit, The filler insertion position in the target language speech information generated by the speech generation unit. The target language filler which is information having the same para language as the para language included in the source language filler information extracted by the filler information extracting unit at the filler insertion position specified by the specifying unit, and is voice information of the filler in the target language A filler information insertion unit that inserts information; and a voice information output unit that outputs target language voice information into which target language filler information is inserted by the filler information insertion unit.

  With such a configuration, the filler can also be translated. Since the target language filler information has the same para language as the filler in the source language, the target language filler information inserted in the target language speech information has the same atmosphere as when the speaker in the source language utters the filler. Will be output. Therefore, it is possible to generate a speech translation result including the emotion of the speaker with a more realistic feeling.

In the speech translation apparatus according to the present invention, the source language filler information is speech information of a filler whose temporal position is specified by the filler time position specifying unit, and the target language filler information is the source language filler information. It may be.
With such a configuration, the source language filler itself can be inserted into the target language speech information.

The speech translation apparatus according to the present invention may further include a filler information generation unit that generates target language filler information having the same para language as the para language included in the source language filler information, and the filler information insertion unit includes the filler The target language filler information generated by the information generation unit may be inserted into the target language voice information.
With such a configuration, a filler having the same para language as the source language filler can be generated, and the generated filler can be inserted into the target language speech information.

  In the speech translation apparatus according to the present invention, the filler information generation unit may generate target language filler information having the same vowel as the vowel of the filler whose temporal position is specified by the filler time position specifying unit. .

  With such a configuration, a filler having a vowel similar to that of the source language filler can be generated and inserted into the target language speech information. This filler may be a different voice than the target language filler, but because the vowels and para-language are in common with the source language filler, the emotions of the source language speaker are sufficiently heard by the filler. It can be considered to be transmitted.

  Further, in the speech translation apparatus according to the present invention, the filler information generation unit generates target language filler information which is speech information of a target language corresponding to a result of speech recognition of the filler of the source language speech information and machine translation. May be.

  With such a configuration, the filler inserted into the target language speech information can be a general filler in the target language, and the inserted filler can be made more natural.

In the speech translation apparatus according to the present invention, the para language may be at least one information selected from frequency, volume, change in frequency, and change in volume.
Further, in the speech translation apparatus according to the present invention, the para-language includes the time length of the filler of the source language speech information, the time length of the speech blank on the start point side of the filler, and the end point side of the filler It may further include the time length of the voice blanking.

  According to the speech translation apparatus and the like according to the present invention, in speech translation, information including the filler para-language can be added to speech information in the target language, and speech translation with a more realistic feeling can be realized. Become.

  Hereinafter, a speech translation apparatus according to the present invention will be described using embodiments. In the following embodiments, components and steps denoted by the same reference numerals are the same or equivalent, and repetitive description may be omitted.

(Embodiment 1)
A speech translation apparatus according to Embodiment 1 of the present invention will be described with reference to the drawings. The speech translation apparatus according to the present embodiment performs speech translation using a para language related to a filler of the source language.

  FIG. 1 is a block diagram showing a configuration of a speech translation apparatus 1 according to this embodiment. The speech translation apparatus 1 according to the present embodiment includes a speech information reception unit 11, a source language speech information storage unit 12, a speech recognition unit 13, a speech recognition result information storage unit 14, a machine translation unit 15, and a translation result. Information storage unit 16, speech generation unit 17, target language speech information storage unit 18, speech information output unit 19, filler time position specification unit 20, filler information extraction unit 21, filler text position specification unit 22 The filler insertion position specifying unit 23, the filler information generating unit 24, and the filler information inserting unit 25 are provided.

  The voice information receiving unit 11 receives source language voice information which is information obtained by collecting the spoken source language voice with a microphone. For example, the voice information receiving unit 11 may receive the source language voice information directly from a microphone, or may receive the source language voice information collected by the microphone once accumulated. The source language audio information is so-called audio signal information.

  For example, the voice information receiving unit 11 may receive source language voice information input from an input device (for example, a microphone), and receives source language voice information transmitted via a wired or wireless communication line. Alternatively, source language audio information read from a predetermined recording medium (for example, an optical disk, a magnetic disk, or a semiconductor memory) may be received. Note that the voice information receiving unit 11 may or may not include a device (for example, a modem or a network card) for receiving. The audio information receiving unit 11 may be realized by hardware, or may be realized by software such as a driver that drives a predetermined device.

  The source language audio information storage unit 12 stores the source language audio information received by the audio information reception unit 11 in a predetermined recording medium. The recording medium is, for example, a semiconductor memory, an optical disk, a magnetic disk, or the like, and may be included in the source language audio information storage unit 12 or may exist outside the source language audio information storage unit 12. . Further, this recording medium may or may not temporarily store source language audio information.

  The speech recognition unit 13 recognizes the source language speech information received by the speech information reception unit 11 and acquires speech recognition result information that is text information corresponding to the source language speech information. The method of speech recognition is already known and will not be described in detail. The voice recognition unit 13 may perform voice recognition by using, for example, an acoustic model, dictionary information, a language model, or the like.

  The voice recognition result information storage unit 14 stores the voice recognition result information acquired by the voice recognition unit 13 in a predetermined recording medium. The recording medium is, for example, a semiconductor memory, an optical disk, a magnetic disk, or the like, and may be included in the voice recognition result information storage unit 14 or may exist outside the voice recognition result information storage unit 14. . Further, this recording medium may or may not temporarily store voice recognition result information.

  The machine translation unit 15 performs machine translation on the speech recognition result information acquired by the speech recognition unit 13, and acquires translation result information that is text information of a target language corresponding to the speech recognition result information. The speech recognition result information is text information in the source language, and the text information having a parallel translation relationship with it is translation result information. The machine translation method is already known, and detailed description thereof is omitted. Further, the source language and the target language may be different languages, and the combination thereof is not limited. For example, the source language may be Japanese, the target language may be English, the source language may be English, and the target language may be French.

  The translation result information accumulation unit 16 accumulates translation result information, which is a result of machine translation by the machine translation unit 15, in a predetermined recording medium. The recording medium is, for example, a semiconductor memory, an optical disk, a magnetic disk, or the like, and may be included in the translation result information storage unit 16 or may exist outside the translation result information storage unit 16. Further, this recording medium may or may not temporarily store the translation result information.

  The voice generation unit 17 generates target language voice information that is voice information of a target language corresponding to the translation result information acquired by the machine translation unit 15. The target language audio information is so-called audio signal information in the target language. A method of generating speech corresponding to the text from the text is already known as a speech synthesis technique, and detailed description thereof is omitted.

  The target language voice information storage unit 18 stores the target language voice information generated by the voice generation unit 17 in a predetermined recording medium. The recording medium is, for example, a semiconductor memory, an optical disk, a magnetic disk, or the like, and may be included in the target language / speech information storage unit 18 or may exist outside the target language / speech information storage unit 18. . The recording medium may or may not temporarily store the target language audio information.

  The voice information output unit 19 outputs target language voice information in which target language filler information is inserted by a filler information insertion unit 25 described later. The processing for inserting the target language filler information and the target language filler information into the target language voice information will be described later.

  Here, this output may be, for example, an audio output by a speaker, transmission via a communication line to a predetermined device, accumulation in a recording medium, or delivery to another component. The audio information output unit 19 may or may not include a device that performs output (for example, a speaker or a communication device). Moreover, the audio | voice information output part 19 may be implement | achieved by hardware, or may be implement | achieved by software, such as a driver which drives those devices.

  The filler time position specifying unit 20 specifies the temporal position of the filler in the source language audio information received by the audio information receiving unit 11. A filler is an utterance inserted between significant utterances. For example, “Utto”, “That ぉ”, “That”, etc.

  Here, a method in which the filler time position specifying unit 20 specifies the time position of the filler will be described.

[How to use voice recognition results (failed to recognize voice)]
Since fillers are not significant utterances, generally, speech recognition processing that does not place importance on recognition accuracy, such as light recognition processing with a small RTF (real time factor), for example, speech recognition is performed for the filler region. There may not be. Therefore, if the speech recognition unit 13 performs such speech recognition, the filler time position specifying unit 20 determines the temporal range of the source language speech information that was not speech-recognized by the speech recognition unit 13, You may specify as a time position of a filler.

[Method of using speech recognition results (results of speech recognition)]
Unlike the above-described case, speech recognition processing with an emphasis on recognition accuracy, for example, speech recognition processing with a large RTF is performed, and a filler speech signal such as a filler dictionary (“Etto” or the like) In the case of performing speech recognition processing using a dictionary for converting to “tto” etc., the filler speech signal included in the source language speech information is also converted to text. Therefore, if the voice recognition unit 13 performs such voice recognition, the filler time position specifying unit 20 preliminarily stores text information corresponding to fillers held in a recording medium (not shown), for example, “e- When the text corresponding to the filler is included in the speech recognition result information using “Tto”, “That”, “That”, etc., it corresponds to the filler text included in the speech recognition result information. By specifying the temporal position in the source language audio information, the temporal position of the filler can be specified.

[Method using waveform pattern matching]
The filler time position specifying unit 20 uses a waveform of an audio signal corresponding to the filler, which is held in advance on a recording medium (not shown), performs pattern matching of the waveform in the source language audio information, and an audio signal corresponding to the filler If there is a temporal region similar to the waveform, the temporal region may be specified as the temporal position of the filler. Waveform pattern matching is already known and will not be described in detail. In this waveform pattern matching, for example, not exact matching, but, for example, waveform envelope matching may be taken as long as matching within a necessary and sufficient range can be achieved.

  In addition, although the three methods for specifying the temporal position of the filler have been described here, it goes without saying that the temporal position of the filler may be specified using other methods.

  Further, specifying the temporal position of the filler may be, for example, accumulating information for specifying the temporal position of the filler in a recording medium (not shown), or determining the temporal position of the filler. The identifiable information may be added to the source language audio information. In the former case, for example, information indicating the start point of the filler and information indicating the end point of the filler may be stored in a recording medium (not shown). The start point and end point of the filler may be indicated by, for example, a time code, may be indicated by the beginning or end point of the source language audio information, or the data capacity (for example, the number of bytes) from a specific position, or It may be indicated by the waveform itself of the source language audio information. In the latter case, in the source language speech information, information indicating the filler start point (for example, a flag) may be added to the filler start point position. Information indicating the end point of the filler may be added to the position of the end point, or information indicating the filler may be added continuously from the start point of the filler to the end point. Needless to say, the filler identification method is not limited to these as long as the temporal position of the filler can be identified.

  The filler information extraction unit 21 extracts source language filler information related to the filler for which the filler time position specifying unit 20 has specified the temporal position in the source language voice information. The source language filler information relating to the filler for which the filler time position specifying unit 20 has specified the temporal position is information including at least a para language relating to the filler for which the filler time position specifying unit 20 has specified the temporal position. Para language is information of prosodic features in speech information. Voice strength, strength change, height (frequency), height change, utterance speed, inflection, tremor, voice color, as well as information during utterance, as well as silence (between) and length of utterance Included in para language. For example, it is difficult to detect the writer's feelings from a single sentence that says “Thank you”, but in the case of voice information that reads the phrase “Thank you”, the speaker ’s feelings (for example, affection, anger, sympathy) Etc.) is usually easier. On the other hand, it is also possible to read aloud with those feelings. This is possible because paralinguistic information exists. The para language related to the filler may be, for example, the filler frequency of the source language voice information, the filler volume, the filler frequency change, or the filler volume change. It may be the length of time of the filler of the source language speech information, or the length of time of the blank space of the speech at the start point of the filler. It may be the length of time of the blank, it may be the speed of the filler's utterance, it may be a combination of any two or more of those information, and the para language related to the filler should be indicated appropriately Any other information may be used as long as the information can be received. The speed of filler utterance can be measured, for example, by the number of mora per unit time or the number of syllables per unit time.

  The filler information extraction unit 21 may extract, as source language filler information, for example, the filler speech information itself, for which the filler time position specifying unit 20 has specified the temporal position, from the source language speech information. Moreover, the filler time position specific | specification part 20 may extract only the para language of the filler which the filler time position specific | specification part 20 specified the time position as source language filler information from source language audio | voice information, for example. As described above, the para language may be at least one information selected from, for example, frequency, volume, change in frequency, change in volume, the length of time of the filler of the source language audio information, The information may further include the time length of the voice blank on the start point side of the filler and the time length of the voice blank on the end point side of the filler. The source language filler information extracted by the filler information extraction unit 21 may be temporarily stored in a recording medium (not shown).

  The frequency may be the frequency of the audio signal of the filler itself in the source language audio information, or the frequency of the audio signal other than the filler and the frequency of the audio signal of the filler in the source language audio information. It may be a difference. The frequency may be averaged over the section of the filler audio signal. The change in frequency can be extracted by, for example, detecting an average frequency for every minute time (for example, 10 ms or 30 ms) in the filler. Further, this frequency may be a so-called basic frequency F0.

  The volume may be the volume (voltage) of the filler audio signal in the source language audio information, or the volume of the audio signal other than the filler in the source language audio information and the volume of the filler audio signal. It may be a difference from the volume. Further, the volume may be averaged over the section of the filler audio signal. The change in frequency can be extracted, for example, by detecting the average of the sound volume for every minute time (for example, 10 ms, 30 ms, etc.) in the filler.

  The filler text position specifying unit 22 specifies a position in the speech recognition result information corresponding to the temporal position of the filler specified by the filler time position specifying unit 20 in the source language voice information.

  For example, when the text in the speech recognition result information is associated with the temporal position in the source language speech information (for example, the time code of the source language speech information is added to the text of the speech recognition result information) In the case), the position of the text in the speech recognition result information corresponding to the temporal position in the source language speech information can be specified. Therefore, in such a case, the filler text position specifying unit 22 sets the text position corresponding to the temporal position of the filler specified by the filler time position specifying unit 20, the text in the speech recognition result information, and the source language. The information can be specified by using information that correlates the temporal position in the audio information.

  When there is no information that associates the text in the speech recognition result information with the temporal position in the source language speech information, for example, the filler text position specifying unit 22 specifies the filler time position in the source language speech information. The unit 20 acquires the audio signal in front of, behind, or both of the filler whose temporal position is specified (this audio signal does not include the filler), and acquires the result of the audio recognition of the audio signal. To do. The position of the text corresponding to the temporal position of the filler specified by the filler time position specifying unit 20 can be specified depending on where the result is located in the speech recognition result information. More specifically, when the result of voice recognition of the voice signal ahead of the filler specified by the filler time position specifying unit 20 is acquired, the filler text position specifying unit 22 determines the position of the result of the voice recognition. It may be specified in the voice recognition result information, and immediately after the specified position may be specified as the text position corresponding to the temporal position of the filler specified by the filler time position specifying unit 20.

  In addition, when the filler position is specified using the fact that voice recognition could not be performed, the filler text position specifying unit 22 sets the position where the voice recognition could not be performed in the voice recognition result information as the filler time position. The position of the text corresponding to the temporal position of the filler specified by the specifying unit 20 may be specified.

  In addition, when the position of the filler is specified using the voice recognition result, the text corresponding to the filler may exist in the voice recognition result information. Therefore, in that case, the filler text position specifying unit 22 is a text corresponding to the temporal position of the filler specified by the filler time position specifying unit 20 in the voice recognition result information. You may specify as a position of.

  Moreover, if the position of the text corresponding to the temporal position of the filler appropriately specified by the filler time position specifying unit 20 can be specified by using other methods, the filler text position specifying unit 22 The position of the text corresponding to the temporal position of the filler specified by the filler time position specifying unit 20 may be specified using the method.

  Further, to specify the position in the speech recognition result information corresponding to the temporal position of the filler specified by the filler time position specifying unit 20 in the source language voice information, for example, the information specifying the position is recorded on a recording medium (not shown). It may be stored, or information that can specify the position may be added to the speech recognition result information. In the former case, when there is no text corresponding to the filler in the speech recognition result information, the information for specifying the position is, for example, from the beginning or the rear end of the speech recognition result information, or a specific position, It may be indicated by the number of characters up to that particular target position, it may be indicated by the data capacity, it may be indicated by text existing in front of the position, or it exists behind the position. It may be shown in text. In the former case, when the text corresponding to the filler exists in the speech recognition result information, the information specifying the position may be information specifying the start point and the end point of the position, for example. Good. For example, the start point and the end point may be indicated by the number of characters from the beginning or the end of the speech recognition result information or a specific position to the specific target position, or may be indicated by a data capacity. It may be indicated by text existing in front of the position and text existing behind the position. In the latter case, in the voice recognition result information, information indicating the position (for example, a flag) may be added to a specific target position. . In addition, if the position in the speech recognition result information corresponding to the temporal position of the filler specified by the filler time position specifying unit 20 in the source language voice information can be appropriately specified, the specifying method is as follows. It goes without saying that it is not limited.

  The filler insertion position specifying unit 23 specifies a filler insertion position that is a temporal position in the target language speech information corresponding to the position in the speech recognition result information specified by the filler text position specifying unit 22. The filler insertion position is specified by, for example, specifying a position in the translation result information corresponding to the position specified in the speech recognition result information, and temporally in the target language voice information corresponding to the position of the specified translation result information. This is done by specifying the correct position.

  First, a method in which the filler insertion position specifying unit 23 specifies the position in the translation result information corresponding to the position specified in the speech recognition result information will be described.

  For example, the filler insertion position specifying unit 23 specifies a chunk following the position specified in the speech recognition result information. Here, the chunk is a certain unitary text, and may be, for example, a morpheme or a word, or may be a plurality of continuous morphemes or a plurality of continuous words. Then, the filler insertion position specifying unit 23 acquires a chunk in the target language corresponding to the specified source language chunk. This can be obtained, for example, by performing machine translation. The machine translation may be performed by the filler insertion position specifying unit 23 or may be performed by the machine translation unit 15. The filler insertion position specifying unit 23 specifies the position of the target language chunk in the translation result information, and the position immediately before the position of the target language chunk in the translation result information corresponding to the position specified in the speech recognition result information. You may specify that it is a position.

  For example, the filler insertion position specifying unit 23 specifies a chunk preceding the position specified in the speech recognition result information, and acquires a target language chunk corresponding to the specified source language chunk. Then, the filler insertion position specifying unit 23 specifies the position of the target language chunk in the translation result information, and the translation result corresponding to the position specified in the speech recognition result information immediately after the position of the target language chunk. You may specify that it is a position in information.

  Further, for example, when the text corresponding to the filler exists in the speech recognition result information, the filler insertion position specifying unit 23 specifies the filler of the target language having a translation relation with the text by the translation result information. The position in the translation result information corresponding to the position specified in the speech recognition result information may be specified. The filler insertion position specifying unit 23 associates a filler in the target language that is in a translation relationship with the filler in the source language, for example, a dictionary that associates the filler in the source language in a translation relationship with the filler in the target language (this dictionary is, for example, May be stored using a recording medium (not shown), or may be acquired from the machine translation unit 15.

  In addition, when the filler insertion position specifying unit 23 specifies the position in the translation result information corresponding to the position specified in the speech recognition result information, the information specifying the position in the translation result information is stored in a recording medium (not shown). Alternatively, information that can specify the position may be added to the translation result information. These methods are the same as the processing in the case of specifying the position in the speech recognition result information corresponding to the temporal position of the filler specified by the filler time position specifying unit 20 in the source language voice information, and detailed description thereof will be given. Omitted.

  Next, a description will be given of a method in which the filler insertion position specifying unit 23 specifies a temporal position (that is, filler insertion position) in the target language speech information corresponding to the position of the specified translation result information.

  For example, when the text in the translation result information is associated with the temporal position in the target language voice information (for example, when the time code of the target language voice information is added to the text of the translation result information) The time position in the target language speech information corresponding to the position of the text in the translation result information can be specified. Therefore, in such a case, the filler insertion position specifying unit 23 determines the position of the target language speech information corresponding to the specified position of the translation result information in terms of the text in the translation result information and the time in the target language speech information. It is possible to specify the information by using information that associates a specific position.

  In addition, when there is no information that associates the text in the translation result information with the temporal position in the target language speech information, for example, the filler insertion position specifying unit 23, in front of the position of the specified translation result information, The text of the back or both is acquired, and the audio | voice signal is produced | generated from the text. Then, the position in the target language speech information corresponding to the position of the identified translation result information can be identified depending on where in the target language speech information the speech signal is located. More specifically, when an audio signal is generated from text ahead of the position in the specified translation result information, the filler insertion position specifying unit 23 specifies the position of the generated audio signal in the target language audio information. Then, the position immediately after the specified position may be specified as a temporal position in the target language speech information corresponding to the position of the specified translation result information.

  When the filler is also recognized and machine-translated, and the text corresponding to the filler exists in the translation result information, the filler insertion position specifying unit 23 corresponds to the filler text in the target language speech information. The position of the audio signal may be specified as a temporal position in the target language audio information corresponding to the position of the specified translation result information.

  Further, specifying the temporal position in the target language speech information corresponding to the position of the specified translation result information may be, for example, accumulating information for specifying the position in a recording medium (not shown), Alternatively, information that can specify the position may be added to the target language audio information. These methods are the same as the processing in the case where the filler time position specifying unit 20 specifies the temporal position of the filler in the source language speech information, and detailed description thereof is omitted.

  Moreover, if the time position in the target language speech information corresponding to the position in the speech recognition result information appropriately specified by the filler text position specifying unit 22 can be specified by using other methods, The filler insertion position specifying unit 23 may specify a temporal position (that is, filler insertion position) in the target language speech information using the method.

  The filler information generation unit 24 generates target language filler information having the same para language as the para language included in the source language filler information extracted by the filler information extraction unit 21. This target language filler information may or may not be recognized as a filler in the target language. For example, when the target language is English, the former target language filler information may be an audio signal of “Well” or “Let me see”, and the latter target language filler information is “e-”. Alternatively, it may be a voice signal of “That ~”.

  For example, the filler information generation unit 24 may generate target language filler information having the same vowel as the vowel of the filler whose temporal position is specified by the filler time position specifying unit 20. Further, the filler information generation unit 24 may generate target language filler information, which is voice information of a target language corresponding to a result obtained by, for example, recognizing the filler of the source language voice information and performing machine translation. At the time of generation, as described above, the target language filler information is generated so as to have the same para language as the para language included in the source language filler information. Therefore, the target language filler information generated by the filler information generation unit 24 is common to the source language filler information and the para language. For example, the frequency, the volume, the change thereof, the filler time length, Between the posts and between the posts.

  Here, processing when the filler information generation unit 24 generates target language filler information having the same vowel as the vowel of the filler whose temporal position is specified by the filler time position specifying unit 20 will be described. First, the filler information generation unit 24 obtains a sequence of vowels and consonants corresponding to the filler by performing the same process as the voice recognition on the filler whose temporal position is specified by the filler time position specifying unit 20. To do. Then, only the acquired vowel is extracted, and an audio signal corresponding to the acquired vowel is synthesized, which has the same paralanguage as the paralanguage included in the source language filler information. At the time of the synthesis, the speech signal may be synthesized from the vowel of the source language (in this case, the speech signal is the source language), or the speech signal of the target language is generated from the vowel of the target language. Synthesis may be performed (in this case, the voice signal is in the target language). In the latter case, information that associates the vowels of the source language with the vowels of the target language is held in a recording medium (not shown), and the filler information generation unit 24 refers to the information, thereby You may convert a vowel into the vowel of a target language. In this way, the same vowel in the “target language filler information having the same vowel as the vowel of the filler whose temporal position is specified by the filler time position specifying unit 20” is exactly the same as the vowel of the source language filler. Or a vowel that has a translation relationship with the vowel.

  In addition, the filler information generation unit 24 uses, for example, voice signal correspondence information that is information that associates a voice signal in the source language of the filler with a voice signal in the target language of the filler corresponding thereto, and uses the filler time position specifying unit 20. Identifies the filler's source language speech signal included in the speech signal correspondence information, similar to the filler's speech signal whose time position is specified, and the filler's target language speech associated with the speech signal The target language filler information may be generated by acquiring the signal and converting the acquired speech signal so as to have the same para language as the para language included in the source language filler information. It is assumed that the audio signal correspondence information is held in a recording medium (not shown) accessible by the filler information generation unit 24. In the audio signal correspondence information, for example, the original language (Japanese) “Etto” audio signal and the target language (English) “Well” audio signal may be associated with each other.

  In addition, the filler information generation unit 24 holds, for example, an audio signal generated as a filler in advance on a recording medium (not shown), and the paralanguage of the audio signal is the same paralanguage as the paralanguage included in the source language filler information. The target language filler information converted so as to have For example, when the target language is English, the filler information generation unit 24 holds an audio signal corresponding to “Well” or “Let me see” in advance as a filler. Then, the filler information generation unit 24 may generate the target language filler information by matching only the para language of the voice signal with the source language filler information. In this case, the filler information generating unit 24 may hold a plurality of filler audio signals in the target language and select the filler audio signal according to the length of the source language filler. . For example, the filler information generation unit 24 may select “Well” when the source language filler is short, and may select “Let me see” when the source language filler is long.

  In addition, the filler information generation unit 24 may perform the processing by the filler information generation unit 24 itself when performing speech recognition, machine translation, and speech synthesis, or other components (for example, a speech recognition unit). 13, machine translation unit 15, voice generation unit 17, etc.) or another device, and may receive the processing result.

In addition, when the source language filler information is the filler speech information itself, the filler information generation unit 24 may perform processing for extracting paralingual information (for example, frequency, volume, etc.) from the source language filler information. Good.
Moreover, the process etc. which change the para language of the audio | voice signal which the filler information production | generation part 24 performs are already well-known, The detailed description is abbreviate | omitted.

  The filler information insertion unit 25 is a parameter included in the source language filler information extracted by the filler information extraction unit 21 at the filler insertion position specified by the filler insertion position specification unit 23 in the target language voice information generated by the voice generation unit 17. The target language filler information, which is information having the same para language as the language and is voice information of the filler in the target language, is inserted. That is, the filler information insertion unit 25 inserts the target language filler information generated by the filler information generation unit 24 into the target language voice information.

  Inserting the target language filler information at the filler insertion position of the target language voice information means that, for example, the target language voice information does not include the voice information corresponding to the filler whose temporal position is specified by the filler time position specifying unit 20. In this case, the target language filler information may be added to the filler insertion position, or the voice information corresponding to the filler for which the filler time position specifying unit 20 specifies the temporal position is added to the target language voice information. Is included, the target language filler information may be overwritten and added to the position of the voice information indicated by the filler insertion position.

  A recording medium in which source language speech information is stored, a recording medium in which speech recognition result information is stored, a recording medium in which translation result information is stored, a recording medium in which target language speech information is stored, etc. It may be realized by a medium, or may be realized by a separate recording medium. In the former case, for example, an area in which source language speech information is stored becomes a recording medium in which source language speech information is stored, and an area in which speech recognition result information is stored is stored in speech recognition result information. Recording medium.

Next, the operation of the speech translation apparatus 1 according to this embodiment will be described using the flowchart of FIG.
(Step S101) The voice information receiving unit 11 determines whether source language voice information has been received. If accepted, the process proceeds to step S102. If not, the process of step S101 is repeated until accepted.

  (Step S102) The source language speech information storage unit 12 stores source language speech information received by the speech information reception unit 11. When the voice information receiving unit 11 receives the source language voice information directly from the microphone in real time, the process of step S101 and step S102 is repeatedly executed, so that the source language of a series of lengths is obtained. Audio information may be accumulated.

  (Step S103) The voice recognition unit 13 performs voice recognition on the source language voice information accumulated by the source language voice information accumulation unit 12, and acquires voice recognition result information corresponding to the source language voice information.

  (Step S104) The voice recognition result information accumulation unit 14 accumulates the voice recognition result information acquired by the voice recognition unit.

  (Step S105) The machine translation unit 15 performs machine translation on the speech recognition result information accumulated by the speech recognition result information accumulation unit 14, and acquires translation result information of a target language corresponding to the speech recognition result information.

  (Step S106) The translation result information accumulation unit 16 accumulates the translation result information acquired by the machine translation unit 15.

  (Step S107) The voice generation unit 17 generates target language voice information corresponding to the translation result information accumulated by the translation result information accumulation unit 16.

  (Step S108) The target language voice information storage unit 18 stores the target language voice information generated by the voice generation unit 17.

  (Step S109) The filler time position specifying unit 20 specifies the temporal position of the filler in the source language speech information accumulated by the source language speech information accumulation unit 12. This specified information may be temporarily stored in a recording medium (not shown).

  (Step S <b> 110) The filler information extraction unit 21 extracts source language filler information related to the filler for which the filler time position specifying unit 20 has specified a temporal position in the source language voice information accumulated by the source language voice information accumulation unit 12. . The extracted source language filler information may be temporarily stored in a recording medium (not shown).

  (Step S111) The filler text position specifying unit 22 specifies the position in the voice recognition result information accumulated by the voice recognition result information accumulating unit 14 corresponding to the temporal position of the filler specified by the filler time position specifying unit 20. . This specified information may be temporarily stored in a recording medium (not shown).

  (Step S112) The filler insertion position specifying unit 23 corresponds to the position in the speech recognition result information specified by the filler text position specifying unit 22, and the temporal position in the target language voice information stored by the target language voice information storage unit 18 The filler insertion position is specified. Information indicating the specified filler insertion position may be temporarily stored in a recording medium (not shown).

  (Step S113) The filler information generation unit 24 generates target language filler information having the same para language as the para language included in the source language filler information extracted by the filler information extraction unit 21. The generated target language filler information may be temporarily stored in a recording medium (not shown).

  (Step S114) The filler information insertion unit 25 inserts the target language filler information generated by the filler information generation unit 24 into the filler insertion position specified by the filler insertion position specification unit 23 in the target language voice information.

(Step S115) The voice information output unit 19 outputs the target language voice information in which the target language filler information is inserted. Then, the process returns to step S101.
In the flowchart of FIG. 2, the process is terminated by powering off or a process termination interrupt. In the flowchart of FIG. 2, the order of processing has a certain degree of arbitraryness. For example, the order of the filler insertion position specifying process (step S112) and the target language filler information generating process (step S113) may be reversed.

Next, the operation of the speech translation apparatus 1 according to this embodiment will be described using a specific example.
In this specific example, a case where a Japanese speaker speaks Japanese and the speech is translated into English will be described.

  First, suppose that the speaker spoke to Mike, saying, “I divided it into three periods for convenience, but there are some overlaps in each period.” Then, the source language voice information is received by the voice information receiving unit 11 and stored by the source language voice information storage unit 12 (steps S101 and S102). The source language voice information of FIG. 3 shows an example of the source language voice information accumulated in this way.

  Next, the speech recognition unit 13 performs speech recognition processing on the accumulated source language speech information, and acquires speech recognition result information that is a recognition result (step S103). It is assumed that the speech recognition result information is "I divided it into three periods for convenience, but there is a slight overlap in each period." It is assumed that the voice recognition unit 13 in this specific example cannot recognize the filler “e˜” and has inserted a <no voice recognition> mark in the voice recognition result information. The acquired voice recognition result information is stored by the voice recognition result information storage unit 14 (step S104). The voice recognition result information in FIG. 3 indicates the voice recognition result information accumulated in this way. As shown in FIG. 3, the time code of the source language speech information is also given to the speech recognition result information, and the correspondence between the speech signal of the source language speech information and the text of the speech recognition result information is taken. Can be done.

  Next, the machine translation unit 15 performs machine translation processing on the accumulated speech recognition result information, and acquires translation result information that is the machine translation result (step S105). Note that the machine translation unit 15 does not perform machine translation for <no speech recognition>. The translation result information is assumed to be "so for practical reason, I divided the era into the eras. The eras are somehow overwrapping." The acquired translation result information is accumulated by the translation result information accumulation unit 16 (step S106). The translation result information of FIG. 3 shows the translation result information accumulated in this way.

  Next, the speech generation unit 17 performs speech synthesis processing on the accumulated translation result information, and generates target language speech information corresponding to the translation result information (step S107). The target language voice information is stored by the target language voice information storage unit 18 (step S108). The target language voice information in FIG. 3 shows an example of the target language voice information accumulated in this way. As shown in FIG. 3, a time code is also given to the target language voice information, but it does not correspond to the time code of the source language voice information. Moreover, since the filler is not translated, it is the target language voice information without the filler. A time code is also given to the translation result information. This is a result of the time code of the target language voice information being added afterwards to the corresponding position of the translation result information after speech synthesis. Note that the translation result information in FIG. 3 shows the case where a time code is assigned every 0.5 seconds, but it goes without saying that the time code may be given more finely.

  Next, the filler time position specifying unit 20 refers to the voice recognition result information in FIG. 3 and specifies a section of <no voice recognition>. Here, it is assumed that the time codes 3.1 to 4.5 are specified as sections where speech recognition is impossible. Since this time code corresponds to the time code of the source language speech information, the range of the time code is directly used as the time position of the filler in the source language speech information. The filler time position specifying unit 20 temporarily stores the specified time code range in a recording medium (not shown) (step S109).

  The filler information extraction unit 21 refers to the time code range specified by the filler time position specification unit 20 and extracts the para language from the source language speech information in the time code range (step S110). The para language is source language filler information. In this specific example, as the para language, the frequency change, volume change, length of the filler of the voice information of the source language, the length of time of the blank of the voice on the start side of the filler (between pre), It is assumed that the time length (between posts) of the voice on the end side is extracted. The filler information extraction unit 21 temporarily stores the extracted source language filler information in a recording medium (not shown).

  The filler text position specifying unit 22 specifies the position in the speech recognition result information corresponding to the temporal position of the source language filler. This specification is performed by specifying the position of <no voice recognition> in the voice recognition result information (step S111). Specifically, the filler text position specifying unit 22 temporarily stores in a recording medium (not shown) a chunk “there is a slight overlap at each time” that exists behind the position of <no voice recognition>. To do that.

  The filler insertion position specifying unit 23 refers to “there is a slight overlap in each period” specified by the filler text position specifying unit 22, and passes the text to the machine translation unit 15, so that the translation result “the eras” are somehow overwrapping. ”. Then, the filler insertion position specifying unit 23 refers to the translation result information accumulated by the translation result information accumulating unit 16 and specifies the position immediately before the translation result “the eras are somehow overlapping”. To do. Since this time code “5.6” corresponds to the time code of the target language voice information, the position of the time code is directly used as the filler insertion position. The filler insertion position specifying unit 23 temporarily stores the time code in a recording medium (not shown) (step S112). In FIG. 3, the filler insertion position is indicated by an arrow (this is for convenience of explanation, and the information on the arrow may not actually be included in the translation result information or the like).

  It is assumed that the filler information generation unit 24 holds a “Well” audio signal in a recording medium (not shown) as a filler in the target language. Then, the filler information generation unit 24 refers to the source language filler information extracted by the filler information extraction unit 21 and matches the para language of the voice signal of “Well” with the para language included in the source language filler information. Target language filler information is generated (step S113). The filler information generation unit 24 temporarily stores the target language filler information in a recording medium (not shown).

  The filler information insertion unit 25 refers to the filler insertion position specified by the filler insertion position specification unit 23, and inserts the target language filler information generated by the filler information generation unit 24 into the filler insertion position (step S114). The target language speech information with the target language filler information inserted in FIG. 3 shows an example of the target language speech information after the target language filler information is inserted. Finally, the speech information output unit 19 outputs the target language speech information with the target language filler information inserted (step S115). That is, a sound corresponding to “so for practical reason, I divided the era into the eras. Well the eras are somehow overlapping” is output. In addition, since the para-language of the “Well” part of the filler matches the para-language of “e-” that the speaker has generated in the original language, the voice that includes the speaker's emotions and the like that is more realistic. Result of translation.

  In this specific example, the case where only one filler is included in the source language speech information has been described, but when two or more fillers are included in the source language speech information, each filler is the same as described above. Processing will be performed. For example, in each process of steps S109 to S114 in the flowchart of FIG. 2, a process for a plurality of fillers is performed.

  Also, of the specific data of FIG. 3 used in this specific example, the portion related to the target language speech information is shown for explanation in this specific example, and may differ from the actual data.

  As described above, according to the speech translation apparatus 1 according to the present embodiment, the filler can also be translated, and the post-translation target language filler and the source language filler can share the para language. Therefore, the filler is information that is normally discarded in speech translation, but by translating the filler, it is possible to realize speech translation that conveys the speaker's emotions with a more realistic feeling. Further, by inserting the target language filler information into the target language voice information, the rhythm of the spoken language in the original language can be given to the target language voice information. Furthermore, since the inserted target language filler information has the same para language as the source language filler information, the air of speech in the source language can be read also with the target language speech information. As a result, a person who is listening to the target language speech information may notice that the context is misunderstood, for example, or may notice a mistranslation due to machine translation.

  In the speech translation apparatus 1 according to the present embodiment, the case where the target language filler information is generated using the source language filler information extracted by the filler information extraction unit 21 has been described, but the generation may not be performed. For example, the source language filler information may be voice information (voice signal) of the filler whose temporal position is specified by the filler time position specifying unit 20, and the target language filler information may be the source language filler information itself. In this case, the source language filler itself is inserted into the target language speech information. Since the filler itself is not significant information, it can be understood to some extent even if the language is different. Therefore, even if the source language filler itself is inserted into the target language speech information, it is considered that this does not hinder the understanding of the listener of the target language speech information. In this case, the speech translation apparatus 1 may not include the filler information generation unit 24. Further, as described above, the filler information insertion unit 25 is information having the same paralanguage as the paralanguage included in the source language filler information at the filler insertion position in the target language speech information, and the filler speech in the target language. If the target language filler information which is information is inserted, the target language filler information generated by the filler information generation unit 24 may not be inserted.

  Moreover, although the case where the speech translation apparatus 1 was a stand-alone was demonstrated in the said embodiment, the stand-alone apparatus may be sufficient as the speech translation apparatus 1, and the server apparatus in a server client system may be sufficient as it. In the latter case, the output unit or the reception unit receives an input or outputs a screen via a communication line.

  In the above embodiment, each process or each function may be realized by centralized processing by a single device or a single system, or may be distributedly processed by a plurality of devices or a plurality of systems. It may be realized by doing.

  In the above embodiment, information related to processing executed by each component, for example, information received, acquired, selected, generated, transmitted, or received by each component In addition, information such as threshold values, mathematical formulas, addresses, etc. used by each component in processing is retained temporarily or over a long period of time on a recording medium (not shown) even when not explicitly stated in the above description. It may be. Further, the storage of information in the recording medium (not shown) may be performed by each component or a storage unit (not shown). Further, reading of information from the recording medium (not shown) may be performed by each component or a reading unit (not shown).

  In the above embodiment, when two or more constituent elements included in the speech translation apparatus 1 have a communication device, an input device, or the like, two or more constituent elements may physically have a single device. Or you may have separate devices.

  In the above embodiment, each component may be configured by dedicated hardware, or a component that can be realized by software may be realized by executing a program. For example, each component can be realized by a program execution unit such as a CPU reading and executing a software program recorded on a recording medium such as a hard disk or a semiconductor memory. The software that realizes the speech translation apparatus 1 in the above embodiment is the following program. In other words, this program uses a computer to obtain a speech information receiving unit that receives source language speech information, which is information obtained by collecting speech of a spoken source language with a microphone, and source language speech information received by the speech information receiving unit. A speech recognition unit that recognizes speech and acquires speech recognition result information that is text information corresponding to the source language speech information, and machine translates the speech recognition result information acquired by the speech recognition unit, and the speech recognition result A machine translation unit that acquires translation result information that is text information of a target language corresponding to information, and a voice that generates target language voice information that is voice information of a target language corresponding to the translation result information acquired by the machine translation unit In the source language speech information received by the generation unit and the speech information reception unit, the temporal position of the filler that is an utterance inserted between significant utterances is specified. A filler time position specifying unit, and a source language that is information including at least a para-language that is non-linguistic information in voice information, regarding the filler whose time position is specified by the filler time position specifying unit in the source language voice information A filler information extracting unit for extracting filler information, and a filler text position specifying unit for specifying a position in the speech recognition result information corresponding to a temporal position of the filler specified by the filler time position specifying unit in the source language speech information A filler insertion position identifying unit that identifies a filler insertion position that is a temporal position in the target language speech information corresponding to the position in the speech recognition result information identified by the filler text position identifying unit, and the speech generation unit The filler insertion position specifying unit in the target language speech information generated by The target language filler information which is the same para language as the para language included in the source language filler information extracted by the filler information extraction unit and is the voice information of the filler in the target language is inserted at the specified filler insertion position. And a filler information insertion unit that functions as a voice information output unit that outputs target language voice information in which target language filler information is inserted by the filler information insertion unit.

  In the above program, the output step for outputting information and the accepting step for receiving information do not include at least processing performed only by hardware, for example, processing performed by a modem or an interface card in the output step. .

  Further, this program may be executed by being downloaded from a server or the like, and a program recorded on a predetermined recording medium (for example, an optical disk such as a CD-ROM, a magnetic disk, a semiconductor memory, or the like) is read out. May be executed by

  Further, the computer that executes this program may be singular or plural. That is, centralized processing may be performed, or distributed processing may be performed.

  FIG. 4 is a schematic diagram showing an example of the appearance of a computer that executes the program and realizes the speech translation apparatus according to the embodiment. The above-described embodiment is realized by computer hardware and a computer program executed on the computer hardware.

  In FIG. 4, a computer system 100 includes a computer 101 including a CD-ROM (Compact Disk Read Only Memory) drive 105, an FD (Flexible Disk) drive 106, a keyboard 102, a mouse 103, and a monitor 104.

  FIG. 5 is a diagram illustrating a computer system. In FIG. 5, in addition to the CD-ROM drive 105 and the FD drive 106, a computer 101 includes a CPU (Central Processing Unit) 111, a ROM (Read Only Memory) 112 for storing a program such as a bootup program, A CPU (Random Access Memory) 113 that is connected to the CPU 111 and temporarily stores application program instructions and provides a temporary storage space, a hard disk 114 that stores application programs, system programs, and data, a CPU 111 and a ROM 112. Etc. to each other. The computer 101 may include a network card (not shown) that provides connection to the LAN.

  A program that causes the computer system 100 to execute the functions of the speech translation apparatus according to the above embodiment is stored in the CD-ROM 121 or FD 122, inserted into the CD-ROM drive 105 or FD drive 106, and transferred to the hard disk 114. May be. Instead, the program may be transmitted to the computer 101 via a network (not shown) and stored in the hard disk 114. The program is loaded into the RAM 113 at the time of execution. The program may be loaded directly from the CD-ROM 121, the FD 122, or the network.

  The program does not necessarily include an operating system (OS) or a third-party program that causes the computer 101 to execute the functions of the speech translation apparatus according to the above embodiment. The program may include only a part of an instruction that calls an appropriate function (module) in a controlled manner and obtains a desired result. How the computer system 100 operates is well known and will not be described in detail.

  Further, the present invention is not limited to the above-described embodiment, and various modifications are possible, and it goes without saying that these are also included in the scope of the present invention.

  As described above, according to the speech translation apparatus or the like according to the present invention, the effect that a more realistic speech translation can be realized by translating the filler is also useful, and is useful as a speech translation system.

1 is a block diagram showing the configuration of a speech translation apparatus according to Embodiment 1 of the present invention. The flowchart which shows the operation | movement of the speech translation apparatus by the embodiment The figure which shows an example of source language audio | voice information etc. by the embodiment Schematic diagram showing an example of the appearance of the computer system in the embodiment The figure which shows an example of a structure of the computer system in the embodiment

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Speech translation apparatus 11 Speech information reception part 12 Source language speech information storage part 13 Speech recognition part 14 Speech recognition result information storage part 15 Machine translation part 16 Translation result information storage part 17 Speech generation part 18 Target language speech information storage part 19 Voice Information output unit 20 Filler time position specifying unit 21 Filler information extracting unit 22 Filler text position specifying unit 23 Filler insertion position specifying unit 24 Filler information generating unit 25 Filler information inserting unit

Claims (9)

A voice information receiving unit that receives source language voice information, which is information obtained by collecting the spoken source language voice with a microphone;
A speech recognition unit that recognizes the original language speech information received by the speech information reception unit and obtains speech recognition result information that is text information corresponding to the source language speech information;
A machine translation unit that machine translates the speech recognition result information obtained by the speech recognition unit, and obtains translation result information that is text information of a target language corresponding to the speech recognition result information;
A speech generation unit that generates target language speech information that is speech information of a target language corresponding to the translation result information acquired by the machine translation unit;
In the source language voice information received by the voice information receiving unit, a filler time position specifying unit that specifies a temporal position of a filler that is an utterance inserted between significant utterances;
Filler information extraction for extracting source language filler information, which is information including at least a para language, which is non-language information in speech information, regarding the filler whose temporal position is specified by the filler time position specifying unit in the source language speech information. And
A filler text position specifying unit for specifying a position in the voice recognition result information corresponding to a temporal position of the filler specified by the filler time position specifying unit in the source language voice information;
A filler insertion position identifying unit that identifies a filler insertion position that is a temporal position in the target language speech information corresponding to the position in the speech recognition result information identified by the filler text position identifying unit;
In the target language voice information generated by the voice generation unit, the filler insertion position specified by the filler insertion position specification unit has the same para language as the para language included in the source language filler information extracted by the filler information extraction unit. A filler information insertion unit that inserts target language filler information that is information and voice information of the filler in the target language;
A speech translation apparatus comprising: a speech information output unit that outputs target language speech information in which target language filler information is inserted by the filler information insertion unit. The source language filler information is voice information of a filler whose temporal position is specified by the filler time position specifying unit,
The speech translation apparatus according to claim 1, wherein the target language filler information is the source language filler information. A filler information generating unit for generating target language filler information having the same para language as the para language included in the source language filler information;
The speech translation apparatus according to claim 1, wherein the filler information insertion unit inserts the target language filler information generated by the filler information generation unit into target language speech information. The speech translation device according to claim 3, wherein the filler information generation unit generates target language filler information having the same vowel as the vowel of the filler whose temporal position is specified by the filler time position specifying unit. 4. The speech translation apparatus according to claim 3, wherein the filler information generation unit generates target language filler information that is speech information of a target language corresponding to a result of speech recognition and machine translation of the filler of the source language speech information. . The speech translation apparatus according to claim 1, wherein the para language is at least one information selected from a frequency, a volume, a change in frequency, and a change in volume. The para language includes the time length of the filler of the source language speech information, the time length of the speech blank on the start point side of the filler, and the time length of the speech blank on the end point side of the filler The speech translation apparatus according to claim 6, further comprising: Voice information reception unit, voice recognition unit, machine translation unit, voice generation unit, filler time position specifying unit, filler information extraction unit, filler text position specifying unit, filler insertion position specifying unit, filler information A speech translation method processed using an insertion unit and a speech information output unit,
A voice information receiving step in which the voice information receiving unit receives source language voice information which is information obtained by collecting voice of the spoken source language with a microphone;
A speech recognition step in which the speech recognition unit recognizes speech information received in the speech information reception step and acquires speech recognition result information that is text information corresponding to the source language speech information;
The machine translation unit machine translates the speech recognition result information obtained in the speech recognition step, and obtains translation result information that is text information of a target language corresponding to the speech recognition result information; and
A voice generation step in which the voice generation unit generates target language voice information that is voice information of a target language corresponding to the translation result information acquired in the machine translation step;
In the source language speech information received in the speech information reception step, the filler time position specification unit, a filler time position specification step for specifying a temporal position of a filler that is an utterance inserted between significant utterances;
Source language filler that is information including at least a para-language that is non-linguistic information in speech information, regarding the filler whose temporal position is identified in the filler time location specification step in the source language speech information by the filler information extraction unit. A filler information extraction step for extracting information;
The filler text position specifying unit specifies a position in the voice recognition result information corresponding to the temporal position of the filler specified in the filler time position specifying step in the source language voice information;
Filler insertion position specifying step in which the filler insertion position specifying unit specifies a filler insertion position that is a temporal position in the target language speech information corresponding to the position in the speech recognition result information specified in the filler text position specifying step. When,
The filler information insertion unit includes the parameter included in the source language filler information extracted in the filler information extraction step at the filler insertion position identified in the filler insertion position identification step in the target language speech information generated in the speech generation step. Filler information insertion step for inserting target language filler information which is information having the same para language as the language and is voice information of the filler in the target language;
A speech translation method comprising: a speech information output step in which the speech information output unit outputs target language speech information in which target language filler information is inserted in the filler information insertion step. Computer
A voice information receiving unit that receives source language voice information, which is information obtained by collecting the spoken source language voice with a microphone;
A speech recognition unit that recognizes the original language speech information received by the speech information reception unit and obtains speech recognition result information that is text information corresponding to the source language speech information;
A machine translation unit that machine translates the speech recognition result information obtained by the speech recognition unit, and obtains translation result information that is text information of a target language corresponding to the speech recognition result information;
A speech generation unit that generates target language speech information that is speech information of a target language corresponding to the translation result information acquired by the machine translation unit;
In the source language voice information received by the voice information receiving unit, a filler time position specifying unit that specifies a temporal position of a filler that is an utterance inserted between significant utterances;
Filler information extraction for extracting source language filler information, which is information including at least a para language, which is non-language information in speech information, regarding the filler whose temporal position is specified by the filler time position specifying unit in the source language speech information. And
A filler text position specifying unit for specifying a position in the voice recognition result information corresponding to a temporal position of the filler specified by the filler time position specifying unit in the source language voice information;
A filler insertion position identifying unit that identifies a filler insertion position that is a temporal position in the target language speech information corresponding to the position in the speech recognition result information identified by the filler text position identifying unit;
In the target language voice information generated by the voice generation unit, the filler insertion position specified by the filler insertion position specification unit has the same para language as the para language included in the source language filler information extracted by the filler information extraction unit. A filler information insertion unit that inserts target language filler information that is information and voice information of the filler in the target language;
A program for functioning as a voice information output unit that outputs target language voice information in which target language filler information is inserted by the filler information insertion unit.

Images