JPWO2011033834A1 - Speech translation system, speech translation method, and recording medium - Google Patents

Abstract

The speech translation system (1) includes a speech recognition unit (11) that recognizes an input speech in a predetermined unit to generate character data, and a recognition result that connects the character data generated by the speech recognition unit (11). A linking unit (12), a sentence determination unit (13) for determining whether or not the character data linked by the recognition result linking unit (12) is established as a sentence, and a translation unit for translating the linked character data (14) and an output unit (15) for outputting a translation result by the translation unit (14). With this configuration, translation results of concatenated character data including those determined to be established as sentences are output, so that translation results can be output in real time and grammatically correct translation results can also be output. Can do.

Description

  The present invention relates to a speech translation system that translates input speech, a speech translation method, and a recording medium.

  In recent years, a speech translation system using a speech recognition system and a machine translation system has been proposed. In general, in a speech recognition system, when there is no device for determining a speech recognition processing unit such as a microphone button, the processing unit for speech recognition processing is determined using a physical phenomenon called speech breaks. In contrast, machine translation systems perform translation processing in units of text called sentences. Therefore, when a speech translation system is created by simply combining these two systems, it is difficult to properly translate the spoken speech because translation processing is performed for each break in speech. In particular, it was difficult to translate between languages such as English and Japanese whose word order was very different.

  In order to solve such problems, a technique has been proposed that accepts ineligible expressions in Japanese spoken language and outputs sequential translation results that can be read unnaturally by those who are native speakers of Japanese. (For example, refer nonpatent literature 1.). In general, spoken language frequently appears with various ineligible expressions that are not found in written words such as repetition, reversal of word order, omission, phrasing, rephrasing, and sayings. Humans can easily understand the meaning of the utterance by accepting the ineligible expression even if the utterance includes the ineligible expression due to their advanced utterance understanding ability. Therefore, the above technique does not eliminate the ineligible expression of the Japanese spoken language, but rather accepts it positively and translates the input other languages sequentially. By adopting such a configuration, the above technique realizes outputting translation results in real time.

Shigeki Matsubara, Satoru Asai, Katsuhiko Toyama, Yasuyoshi Inagaki, Progressive English-Japanese Spoken Language Translation Method Using Inappropriate Expressions, IEEJ Transactions, Vol.118-C, No.1, pp.71-78 (1998.1)

  However, with the technology that tolerates ineligible expressions, the translation result is output even if the written sentence is not a correct sentence. For this reason, the translation results were good when read in real time, but were very difficult to read when reread later.

  Accordingly, an object of the present invention is to propose a speech translation system, a speech translation method, and a recording medium capable of outputting a translation result in real time and outputting a grammatically correct translation result.

  In order to solve the problems as described above, a speech translation system according to the present invention includes a speech recognition unit that recognizes input speech in a predetermined unit to generate character data, and character data generated by the speech recognition unit. A recognition result linking unit for linking the character data, a sentence determination unit for determining whether the character data linked by the recognition result linking unit is established as a sentence, a translation unit for translating the linked character data, and this translation An output unit for outputting a translation result by the unit, and the recognition result connecting unit further connects the character data to the connected character data determined not to be established as a sentence by the sentence determining unit. It is.

  The speech translation method according to the present invention includes a speech recognition step for recognizing input speech in a predetermined unit to generate character data, and a recognition result connecting step for connecting character data generated by the speech recognition step. A sentence determination step for determining whether or not the character data concatenated by the recognition result concatenation step is established as a sentence, a translation step for translating the concatenated character data, and an output for outputting a translation result by the translation step The recognition result connecting step is characterized in that character data is further connected to the connected character data determined not to be established as a sentence by the sentence determining step.

  The recording medium according to the present invention includes a speech recognition step for generating character data by recognizing input speech in a predetermined unit to a computer, and a recognition result connection for connecting the character data generated by the speech recognition step. A step for determining whether or not the character data concatenated in this recognition result concatenation step is established as a sentence, a translation step for translating the concatenated character data, and outputting a translation result by this translation step The recognition result connecting step further connects character data to the connected character data determined not to be established as a sentence by the sentence determining step. It is characterized by this.

  According to the present invention, character data is generated by recognizing input speech in a predetermined unit, the generated character data is concatenated, it is determined whether or not the character data is established as a sentence, Concatenated text data that has been determined to be established as a sentence by further connecting character data to the connected text data determined not to be established, translating the linked text data, and outputting the translation result The translation result of the character data is output. Therefore, translation results can be output in real time, and grammatically correct translation results can also be output.

FIG. 1 is a block diagram showing a configuration of a speech translation system according to the first embodiment of the present invention. FIG. 2 is a block diagram showing a configuration of a speech translation system according to the second embodiment of the present invention. FIG. 3 is a flowchart showing the operation of the speech translation system according to the second embodiment of the present invention. FIG. 4 is a display example of character data and translation data in the output device. FIG. 5 is a display example of character data and translation data in the output device. FIG. 6 is a block diagram showing a configuration of a speech translation system according to the third embodiment of the present invention. FIG. 7 is a display example of character data and translation data in the output device. FIG. 8 is a display example of character data and translation data in the output device. FIG. 9 is a display example of character data and translation data in the output device.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

[First Embodiment]
First, the speech translation system according to the first embodiment of the present invention will be described.

  As shown in FIG. 1, the speech translation system 1 according to this embodiment includes a speech recognition unit 11 that recognizes an input speech in a predetermined unit to generate character data, and the speech recognition unit 11 generates the speech data. A recognition result linking unit 12 for linking character data, a sentence determination unit 13 for determining whether or not the character data linked by the recognition result linking unit 12 is established as a sentence, and a translation for translating the linked character data A unit 14 and an output unit 15 for outputting a translation result by the translation unit 14 are provided.

  Here, the recognition result connecting unit 12 further connects character data to the connected character data determined not to be established as a sentence by the sentence determining unit 13.

  Such a speech translation system 1 detects an input of information from the outside, such as an arithmetic device such as a CPU, a storage device such as a memory or an HDD (Hard Disk Drive), and a keyboard, mouse, pointing device, button, touch panel, or the like. An input device, an I / F device that transmits and receives various types of information via a communication line such as a LAN (Local Area Network) and a WAN (Wide Area Network), and a CRT (Cathode Ray Tube), LCD (Liquid Crystal Display) or The computer includes a display device such as a field emission display (FED) and a program installed in the computer. That is, the hardware device and software cooperate to control the above hardware resources by a program, and the above-described speech recognition unit 11, recognition result connection unit 12, sentence determination unit 13, translation unit 14, and output unit 15. Is realized. The program may be provided in a state of being recorded on a recording medium such as a flexible disk, a CD-ROM, a DVD-ROM, a memory card, and an IC memory.

  By having such a configuration, in the present embodiment, the input speech is recognized by a predetermined unit to generate character data, the generated character data is concatenated, and the character data is established as a sentence. It is established as a sentence by connecting further character data to the linked character data determined not to be established as a sentence, translating the linked character data, and outputting the translation result. Then, the translation result of the connected character data including the determined one is output. Therefore, translation results can be output in real time, and grammatically correct translation results can also be output.

[Second Embodiment]
Next, a speech translation system according to the second embodiment of the present invention will be described.

<Configuration of speech translation system>
As shown in FIG. 2, the speech translation system 2 according to the present embodiment includes an input device 21 to which the user's speech is input, and data that translates words uttered by the user from the speech input to the input device 21. A processing device 22, a data storage device 23 that stores data used for information processing in the data processing device 22, and an output device 24 that outputs information processing results by the data processing device 22 are provided.

≪Configuration of input device≫
The input device 21 includes a known sound detection device that converts detected sound such as a microphone into an electric signal.

≪Data processing device configuration≫
The data processing device 22 is constituted by an information processing device that translates characters and words emitted by a user from voices emitted by the user by performing information processing on the electrical signal input from the input device 21. Such a data processing device 22 includes a speech recognition unit 221, a first translation unit 222, a recognition result connection unit 223, a sentence determination unit 224, a second translation unit 225, and a translation result output unit 226. Is provided.

  The voice recognition unit 221 is a functional unit that performs a voice recognition process of analyzing an electric signal (hereinafter, referred to as voice data) input from the input device 1 and extracting voice uttered by the user as character data. Note that the voice recognition processing by the voice recognition unit 221 is performed for each predetermined unit, for example, the voice data is divided for each voice break in which silence continues for 3 seconds or more.

  The 1st translation part 222 is a function part which translates the character data output from the speech recognition part 221 using a well-known machine translation technique.

  The recognition result connection unit 223 is a functional unit that generates a character data string obtained by connecting the character data output from the speech recognition unit 221 in time series and outputs the character data string to the sentence determination unit 224. Here, for the character data first output from the speech recognition unit 221, only this character data is output to the sentence determination unit 224 as a character data string. On the other hand, for a character data string determined by the sentence determination unit 224, which will be described later, as a sentence, the character data that continues in time series with the last character data included in the character data string is connected to create a new character. Output as a data string.

  The sentence determination unit 224 is a functional unit that determines whether the character data string output from the recognition result connection unit 223 is a sentence based on a sentence determination model 231 described later stored in the data storage device 23. is there.

  The second translation unit 225 is a functional unit that translates a character data string that is determined as a sentence by the sentence determination unit 224 using a known machine translation technique.

  The translation result output unit 226 outputs the translation result of the first translation unit 222 (hereinafter referred to as the first translation result) and the translation result of the second translation unit 225 (hereinafter referred to as the second translation result). This is a functional unit that outputs to the device 24.

  Such a data processing device 22 includes an arithmetic device such as a CPU, a storage device such as a memory and an HDD, an input device that detects input of information from the outside such as a keyboard, a mouse, a pointing device, a button, and a touch panel, and a LAN. And an I / F device that transmits and receives various types of information via a communication line such as a WAN, a computer having a display device such as a CRT, LCD, or FED, and a program installed in the computer. That is, the hardware device and software cooperate to control the above hardware resources by a program, and the above-described speech recognition unit 221, first translation unit 222, recognition result connection unit 223, sentence determination unit 224, A second translation unit 225 and a translation result output unit 226 are realized. The program may be provided in a state of being recorded on a recording medium such as a flexible disk, a CD-ROM, a DVD-ROM, a memory card, and an IC memory.

≪Configuration of data storage device≫
The data storage device 23 includes a known magnetic storage device, and stores a sentence determination model 231 including information for determining whether or not the character data is a sentence. The sentence determination model 231 includes a database including a large number of models learned by the N-gram model, for example. As the model, for example, in the case of a database learned with a 3-gram model of N = 3, it consists of a combination of three character strings or words. Such a sentence determination model 231 is constructed in advance.

≪Configuration of output device≫
The output device 24 includes a known display device such as a CRT, LCD, or FED, a known voice output device such as a speaker, and the like, and outputs the translation result input from the translation result output unit 226 of the data processing device 22.

  The speech translation system 2 according to the present embodiment includes an input device 21, a data processing device 22, a data storage device 23, and an output device 24, which are integrated into a computer and a program installed in the computer. Needless to say, it may be possible to do so.

<Operation of speech translation system>
Next, the operation of the speech translation system 2 according to the present embodiment will be described with reference to FIG.

  First, when a user's voice is input, the input device 21 converts the voice into an electric signal (voice data) and inputs the electric signal to the voice recognition unit 221 of the data processing device 22 (step S1).

  When the voice data is input, the voice recognition unit 221 recognizes the voice data and generates character data for each predetermined unit (step S2).

  When character data of a predetermined unit is generated, the first translation unit 222 translates the character data (step S3). The translated character data (first translation result) is sent to the translation result output unit 226, and the process of step S7 described later is performed.

  In addition, the recognition result connecting unit 223 connects the character data generated by the speech recognition unit 221 in time series to generate a character data string (step S4).

  When the character data string is generated, the sentence determination unit 224 determines whether or not the character data string holds as a sentence based on the sentence determination model 231 of the data storage device 23 (step S5).

  When the character data string does not hold as a sentence (step S5: NO), the sentence determination unit 224 sends the character data string to the recognition result connection unit 223. Then, the recognition result concatenation unit 223 generates a new character data string obtained by concatenating character data that continues in time series to the end of the received character data string.

  On the other hand, when the character data string is established as a sentence (step S5: YES), the second translation unit 225 translates the character data string (step S6). This translated character data string (second translation result) is sent to the translation result output unit 226.

  When the translation is performed by the first translation unit 222 and the second translation unit 225, the translation result output unit 226 sends the first translation result and the second translation result to the output device 24. Then, the output device 24 displays the received first translation result and second translation result on the display screen or outputs them from the speaker (step S7).

  Such a series of operations is performed until there is no sound input to the input device 21.

  As described above, according to the present embodiment, by sequentially translating the character data speech-recognized in a predetermined unit, the speech recognition result of the predetermined unit can be understood in real time, and the predetermined unit It is easy to read when the translation result is reviewed later by translating and outputting the character data string that is judged to be a sentence by linking the character data recognized by the voice in the above. Translation results can be provided.

<Operation example of speech translation system>
For example, a voice “I live” is input from the input device 21, and the voice recognition unit 221 detects that there is no voice information immediately after the utterance of “I live”, and the character data “I live”. Is output.

  The first translation unit 222 translates the character data and outputs the first translation result to the translation result output unit 226 as “I live”. Then, as shown in FIG. 4, in the area of the sequential output unit of the output device 24 composed of a display, the character data “I live” and the first “I live” translated this character data. The translation result of is displayed.

  On the other hand, since the first voice is “I live”, the recognition result connecting unit 223 sends “I live” to the sentence determination unit 224 as it is. The sentence determination unit 224 uses the sentence determination model 231 to determine whether or not the sent character data string “I live” holds as a sentence. Here, the sentence determination model 231 is composed of, for example, a database learned with a 3-gram model of N = 3, and only “I”, “live”, and “end of sentence” models exist in this database. To do. In such a case, the sentence determination unit 224 determines that the character data string “I live” does not hold as a sentence because there is no word or character string corresponding to “end of sentence” at the end, and the second translation unit No character data string is sent to H.225. Therefore, as shown in FIG. 4, nothing is displayed in the area of the sentence output unit of the output device 24 composed of a display.

  Next, in this state, it is assumed that voice is input to the input device 21 and the voice recognition unit 221 outputs character data “in a university dormitory”.

  Then, the first translation unit 222 translates the character data and outputs the first translation result “in the university dormitory” to the translation result output unit 226. Then, as shown in FIG. 5, the characters “in a university dormitory” are displayed in the sequential output unit area of the output device 24 under the above-mentioned indications “I live” and “I live”. The data and the first translation result “in the university dormitory”, which is a translation of this character data, are displayed.

  On the other hand, the recognition result concatenation unit 221 concatenates the character data string “I live” that was not determined as a sentence by the previous sentence determination unit 224 and the character data “in a university dormitory” input this time, Create a character string “I live in a university dormitory”. If there is a model of “university”, “dormitory”, and “end of sentence” in the sentence determination model 231 for this character data string, the sentence determination unit 224 determines that the character data string holds as a sentence, and “I A character data string “live in a university dormitory” is sent to the second translation unit 225. The second translation unit 225 outputs, to the translation result output unit 226, the second translation result “I live in a university dormitory” translated from the character data string. Then, as shown in FIG. 5, in the sentence output area of the output device 24, a character data string “I live in a university dormitory” and “I live in a university dormitory” The second translation result is displayed.

  As described above, according to the present embodiment, it is possible to understand the speech recognition result in real time by sequentially translating the character data speech-recognized in a predetermined unit. Moreover, it is determined whether or not the character data is concatenated to be established as a sentence, and the character data is translated in a unit that is established as a sentence, so that it is possible to understand in a unit.

  In the present embodiment, the sentence determination model 231 is described as an N-gram model with N = 3. However, the value of N is not limited to 3, and can be set as appropriate.

  In this embodiment, the case where the N-gram model is applied as the sentence determination model 231 has been described as an example. However, the sentence determination model 231 is not limited to the N-gram model, and for example, for parsing a character string. Various models such as a syntax analysis model such as the CFG rule used in the above can be freely applied as appropriate. For example, the verb “live” is often accompanied by a subject and a prepositional phrase, so that “I live” alone is not determined as a sentence by the sentence determination unit 224, but “I live in a university dormitory” is determined as a sentence for the first time. The

  In this embodiment, the case where the speech recognition unit 221 generates speech data in a predetermined unit by separating speech data at speech breaks has been described as an example. However, positions where speech data is separated are limited to speech breaks. Instead, for example, the pitch of the sound, the intensity, the specific sound, etc. can be set as appropriate.

  In the present embodiment, the case where the sentence determination unit 224 determines whether or not the character data is established as a sentence in the unit recognized by the voice recognition unit 221 is described as an example. You may make it determine as a sentence. For example, if the first utterance is “I live” and the second utterance is “in a university dormitory can I use”, the sentence up to “I live in a university dormitory” is determined as a sentence, and the remaining “can I “use” may be determined as a part of the next sentence. The sentence determination model 231 in this case can realize more accurate determination by using the above-described N-gram and syntax analysis result.

  Further, in the present embodiment, the case has been described as an example in which it is determined whether or not the text data received by the text determination unit 224 is satisfied as a text. However, when the determination is difficult, the text data of the next utterance is determined. You may make it determine after receiving. For example, if the first utterance “I live” alone cannot determine whether or not the sentence is established, it is determined that the sentence is not established, and the next utterance “in a university dormitory can I use” is awaited. Therefore, it may be determined that “I live” does not hold as a sentence but “I live in a university dormitory” is a sentence.

  In the present embodiment, the output device has been described as an example in which the display 24 is configured by a display. For example, only the first translation result displayed in the area of the sequential output unit is synthesized by speech, and the speaker May be output.

[Third Embodiment]
Next, a third embodiment according to the present invention will be described. In the present embodiment, components that are the same as those in the second embodiment described above are given the same names and reference numerals, and descriptions thereof are omitted as appropriate.

<Configuration of speech translation system 3>
As shown in FIG. 6, the speech translation system 3 according to the present embodiment includes an input device 21 to which the user's speech is input, and data that translates words uttered by the user from the speech input to the input device 21. A processing device 22, a data storage device 23 that stores data used for information processing in the data processing device 22, and an output device 24 that outputs information processing results by the data processing device 22 are provided.

  The data processing device 22 includes an information processing device that translates words uttered by a user from speech uttered by the user by performing information processing on an electrical signal input from the input device 21. Such a data processing device 22 includes a voice recognition unit 221, a recognition result connection unit 223, a sentence determination unit 224, a translation unit 301, and a display control unit 302.

  Here, the sentence determination unit 224 determines whether or not the character data string output from the recognition result connection unit 223 is established as a sentence based on a sentence determination model 231 described later stored in the data storage device 23. Part. The determination result of the sentence determination unit 224 is sent to the translation unit 301 and the display control unit 302.

  The translation unit 301 is a functional unit that translates the character data string determined by the sentence determination unit 224. Specifically, when the concatenated character data string is established as a sentence, the translation unit 301 translates the entire character data string. On the other hand, even when the sentence is not established, the character data string is translated. The translation result of the character data string by the translation unit 301 is sent to the display control unit 302.

  The display control unit 302 displays the character data string and the translation result on the output device 24 including a display, and controls the display position of the character data string and the translation result on the output device 24.

  In the speech translation system 3 having such a configuration, if it is determined that the concatenated character data string is established as a sentence, the character data string and the translation result are displayed at the display position of the corresponding output device 24, and the display position is displayed. The remaining character data string and translation result are displayed while moving.

<Operation example of speech translation system>
Next, an operation example of the speech translation system 3 according to the present embodiment will be described.

  As in the second embodiment, the utterance “I live” is input to the input device 21 and the speech recognition unit 221 detects that there is no speech information immediately after the utterance of “I live”. Assume that the character data “I live” is output.

  Since the first voice is “I live”, the recognition result connecting unit 223 sends the character data “I live” to the sentence determining unit 224 as a character data string. Then, the sentence determination unit 224 uses the sentence determination model 231 to determine whether or not the sent character data string “I live” holds as a sentence, as in the second embodiment. It is determined that there is no.

  Then, the translation unit 301 directly translates the character data string “I live” that does not hold as a sentence, and outputs the translation result “I live” to the display control unit 302. As shown in FIG. 7, the display control unit 302 displays a character data string “I live” and a translation result “I live” in the upper left area of the display device 24 set as the first display position. Is displayed. At this time, the display control unit 302 determines that the sentence determination unit 224 does not hold the sentence, and thus fixes the display position of the character data string and the translation result in the upper left area.

  In such a state, next, a voice “in a university dormitory can I use” is input to the input device 21, and voice data is immediately after the voice recognition unit 221 speaks “in a university dormitory can I use”. It is assumed that the character data “in a university dormitory can I use” is output.

  The recognition result concatenation unit 221 concatenates the character data “in a university dormitory can I use” input this time to the character data string “I live” that was not determined to be a sentence by the previous sentence determination unit 224, Create a character data string “I live in a university dormitory can I use”. As in the second embodiment, the sentence determination unit 224 determines that the part of “I live in a university dormitory” is established as a sentence by matching using the 3-gram model, as in the second embodiment. It is assumed that the remaining “can I use” is determined not to be established as a sentence.

  The translation unit 301 translates the part of the character data string “I live in a university dormitory” that is determined to be a sentence by the sentence determination unit 224 to “I live in a university dormitory”, and the rest Assume that the character data “can I use” is translated as “I can use it”.

  The display control unit 302 translates the character data string “I live in a university dormitory” determined to be a sentence by the sentence determination unit 224 and the translation result “I live in a university dormitory” "Is displayed in the upper left area of the output device 24 as shown in FIG. That is, the character data string “I live in a university dormitory” and the translation result “I live in a university dormitory” are displayed in the area where the character data string “I live” and the translation result “I live” are displayed. Overwrite "

In addition, the display control unit 302 moves the display position to the next line because the sentence determination unit 224 determines that the sentence is established as a sentence.
As for the character data string “can I use” determined not to be a sentence and the translation result “Is it usable” translated from the character data string, as shown in FIG. And is not displayed in the display position described above, that is, in the next line.

  By repeating such processing, the character data string that is voice-recognized in a predetermined unit on one display is sequentially translated, so that it is possible to understand the voice recognition result in real time and to review it later. Since the translation result of the part determined to be formed as a sentence by connecting the character data is displayed, the sentence can be understood in a unit.

  As described above, according to the present embodiment, since the character data string determined not to be established as a sentence is displayed as it is, the character data determined to be satisfied as a sentence can be understood in real time. Since the translation result for each sentence is displayed in the column, real-time translation with one display screen and sentence-by-sentence translation are displayed when reviewed later, making it easier to understand the sentence. .

  It is assumed that “I live in a university dormitory”, which is determined as a sentence by the sentence determination unit 224, is translated by the translation unit 301 as “I live in a university dormitory”. The display control unit 302 matches the character data string of the translation result “I live in” and the next translation result “I live in the university dormitory” at the display position at that time, and changes The translation result portion may be extracted, and the changed portion may be underlined or the color may be changed as shown in FIG. As a result, the character data added sequentially becomes easy to understand, and as a result, it becomes easy to understand in real time.

  It should be noted that the present invention is not limited to the first to third embodiments described above.

  This application claims the priority on the basis of Japanese application Japanese Patent Application No. 2009-216803 for which it applied on September 18, 2009, and takes in those the indications of all here.

  The present invention can be applied to a speech recognition device that generates a character string from speech, a speech translation device that translates the character string, and the like.

  DESCRIPTION OF SYMBOLS 1 ... Speech translation system, 11 ... Speech recognition part, 12 ... Recognition result connection part, 13 ... Sentence determination part, 14 ... Translation part, 15 ... Output part.

In order to solve the problems as described above, a speech translation system according to the present invention includes a speech recognition unit that recognizes input speech in a predetermined unit to generate character data, and character data generated by the speech recognition unit. A recognition result linking unit for linking, a sentence determination unit for determining whether or not the character data linked by the recognition result linking unit includes a sentence , a translation unit for translating the linked character data, and this translation unit An output unit that outputs the translation result of the recognition result, and the recognition result connecting unit further connects the character data to the connected character data determined not to include a sentence by the sentence determining unit. is there.

The speech translation method according to the present invention includes a speech recognition step for recognizing input speech in a predetermined unit to generate character data, and a recognition result connecting step for connecting character data generated by the speech recognition step. A sentence determination step for determining whether the character data concatenated by the recognition result concatenation step includes a sentence , a translation step for translating the concatenated character data, and an output step for outputting a translation result by the translation step The recognition result connecting step is characterized in that the character data is further connected to the connected character data determined not to include a sentence in the sentence determining step.

The recording medium according to the present invention includes a speech recognition step for generating character data by recognizing input speech in a predetermined unit to a computer, and a recognition result connection for connecting the character data generated by the speech recognition step. A step, a sentence determination step for determining whether or not the character data concatenated by the recognition result concatenation step includes a sentence , a translation step for translating the concatenated character data, and a translation result by the translation step is output. In the recording medium recording the program for executing the output step, the recognition result concatenation step further concatenates character data to the concatenated character data determined not to include a sentence by the sentence determination step. It is characterized by.

Claims (6)

A speech recognition unit that recognizes input speech in a predetermined unit and generates character data;
A recognition result concatenation unit that concatenates the character data generated by the speech recognition unit;
A sentence determination unit that determines whether or not the character data connected by the recognition result connection unit is established as a sentence;
A translation unit for translating the concatenated character data;
An output unit for outputting the translation result of the translation unit,
The speech recognition system, wherein the recognition result connecting unit further connects the character data to the connected character data determined not to be established as a sentence by the sentence determining unit. The translation unit
A first translation unit that translates the character data generated by the voice recognition unit for each predetermined unit;
A second translation unit that translates the linked character data determined to be established by the sentence determination unit;
The speech translation system according to claim 1, wherein the output unit outputs a translation result of the first translation unit and a translation result of the second translation unit. The output unit includes a display screen, displays the linked character data and the translation result of the translation unit at a predetermined position on the display screen, and determines that the sentence determination unit establishes the sentence. 2. The speech translation system according to claim 1, wherein the character data following the linked character data and the translation result of the translation unit are displayed at a position different from the predetermined position on the display screen. The output unit compares the translation result of the translation unit displayed at the predetermined position with the translation result of the translation unit to be displayed next at the predetermined position, and calculates a difference detected by the comparison. The speech translation system according to claim 3, wherein the translation result is displayed at the predetermined position with emphasis. A speech recognition step of generating input data by recognizing input speech in a predetermined unit;
A recognition result concatenation step for concatenating the character data generated by this speech recognition step;
A sentence determination step for determining whether or not the character data connected by the recognition result connection step is established as a sentence;
A translation step of translating the concatenated character data;
An output step for outputting the translation result of this translation step, and
The speech translation method, wherein the recognition result connection step further connects the character data to the connected character data determined not to be established as a sentence by the sentence determination step. On the computer,
A speech recognition step of generating input data by recognizing input speech in a predetermined unit;
A recognition result concatenation step for concatenating the character data generated by this speech recognition step;
A sentence determination step for determining whether or not the character data connected by the recognition result connection step is established as a sentence;
A translation step of translating the concatenated character data;
A recording medium on which a program for executing the output step of outputting the translation result of the translation step is recorded,
The recording medium characterized in that the recognition result connecting step further connects the character data to the connected character data determined not to be established as a sentence by the sentence determining step.

Images