JP6702119B2 - Speech recognition result creating device, method and program - Google Patents

Description

  The present invention relates to a speech recognition result creating device, method and program.

  Conventionally, a voice recognition technology for recognizing voice is known. Speech recognition technology converts speech into text, divides the obtained text into multiple time segments, selects the phrase corresponding to each segmented text, and concatenates the selected phrases to obtain the speech recognition result. (Sentence corresponding to voice) is created. Generally, in speech recognition technology, dictionary data prepared for each phrase is used to convert each segmented text into a corresponding phrase. Thereby, the accuracy of the voice recognition result can be improved.

  On the other hand, in recent years, a voice recognition service has been provided on the Internet. Some voice recognition services provide the user with voice recognition data including a voice recognition result and word and phrase candidates corresponding to each segmented text. The user can obtain voice recognition data corresponding to the voice data by inputting the voice data to the voice recognition service.

  However, in the above-mentioned conventional voice recognition service, the user cannot prepare the dictionary data independently. Therefore, there is a problem that the accuracy of the voice recognition result obtained from the voice recognition service is low.

  The present invention has been made in view of the above problems, and an object of the present invention is to make it possible to create a highly accurate voice recognition result based on voice recognition data.

  A speech recognition result creating device according to an embodiment is an input unit for receiving speech recognition data including candidate words and phrases corresponding to segment texts and evaluation values of the candidates, and a dictionary data storage unit that stores dictionary data. And a search unit that searches a matching portion of the dictionary data and the voice recognition data, and a creation unit that creates a voice recognition result based on the search result and the voice recognition data. ..

  According to each embodiment of the present invention, a highly accurate voice recognition result can be created based on voice recognition data.

The figure which shows an example of a function structure of a speech recognition result preparation device. The figure which shows an example of voice recognition data. The figure which shows an example of text data (dictionary data). The figure which shows an example of the hardware constitutions of a speech recognition result preparation device. The flowchart which shows the outline|summary of operation|movement of a speech recognition result preparation apparatus. The flowchart which shows an example of the production|generation process of the speech recognition result in 1st Embodiment. The figure which shows an example of the voice recognition data after the update of an evaluation value. The figure which shows an example of homophone data (dictionary data). The flowchart which shows an example of the production|generation process of the speech recognition result in 2nd Embodiment. The figure which shows an example of the data (dictionary data) for substitution. The flowchart which shows an example of the production|generation process of the speech recognition result in 3rd Embodiment. The flowchart which shows an example of the production|generation process of the speech recognition result in 4th Embodiment.

  Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In addition, regarding the description of the specification and the drawings according to each embodiment, components having substantially the same functional configurations are denoted by the same reference numerals and overlapping description will be omitted.

(First embodiment)
A speech recognition result creating apparatus (hereinafter referred to as “creating apparatus”) 1 according to the first embodiment will be described with reference to FIGS. 1 to 7. First, the functional configuration of the creation device 1 will be described. FIG. 1 is a diagram illustrating an example of a functional configuration of the creation device 1. The creation device 1 of FIG. 1 includes an input unit 11, a voice recognition data storage unit 12, a dictionary data storage unit 13, a search unit 14, and a creation unit 15.

  The input unit 11 receives voice recognition data output by an external voice recognition service. The input unit 11 stores the input voice recognition data in the voice recognition data storage unit 12.

  The voice recognition data storage unit 12 stores voice recognition data. Here, the voice recognition data will be described. When a voice to be recognized is input, the voice recognition service converts the voice into text. The speech recognition service then segments the resulting text into multiple time segments. Each segmented text is called segment text. Then, the speech recognition service outputs the candidate words and phrases corresponding to each segment text together with the evaluation value of each candidate as the speech recognition data. That is, the voice recognition data is data including candidate words and phrases corresponding to each segment text and an evaluation value of each candidate. The evaluation value here is a value indicating the likelihood of the candidate, that is, the high probability of being a correct answer. The correct answer means matching with the word or phrase intended by the speaker of the voice. Below, it is assumed that the higher the evaluation value, the higher the probability that the answer is correct. In this case, the speech recognition result obtained by the speech recognition service is a concatenation of the candidates having the highest evaluation value among the candidate words of each segment text.

  FIG. 2 is a diagram showing an example of the voice recognition data stored in the voice recognition data storage unit 12. The voice recognition data in FIG. 2 is output for the voice "Ricoh no Kakuda". It is assumed that the speaker intends to say "Ricoh's project."

  In the example of FIG. 2, the voice recognition service converts the voice "Ricoh no Kakuda" into the text "Iko no Kakuda" and divides it into four segment texts "Iko", "No", "Kikaku", and "Da". There is. "Let's go" and "transition" are output as the candidates for the word corresponding to "Iko". The evaluation value of "Let's go" is 0.3, and the evaluation value of "Transition" is 0.1. In the case of the voice recognition data shown in FIG. 2, the voice recognition result obtained by the voice recognition service is "the standard for going".

  The dictionary data storage unit 13 stores dictionary data prepared by the user. The dictionary data is data for improving the accuracy of the voice recognition result. In this embodiment, sentence data is stored as dictionary data. The sentence data is preferably related to the utterance of the speaker.

  FIG. 3 is a diagram showing an example of sentence data (dictionary data) stored in the dictionary data storage unit 13. In the example of FIG. 3, as the dictionary data, three sentence data of “Develop a service of Ricoh”, “The plan has already been considered”, and “It is necessary to formulate a future plan immediately”. Is remembered.

  The search unit 14 matches the voice recognition data and the dictionary data. That is, the search unit 14 searches for a matching portion between the voice recognition data and the dictionary data. The search method by the search unit 14 will be described later in detail.

  The creation unit 15 creates a voice recognition result based on the voice recognition data and the search result by the search unit 14. The method of creating the voice recognition result by the creating unit 15 will be described in detail later.

  Next, the hardware configuration of the creation device 1 will be described. FIG. 4 is a diagram illustrating an example of the hardware configuration of the creation device 1. The creation device 1 of FIG. 4 includes a CPU (Central Processing Unit) 101, a ROM (Read Only Memory) 102, a RAM (Random Access Memory) 103, and an HDD (Hard Disk Drive) 104. Further, the creation device 1 includes an input device 105, a display device 106, a communication interface 107, and a bus 108.

  By executing the program, the CPU 101 controls the entire creation device 1 and realizes the above-described functional configurations. The ROM 102 stores various data including a program executed by the CPU 101. The RAM 103 provides the CPU 101 with a work area. The HDD 104 stores various data including programs executed by the CPU 101. The input device 105 inputs information according to a user operation to the creation device 1. The input device 105 is, for example, a keyboard, a mouse, a touch panel, or the like. The display device 106 displays video and images. The display device 106 is, for example, a liquid crystal display or an organic EL (Electro Luminescence) display. The communication interface 107 is an interface for connecting the creating apparatus 1 to an external network. The bus 108 connects the CPU 101, the ROM 102, the RAM 103, the HDD 104, the input device 105, the display device 106, and the communication interface 107.

  Next, the operation of the creating device 1 according to the present embodiment will be described. FIG. 5 is a flowchart showing an outline of the operation of the creation device 1. When the voice recognition data is input to the input unit 11, the creation device 1 starts the operation of FIG.

  First, the input unit 11 stores the input voice recognition data in the voice recognition data storage unit 12 (step S101). As a result, the voice recognition data as shown in FIG. 2 is stored in the voice recognition data storage unit 12.

  Next, the search unit 14 matches the voice recognition data and the dictionary data. Then, the creation unit 15 creates a voice recognition result based on the search result by the search unit 14 and the voice recognition data (step S102).

  FIG. 6 is a flowchart showing an example of a voice recognition result creation process in this embodiment. The flowchart of FIG. 6 corresponds to the internal processing of step S102 of FIG. In the following, it is assumed that the voice recognition data storage unit 12 stores the voice recognition data of FIG. 2 and the dictionary data storage unit 13 stores the sentence data of FIG.

  First, the search unit 14 selects one segment text from the segment texts included in the voice recognition data (step S201). Here, it is assumed that "Iko" is selected.

  Next, the search unit 14 selects one candidate from the candidates of the phrase of the selected segment text (step S202). Here, it is assumed that “Let's go” is selected.

  Subsequently, the search unit 14 searches the sentence data (dictionary data) using the selected candidate as a search key (step S203). The search unit 14 outputs the number of matches between the search key (selected candidate) and the text data as a search result. Since the text data in FIG. 3 has no part that matches "Let's go", 0 is output as the search result.

  When the search unit 14 outputs the search result, the creation unit 15 updates the evaluation value of the selected candidate based on the output search result (step S204). In the present embodiment, the creation unit 15 updates the evaluation value such that the evaluation value increases as the number of matches with the text data increases. The method of updating the evaluation value is arbitrary. In the following, the evaluation value is updated by adding the number of matches to the original evaluation value. In this case, since the original evaluation value of "Let's go" is 0.3 and the number of matching cases is 0, the updated evaluation value is 0.3 (=0.3+0).

  When the search is completed, the search unit 14 confirms whether all the segment text candidates selected in step S201 have been selected (whether there are unselected candidates) (step S205). If there are unselected candidates (NO in step S205), the process returns to step S202. Then, the search unit 14 selects the next candidate from the unselected candidates (step S202). As a result, "migration" is selected.

  On the other hand, when all candidates have been selected (YES in step S205), the search unit 14 confirms whether all segment texts included in the voice recognition data have been selected (whether there is any unselected segment text) (step S206). ).

  If there is an unselected segment text (NO in step S206), the process returns to step S201. Then, the search unit 14 selects the next segment text from the unselected segment texts (step S201). As a result, "No" is selected.

  On the other hand, when all segment texts are selected (YES in step S206), the creating unit 15 selects the candidate having the highest evaluation value for each segment text and connects the selected candidates (step S207). As a result, a voice recognition result is created.

  FIG. 7 is a diagram showing an example of the voice recognition data stored in the voice recognition data storage unit 12. The voice recognition data of FIG. 7 corresponds to the updated evaluation value of the voice recognition data of FIG. In the case of the voice recognition data of FIG. 7, "Let's go" is selected as a candidate of "Iko", "No" is selected as a candidate of "No", "Plan" is selected as a candidate of "Kikaku", and " "Da" is selected as a candidate for "Da". As a result, the voice recognition result of "I'm planning to go" is created.

  The voice recognition result "I plan to go" created by the creating unit 15 is "Ricoh's plan" that the speaker intended compared to the voice recognition result "It's a standard to go" obtained by the voice recognition service. It is close to the statement "No." This is because, as a result of updating the evaluation value based on the text data, the evaluation value of "plan" becomes higher than the evaluation value of "standard", and "plan" is selected as the phrase corresponding to the segment text "Kikaku". Is.

  As described above, according to this embodiment, the accuracy of the voice recognition result can be improved based on the voice recognition data output by the voice recognition service and the sentence data (dictionary data) prepared by the user. In other words, according to the present embodiment, it is possible to create a highly accurate voice recognition result as compared with the voice recognition service.

  Further, according to the present embodiment, since the dictionary data is text data, it can be easily prepared. That is, dictionary data can be prepared without analyzing or processing the data.

  In the present embodiment, selection of segment text having only one word candidate may be omitted. This is because if there is only one candidate, that candidate is selected regardless of whether or not the evaluation value has been updated.

(Second embodiment)
The creation device 1 according to the second embodiment will be described with reference to FIGS. 8 and 9. The functional configuration and hardware configuration of the creation device 1 according to this embodiment are the same as those in the first embodiment. However, in this embodiment, sentence data and homophone data are stored as dictionary data, and a voice recognition result is created using the homophone data. The text data is as described above.

  The homophone word data is data indicating a set of a plurality of homophone words (hereinafter referred to as “homophone set”). The homophone data includes one or a plurality of homophone sets, and each homophone set includes a plurality of homophones.

  FIG. 8 is a diagram showing an example of homophone data (dictionary data) stored in the dictionary data storage unit 13. In the example of FIG. 8, the homophone word data includes three homophone word sets corresponding to “Kikaku”, “Hashi” and “jisho”. For example, a homophone word set corresponding to "Kikaku" includes three homophone words of "plan", "standard", and "its corner".

  Here, the operation of the creating apparatus 1 according to the present embodiment will be described. The outline of the operation of the creation device 1 according to this embodiment is the same as that of the first embodiment. FIG. 9 is a flowchart showing an example of a voice recognition result creation process according to the present embodiment. In the following, it is assumed that the voice recognition data storage unit 12 stores the voice recognition data of FIG. 2, and the dictionary data storage unit 13 stores the sentence data of FIG. 3 and the homophone data of FIG.

  First, the search unit 14 selects one segment text from the segment texts included in the voice recognition data (step S301). Here, it is assumed that "Kikaku" is selected.

  Next, the search unit 14 selects one candidate from the candidates of the phrase of the selected segment text (step S302). Here, it is assumed that “plan” is selected.

  Subsequently, the search unit 14 searches for homophone data (dictionary data) using the selected candidate as a search key (step S303). The search unit 14 outputs a homophone set including the search key (selected candidate) as a search result.

  When there is no homophone set including the search key (NO in step S304), the process proceeds to step S307.

  On the other hand, when there is a homophone set including the search key (YES in step S304), the search unit 14 searches the sentence data (dictionary data) using the selected homophone as a search key (step S305). The search unit 14 outputs the number of matches between the search key (selected candidate) and the text data as a search result. Since the text data in FIG. 3 has one part that matches “plan”, one result is output as the search result.

  When the search unit 14 outputs the search result, the creation unit 15 updates the evaluation value of the selected candidate based on the output search result (step S306). The method of updating the evaluation value is the same as in the first embodiment. In this case, since the original evaluation value of “plan” is 0.4 and the number of matching cases is 1, the updated evaluation value is 1.43 (=0.4+1).

  When the search is completed, the search unit 14 confirms whether all candidates of the segment text selected in step S301 have been selected (whether there are unselected candidates) (step S307). If there are unselected candidates (NO in step S307), the process returns to step S302. Then, the search unit 14 selects the next candidate from the unselected candidates (step S302). As a result, the "standard" is selected.

  On the other hand, when all candidates are selected (YES in step S307), the search unit 14 confirms whether all segment texts included in the voice recognition data have been selected (whether there is any unselected segment text) (step S308). ).

  If there is an unselected segment text (NO in step S308), the process returns to step S301. Then, the search unit 14 selects the next segment text from the unselected segment texts (step S301). As a result, “D” is selected.

  On the other hand, when all segment texts are selected (YES in step S308), the creating unit 15 selects the candidate having the highest evaluation value for each segment text and connects the selected candidates (step S309). As a result, a voice recognition result is created. When the text data in FIG. 3 and the homophone data in FIG. 8 are used, the voice recognition data after the evaluation value is updated becomes the same as the voice recognition data in FIG. 7. As a result, the voice recognition result of "I'm planning to go" is created.

  As described above, according to the present embodiment, when the segment text candidate is included in the homophone set, the evaluation value of the candidate is updated. In other words, when the segment text candidate is not included in the homophone set, the evaluation value of the candidate is not updated. As a result, it is possible to suppress excessive updating of the evaluation value and create a highly accurate voice recognition result.

(Third Embodiment)
The creation device 1 according to the third embodiment will be described with reference to FIGS. 10 and 11. The functional configuration and hardware configuration of the creation device 1 according to this embodiment are the same as those in the first embodiment. However, in the present embodiment, replacement data is stored as dictionary data, and a speech recognition result is created using this replacement data.

  The replacement data is data indicating a correspondence relationship between a word/phrase to be replaced (hereinafter referred to as “first word/phrase”) and a word/phrase to be replaced (hereinafter referred to as “second word/phrase”). The replacement data is prepared, for example, based on the erroneous conversion in the past voice recognition result obtained by the voice recognition service.

  FIG. 10 is a diagram showing an example of replacement data (dictionary data) stored in the dictionary data storage unit 13. In the example of FIG. 10, the first word “Ricoh” is associated with the second words “Go” and “Transition”. Further, the second phrase "always comes out" is associated with the first phrase "topic model".

  Here, the operation of the creating apparatus 1 according to the present embodiment will be described. The outline of the operation of the creation device 1 according to this embodiment is the same as that of the first embodiment. FIG. 11 is a flowchart showing an example of a voice recognition result creation process in this embodiment. The flowchart of FIG. 11 corresponds to the internal processing of step S102 of FIG. In the following, it is assumed that the voice recognition data storage unit 12 stores the voice recognition data of FIG. 2 and the dictionary data storage unit 13 stores the replacement data of FIG.

  First, the creating unit 15 refers to the voice recognition data, selects a candidate having the highest evaluation value for each segment text, and connects the selected candidates to create a voice recognition result (step S401). This voice recognition result corresponds to the voice recognition result obtained by the voice recognition service. In other words, the voice recognition result "It is a standard for going" is created.

  Next, the search unit 14 selects one first word/phrase from the first word/phrase text included in the replacement data (step S402). Here, it is assumed that “Ricoh” is selected.

  Subsequently, the search unit 14 selects one second phrase from the second phrases corresponding to the selected first phrase (step S403). Here, it is assumed that “Let's go” is selected.

  The search unit 14 searches for the voice recognition result created by the creating unit 15 using the selected second phrase as a search key (step S404). The search unit 14 outputs the matching portion between the search key (selected second phrase) and the voice recognition result as the search result.

  When the search unit 14 outputs the search result, the creation unit 15 updates the voice recognition result based on the search result. Specifically, when the voice recognition result includes a portion that matches the second word (YES in step S405), that is, when the voice recognition result includes the second word, the creating unit 15 handles the second word. It is replaced with the first phrase to be executed (step S406). As a result, “let's go” included in the voice recognition result is replaced with “Ricoh”. Then, a process progresses to step S407.

  On the other hand, if the voice recognition result does not have a portion that matches the second word (NO in step S405), that is, if the voice recognition result does not include the second word, the process proceeds to step S407.

  Upon completion of the search, the search unit 14 confirms whether all the second phrases corresponding to the first phrase selected in step S402 have been selected (whether there is any unselected second phrase) (step S407). If there is an unselected second word/phrase (NO in step S407), the process returns to step S403. Then, the search unit 14 selects the next second phrase from the unselected second phrases (step S403). As a result, "migration" is selected.

  On the other hand, when all the second words are selected (YES in step S407), the search unit 14 confirms whether all the first words included in the replacement data have been selected (whether there is any unselected first word). Yes (step S408).

  If there is an unselected first word/phrase (NO in step S408), the process returns to step S402. Then, the search unit 14 selects the next first phrase from the unselected first phrases (step S402). As a result, the "topic model" is selected.

  On the other hand, if all the first words have been selected (YES in step S408), the voice recognition result creation process ends. The voice recognition result held by the creating unit 15 at this point becomes the voice recognition result obtained by the creating device 1. As a result, a voice recognition result "Ricoh's standard" is created.

  The voice recognition result "Ricoh's standard" created by the creating unit 15 is "Ricoh's plan" that the speaker intended, compared to the voice recognition result "It's a standard for going" obtained by the voice recognition service. It is understood that it is close to the remark. This is because “goko” was replaced with “Ricoh” by replacing the phrase based on the replacement data.

  As described above, according to the present embodiment, the accuracy of the voice recognition result can be improved based on the voice recognition data output by the voice recognition service and the replacement data (dictionary data) prepared by the user. .. In other words, according to the present embodiment, it is possible to create a highly accurate voice recognition result as compared with the voice recognition service.

  Further, according to this embodiment, the segment text can be converted into a phrase that is not included in the speech recognition data as a phrase candidate.

(Fourth Embodiment)
The creation device 1 according to the fourth embodiment will be described with reference to FIG. The functional configuration and hardware configuration of the creation device 1 according to this embodiment are the same as those in the first embodiment. However, in the present embodiment, the sentence data and the replacement data are stored as the dictionary data, and the speech recognition result is created using the sentence data and the replacement data. The text data and the replacement data are as described above.

  Here, the operation of the creating apparatus 1 according to the present embodiment will be described. The outline of the operation of the creation device 1 according to this embodiment is the same as that of the first embodiment. FIG. 12 is a flowchart showing an example of a voice recognition result creation process in this embodiment. The flowchart of FIG. 12 is obtained by adding steps S409 and S410 to the flowchart of FIG. Hereinafter, steps S409 and S410 will be described. It is assumed that the voice recognition data storage unit 12 stores the voice recognition data of FIG. 2, and the dictionary data storage unit 13 stores the sentence data of FIG. 3 and the replacement data of FIG.

  In the present embodiment, when the voice recognition result has a portion that matches the second word (YES in step S405), the search unit 14 uses the first neighborhood character string and the second neighborhood character string as search keys to retrieve the text data. A search is performed (step S409).

  The second neighboring character string is a character string including the second word and its neighboring character string in the voice recognition result. The neighborhood character string of the second word is, for example, a character string of several characters immediately before or after the second word. The first neighbor character string is a character string in which the second word/phrase included in the second neighbor character string is replaced with the corresponding first word/phrase. Here, the second neighborhood character string is assumed to be a character string consisting of the second word and one character immediately after it.

  For example, when the first word is "Ricoh", the second word is "Let's go", and the voice recognition result is "Goko's standard", the second neighborhood character string is "Gokono". , The first neighborhood character string is “Ricoh's”.

  The search unit 14 outputs the number of matches R2 between the search key (first neighborhood character string) and the text data as a search result. Since the text data in FIG. 3 has one part that matches "Ricoh's", one case is output as the search result.

  Further, the search unit 14 outputs the number of matches R2 between the search key (the second neighborhood character string) and the text data as the search result. Since the text data in FIG. 3 does not have a part that matches “Yuko”, 0 is output as the search result.

  When the number of matches R1 is larger than the number of matches R2 (YES in step S410), the creation unit 15 replaces the second word with the corresponding first word (step S406). Then, a process progresses to step S407. On the other hand, if the number of matches R1 is less than or equal to the number of matches R2 (NO in step S410), the process proceeds to step S407. In the present embodiment, the number of matching cases R1 is larger than the number of matching cases R2, and thus “let's go” is replaced with “Ricoh”, and a speech recognition result “Ricoh's standard” is created as in the third embodiment. ..

  As described above, according to the present embodiment, when the number of matches R1 of the first neighborhood character string with respect to the sentence data is larger than the number of matches R2 of the second neighborhood character string, the second word is replaced with the first word. It In other words, if the number of matches R1 of the first neighborhood character string to the text data is less than or equal to the number of matches R2 of the second neighborhood character string, the second word is not replaced with the first word. Thereby, the creation device 1 can suppress excessive substitution of the second word and create a highly accurate voice recognition result.

  It should be noted that the present invention is not limited to the configurations shown here, such as the combination of the configurations described in the above embodiments with other elements. These points can be changed without departing from the spirit of the present invention, and can be appropriately determined according to the application form.

1: Voice recognition result creation device 11: Input unit 12: Voice recognition data storage unit 13: Dictionary data storage unit 14: Search unit 15: Creation unit

JP, 2004-333703, A

Claims (8)

An input unit for inputting voice recognition data including candidate words and phrases corresponding to each segment text and evaluation values of the respective candidates,
A dictionary data storage unit for storing dictionary data prepared in advance by the user,
A search unit that searches a matching portion of the dictionary data and the voice recognition data;
A creation unit that creates a voice recognition result based on the search result and the voice recognition data;
A voice recognition result creating device including. The speech recognition result creation according to claim 1, wherein the creation unit creates the speech recognition result by selecting the candidate based on the evaluation value for each of the segment texts and connecting the selected candidates. apparatus. The dictionary data includes text data,
The search unit searches the sentence data using the candidate as a search key,
The speech recognition result creation device according to claim 1, wherein the creation unit updates the evaluation value of the candidate based on the number of matches between the candidate and the text data. The dictionary data includes homophone data including at least one homophone set including a plurality of homophones,
The voice according to claim 3, wherein the search unit searches the homonym data using the candidate as a search key, and if the homophone set including the candidate exists, searches the sentence data using the candidate as a search key. Recognition result creation device. The dictionary data includes replacement data indicating a correspondence relationship between a first word to be replaced and a second word to be replaced,
The creating unit selects the candidate based on the evaluation value for each segment text, and connects the selected candidates to create the voice recognition result,
The search unit searches the voice recognition result using the second word or phrase as a search key,
The speech recognition result creating apparatus according to claim 1, wherein, when the speech recognition result includes the second word/phrase, the creating unit replaces the second word/phrase with the corresponding first word/phrase. The dictionary data includes text data,
The search unit searches the sentence data using a first neighborhood character string including the first word and a second neighborhood character string including the second word as a search key.
When the number of matches of the first neighborhood character string and the sentence data is greater than the number of matches of the second neighborhood character string and the sentence data, the creation unit replaces the second phrase with the corresponding first phrase. The speech recognition result creation device according to claim 5. An input step of inputting speech recognition data including candidate words and phrases corresponding to each segment text and evaluation values of the respective candidates;
A search step of searching a matching portion of the dictionary data prepared in advance by the user and the voice recognition data;
A creation step of creating a voice recognition result based on the search result and the voice recognition data;
A method for creating a voice recognition result including. An input step of inputting speech recognition data including candidate words and phrases corresponding to each segment text and evaluation values of the respective candidates;
A search step of searching a matching portion of the dictionary data prepared in advance by the user and the voice recognition data;
A creation step of creating a voice recognition result based on the search result and the voice recognition data;
A program that causes a computer to execute.

Images