JP4515186B2 - Speech dictionary creation device, speech dictionary creation method, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To efficiently create a voice dictionary suitable for practical use. <P>SOLUTION: This voice dictionary creation device has a reading processing part connected to the Internet network. The reading processing part reads a site to which character input is attempted to be performed or a site linked from the site or other site which a user specifies as especially having deep relation, transfers the data to a character string extraction part. The character string extraction part analyses the transferred data as character strings to be used for voice recognition of words, etc., and transfers its result to the voice dictionary creation part. The voice dictionary creation part creates the voice dictionary by performing processing such as adjustment of use frequency of the words and addition of unknown words according to the data transferred from the character string extraction part. <P>COPYRIGHT: (C)2005,JPO&amp;NCIPI

Description

  The present invention relates to a speech dictionary creation device, a speech dictionary creation method, and a program for creating speech dictionary data used when speech is recognized and converted into a character string.

  The speech recognition device is used in part or separately with the speech dictionary creation device. This is because, for example, depending on the application, the same pronunciation “Koukai” can be “public”, “regret”, “voyage”, etc. This is because if the user side does not specify which example to use, the intended character as a result of speech recognition cannot be obtained. In some cases, such a speech dictionary is prepared in advance, but it is still necessary to operate the speech dictionary because it is difficult to obtain a desired result unless the user adjusts the priority (selection order). In addition, if an "unknown word" that is a word that is unknown to the speech recognition device is included in the first place, it will interfere with speech recognition. Therefore, such an unknown word is added by the speech dictionary creation device and the speech recognition device recognizes it. Increase the rate.

  In the case of creating such a voice dictionary, humans have conventionally registered a dictionary for each word, and it took a lot of man-hours to create a practical dictionary. Further, even when a dictionary prepared in advance is used, the priority order of the words recognized by the application is different, so that it takes a lot of labor to operate the speech dictionary by trial and error by the user. In addition, attempts have been made to increase efficiency by recognizing large, long sentences, but since the prepared text does not always approximate what the speaker intended, there are still many man-hours for correction. Was needed.

In the process of converting speech into a character string by speech recognition, if there are multiple character string candidates to be converted, an appropriate candidate is selected from a plurality of candidates based on the history of character strings input in the past A speech recognition apparatus and a speech recognition method are disclosed in Patent Document 1.
JP 2003-108189 A

  As described above, it takes a lot of man-hours to create a speech dictionary that can withstand practical use by registering the priorities for converting speech into character strings one by one by the user of the speech recognition apparatus. Further, even when a speech dictionary provided in advance in the speech dictionary creation device is used, it takes a great deal of time to adjust the priority order according to the user's application. Even when the speech recognition apparatus and speech recognition method disclosed in Patent Document 1 are used, an appropriate character string candidate cannot be selected if there is no history of character strings input in the past.

  The present invention has been devised in view of the above problems, and when a user has an Internet connection means and wants to input characters using a speech recognition apparatus in a site where characters can be input provided through the Internet connection means, The purpose is to enable the creation of a speech dictionary that can withstand practical use.

  The reading processing unit connected to the Internet network reads the site where the character input is to be performed, the site linked from the site, or the other site specified by the user as being particularly relevant, and the data is read to the character string extracting unit. Deliver. The character string extraction unit analyzes the received data as a character string used for speech recognition of a word or the like, and passes the result to the speech dictionary creation unit. The speech dictionary creation unit creates a speech dictionary by adjusting the frequency of use of words and adding unknown words according to the data passed from the character string extraction unit, and creates a speech dictionary for the speech recognition device. To do.

  Conventionally, a voice dictionary is created manually by trial and error one by one. However, in the present invention, data for creating a voice dictionary is obtained from the Internet, and the range is to be used for voice input. The difference is that the site, the link destination, or the site designated by the user as being particularly relevant is targeted for data acquisition. Also, rather than simply using a long sentence to create a speech dictionary, the scope of the data acquisition targets sites that are intended to be used for speech input, links to them, or sites that the user has specified as being particularly relevant. The point is different.

  Utilizing the nature of the Internet, a voice dictionary is created from the site or the link destination where the voice is to be input. In addition, when the amount of data is insufficient as described above, a user can arbitrarily add a site that seems to be closely related to the read processing unit. By such means, means for efficiently creating a speech dictionary is provided.

  There is a character input site that accepts character input on the Internet. An Internet connection device connected to the Internet network is connected to this character input site, and a voice recognition device is added to the Internet connection device. When trying to send a voice-recognized character to a site, the user may consider that the reading processing unit connected to the Internet network attempts to input the character, the site linked from that site, or other particularly relevant The specified site is read and the data is passed to the character string extraction unit. The character string extraction unit analyzes the received data as a character string used for speech recognition of words and the like, and the result is a voice dictionary creation unit. The voice dictionary creation unit uses words according to the data received from the character string extraction unit. Create a speech dictionary by adjusting the degree or adding unknown words to create a speech dictionary for the speech recognition device. The speech recognition device uses this speech dictionary for efficient speech recognition processing. Perform character input processing.

  The speech dictionary creation device of the present invention includes a CPU, a memory, a communication interface for connecting to the Internet, and a first speech dictionary storage unit and a second speech dictionary storage unit which are storage areas provided in the memory. A read URL storage unit, an appearance frequency storage unit, and a read processing unit, a character string extraction unit, and a speech dictionary creation unit realized by the CPU executing a program stored in the memory. The first voice dictionary storage unit stores character string data indicating a character string such as a word and reading data indicating a reading of the character string in association with each other, and the reading URL storage unit stores The URL of the Web page is stored, and the read processing unit accesses the Web page specified by the URL stored in the read URL storage unit to a Web server on the Internet. The character string extraction unit extracts the character string data included in the data constituting the Web page acquired by the read processing unit, and appears for each character string data. The frequency is obtained, the character string data and the appearance frequency data which is data indicating the appearance frequency are associated with each other and stored in the appearance frequency storage unit, and the speech dictionary creation unit is stored in the appearance frequency storage unit The reading data corresponding to the character string data is acquired from the first speech dictionary storage unit, and the acquired reading data is obtained by speech recognition for each of the plurality of character string data having the same reading data. The selection order of the character string data when converting the data into one character string data selected from the plurality of character string data is determined based on the appearance frequency data The character string data, the reading data, and the selection order data that is the data indicating the selection order are associated with each other to generate speech dictionary data that is data used when the speech is recognized and converted into a character string. The speech dictionary data is stored in the second speech dictionary storage unit.

  The program of the present invention includes a CPU, a memory, a communication interface for connecting to the Internet, character string data indicating a character string such as a word, which is a storage area provided in the memory, Stored in the read URL storage unit in a computer comprising a first voice dictionary storage unit storing reading data indicating readings in association with each other and a read URL storage unit storing a URL of a Web page A procedure for acquiring a Web page specified by the URL that is specified by accessing a Web server on the Internet, and extracting the character string data included in the acquired Web page to extract the character string data Associating the procedure for obtaining the appearance frequency for each character with the character string data and the appearance frequency data that is the data indicating the appearance frequency The procedure of storing in the appearance frequency storage unit, which is a storage area provided in the memory, and the reading data corresponding to the character string data stored in the appearance frequency storage unit from the first speech dictionary storage unit One character selected from among the plurality of character string data for the reading data obtained by voice recognition for each of the plurality of character string data having the same reading data and the acquired reading data Corresponds to the procedure for determining the selection order of the character string data when converting to column data based on the appearance frequency data, and the selection order data that is the data indicating the character string data, the reading data, and the selection order In addition, voice dictionary data, which is data used when recognizing voice and converting it into a character string, is generated, and the voice dictionary data is stored in the memory. Is intended for executing the instructions stored in the second speech dictionary storage unit is a region.

  With the development of speech recognition technology, speech is converted into a character string using a speech recognition program. For example, it is used for character input to a Web page that requires character input such as a daily business report, or creation of document data such as academic papers. In such a speech recognition program, a speech dictionary for converting speech into a character string is used. For example, a standard speech dictionary (hereinafter referred to as “standard dictionary”, which corresponds to the “first speech dictionary” of the present invention) included in the speech recognition program is “ Character strings such as “regret”, “public”, and “voyage” are stored together with the priority order to be converted. The voice recognition program determines a character string corresponding to the recognized voice according to the priority order of the standard dictionary.

  Therefore, the user of the voice recognition program needs to adjust the priority order set in the standard dictionary so that the voice is converted into the character intended by the user. For example, if the target is the marine field, work such as raising the priority of “voyage” is required.

  The program of the present invention is installed and used in a computer (voice dictionary creation device) together with a voice recognition program. If such a speech dictionary creation device is used, the URL of a web page of a site related to the field in which character recognition is performed by voice recognition is simply stored in the read URL storage unit. A speech dictionary (hereinafter referred to as “user dictionary”, which corresponds to the “second speech dictionary” of the present invention) in which priorities are set in accordance with the appearance frequency of the character string in FIG.

  Note that the user dictionary may be created as a new dictionary different from the standard dictionary, or may be created by updating the standard dictionary. Further, the first speech dictionary is not limited to the standard dictionary of the speech recognition program, but may be any one as long as a character string and a reading are associated with each other. For example, a kana-kanji conversion dictionary provided in the kana-kanji conversion program can be used as the first speech dictionary. The generated user dictionary can be used not only for character conversion in speech recognition but also for kana-kanji conversion.

  Furthermore, the program according to the present invention is configured to store the character string data that is not stored in the first phonetic dictionary storage unit among the character string data stored in the appearance frequency storage unit. Receiving the reading data indicating the reading of the character string represented by the procedure of outputting to the output device and the character string data output from the predetermined input device included in the computer and output to the output device; And a procedure for associating the column data and the reading data with each other and storing them in the reading data storage unit that is a storage area provided in the memory, and corresponding to the character string data stored in the appearance frequency storage unit The step of acquiring the reading data includes the reading data corresponding to the character string data stored in the appearance frequency storage unit. Of it may be a procedure for acquiring a speech dictionary storage unit or the read data storage unit.

  If there is a character string that is not registered in the standard dictionary (hereinafter referred to as “unregistered word”) in the acquired Web page, it is necessary to register the reading of the character string. Therefore, the unregistered word is output to an output device such as a display by the speech dictionary creation device, and the user inputs reading of the unregistered word from an input device such as a keyboard or a microphone. Then, the speech dictionary creation device creates a user dictionary in which the priority order of homonyms including unregistered words is set. In this way, even in a new technical field where there are many unregistered words, it is possible to obtain a user dictionary with a priority set according to the appearance frequency of the character string, and to convert the speech into a character string intended Can increase the probability of being able to.

  The program of the present invention is stored in a URL storage unit, which is a storage area for storing a URL of a Web page designated by a user by a Web browser, which is software that displays an image, a character string, or the like corresponding to the Web page on a display. It is also possible to further include a procedure for acquiring the URL that has been stored and a procedure for storing the URL acquired from the URL storage unit in the read URL storage unit.

  In general, a Web browser includes a storage unit (hereinafter referred to as “URL storage unit”) that stores a URL of a specific Web page as designated by a user. In many cases, a user registers a URL of a Web page related to his / her work or research in the URL storage unit. Therefore, by storing the URL stored in the URL storage unit in the read URL storage unit according to the present invention, the user does not register a separate URL for creating a speech dictionary, but in a field related to business or research. It is possible to obtain a speech dictionary in which priorities are set according to the appearance frequency of character strings on a Web page.

  Furthermore, the program of the present invention further includes a procedure of accepting the URL specified by the user among the plurality of URLs stored in the read URL storage unit, and acquires the URL stored in the URL storage unit The procedure to be performed may be a procedure for acquiring, from the URL storage unit, the URL specified by a user among the URLs stored in the URL storage unit.

  There are cases where URLs of Web pages in various fields are registered in the URL storage unit. Therefore, the user designates the URL of the Web page that matches the field in which the user is trying to input characters from the URLs registered in the URL storage unit. When the user has a folder for each field in the URL storage unit, a plurality of URLs stored in the folder can be specified by selecting the folder.

  Thus, even when URLs of Web pages in various fields are registered in the URL storage unit provided in the Web browser, a speech dictionary in which priorities are set according to the fields in which the user is trying to input characters. Can be obtained. In addition, when the user organizes URLs by field and registers them in the URL storage unit, a speech dictionary corresponding to the target field can be easily obtained.

  In addition, link hierarchy data indicating the number of hierarchies that follow links to other Web pages set in the Web page is stored in a link hierarchy storage unit that is a storage area provided in the memory, and stored in the read URL storage unit In the procedure for acquiring the Web page specified by the URL that is specified, a link destination that is a URL indicating a link to another Web page included in the data constituting the acquired Web page The process of extracting the URL and acquiring the other Web page specified by the link destination URL is repeatedly executed according to the number of levels indicated by the link level data stored in the link level storage unit. Procedures may be included.

  In many cases, a web page link related to the web page is set in the web page. Therefore, when acquiring the Web page indicated by the URL stored in the read URL storage unit, the speech dictionary creating apparatus acquires another Web page by following the link set for the Web page. Here, if the link is traced infinitely, Web pages in all fields are acquired, and a speech dictionary corresponding to the target field cannot be created. Conversely, if the number of layers following the link is reduced, the number of character strings that can be acquired is limited, and a practical speech dictionary cannot be created. Therefore, when the user sets the number of layers following the link in the link layer storage unit, a speech dictionary corresponding to the field intended by the user can be created.

  Further, the link hierarchy storage unit stores the link hierarchy data for each URL of the Web page stored in the read URL storage unit, and the process of acquiring the other Web page is performed according to the number of layers. The procedure of repeatedly executing the process according to the process repeats the process of acquiring the other Web page according to the number of levels indicated by the link level data corresponding to the URL stored in the URL storage unit. It may be a procedure to be executed.

  The number of hierarchies to follow the link can be designated according to the characteristics of the Web page indicated by the URL stored in the read URL storage unit. For example, in the case of a web page of a glossary of a certain field, there are cases where the web page includes a large number of character strings in the field targeted by the user, and there is no need to follow the link further. On the other hand, for example, when it is a Web page (top page) at the entrance of a general site in a certain field, the user can obtain a character string in the target field by following the link set in the Web page. Can do. In this way, by making it possible to specify the number of layers following the link for each Web page indicated by the URL stored in the read URL storage unit, it is possible to create a speech dictionary that further matches the target field. .

  In addition, the program of the present invention acquires a URL input to an area for inputting a URL of a Web page to be displayed provided in software for displaying the Web page, and stores the acquired URL in the read URL storage unit. Further, it may be provided.

  For example, when a Web page on which a user is trying to input characters using voice recognition is an input page for a daily business report, the character string that the user is trying to input is a past text included in the Web page. It is often included in business daily reports. In such a case, the URL input in the URL input area of the Web browser is stored in the read URL storage unit, so that when the user tries to input characters, the URL to be read is not specified separately. A voice dictionary can be created according to

  Rather than creating a speech dictionary that can withstand practical use by performing trial and error on a case-by-case basis, a speech dictionary that can withstand practical use can be created efficiently. In addition, there has been a means for creating a speech dictionary based on an appropriate long sentence, but it is possible to create a speech dictionary with higher accuracy by utilizing the nature of the Internet. Furthermore, since a quick voice dictionary can be created, a voice recognition function can be easily added to an existing site.

  Thus, according to the present invention, it is possible to provide a speech dictionary creation device, a speech dictionary creation method, and a program that efficiently create a speech dictionary that can be practically used.

=== First Embodiment ===
FIG. 1 is a configuration diagram of an embodiment of the present invention, where there is a character input site that accepts character input on the Internet, and an Internet connection device connected to the Internet network is connected to the character input site. A site where a speech recognition device is added to the Internet connection device so that the speech-recognized characters can be sent to the character input site, and the site where the reading processing unit connected to the Internet network attempts to input characters or the site Sites linked from or other sites that are specified by the user as particularly relevant are read, and the data is passed to the character string extractor, which uses the received data for speech recognition of words, etc. And the result is passed to the speech dictionary creation unit. The speech dictionary creation unit Characters by efficient speech recognition by creating a speech dictionary by adjusting the frequency of use of words or adding unknown words according to the passed data and creating a speech dictionary of the speech recognition device Input processing is performed.

  In general, the problem that arises when trying to input speech is that the usage of frequently used words, industry terms, idioms, etc. differ from site to site. However, these words are mostly written on the site where the voice input is to be made or the linked site. For example, in the marine industry, “Koukai” should be used as “voyage” and “high seas” and there should be fewer examples such as “open” and “renewal”. Therefore, if it is “Like”, it will be “Navigating” and “Going to sea” will be strange, so it can be judged from the analysis of the dependency that the person who has a high probability of “Navigating” is correct speech recognition. In addition, due to the nature of the Internet, if it is a site where characters are already input or a linked site, the contents related to the ship industry should be described, and the above example can be easily found. In addition, if there is a famous site in the ship industry, such a site is also useful for creating a speech dictionary, so that data should be acquired as necessary. In addition, in a site such as a bulletin board, there should be many cases where the content is written in the form of receiving the previous writing of the person trying to write, and the words necessary for speech recognition are already already in the bulletin board site. It seems to have appeared in. Utilizing such properties of the Internet itself, the reading processing unit acquires data from the Internet, and as a result, an efficient speech dictionary can be created.

=== Second Embodiment ===
== Overall structure ==
Next, a second embodiment of the present invention will be described. FIG. 2 is a diagram illustrating a network configuration according to the second embodiment. The speech dictionary creation device 1 is a device that executes the speech dictionary creation program according to the present invention, and is a computer such as a personal computer or a dedicated device used for speech recognition. The speech dictionary creation device 1 can communicate with a plurality of Web servers 3a, 3b, 3c via a network 2 such as the Internet or an intranet.

  The Web server 3a stores a Web page 4a that is data described in a markup language such as HTML (Hyper Text Markup Language) or XML (eXtensible Markup Language). Then, the speech dictionary creation device 1 can acquire the Web page 4a by accessing the Web server 3a via the network 2 in accordance with HTTP (Hyper Text Transfer Protocol). FIG. 3 shows an example of the Web page 4a. The Web page 4a includes tags such as “<HTML>” and “<TITLE>”, document data 5 that is a document displayed on the Web browser, and the like. Further, in the Web page 4a, a link 6 to another Web page is set using a tag "<A HREF="@@@"> </A>". The part “@@@” indicates the URL of the linked web page.

  A web browser is software that displays a web page. FIG. 4 is a diagram showing the web browser 7 displaying the web page 4a. The Web browser 7 is provided with a URL input area 8 for inputting the URL of a Web page to be displayed and a display area 9 for displaying document data 5 and image data. When the URL of the Web page 4a is input to the URL input area 8, the speech dictionary creation device 1 acquires the Web page 4a from the Web server 3a and displays the document data 5 and the like included in the Web page 4a in the display area 9. . A link 6 is also displayed in the display area 9, and when the user clicks on the link 6, the speech dictionary creation device 1 acquires a linked Web page.

  In this way, a plurality of Web pages can be acquired by following the links set in the Web pages. In the example of FIG. 2, a link to the Web page 4b stored in the Web server 3b is set in the Web page 4a, and a link to the Web page 4c stored in the Web server 3c is set in the Web page 4b. It shows that.

== Hardware configuration ==
FIG. 5 is a diagram illustrating a hardware configuration of the speech dictionary creation device 1. The speech dictionary creation device 1 includes a CPU 10, a memory 11, a storage device 12, a recording medium reading device 13, a communication interface 14, an input device 15, an output device 16, and a microphone 17.

  The CPU 10 controls the entire speech dictionary creation device 1, and can sequentially store programs stored in the storage device 12 in the memory 11 and execute the programs stored in the memory 11. The storage device 12 is a hard disk drive or the like, and stores programs and data. The recording medium reading device 13 can read a program recorded on a recording medium 18 such as a CD-ROM and store it in the storage device 12.

  The communication interface 14 is an interface for transmitting / receiving data to / from the Web servers 3a, 3b, 3c via the network 2. The input device 15 is a device that receives data input from a user, and is, for example, a keyboard or a mouse. The output device 16 is a device that outputs a processing result of the CPU 10, and is, for example, a display or a printer. The microphone 17 is a device that acquires sound generated by the user's utterance. The input device 15, the output device 16, and the microphone 17 may be configured separately from the speech dictionary creation device 1 and connected to the speech dictionary creation device 1 using a cable or the like.

== Functional configuration ==
FIG. 6 is a diagram illustrating a configuration of functions provided in the speech dictionary creation device 1. The speech dictionary creation device 1 includes a speech recognition function unit 20, a speech dictionary creation function unit 30, and a storage unit 40.

  The voice recognition function unit 20 is realized by the CPU 10 executing a voice recognition program stored in the memory 11 of the voice dictionary creation device 1. The voice acquisition unit 21, the voice recognition unit 22, and the character string conversion unit 23 are realized. It has.

  The speech dictionary creation function unit 30 is realized by the CPU 10 executing the speech dictionary creation program stored in the memory 11 of the speech dictionary creation device 1. The speech processing unit 31, the character string extraction unit 32, A registered word output unit 33, a reading data reception unit 34, a speech dictionary creation unit 35, and an input URL acquisition unit 36 are provided.

  The storage unit 40 is a storage area provided in the memory 11, and includes a standard dictionary storage unit (corresponding to the “first speech dictionary storage unit” of the present invention) 41, a user dictionary storage unit (“ Corresponding to the “second voice dictionary storage unit”.

  The voice acquisition unit 21 acquires voice data input from the microphone 17. The voice recognition unit 22 recognizes a character indicated by the voice data acquired by the voice acquisition unit 21 and generates character data. The character string conversion unit 23 generates a hiragana character string from a plurality of character data generated by the speech recognition unit 22 and converts the hiragana character string into a character string such as kanji or katakana.

  The character string conversion unit 23 can use a standard dictionary stored in the standard dictionary storage unit 41 or a user dictionary stored in the user dictionary storage unit 42 when converting a character string. A standard dictionary is a basic speech dictionary created when a speech recognition program is installed. The user dictionary is selected when a user adds or changes a character string or when there are a plurality of candidate kanji or katakana character strings corresponding to a hiragana character string. This is a speech dictionary that can be changed in order (priority order). It is also possible to create a user dictionary by overwriting and updating the standard dictionary.

  FIG. 7 is a diagram showing the structure of the standard dictionary storage unit 41 and the user dictionary storage unit 42. The standard dictionary storage unit 41 and the user dictionary storage unit 42 store voice dictionary data, which is data in which character string data, reading data, and priority order data are associated with each other. A character string of kanji or katakana is set in the character string data, and a hiragana character string indicating the reading of the character string is set in the reading data. In the priority order data, data indicating the priority order when the character string conversion unit 23 converts a hiragana character string into a kanji or katakana character string is set.

  In the example shown in FIG. 7, the priority of “voyage”, “public”, and “regret” is “1”, “2”, and “3” in order with respect to the reading “Koukai”. In this case, the character string conversion unit 23 converts the character string of hiragana “Koukai” into the character string of kanji “voyage” having the highest priority.

  The read processing unit 31 acquires the URL of the Web page stored in the read URL storage unit 43, and acquires the Web page indicated by the URL from the Web server. FIG. 8 is a diagram illustrating the structure of the read URL storage unit 43. The read URL storage unit 43 stores the URL of the Web page designated by the user. The URL stored in the read URL storage unit 43 may be one or plural.

  Further, the read processing unit 31 acquires a link set in the acquired Web page, and acquires another Web page indicated by the URL set in the link. In this manner, the read processing unit 31 acquires a plurality of Web pages by following the links set in the Web page. FIG. 9 is a diagram illustrating the structure of the link hierarchy storage unit 44. In the link hierarchy storage unit 44, for each URL stored in the read URL storage unit 43, link hierarchy data, which is data indicating the number of layers following the link, is set. Note that the link hierarchy storage unit 44 may store only one link hierarchy data common to all URLs stored in the read URL storage unit 43.

  The character string extraction unit 32 extracts character string data decomposed into meaningful units from document data obtained by removing tags from the data constituting the Web page acquired by the read processing unit 31. As a method for extracting character string data from document data, for example, a word division method disclosed in Japanese Patent Laid-Open No. 2000-259673 can be used. Then, the character string extraction unit 32 counts the number of appearances for each character string data, and stores the character string data in the appearance frequency storage unit 45 in association with appearance frequency data that is data indicating the number of appearances. FIG. 10 is a diagram illustrating the structure of the appearance frequency storage unit 45.

  The unregistered word output unit 33 outputs character string data that is not registered in the standard dictionary among the character string data stored in the appearance frequency storage unit 45 to the output device 16 such as a display. The reading data receiving unit 34 receives reading data indicating reading of character string data output to the output device 16 input from the input device 15 or the microphone 17 by the user, and associates the character string data with the reading data. At the same time, it is stored in the reading data storage unit 46. FIG. 11 is a diagram illustrating the structure of the reading data storage unit 46.

  The speech dictionary creation unit 35 acquires reading data corresponding to the character string data stored in the appearance frequency storage unit 45 from the standard dictionary storage unit 41 or the reading data storage unit 46, and character string data having the same reading data. The priority order is determined based on the appearance frequency data. That is, the speech dictionary creation unit 35 sets a higher priority for character string data having the same reading data in descending order of the number of appearances. Then, the speech dictionary creation unit 35 generates speech dictionary data by associating the character string data, the reading data, and the priority order data that is the data indicating the hot water ranking, and stores the speech dictionary data in the user dictionary storage unit 42. To do.

  The input URL acquisition unit 36 acquires a URL input to the URL input area 8 of the Web browser 7 and stores the URL in the read URL storage unit 43. That is, the input URL acquisition unit 36 is used when the priority order of the character strings is set based on the Web page that the user is trying to display on the Web browser 7.

== Voice dictionary creation process ==
Next, the speech dictionary creation process in the speech dictionary creation device 1 will be described. FIG. 12 is a flowchart of the voice dictionary creation process. First, the read processing unit 31 acquires one URL stored in the read URL storage unit 43 (S1201). Further, the read processing unit 31 acquires the link hierarchy data corresponding to the URL from the link hierarchy storage unit 44 (S1202), and sets 0 in the work hierarchy data that is work data for counting the link hierarchy. (S1203).

  Subsequently, the read processing unit 31 acquires the Web page indicated by the acquired URL from the Web servers 3a, 3b, and 3c (S1204). Then, the read processing unit 31 confirms whether the work hierarchy data is equal to the link hierarchy data (S1205).

  When the work hierarchy data is not equal to the link hierarchy data (S1205: No), the read processing unit 31 acquires a link destination URL included in the data constituting the acquired Web page (S1206), and the link destination URL indicates A web page is acquired (S1207). Then, the read processing unit 31 adds 1 to the work hierarchy data (S1208), and returns to the step of confirming whether the work hierarchy data is equal to the link hierarchy data (S1205). In other words, the read processing unit 31 repeats the processes of S1205 to S1208, so that a Web page within the link hierarchy indicated by the link hierarchy data is displayed from the Web page indicated by the URL stored in the read URL storage unit 43. get.

  When the work hierarchy data becomes equal to the link hierarchy data (S1205: Yes), the read processing unit 31 checks whether there is an unacquired URL in the read URL storage unit 43 (S1209). While there is an unacquired URL (S1209: Yes), the read processing unit 31 acquires one unacquired URL and repeatedly executes a series of processes (S1201 to S1208) for acquiring a Web page.

  When all the URLs in the read URL storage unit are acquired by the read processing unit 31 (S1209: No), the character string extraction unit 32 converts the character string data included in the data constituting the Web page acquired by the read processing unit 31. Extract (S1210). Then, the character string extraction unit 32 counts the number of appearances for each extracted character string data, and stores the character string data and the appearance frequency data indicating the number of appearances in association in the appearance frequency storage unit 45 (S1211).

  Subsequently, the unregistered word output unit 33 extracts character string data not stored in the standard dictionary storage unit 41 from the character string data stored in the appearance frequency storage unit 45, and displays the character string data on the display. To the output device 16 (S1212). The reading data receiving unit 34 receives reading data for the character string data input from the input device 15 such as a keyboard or the microphone 17 by the user, and stores the reading data in association with the character string data and the reading data. The information is stored in the unit 46 (S1213).

  Next, the speech dictionary creation unit 35 acquires reading data corresponding to the character string data stored in the appearance frequency storage unit 45 from the standard dictionary storage unit 41 or the reading data storage unit 46 (S1214). Then, the voice dictionary creation unit 35 determines the priority order for a plurality of character string data having the same reading data based on the appearance frequency data stored in the appearance frequency storage unit 45 (S1215). Finally, the speech dictionary creation unit 35 generates speech dictionary data by associating the character string data, the reading data, and the priority order data that indicates the priority order, and stores the speech dictionary data in the user dictionary storage unit 42. Store (S1216).

  Here, an example of a method for determining the priority order based on the appearance frequency data will be described. FIG. 13 is a diagram illustrating a first example of a method for determining a priority order. First, the speech dictionary creation unit 35 stores, in the user dictionary storage unit 42, speech dictionary data in which the priority order of each character string data is set in descending order of the appearance frequency data stored in the appearance frequency storage unit 45 (S1301). . Then, the voice dictionary creation unit 35, for the character string data that does not exist in the appearance frequency storage unit 45 among the character string data stored in the standard dictionary storage unit 41, is later than the character string data stored in S1301. Priorities are set and stored in the user dictionary storage unit 42 (S1302).

  FIG. 14 is a diagram illustrating a second example of the method for determining the priority order. First, the speech dictionary creation unit 35 stores the speech dictionary data in which the priority order of the character string data stored in the standard dictionary storage unit 41 is increased by the number of appearances stored in the appearance frequency storage unit 45. Store in the unit 42 (S1401). In the example of FIG. 14, since the number of appearances of “voyage” is two, the priority of “voyage” in the standard dictionary increases by “2” from “3” to “1”. Accordingly, the priority order of “regret” and “public” is lowered by one.

  Next, the speech dictionary creation unit 35 stores the speech dictionary data in which the lowest priority is set for unregistered words in the user dictionary storage unit 42 (S1402). In the example of FIG. 14, the lowest priority “4” is set for the unregistered word “high seas” and added to the user dictionary. Then, the speech dictionary creation unit 35 increases the priority of the added unregistered word by the number of appearances stored in the appearance frequency storage unit 45 (S1403). In the example of FIG. 14, since the appearance frequency of “high seas” is one, the priority of “high seas” increases by “1” from “4” to “3”. Then, the priority of “public” is lowered from “3” to “4”.

  As illustrated in FIGS. 13 and 14, the speech dictionary creation unit 35 is based on the appearance frequency data stored in the appearance frequency storage unit 45 for the priority order data for each of a plurality of character string data having the same reading data. The voice dictionary data that is generated and associated with the character string data, the reading data, and the priority order data can be stored in the user dictionary storage unit 42.

== Use URL storage part of Web browser ==
Next, a case where the URL storage unit included in the Web browser 7 is used as a URL to be read will be described. FIG. 15 is a diagram showing the registration URL menu 50 of the Web browser 7. When there is a web page that the user likes, when the web page is displayed on the web browser 7, the user selects the add button 51 from the registration URL menu 50. Then, the title 52 of the Web page is added to the registration URL menu 50. The user can also create a folder 53 in the registration URL menu 50 and add a Web page title 54 to the folder 53.

  FIG. 16 is a diagram showing the structure of the URL storage unit 55 that stores the URL of the Web page registered by the user using the registration URL menu 50. The URL storage unit 55 is a storage area provided in the memory 11. The URL storage unit 55 stores a folder 53, Web page titles 52 and 54, and Web page URLs in association with each other.

  The read processing unit 31 can acquire the URL stored in the URL storage unit 55 and store it in the read URL storage unit 43. Further, the read processing unit 31 receives the titles 52 and 54 and the folder 53 of the Web page registered in the registration URL menu 50 designated by the user, and reads the URLs corresponding to the titles 52 and 54 and the folder 53 by reading the URL. It can also be stored in the unit 43.

== Use URL entered by Web browser ==
Next, a case where the URL input in the URL input area 8 of the Web browser 7 is used as the URL to be read will be described. FIG. 17 is a flowchart showing the processing of the input URL acquisition unit 36. First, the input URL acquisition unit 36 acquires a URL input by the user to the URL input area 8 (S1701). Then, the input URL acquisition unit 36 stores the acquired URL in the read URL storage unit 43 (S1702), and starts the speech dictionary creation process shown in FIG. 12 (S1703). As a result, a user dictionary is generated based on the Web page indicated by the URL input in the URL input area 8 and other Web pages acquired by following links set in the Web page.

  The operation of the speech dictionary creation device 1 that executes the speech dictionary creation program has been described above. As described above, if the speech dictionary creation device 1 in which the speech dictionary creation program of the present invention is installed is used, the URL of a Web page of a site related to the field in which speech recognition is to be performed is read and URL storage is performed. By simply storing it in the unit 43, it is possible to obtain a user dictionary in which priorities are set in accordance with the appearance frequency of character strings on the Web page. The generated user dictionary can be used not only for character conversion in speech recognition but also for character conversion in kana-kanji conversion.

  In addition, the voice dictionary creation device 1 outputs unregistered words to the output device 16 such as a display. Then, the user inputs reading of the unregistered word from an input device such as a keyboard or a microphone. The speech dictionary creation device 1 generates a user dictionary in which priorities of homophones including unregistered words are set. In this way, even when there are unregistered words, it is possible to obtain a speech dictionary in which priorities according to the appearance frequency of character strings are set.

  The voice dictionary creation device 1 can also acquire a URL from the URL storage unit 55 in which a URL registered by the user from the registration URL menu 50 of the Web browser 7 is stored and store it in the read URL storage unit 43. In many cases, the user registers the URL of a Web page related to his / her work or research in the URL storage unit 55. Therefore, by storing the URL stored in the URL storage unit 55 in the read URL storage unit, the user does not register a separate URL for creating a speech dictionary, and the web page in a field related to business or research. A speech dictionary in which priorities are set in accordance with the appearance frequency of the character string can be obtained.

  In this case, URLs of Web pages in various fields may be registered in the URL storage unit 55. Therefore, the speech dictionary creation device 1 can store only the URL designated by the user from the URLs registered in the URL storage unit 55 in the read URL storage unit 43. Note that, when a folder for each field is provided in the URL storage unit 55, the user can designate a plurality of URLs stored in the folder by selecting the folder.

  As described above, even when URLs of Web pages in various fields are registered in the URL storage unit 55 provided in the Web browser, the voice in which the priority order is set according to the field in which the user intends to input characters. You can get a dictionary. In addition, when the user organizes URLs by field and registers them in the URL storage unit 55, a speech dictionary corresponding to the target field can be easily obtained.

  In addition, as a feature of a Web page, a link to another related Web page is often set in the Web page. Therefore, when acquiring the Web page indicated by the URL stored in the read URL storage unit 43, the voice dictionary creating apparatus 1 acquires another Web page by following the link set for the Web page. Here, if the link is traced infinitely, Web pages in all fields are acquired, and a speech dictionary corresponding to the target field cannot be created. Conversely, if the number of layers following the link is reduced, the number of character strings that can be acquired is limited, and a practical speech dictionary cannot be created. Therefore, when the user sets the number of hierarchies following the link in the link hierarchy storage unit 44, an audio dictionary corresponding to the field intended by the user can be created.

  In this case, it is possible to specify the number of hierarchies to follow the link according to the characteristics of the Web page indicated by the URL stored in the read URL storage unit 43. For example, when a web page is a glossary web page in a certain field, the web page includes a large number of character strings in the field targeted by the user, and there is a case where it is not necessary to follow the link further. is there. On the other hand, for example, when it is a Web page (top page) at the entrance of a general site in a certain field, the user can obtain a character string in the target field by following the link set in the Web page. Can do. As described above, for each Web page indicated by the URL stored in the read URL storage unit 43, it is possible to specify the number of layers to follow the link, thereby creating a speech dictionary that further matches the target field. it can.

  The voice dictionary creation device 1 can also acquire the URL input in the URL input area 8 of the Web browser 7 and store the acquired URL in the read URL storage unit 43. For example, when a Web page on which a user is trying to input characters using voice recognition is an input page for a daily business report, the character string that the user is trying to input is a past text included in the Web page. It is often included in business daily reports. In such a case, by storing the URL input in the URL input area 8 of the Web browser 7 in the read URL storage unit 43, when the user tries to input characters, the URL to be read is specified separately. The voice dictionary according to the purpose can be created.

  Although the embodiment of the present invention has been described above, the above description of the embodiment is intended to facilitate understanding of the present invention and is not intended to limit the present invention. It goes without saying that the present invention can be changed and improved without departing from the gist thereof, and that the present invention includes equivalents thereof.

It is a block diagram of the Example of this invention. It is a figure which shows the network structure in 2nd embodiment of this invention. It is a figure which shows an example of a web page. It is a figure which shows the web browser which displayed the web page. It is a figure which shows the hardware constitutions of an audio dictionary creation apparatus. It is a figure which shows the structure of the function with which an audio dictionary creation apparatus is provided. It is a figure which shows the structure of a standard dictionary memory | storage part and a user dictionary memory | storage part. It is a figure which shows the structure of a reading URL memory | storage part. It is a figure which shows the structure of a link hierarchy memory | storage part. It is a figure which shows the structure of an appearance frequency memory | storage part. It is a figure which shows the structure of a reading data storage part. It is a flowchart of an audio dictionary creation process. It is a figure which shows the 1st example of the method of determining a priority. It is a figure which shows the 2nd example of the method of determining a priority. It is the figure which displayed the registration URL menu of the Web browser. It is a figure which shows the structure of the URL memory | storage part which memorize | stored URL of the web page which the user registered using the registration URL menu. It is a flowchart which shows the process of an input URL acquisition part.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Speech dictionary creation apparatus 2 Network 3a, 3b, 3c Web server 4a, 4b, 4c Web page 5 Document data 6 Link 7 Web browser 8 URL input area 9 Display area 10 CPU
DESCRIPTION OF SYMBOLS 11 Memory 12 Storage device 13 Recording medium reader 14 Communication interface 15 Input device 16 Output device 17 Microphone 18 Recording medium 20 Voice recognition function part 21 Voice acquisition part 22 Voice recognition part 23 Character string conversion part 30 Voice dictionary creation function part 31 Reading Processing unit 32 Character string extraction unit 33 Unregistered word output unit 34 Reading data reception unit 35 Speech dictionary creation unit 36 Input URL acquisition unit 40 Storage unit 41 Standard dictionary storage unit 42 User dictionary storage unit 43 Read URL storage unit 44 Link hierarchy storage Unit 45 Appearance frequency storage unit 46 Reading data storage unit 50 Registration URL menu 51 Add button 52, 54 Web page title 53 Folder 55 URL storage unit

Claims (10)

A communication interface for connecting to a CPU, memory, and the Internet;
A first voice dictionary storage unit, a second voice dictionary storage unit, a read URL storage unit, and an appearance frequency storage unit, which are storage areas provided in the memory;
A read processing unit, a character string extraction unit, and a speech dictionary creation unit, which are realized by the CPU executing a program stored in the memory;
With
In the first speech dictionary storage unit, character string data indicating a character string such as a word and reading data indicating a reading of the character string are stored in association with each other,
The read URL storage unit stores a URL of a Web page,
The read processing unit acquires a Web page specified by the URL stored in the read URL storage unit by accessing a Web server on the Internet,
The character string extraction unit
Extracting the character string data included in the data constituting the web page acquired by the read processing unit to obtain the appearance frequency for each character string data;
Storing the character string data and the appearance frequency data, which is data indicating the appearance frequency, in association with each other in the appearance frequency storage unit;
The voice dictionary creation unit
Obtaining the reading data corresponding to the character string data stored in the appearance frequency storage unit from the first speech dictionary storage unit;
For each of the plurality of character string data having the same acquired reading data, the reading data obtained by voice recognition is converted into one character string data selected from the plurality of character string data. The selection order of the character string data at the time is determined based on the appearance frequency data,
Associating the character string data, the reading data, and selection order data that is data indicating the selection order, and generating speech dictionary data that is data used when recognizing speech and converting it to a character string; Storing the voice dictionary data in the second voice dictionary storage unit;
Voice dictionary creation device characterized by. A communication interface for connecting to a CPU, memory, and the Internet;
A storage area provided in the memory;
A first voice dictionary storage unit in which character string data indicating a character string such as a word and reading data indicating a reading of the character string are stored in association with each other;
A read URL storage unit in which the URL of the Web page is stored;
On a computer with
Obtaining a Web page specified by the URL stored in the read URL storage unit by accessing a Web server on the Internet;
A procedure for extracting the character string data included in the acquired web page and obtaining an appearance frequency for each character string data;
A procedure for associating the character string data and appearance frequency data that is data indicating the appearance frequency and storing them in an appearance frequency storage unit that is a storage area provided in the memory;
A procedure for acquiring the reading data corresponding to the character string data stored in the appearance frequency storage unit from the first speech dictionary storage unit;
For each of the plurality of character string data having the same acquired reading data, the reading data obtained by voice recognition is converted into one character string data selected from the plurality of character string data. A procedure for determining the selection order of the character string data at the time based on the appearance frequency data;
Associating the character string data, the reading data, and selection order data that is data indicating the selection order, and generating speech dictionary data that is data used when recognizing speech and converting it to a character string; A procedure for storing the speech dictionary data in a second speech dictionary storage unit that is a storage area provided in the memory;
A program for running The program according to claim 2,
Among the character string data stored in the appearance frequency storage unit, a procedure for outputting the character string data not stored in the first speech dictionary storage unit to a predetermined output device such as a display or a printer;
The reading data indicating the reading of the character string represented by the character string data output to the output device, which is input from a predetermined input device included in the computer, is received, and the character string data and the reading data are associated with each other. In addition, a procedure for storing in a reading data storage unit which is a storage area provided in the memory;
Further comprising
The procedure of acquiring the reading data corresponding to the character string data stored in the appearance frequency storage unit includes the reading data corresponding to the character string data stored in the appearance frequency storage unit. A procedure to obtain from one voice dictionary storage unit or the reading data storage unit,
A program characterized by A CPU, a memory, an input device, an output device, and a communication interface connected to the Internet;
A storage area provided in the memory;
A first voice dictionary storage unit in which character string data indicating a character string such as a word and reading data indicating a reading of the character string are stored in association with each other;
A read URL storage unit in which the URL of the Web page is stored;
On a computer with
Obtaining a Web page specified by the URL stored in the read URL storage unit by accessing a Web server on the Internet;
A procedure for extracting the character string data included in the data constituting the acquired web page and obtaining an appearance frequency for each character string data;
A procedure for associating the character string data and appearance frequency data that is data indicating the appearance frequency and storing them in an appearance frequency storage unit that is a storage area provided in the memory;
A procedure for outputting the extracted character string data to the output device;
The reading data indicating the reading of the character string represented by the character string data output from the input device and output to the output device is received, and the character string data and the reading data are associated with each other in the memory. A procedure of storing in a reading data storage unit which is a storage area provided;
A procedure for acquiring the reading data corresponding to the character string data stored in the appearance frequency storage unit from the reading data storage unit;
For each of the plurality of character string data having the same read data, the read data obtained by voice recognition is selected from the plurality of character string data having the same read data. A procedure for determining the selection order of the character string data when converting to the character string data based on the appearance frequency data;
Associating the character string data, the reading data, and selection order data that is data indicating the selection order, and generating speech dictionary data that is data used when recognizing speech and converting it to a character string; A procedure for storing the speech dictionary data in a second speech dictionary storage unit that is a storage area provided in the memory;
A program for running The program according to claim 2 or 4,
A Web browser, which is software that displays an image, a character string, or the like corresponding to the Web page on the display, acquires the URL stored in the URL storage unit that is a storage area for storing the URL of the Web page specified by the user. Procedure and
Storing the URL obtained from the URL storage unit in the read URL storage unit;
A program characterized by further comprising: The speech dictionary creation program according to claim 5,
A step of accepting the URL specified by the user among the plurality of URLs stored in the read URL storage unit;
The procedure for acquiring the URL stored in the URL storage unit is a procedure for acquiring, from the URL storage unit, the URL specified by the user among the URLs stored in the URL storage unit. thing,
A program characterized by The program according to claim 2 or 4,
Link hierarchy data indicating the number of hierarchies that follow links to other web pages set in a web page is stored in a link hierarchy storage unit that is a storage area provided in the memory,
In the procedure of acquiring the Web page specified by the URL stored in the read URL storage unit,
A link destination URL that is a URL indicating a link to another Web page included in the data constituting the acquired Web page is extracted, and the other Web page specified by the link destination URL is acquired. Including a procedure of repeatedly executing the processing to be performed according to the number of levels indicated by the link level data stored in the link level storage unit,
A program characterized by The program according to claim 7,
The link hierarchy storage unit stores the link hierarchy data for each URL of the Web page stored in the read URL storage unit,
The procedure of repeatedly executing the process of acquiring the other Web page according to the number of hierarchies corresponds to the URL stored in the URL storage unit of the process of acquiring the other Web page. A procedure that is repeatedly executed according to the number of layers indicated by the link layer data.
A program characterized by The program according to claim 2 or 4,
A program further comprising a procedure of acquiring a URL input in an area for inputting a URL of a Web page to be displayed provided in software for displaying the Web page, and storing the acquired URL in the read URL storage unit . A communication interface for connecting to a CPU, memory, and the Internet;
A storage area provided in the memory;
A first voice dictionary storage unit in which character string data indicating a character string such as a word and reading data indicating a reading of the character string are stored in association with each other;
A read URL storage unit in which the URL of the Web page is stored;
A computer comprising
Obtaining a Web page specified by the URL stored in the read URL storage unit by accessing a Web server on the Internet;
Extracting the character string data contained in the acquired web page and determining the appearance frequency for each character string data;
The character string data and the appearance frequency data which is the data indicating the appearance frequency are associated with each other and stored in the appearance frequency storage unit which is a storage area provided in the memory,
Obtaining the reading data corresponding to the character string data stored in the appearance frequency storage unit from the first speech dictionary storage unit;
For each of the plurality of character string data having the same acquired reading data, the reading data obtained by voice recognition is converted into one character string data selected from the plurality of character string data. The selection order of the character string data at the time is determined based on the appearance frequency data,
Associating the character string data, the reading data, and selection order data that is data indicating the selection order, and generating speech dictionary data that is data used when recognizing speech and converting it to a character string; Storing the speech dictionary data in a second speech dictionary storage unit which is a storage area provided in the memory;
A voice dictionary creation method characterized by

Images