JP2009086597A - Text-to-speech conversion service system and method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To solve the problem wherein listeners will hear inappropriate terms uttered if parts of text are converted to create audio data and a playback order of the audio data is created to match utterances of inappropriate terms. <P>SOLUTION: When text data is converted into audio data, text data indicating the reading of the audio data is also generated. If a change in the playback order of text data indicating the reading of audio data corresponding to a plurality of partial text data makes the changed text data indicating reading match a preset reading-prohibited term, the audio data corresponding to the partial text data are replaced with predetermined audio data. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  In the present invention, a blog site (a site that acts as a proxy for publishing and operating a blog (Weblog) on a Web server) or an SNS site (Social Networking Service: a community-type Web site) The present invention relates to a text-to-speech conversion service system that inputs via a network, converts it into voice data, publishes it, and outputs a sentence by voice when viewed by another user.

  A speech synthesizer that accumulates and analyzes human voices, extracts feature data representing features, converts any text data into speech data using the feature data, and utters it with natural accents like a human Has emerged. As these speech synthesizers, for example, Hideyuki Mizuno, et al., Text-to-Speech Synthesis Technology Corpus-Based Approach, NTT Technical Review, Vol. 2, no. 3, pp. 70-75, March 2004 (Non-Patent Document 1).

  Using such a speech synthesizer, depending on the listening environment of the viewer, it may sound as if the person who provided the voice from which the feature data was based (hereinafter referred to as the former speaker) is actually speaking There is sex. The former speaker may be an actor or a voice actor who dubbes the voice of an animated character (called a character). When such a speech synthesizer is used in front of the public, it is possible to utter inappropriate words such as rough words that cannot be spoken by these actors and characters. In this case, depending on the contents converted into audio data, the image of these actors and voice actors may be damaged.

  In order to solve this problem, several techniques have been developed.

  For example, in the technique described in Japanese Patent Application Laid-Open No. 5-165486 (Patent Document 1), in a text-to-speech conversion device that converts an input sentence composed of text data into a speech signal and outputs a pronunciation, prohibition of reading out a reading-prohibited term. A reading prohibition term determining means for extracting a table and an input sentence in units of words, searching a reading prohibition table to determine whether or not a word included in the input sentence is a reading prohibition term, and a reading prohibition term determination means Based on the determination result, pronunciation prohibiting means for prohibiting pronunciation of a word corresponding to the reading prohibition term is provided.

  In the technique disclosed in Patent Document 1, in a text-to-speech conversion device that converts an input sentence composed of text data into a speech signal and outputs a pronunciation, it reads out a pair of a reading-prohibited term and a replacement expression that expresses the reading-prohibited term. A prohibition table, input sentence is cut out in units of words, a reading prohibition table is searched to determine whether a word included in the input sentence is a read prohibition term, and this reading prohibition term determination means On the basis of the determination result, a reading prohibition term replacement means for converting a reading prohibition term into a replacement expression and generating a pronunciation is provided.

  Furthermore, in the technique described in Japanese Patent Application Laid-Open No. 2004-271727 (Patent Document 2), speech synthesis processing is performed using the utterance content of the voice message designated by the orderer and the voice feature data of a specific speaker, The voice data providing system that provides the voice synthesis data obtained as a voice data, and the order receiving side that receives an order for voice data creation from the orderer receives the utterance content of the voice message designated by the orderer, When the selection information of the speaker who utters the voice message is received, it is determined whether or not the utterance content of the received voice message contains an inappropriate expression as the utterance content of the selected speaker. If it is determined that the speech expression is not included, the speech synthesis process is performed using the utterance content of the voice message and the voice feature data of the selected speaker. And it had been the voice synthesis data to be provided as audio data.

Hideyuki Mizuno et al., Text-to-Speech Technology Technology Using Corpus-Based Approach, NTT Technical Review, Vol. 2, no. 3, pp. 70-75, March 2004 JP-A-5-165486 JP 2004-271727 A

  With the spread of the Internet, it has become more common to provide Web pages for individuals to send information. A typical example is a blog. A blog is a set of web pages that describe what happened daily or a specific topic by an individual or group. Many blogs publish photos, music, and video files, and people other than those who have written can access them using a web browser, read the posted text, watch photos and videos, I can listen to music. The voice data created by the voice synthesizer can be made public on a blog site.

  With the above-mentioned technology, when the voice data is made public on a blog site, etc., the voice data is created by converting a part of the sentence, and the playback order of the voice data is created so that it is the same as the utterance of inappropriate terms. For example, the person who hears this may sound as if he / she uttered an inappropriate term, which may damage the image of the actor and voice actor of the former speaker.

  The text-to-speech conversion service system and method of the present invention solve the above-mentioned problems in the following manner. A plurality of partial text data is extracted from the text data received from the connected terminal. The extracted partial text data is converted into voice data, and text data representing the reading of the voice data is generated. When text data representing the reading of audio data corresponding to multiple partial text data is made continuous (when the playback order of the audio data is changed), the text data representing the continuous reading corresponds to a preset reading-prohibited term In this case, the voice data corresponding to the partial text data is replaced with predetermined voice data.

  The aspect of the present invention is further efficient by making each of the speech synthesis site for speech synthesis and the content examination site for checking the reading prohibition term independent sites and sharing them with other blog sites.

  According to the present invention, by using the text data of the reading corresponding to the voice data, it is possible to prohibit the utterance of an inappropriate term accompanying the voice synthesis of the partial text data.

  An embodiment in which text data is converted into audio data and reproduced at a blog site or the like will be described below.

  The first embodiment will be described in detail with reference to FIGS. The first embodiment is composed of three parts: a personal computer having a Web browser and a voice reproduction function, a blog site, and a voice synthesis site.

  In the first embodiment, when a user accesses a blog site from a personal computer and posts a sentence, a part of the sentence can be replaced with a voice approximate to the voice of an actor or animation character. When another user accesses the blog site from another personal computer and browses the text, a part of the text can be reproduced as if an actor or an animated character is speaking.

  FIG. 1 is a configuration diagram of the first embodiment. Reference numerals 1 and 2 denote personal computers (PCs) equipped with an operating system. Reference numerals 3 and 4 denote display devices that display characters and figures so that humans can recognize them. Reference numerals 5 and 6 are keyboards for the user to input characters. Reference numerals 7 and 8 denote mouses (pointing devices) with buttons indicating the locations of figures and characters displayed on the display devices 3 and 4, respectively. 9 and 10 are programs that operate on a personal computer, Web browsers that display text data described using HTML on the displays 3 and 4, and 11 and 12 are programs that operate on a personal computer. Audio reproduction units 13 and 14 for reproducing audio data recorded in the WAVE format with audio are speakers that convert the outputs of the audio reproduction units 11 and 12 into audio that can be recognized by humans, respectively.

  The WAVE format is an audio file format, and is a storage format for recording an audio signal converted into digital data. HTML (HyperText Markup Language) is a markup language for describing sentences on the Web.

  The blog site 20 includes a Web server 22 that communicates with a PC using HTTP (Hyper Text Transfer Protocol), an editing unit 24 that edits HTML text data described in HTML, and an HTML text database that stores HTML text data described in HTML. 26, an audio text database 28 for storing audio data and text data, a conversion requesting unit 30, a reading-inhibited term database 32, and a content examination unit 34. The conversion requesting unit 30 requests to convert a part of the text data stored in the HTML text database 26 into speech, issues an instruction to store the converted speech data and the read-out text data in the speech text database 28, A part of the text data stored in the HTML text database 26 is replaced. The reading-prohibited term database 32 is a reading-prohibited term database that stores reading-prohibited terms that are not suitable for voice reading. The content examination unit 34 refers to the HTML text database 26, the speech text database 28, and the reading-prohibited term database 32, and when the term to be read is a reading-prohibited term, changes the speech data corresponding to the term to be read.

  The blog site 20 is a website on the Internet that acts as a blog provider, and includes a computer such as a server and software for the website. By inputting a URI (Uniform Resource Identifier) uniquely corresponding to the Web site with the Web browser 9 of the PC 1, the user can access the blog site 20 as the Web site and browse the blog Web page. The blog site 20 has a login function for authenticating a user and a search function for searching a Web page.

  The editing unit 24, the conversion requesting unit 30, and the content examination unit 34 are programs that operate on the blog site 20, and the HTML text database 26, the voice dext database 28, and the reading-prohibited term database 32 are used on the blog site 20. It is a database. These are realized by hardware (computer) constituting the blog site 20, an operating system thereof, a file system provided by them, and the like.

  The speech synthesis site 40 receives a URI 42 and text data from the blog site 20, inputs text data mixed with kanji and kana, and speaks speech data in the form of speech files such as WAVE and utterances written in Roman characters, etc. Are a speech synthesizer 44 that outputs a reading text that expresses the data, and a transmitter 46 that transmits data to the blog site 20 using a designated URI. A technique for implementing a speech synthesis function for converting text data mixed with kanji and kana into speech data is described in detail in Non-Patent Document 1.

  In this embodiment, “aho” and “baka” are utterances of inappropriate terms, and the person who hears them sounds as if an inappropriate term was uttered.

  Hereinafter, the operation of the editing unit 24 of the blog site 20 in accordance with an end user operation on the personal computer 1 will be described with reference to FIGS. 2 and 3.

  FIG. 2 is an example of an input screen 200 displayed on the display device 3 of the personal computer 1 when the end user writes a sentence on the blog.

  The input screen 200 for inputting a sentence includes a title input part 202 for inputting a title of the sentence, a body input part 204 for inputting the text of the sentence, and an audio for displaying the correspondence between the actor or animation character that reads the converted voice and the pictogram. “Cancel button” 208 for instructing to end the text written in the pictogram corresponding display portion 206, the title input portion 202 and the text input portion 204 without storing them in the HTML text database 26 of the blog site 20, the title input portion 202 And a “write button” 210 for instructing to store the text written in the text input portion 204 in the HTML text database 26 of the blog site 20.

  The voice-pictogram correspondence display portion 206 converts the text surrounded by the star, which is a pictogram, into the voice of the actor 1 in the body input portion 204, and converts the text enclosed by the triangular mark, which is a pictogram, into the voice of the actor 2. It is displayed to convert. In this example, two types of sound can be selected, but two or more types may be selected. The display of the voice-pictogram correspondence display portion 206 may be different for each user. For example, a square as a pictogram may be the actor 1.

  FIG. 3 is a process flowchart of the editing unit 24. Now, assume that the Web browser 9 of the personal computer 1 accesses the blog site 20 and the editing unit 24 starts processing through the Web server 22.

  The editing unit 24 starts the process (step 300), creates a screen in which the title input part 202 and the text input part 204 in FIG. 2 are blank (step 302), and issues a command for displaying them to the personal computer through the web server 22. When it is transmitted to the first Web browser 9, the Web browser 9 displays it on the display 3 (step 304) and waits for an input from the user (step 306).

  If there is an input of text, the editing unit 24 transmits an instruction to display the input text to the Web browser 9 of the personal computer 1 through the Web server 22, and the Web browser 9 displays it on the display device 3. By repeating steps 304 and 306, the text is entered and displayed. When the “cancel button” 208 or “write button” 210 is selected (input) instead of text input in step 306, the process branches depending on the selected button (step 308). When the “cancel button” 208 is selected, the title input part 202 and the text input part 204 are changed to blanks (step 310), and the process ends (step 314). When the “write button” 210 is selected, the contents in the title input part 202 and the text input part 204 are stored in the HTML text database 26 with an appropriate unique URI (step 312), and the process is terminated (step 312). 314).

  Here, the user uses the keyboard 5 and the mouse 7 from the Web browser 9 as shown in FIG. 2, the title input portion 200 is “Neighborhood Park”, and the text input portion 204 is “an interesting place. Is entered "(steps 300 to 304).

  Next, after that, in order to convert “place” and “ka” into audio data, the user inserts an asterisk as a pictograph immediately before and after “place” and “ka”, respectively. To do. In the example of FIG. 2, a screen instructing to convert “place” and “ka” into the synthesized voice of actor 1 is shown.

  When the user does not want to store this sentence, the user presses the cancel button 208. The input data is erased in step 310, and the web browser 9 of the personal computer 1 changes the title input part 202 and the text input part 204 to blanks in FIG. 2, displays them on the display device 3, and ends the processing. (Steps 308 to 314).

  Here, the user selects the write button 210 with the mouse 7 and stores the texts written in the title input part 202 and the text input part 204 in the HTML text database 26 of the blog site 20 with appropriate URIs stored as files. (Steps 310 to 314). Here, the attached URI is “http://blog1.com/u1/10/honbun.html”.

  FIG. 4 shows an HTML sentence stored in the HTML text database 26 at this time. FIG. 4 shows an HTML sentence 400 that is one file. The numbers 4010 to 4090 at the left end of FIG. 4 are line numbers added for explaining this embodiment, and are not included in the original HTML text.

  The conversion request unit 30 periodically refers to the HTML text database 26 to detect HTML text that is a newly stored file, and extracts a text data portion surrounded by pictograms such as stars and triangles. For each extracted text data, a unique URI is generated, the text data and the generated URI are transmitted to the speech synthesis site 40, the voice data of the file name corresponding to the URI, and the last of the URI The text data of the voice data reading obtained by replacing the extension “.wav” with “.txt” is obtained.

Here, one of the data transmitted to the speech synthesis site 40 is text data “place”, and the URI generated corresponding to the data is “
“Http://blog1.com/u1/10/01.wav”
The other one is that the text data is “ka”, and the corresponding URI generated is
“Http://blog1.com/u1/10/02.wav”
Suppose that

  Here, the generated URI is obtained by adding a unique number to the path name of the HTML text stored in the HTML text database 26, but may be unrelated to the path name of the HTML text.

  The reception unit 42 of the speech synthesis site 40 that has received the above data outputs the text data to the speech synthesis unit 44 and the URI to the transmission unit 46. The speech synthesizer 44 converts the text data into speech data, and outputs reading text data and speech data that express the utterance of the speech data.

  The transmitter 46 obtains voice data and reading text data from the voice synthesizer 44, obtains a URI from the receiver 42, and obtains a voice data file corresponding to the URI and an extension “. The read text data corresponding to the URI in which “wav” is replaced with “.txt” is returned to the conversion request unit 30 of the blog site 20. Here, it is assumed that the text data of the reading is converted from “ba” to a utterance sound “ba” and “ka” is converted to a utterance sound “ka”, and the text data is “ba” and “ka”.

At this stage, the URI of the text data of the reading corresponding to “place” is
“Http://blog1.com/u1/10/01.txt”
The content of the file specified by this URI is “ba” when interpreted as characters. The URI of the text data of the reading corresponding to “ka” is
“Http://blog1.com/u1/10/02.txt”
The content of the file specified by this URI is “ka” when interpreted as characters.

  Upon receiving these data, the conversion requesting unit 30 stores the received data in the speech text database 28, accesses the HTML text database 26, and is surrounded by speech data and pictorial characters of the HTML sentence from which the text data was obtained. Is replaced with an appropriate tag including a URI so that the Web browser can reproduce it.

  FIG. 5 shows the HTML sentence 500 after replacement. The numbers 5010 to 5090 at the left end are line numbers added for explaining this embodiment, and are not included in the original HTML text.

  Since the portion of the HTML text obtained by reading the text data is replaced with the speech data, the portion of the text data surrounded by the star is replaced with an appropriate tag including the URI of the speech data. Here, it is assumed that the Web browser interprets the end of the URI of the link destination as an extension and automatically starts the application program corresponding to the extension, and the “<a” tag and the URI indicating the link destination , And “</ a” tags.

Therefore, the portion “field” surrounded by the star on line 4060 in FIG. 4 is “interesting <a href =“ http: // blog1. com / u1 / 10/01. wav ">
</a>
It was a place. "
Is replaced by In addition, as shown in the lines 5070 to 5076 in FIG.
“Also, let ’s say <a href =“ http: // blog1. com / u1 / 10/02. wav ">
</a>
Yeah. "
Is replaced by Thereafter, the HTML text shown in FIG. 5 is stored in the HTML text database 26 instead of the HTML text shown in FIG.

  FIG. 6 is a diagram for explaining an example of voice data and reading text data stored in the voice text database 28. FIG. 6 shows Wave data 602 and 606 in Wave format and text data 604 and 608 representing the utterance sound of the audio data.

  The content examining unit 34 refers to all the HTML texts in the HTML text database 26, the speech data and text data in the speech text database 28, and processes them using the speech-prohibited terminology stored in the speech-prohibited term database 32.

  Hereinafter, the processing of the content examining unit 34 will be described in detail with reference to FIGS. FIG. 7 is a process flowchart of the content examination unit 34. FIG. 8 shows examples of prohibited words to be read stored in the prohibited words database 32.

  The content examination unit 34 periodically performs the processing from step 700 to step 730 in FIG. 7 for the HTML texts in the HTML text database 26. When the process is started (step 700), the content examining unit 34 empties $ F, which is a variable for storing a list of URIs, and $ S, which is a variable for storing a character string (step 702). Next, one HTML sentence in the HTML text database 26, that is, one line from the head of one file specified by the URI is read (step 704), and it is determined whether the end of file (EOF) or not (step 706). . If it is completed, the process proceeds to step 714, and if not completed, the process proceeds to step 708 to perform pattern matching, extract a file (Wave file) specified by the URI of the audio data, and proceed to step 710.

  If the pattern match is successful and there is a Wave format URI in step 710, the process proceeds to step 712, where the URI is added to $ F and the extension of the URI is replaced from “.wav” to “.txt”. The text data stored in the file specified by the URI is added to $ S, the process returns to step 704, and steps 704 to 712 are repeated. If there is no Wave format URI, the process returns to step 704 and steps 704 to 712 are repeated.

  Here, it is assumed that the content examining unit 34 processes the HTML text of FIG. 5 and repeats the processing from steps 702 to 712. Then, line 5062 of the HTML sentence 500 in FIG. 5 is pattern-matched at step 708, and “http://blog1.com/u1/10/01.wav” is added to $ F at step 712. The URI replaced from “.wav” to “.txt” is “http://blog1.com/u1/10/01.txt”, and the text data “ba” stored therein is $ S. To be added.

  Returning to step 704, the above-mentioned steps 704 to 712 are repeated, and line 5072 of the HTML sentence 500 of FIG. 5 is again pattern-matched in step 708. In step 712, “http://blog1.com/u1/ 10 / 02.wav "is added to $ F. The URI replaced from “.wav” to “.txt” is “http://blog1.com/u1/10/02.txt”, and the text data stored in the URI “ka” is $ S. To be added.

  Therefore, $ F is [http: // blog1. com / u1 / 10/01. wav, http: // blog1. com / u1 / 10/02. wav] and $ S becomes [ba, ka].

  Proceeding to step 714, if $ S is blank, the process ends (730), and if it is not blank, pattern matching is performed between $ S and the prohibited words to be read in the prohibited words database 28 (step 716). Next, the process proceeds to step 718. If the pattern match is successful, the corresponding audio data is replaced with predetermined audio data (step 720), and $ F and $ S are shifted to the left by one element (step 720). Step 722). If the pattern match is not successful, the process proceeds to step 722. After step 722 ends, step 714 to step 722 are repeated, and when $ S is blank, it ends (step 730).

  Here, as shown in FIG. 8, it is assumed that the prohibited term database 28 stores two reading-prohibited terms 802 and 804 and the contents are “aho” and “baka”, respectively. Here, in step 714, $ S is [ba, ka] and is not blank, so the process proceeds to step 716. The reading prohibition term 802 is “aho” and the pattern match does not succeed, but the reading prohibition term 804 is “baka” and the pattern match succeeds. Accordingly, in step 720, the content of the corresponding audio data is replaced with predetermined audio data. Here, it is assumed that the audio data is replaced with silence in the audio data file specified by the URI at the top (left side) of $ F.

  At this stage, $ F is [http: // blog1. com / u1 / 10/01. wav, http: // blog1. com / u1 / 10/02. wav], the content of “http://blog1.com/u1/10/01.wav”, which is the audio data 602 of FIG. 6, is not a sound corresponding to “ba” but is silent.

  In step 722, $ F and $ S are shifted to the left by one element, and $ F is changed to [http: // blog1. com / u1 / 10/02. wav] and $ S becomes [ka]. Since there is no pattern match from step 714 to step 718, the process proceeds to step 722, $ F and $ S are shifted to the left by one element, and again in step 714, both $ F and $ S are left blank. Thus, the processing of the content examination unit 34 is terminated (step 730).

  Assume that the Web browser 10 of the personal computer 2 accesses the blog site 20 and browses the HTML text (FIG. 5) in the HTML text database 26 via the Web server 20 in the above state. At this time, it is displayed on the display 4 of the personal computer 2 as shown in FIG.

  FIG. 9 is an example of a display screen displayed on the display 4 of the personal computer 2, 900 is a display screen, 902 is a sentence title, 904 is a sentence body, and 906 and 908 are links. This is an underline. In the Web browser, when the underlined text is clicked with the mouse, an application program corresponding to the extension of the link written in the <a> tag is started, and the file indicated by the link is read and played, and the speaker 14 Output as audio.

  Here, it is assumed that the user operating the personal computer 2 selects the underlines 906 and 908 in this order with the mouse. The links corresponding to the underlines 906 and 908 are the lines 5062 to 5064 and lines 5072 to 5074 in FIG. 5, respectively, “http://blog1.com/u1/10/01.wav” and “http: //Blog1.com/u1/10/02.wav ". Since the extension is “.wav”, the audio playback unit 12 of the personal computer 2 uses “http://blog1.com/u1/10/01.wav” and “http://blog1.com/u1/”. "10 / 02.wav" is attempted to be played back. The voice reproduction unit 12 accesses the voice text database 28 via the Web server 22 of the blog site 20, and “http://blog1.com/u1/10/01.wav” and “http://blog1.com”. Audio files 602 and 606 corresponding to “/u1/10/02.wav” are downloaded to the personal computer 2 and played back.

  Here, as described above, the content of “http://blog1.com/u1/10/01.wav” is silent, so when the underline 906 is selected with the mouse, the “ba” The sound is not reproduced, the speaker 14 remains silent, and the sound “ka” is reproduced only when the underline 908 is selected with the mouse, so that the speaker 14 does not hear “baka”.

  In the present embodiment, in step 720, the voice data is replaced by silence in the voice data file corresponding to the word on the left. However, instead of silence, a sound fixed in advance may be used.

  In the present embodiment, an example has been described in which the user of the personal computer 1 browses the Web page including the reproduction of the stored synthesized speech as it is in the processing of steps 300 to 314, but the user of the personal computer 2 browses the web page as it is. Even if the user edits the Web page once stored and changes the position and order of links to the audio file to be played back, the content examining unit 34 periodically performs the processing of steps 700 to 730. Therefore, inappropriate utterances can be detected.

  In the first embodiment described above, before the text data received by the receiving unit 42 is input to the voice synthesizing unit 44 at the speech synthesizing site 40, the contents are examined with the text data, and if it is determined to be inappropriate, the change is made. The text data thus sent may be text data sent to the speech synthesizer 44.

  In this embodiment, the text data 604 and 608 stored in the speech text database 28 are accessed only from the conversion requesting unit 30 and the content examining unit 34, and are not accessed from other programs such as the editing unit 24. Once the voice file is created and stored in the voice text database 28, only the text data in the voice text database 28 is changed, and the content judging unit 34 determines that it is appropriate, and changes to utter an inappropriate term. It has the effect that it cannot be done.

  In this embodiment, the case of Japanese is used, but other languages such as English and Chinese can also be handled in the same manner because text data expressing utterances can be used.

  In the first embodiment, the converted text data is directly stored in a file as text data, and the text data is transmitted to the blog site together with the voice data. Instead of storing the text data in the file as text data, the text data is stored in the voice data. It may be embedded with a digital watermark and only the audio data with the digital watermark may be transmitted to the blog site.

  Hereinafter, this case will be described as a second embodiment with reference to FIGS. 2 to 4 and FIGS. 8 to 13. 2 to 4 and FIGS. 8 to 9 are the same as those of the first embodiment.

  The second embodiment is composed of three parts: a personal computer having a web browser and a voice reproduction function, a blog site A having a configuration different from that of the first embodiment, and a speech synthesis site A having a configuration different from that of the first embodiment. . The configuration of Example 2 is shown in FIG. 10, the same components as those in FIG. 1 are denoted by the same reference numerals.

  The main differences between the blog site A1000 and the first embodiment are the voice database 1002, the conversion request unit A1004, the digital watermark detection unit 1006, and the content examination unit A1008. The voice database 1002 stores voice data with a digital watermark into which a digital watermark has been inserted. The conversion request unit A1004 requests to convert a part of the text data stored in the HTML text database 26 into digital watermarked voice, and issues an instruction to store the converted digital watermarked voice data in the voice database 1002. , A part of the text data stored in the HTML text database 26 is replaced. The digital watermark detection unit 1006 extracts text data inserted by digital watermark from the voice data with digital watermark stored in the voice database 1002. The content examination unit A1008 refers to the HTML text database 26, the reading-prohibited term database 32, and the text data obtained from the digital watermark detection unit 1006, determines whether or not it is a reading-prohibited term, and determines that it is a reading-prohibited term. If so, the corresponding audio data stored in the audio database 1002 is changed.

  The voice database 1002, the conversion request unit A1004, the digital watermark detection unit 1006, and the content examination unit A1008 are programs that operate on the blog site A1000. The hardware (computer) that configures the blog site A1000, its operating system, and the like This is realized by the file system provided by.

  The speech synthesis site A1010 corresponds to the reception unit A1012 that receives the URI and text data from the blog site A1000, the digital watermark insertion unit 1014 that inserts the digital watermark of the text data into the voice data, and the voice data with the digital watermark corresponding to the URI. The transmission unit A1016 transmits the file name audio data file to the blog site A1000.

  Also in the present embodiment, “aho” and “baka” are utterances of inappropriate terms, and the person who hears them sounds as if they have uttered inappropriate terms.

  The operation flow of the end user on the personal computer 1 and the processing flow of the editing unit 24 are the same as those in the first embodiment (FIGS. 2 and 3). Here, the end user operates the write button 210 with the mouse 7, and stores the text written in the title input part 202 and the text input part 204 as a file in the HTML text database 26 of the blog site A1000 with an appropriate URI. (Steps 310 to 314).

  Here, the URI attached is “http://blog1.com/u1/10/honbun.html” as in the first embodiment, and the HTML text stored in the HTML text database 26 is the same as in the first embodiment. Similarly, assume that FIG.

  The conversion request unit A1004 periodically refers to the HTML text database 26, detects HTML text that is a newly stored file, extracts a text data portion surrounded by pictograms such as stars and triangles, A unique URI is generated for each extracted text data, and the text data and the URI are transmitted to the speech synthesis site A 1010 to obtain digital watermarked speech data having a file name corresponding to the URI.

  Here, it is assumed that one of the data output to the speech synthesis site is that the text data is “place” and the URI is “http://blog1.com/u1/10/03.wav” One is that the text data is “ka” and the URI is “http://blog1.com/u1/10/04.wav”.

  Receiving unit A1012 of speech synthesis site A1010 that has received the above data outputs text data to speech synthesis unit 44 and outputs a URI to transmission unit A1016. The voice synthesizer 44 converts the text data into voice data, and outputs the text data expressing the voice of the voice data and the voice data to the digital watermark insertion unit 1014.

  The digital watermark insertion unit 1014 inserts text data as a digital watermark into the voice data input from the voice synthesis unit 44 and outputs the text data to the transmission unit A1016. Japanese Patent Laid-Open No. 2003-99077 describes the insertion of digital watermarks into audio data and the detection and extraction of digital watermarks.

  The transmission unit A1016 returns the audio data with digital watermark obtained from the digital watermark insertion unit 1014 to the conversion request unit A1004 of the blog site A1000 as a file of audio data having a file name corresponding to the URI obtained from the reception unit A1012. Here, it is assumed that the text data is obtained by converting “place” into a sound “ba” and converting “ka” into a sound “ka”.

  One “ba” is a watermarked audio data file inserted with a digital watermark. The URI is “http://blog1.com/u1/10/03.wav” and the other “ka” However, a file of audio data with a digital watermark inserted with a digital watermark has a URI “http://blog1.com/u1/10/04.wav”.

  Upon receiving these data, the conversion request unit A1004 stores the received data in the speech database 1002, accesses the HTML text database 26, and converts the speech data and the portion surrounded by pictographs of the HTML text from which the text data was obtained. Replace with an appropriate tag containing a URI so that the web browser can play it.

  FIG. 11 shows the HTML sentence 1100 after replacement. The numbers 11010 to 11090 at the left end are line numbers added for explaining the present embodiment, and are not included in the original HTML text.

  Since the portion surrounded by the pictographs of the HTML text obtained from the voice data and the text data is replaced, the portion of the text data surrounded by the star is replaced with an appropriate tag including the URI of the voice data. Here, it is assumed that the Web browser interprets the end of the URI of the link destination as an extension and automatically starts the application program corresponding to the extension, and the “<a” tag and the URI indicating the link destination , Replace with “</ a” tag.

Therefore, the portion “field” surrounded by the star on line 4060 in FIG. 4 is shown on lines 11060 to 11066 in FIG.
“Interesting <a href =“ http: // blog1. com / u1 / 10/03. wav ">
</a>
It was a place. "
Is replaced by As shown in the 11070th line to the 11076th line in FIG.
“Also, let ’s say <a href =“ http: // blog1. com / u1 / 10/04. wav ">
</a>
Yeah. "
Is replaced by Thereafter, the HTML text shown in FIG. 11 is stored in the HTML text database 26 instead of the HTML text shown in FIG. FIG. 12 shows examples 1202 and 1204 of voice data with digital watermarks stored in the voice database 1002.

  The content examination unit A 1008 refers to the HTML text in the HTML text database 26 and the digital watermarked voice data in the voice database 1002 and performs processing using the read-inhibited terminology stored in the read-in prohibition term database 32.

  Hereinafter, the processing of the content examination unit A1008 will be described in detail with reference to FIGS. 8 and 11 to 13. FIG. 13 is a flowchart showing the processing of the content examination unit A1008.

  The content examination unit A1008 periodically performs the processing from step 1300 to step 1330 in FIG. 13 for each HTML sentence that is a file in the HTML text database 26.

  When the content examination unit A1008 starts processing (step 1300), $ F which is a variable for storing a list of URIs and $ S which is a variable for storing character strings are emptied (step 1302). Next, one HTML sentence in the HTML text database 26, that is, one line from the head of one file specified by the URI is read (step 1304), and it is determined whether the end of file (EOF) or not (step 1306).

  If it is finished, the process proceeds to step 1314. If it is not finished, the process proceeds to step 1308 to perform pattern matching, extract the file specified by the URI of the audio data, and proceed to step 1310.

  If the pattern match succeeds and there is a URI of the voice data in step 1310, the process proceeds to step 1311, where the voice data of the voice data file specified by the pattern matched URI is output to the digital watermark detection unit 1006, The text data inserted by the watermark is extracted, and the process proceeds to step 1312. The pattern-matched URI is added to $ F, the text data is added to $ S, the process returns to step 1304, and steps 1304 to 1312 are repeated. If there is no audio data URI, the process returns to step 1304 and steps 1304 to 1312 are repeated.

  Here, it is assumed that the content examination unit A1008 processes the HTML text of FIG. 11 and repeats the processing of steps 1302 to 1312. Then, the 11062st line of the HTML sentence 1100 in FIG. 11 is pattern-matched in Step 1308, and in Step 1311, the text inserted from “http://blog1.com/u1/10/03.wav” with a digital watermark is inserted. Data “ba” is extracted. In step 1312, “http://blog1.com/u1/10/03.wav” is added to $ F, and “ba” is added to $ S.

  Returning to step 1304, the above steps 1304 to 1312 are repeated, and the 11072st line of the HTML sentence 1100 in FIG. 11 is again pattern matched in step 1308. In step 1311, “http://blog1.com/u1/ The text data “ka” inserted by the digital watermark is extracted from “10 / 04.wav”. In step 1312, “http://blog1.com/u1/10/04.wav” is added to $ F, and “ka” is added to $ S.

  Therefore, $ F is [http: // blog1. com / u1 / 10/03. wav, http: // blog1. com / u1 / 10/04. wav] and $ S becomes [ba, ka].

  Proceeding to step 1314, if $ S is blank, the process ends (1330). If $ S is not blank, pattern matching is performed between $ S and the prohibited words to be read in the prohibited words database 28 (step 1316). Next, the process proceeds to step 1318. If the pattern match is successful, the content of the corresponding audio data is replaced with predetermined audio data (step 1320), and $ F and $ S are shifted to the left by one element (step 1320). Step 1322). If the pattern match is not successful, the process proceeds to step 1322. After step 1322 is completed, steps 1314 to 1322 are repeated, and when $ S is blank, the process ends (step 1330).

  Here, as shown in FIG. 8, it is assumed that the prohibited term database 28 stores two reading-prohibited terms 802 and 804 and the contents are “aho” and “baka”, respectively. In step 1314, $ S is [ba, ka] and is not blank, so the process proceeds to step 1316. The reading prohibition term 802 is “aho” and the pattern match does not succeed, but the reading prohibition term 804 is “baka” and the pattern match succeeds. Accordingly, in step 1320, the content of the corresponding audio data is replaced with predetermined audio data. Here, it is assumed that the audio data is replaced with silence in the audio data file specified by the URI at the top (left side) of $ F.

  At this stage, $ F is [http: // blog1. com / u1 / 10/03. wav, http: // blog1. com / u1 / 10/04. wav], the content of “http://blog1.com/u1/10/03.wav”, which is the audio data 1202 in FIG. 12, is not a sound corresponding to “ba” but a silence.

  Next, in step 1322, $ F and $ S are shifted to the left by one element, and $ F is changed to [http: // blog1. com / u1 / 10/04. wav] and $ S becomes [ka]. Since there is no pattern match from step 1314 to step 1318, the process proceeds to step 1322, where $ F and $ S are shifted to the left by one element, and again in step 1314, both $ F and $ S are left blank. Thus, the processing of the content examination unit A1008 is terminated (step 1330).

  In this state, the user of the personal computer 2 accesses the blog site 1000 using the Web browser 10 and browses the HTML text (FIG. 11) in the HTML text database 26 via the Web server 22. . At this time, it is displayed on the display 4 of the personal computer 2 as shown in FIG. Here, it is assumed that the user operating the personal computer 2 selects the underlines 906 and 908 in this order with the mouse.

  The links corresponding to the underlines 906 and 908 are the lines 11062 to 11064 and the lines 11072 to 11074 in FIG. 11, respectively, “http://blog1.com/u1/10/03.wav” and “http: // /Blog1.com/u1/10/04.wav ". Since the extension is “.wav”, the sound reproduction unit 12 of the personal computer 2 can execute “http://blog1.com/u1/10/03.wav” and “http: //” as in the first embodiment. blog1.com/u1/10/04.wav "is attempted.

  Here, as described above, since the content of “http://blog1.com/u1/10/03.wav” is silent, when the underline 906 is selected with the mouse, “ba” The sound is not reproduced, the speaker 14 remains silent, and the sound “ka” is reproduced only when the underline 908 is selected with the mouse, so that the speaker 14 does not hear “baka”.

  In this embodiment, in the voice data replacement in step 1320, the voice data file corresponding to the word on the left is silenced. However, instead of silence, a sound fixed in advance may be used.

  In this embodiment, before inputting the text data output from the receiving unit A1012 to the speech synthesizing unit 44 at the speech synthesizing site A1010, the contents are examined with the text data and the text data to be sent to the speech synthesizing unit 44 is changed. May be.

  In this embodiment, the case of Japanese is used, but other languages such as English and Chinese can also be handled in the same manner because text data expressing utterances can be used.

  In the first embodiment, when there are a plurality of blog sites, it is necessary to update the reading-prohibited term database of the plurality of blog sites when registering a new reading-prohibited term. If you set up the content review department and the database for prohibited reading aloud on a separate site outside the blog site as a content review site, and share the content review site with multiple blog sites, register new prohibited words for reading aloud When you do this, you only need to update the read-only words database on the content review site that you share, saving you time and effort.

  Hereinafter, Embodiment 3 in this case will be described with reference to FIGS. 2 to 9 and FIG. This embodiment is composed of four parts: a personal computer having a Web browser and a voice reproduction function, a blog site having a configuration different from that of the first embodiment, the same voice synthesis site as that of the first embodiment, and a content examination site. The operation of each component in FIGS. 2 to 9 is the same as the operation of each component in FIGS. 2 to 9 of the first embodiment. The content examination site is a website on the Internet, and is composed of a computer such as a server and software.

  FIG. 14 shows the configuration of the third embodiment. In FIG. 14, the same components as those in FIG.

  The blog site B1400 is different from the first embodiment in that the blog site B1400 is not provided with a configuration relating to content examination, and the configuration relating to the content examination is provided in the content examination site 1402, which is a site different from the blog site B1400.

  The content review site 1402 has a reading-prohibited term database 1404 and a content review unit 1406. The reading-prohibited term database 1404 stores the reading-prohibited terms that are not suitable for voice reading. The content examination unit 1406 refers to the HTML text database 26 and the speech text database 28 of the blog site B 1400, and the reading-prohibited term database 1404 to determine whether or not it is a reading-prohibited term. Change the audio data to be used.

  The blog site B1400 has a login function for authenticating a user and a search function for searching for a web page, which are also omitted in this embodiment.

  The reading-prohibited term database 1404 and the content review unit 1406 are a database and a program that operate on the content review site 1402, respectively, and the hardware (computer) that configures the content review site 1402, its operating system, and they provide Realized with a file system.

  Also in the present embodiment, “aho” and “baka” are utterances of inappropriate terms, and the person who hears them sounds as if they have uttered inappropriate terms.

  Hereinafter, as in the first embodiment, it is assumed that the user inputs the personal computer 1 as shown in FIG. 2, and the editing unit 24 of the blog site 1400 executes the processing shown in FIG. The HTML text stored in the HTML text database 26 at this time is the same as that shown in FIG. 4, and the file name assigned is “http://blog1.com/u1/10/honbun. html ".

  As in the first embodiment, the conversion request unit 30 of the blog site B1400 periodically refers to the HTML text database 26, detects HTML text that is a newly stored file, and uses pictographs such as stars and triangles. A portion of the enclosed text data is extracted, a unique URI is generated for each extracted text data, the text data and the URI are transmitted to the speech synthesis site 40, and voice data having a file name corresponding to the URI The text data of the reading of the file name is obtained by replacing “.wav” at the end of the URI with “.txt”. Here, as in the first embodiment, one of the data transmitted to the speech synthesis site is text data “place”, and the generated URI is “http://blog1.com/u1/10/01”. .Wav ”, and the other one is that the text data is“ ka ”and the generated URI is“ http://blog1.com/u1/10/02.wav ”.

  The voice synthesis site 40 that has received the above data operates in the same manner as in the first embodiment, and the voice data file having the file name corresponding to the URI and the last “.wav” of the URI are changed to “.txt”. The text data is returned to the conversion request unit 30 of the blog site 3 1400 with the replaced URI. Here, it is assumed that the text data is converted from “ba” to a sound “ba” and “ka” is converted to a sound “ka”, as in the first embodiment, and the text data is “ba”. “Ka”.

  At this stage, as in the first embodiment, the URI of the text data corresponding to “place” is “http://blog1.com/u1/10/01.txt”, and the file specified by this URI The content is interpreted as characters and is “ba”. The URI of the text data corresponding to “ka” is “http://blog1.com/u1/10/02.txt”. The contents of the file specified by this URI are interpreted as characters, and “ka "

  Upon receiving these data, the conversion request unit 30 stores the received data in the speech text database 28 as in the first embodiment, accesses the HTML text database 26, and obtains the speech data and the HTML text obtained from the text data. The part surrounded by pictograms is replaced with an appropriate tag including a URI so that the Web browser can reproduce it. The HTML text after replacement is the same as that shown in FIG. 5, and the HTML text shown in FIG. 5 is stored in the HTML text database 26 instead of the HTML text shown in FIG. 4. The voice data and the text data of reading stored in the voice text database 28 are as shown in FIG.

  The content review unit 1406 of the content review site 1402 refers to the HTML text in the HTML text database 26 of the blog site B 1400, the speech data and text data in the speech text database 28, and the speech prohibition term stored in the speech prohibition term database 1404. To process.

  The processing of the content examination unit 1406 is the same as that of the content examination unit 34 of the first embodiment. For each HTML sentence that is a file in the HTML text database 26, the processing from step 700 to step 730 in FIG. Do. Here, it is assumed that the content examination unit 34 performs processing on the HTML text of FIG. 5 and repeats the processing from steps 702 to 712.

  Then, line 5062 of the HTML sentence 500 in FIG. 5 is pattern-matched at step 708, and “http://blog1.com/u1/10/01.wav” is added to $ F at step 712. The URI replaced from “.wav” to “.txt” is “http://blog1.com/u1/10/01.txt”, and the text data “ba” stored therein is $ S. To be added.

  Returning to step 704, the above-mentioned steps 704 to 712 are repeated, and line 5072 of the HTML sentence 500 of FIG. 5 is again pattern-matched in step 708. In step 712, “http://blog1.com/u1/ 10 / 02.wav "is added to $ F. The URI replaced from “.wav” to “.txt” is “http://blog1.com/u1/10/02.txt”, and the text data stored in the URI “ka” is $ S. To be added.

  As a result, $ F becomes [http: // blog1. com / u1 / 10/01. wav, http: // blog1. com / u1 / 10/02. wav] and $ S becomes [ba, ka].

  Next, the process proceeds to step 714. If $ S is blank, the process ends (730). If $ S is not blank, pattern matching is performed between $ S and the prohibited words to be read in the prohibited words database 28 (step 716).

  Next, the process proceeds to step 718. If the pattern match is successful, the corresponding audio data is replaced with predetermined audio data (step 720), and $ F and $ S are shifted to the left by one element (step 720). Step 722). If the pattern match is not successful, the process proceeds to step 722. After step 722 ends, step 714 to step 722 are repeated, and when $ S is blank, it ends (step 730).

  The reading-prohibited term database 1404 stores two reading-prohibited terms 802 and 804 shown in FIG. 8, and the contents are “aho” and “baka”, respectively.

  Here, in step 714, $ S is [ba, ka] and is not blank, so the process proceeds to step 716. The reading prohibition term 802 is “aho” and the pattern match does not succeed, but the reading prohibition term 804 is “baka” and the pattern match succeeds. In step 720, the content of the corresponding audio data is replaced with predetermined audio data. Here, it is assumed that the audio data is replaced by silence in the audio data file specified by the URI at the top (left side) of $ F.

  At this stage, $ F is [http: // blog1. com / u1 / 10/01. wav, http: // blog1. com / u1 / 10/02. wav], the content of “http://blog1.com/u1/10/01.wav”, which is the audio data 602 in FIG. 6, is not a sound corresponding to “ba” but is silent.

  Next, in step 722, $ F and $ S are shifted to the left by one element, and $ F is changed to [http: // blog1. com / u1 / 10/02. wav] and $ S becomes [ka]. Since there is no pattern match from step 714 to step 718, the process proceeds to step 722, $ F and $ S are shifted to the left by one element, and again in step 714, both $ F and $ S are left blank. Thus, the processing of the content examination unit 1406 is terminated (step 730).

  In the above state, the user of the personal computer 2 accesses the blog site B 1400 using the Web browser 10 and browses the HTML text (FIG. 5) in the HTML text database 26 via the Web server 20. To do. At this time, the display 4 of the personal computer 2 is displayed as shown in FIG.

  Here, it is assumed that the user operating the personal computer 2 selects the underlines 906 and 908 with the mouse in this order as in the first embodiment. Here, as described above, the content of “http://blog1.com/u1/10/01.wav” is silent, so when the underline 906 is selected with the mouse, the “ba” The sound is not reproduced, the speaker 14 remains silent, and the sound “ka” is reproduced only when the underline 908 is selected with the mouse, so that the speaker 14 does not hear “baka”.

  In this embodiment, in step 720, the voice data is replaced by silence in the voice data file corresponding to the left word. However, as in the first embodiment, instead of silence, Anyway.

  In the present embodiment, an example has been described in which the user of the personal computer 1 browses the Web page including the reproduction of the stored synthesized speech as it is in the processing of steps 300 to 314, but the user of the personal computer 2 browses the web page as it is. Even if the user edits the Web page once stored and changes the position and order of links to the audio file to be played back, the content examining unit 34 periodically performs the processing of steps 700 to 730. Therefore, inappropriate utterances can be detected.

  In Example 3 described above, before the text data output from the receiving unit 42 is input to the speech synthesizing unit 44 at the speech synthesizing site 40, the contents are examined with the text data, and the text data is determined to be inappropriate. The data may be changed, and the changed text data may be sent to the speech synthesizer 44 as text data.

  In this embodiment, the text data 604 and 608 stored in the speech text database 28 are accessed only from the conversion request unit 30 and the content examination site 1402 and are not accessed from other programs such as the editing unit 24. Once the voice file is created and stored in the voice text database 28, only the text data in the voice text database 28 is changed, and the content review site 1402 determines that it is appropriate, and changes to utter inappropriate terms. It has the effect that it cannot be done.

  In this embodiment, the case of Japanese is used, but other languages such as English and Chinese can also be handled in the same manner because text data expressing utterances can be used.

  In the second embodiment, the content review unit, the digital watermark detection unit, and the reading prohibition term database are provided in the blog site. However, the content review unit, the digital watermark detection unit, and the read prohibition term database are separately provided outside the blog site. This site may be set up as a content review site, and the content review site may be shared by a plurality of blog sites.

  Hereinafter, Embodiment 4 in this case will be described with reference to FIGS. 2 to 4, 8, 9, 11 to 13, and 15. This embodiment has four parts: a personal computer having a web browser and a voice reproduction function, a blog site having a different configuration from that of the second embodiment, the same voice synthesis site as that of the second embodiment, and a content examination site having a different configuration from that of the third embodiment. Consists of

  The content examination site is a website on the Internet, and is composed of a computer such as a server and software.

  The operation of each component in FIGS. 4, 8, 9, and 11 to 13 is the same as the operation of each component in the second embodiment.

  FIG. 15 shows the configuration of the fourth embodiment. In FIG. 15, the same components as those in FIG. 10 are denoted by the same reference numerals.

  The blog site C1500 is different from the second embodiment in that it does not have a configuration relating to content examination, and a configuration relating to content examination is provided in the content examination site 1502, which is a site different from the blog site C1500.

  The content review site 1502 includes a digital watermark detection unit 1504, a reading prohibited term database 1506, and a content review unit A 1508. The digital watermark detection unit 1504 is the same as the digital watermark detection unit 1006 in FIG. The reading-prohibited term database 1506 is the same as the reading-prohibited term database 32 in FIG. The content examination unit A1508 refers to the voice database 1002 of the blog site C1500, obtains text data from the digital watermark detection unit 1504, refers to the HTML text database 26 of the blog site C1500, and the reading-prohibited term database 1506, and is prohibited from reading. It is determined whether or not the term is a term, and when it is determined that the term is a reading-prohibited term, the corresponding speech data stored in the speech database 1002 is changed.

  The blog site has a login function for authenticating a user and a search function for searching for a Web page, which are also omitted in this embodiment.

  The reading-prohibited term database 1506, the digital watermark detection unit 1504, and the content screening unit A 1508 are a database and a program that operate on the content screening site A 1502, hardware (computer) that constitutes the content screening site A 1502, an operating system thereof, And a file system provided by them.

  Also in the present embodiment, “aho” and “baka” are utterances of inappropriate terms, and the person who hears them sounds as if they have uttered inappropriate terms.

  The flow of operation of the end user on the personal computer 1 and the flow of processing in the editing unit 24 are the same as those in the first to third embodiments (FIGS. 2 and 3).

  Here, the end user operates the write button 210 with the mouse 7, assigns a proper file name to the HTML text database 26 of the blog site C 1500, and writes the text written in the title input part 202 and the text input part 204. (Steps 310 to 314). Here, it is assumed that the assigned file name is “http://blog1.com/u1/10/honbun.html” as in the first to third embodiments. Assume that the HTML text stored in the HTML text database 26 at this time is FIG. 4 as in the first embodiment.

  As in the first to third embodiments, the conversion request unit A1004 periodically refers to the HTML text database 26, detects HTML text that is a newly stored file, and uses pictograms such as stars and triangles. Extracts the enclosed text data part, generates a unique URI for each extracted text data, sends the text data and the URI to the speech synthesis site A1010, and includes an electronic watermark with a file name corresponding to the URI Get audio data.

  Here, it is assumed that one of the data transmitted to the speech synthesis site A 1010 is that the text data is “place”, and the URI is “http://blog1.com/u1/10/03.wav”, and the other One of them is that the text data is “ka” and the URI is “http://blog1.com/u1/10/04.wav”.

  Receiving unit A1012 of voice joint site A1010 that has received the above data outputs text data to voice synthesizing unit 44 and outputs a URI to transmitting unit A1016. The voice synthesis unit 44 converts the text data into voice data, and outputs the read text data and the voice data to the digital watermark insertion unit 1014. The digital watermark insertion unit 1014 inserts the text data of the reading into the voice data as the digital watermark into the voice data input from the voice synthesis unit 44, and outputs the voice data to the transmission unit A 1016. Insertion of digital watermark into audio data, detection and extraction of digital watermark are the same as in the second embodiment.

  Similarly to the second embodiment, the transmission unit A1016 transmits the audio data with digital watermark obtained from the digital watermark insertion unit 1014 as a file of audio data having a file name corresponding to the URI obtained from the reception unit A1012. Return to conversion request unit A1004. Here, it is assumed that the text data is converted from “ba” into a sound “ba” and “ka” into a sound “ka”.

  The file of audio data with a digital watermark in which one “ba” is inserted with a digital watermark has a URI “http://blog1.com/u1/10/03.wav” and the other “ka” The audio data file with the digital watermark inserted by the digital watermark has a URI “http://blog1.com/u1/10/04.wav”.

  Upon receiving these data, the conversion request unit A1004 stores the received data in the speech database 1002, accesses the HTML text database 26, and converts the speech data and the portion surrounded by pictographs of the HTML text from which the text data was obtained. Replace with an appropriate tag containing a URI so that the web browser can play it. The HTML text after the replacement is the same as that in FIG. Thereafter, the HTML text shown in FIG. 11 is stored in the HTML text database 26 instead of the HTML text shown in FIG. The voice data with digital watermark stored in the voice database 1002 is the same as that of FIG.

  The content review unit A1508 of the content review site A1502 refers to the HTML text in the HTML text database 26 of the blog site C1500 and the digital watermark-added speech data in the speech database 1002, and the speech-prohibited terminology stored in the speech-prohibited terminology database 1506 Process using.

  The processing of the content screening unit A 1508 is the same as the processing of the content screening unit of the second embodiment. For each HTML sentence that is a file in the HTML text database 26 periodically, the processing from step 1300 to step 1330 in FIG. Execute. Here, it is assumed that the content examination unit 2 1508 performs processing on the HTML text of FIG. 11 and repeats the processing of steps 1302 to 1312.

  Then, the 11062st line of the HTML sentence 1100 in FIG. 11 is pattern-matched in Step 1308, and in Step 1311, the text inserted from “http://blog1.com/u1/10/03.wav” with a digital watermark is inserted. Data “ba” is extracted. In step 1312, “http://blog1.com/u1/10/03.wav” is added to $ F, and “ba” is added to $ S.

  Returning to step 1304, the above steps 1304 to 1312 are repeated, and the 11072st line of the HTML sentence 1100 in FIG. 11 is again pattern matched in step 1308. In step 1311, “http://blog1.com/u1/ The text data “ka” inserted by the digital watermark is extracted from “10 / 04.wav”. In step 1312, “http://blog1.com/u1/10/04.wav” is added to $ F, and “ka” is added to $ S.

  As a result, $ F is [http: // blog1. com / u1 / 10/03. wav, http: // blog1. com / u1 / 10/04. wav] and $ S becomes [ba, ka].

  Next, the process proceeds to step 1314. If $ S is blank, the process is terminated (1330). If $ S is not blank, pattern matching is performed between $ S and the prohibited words to be read in the prohibited words database 1506 (step 1316).

  Next, the process proceeds to step 1318. If the pattern match is successful, the content of the corresponding audio data is replaced with predetermined audio data (step 1320), and $ F and $ S are shifted to the left by one element (step 1320). Step 1322). If the pattern match is not successful, the process proceeds to step 1322. After step 1322 is completed, steps 1314 to 1322 are repeated, and when $ S is blank, the process ends (step 1330).

  Here, as shown in FIG. 8, the prohibited term database 1506 stores two reading prohibited terms 802 and 804 as shown in FIG. 8, and the contents are “aho” and “baka”, respectively. To do.

  In step 1314, since $ S is [ba, ka] and is not blank, the process proceeds to step 1316. The reading prohibition term 802 is “aho” and the pattern match does not succeed, but the reading prohibition term 804 is “baka” and the pattern match succeeds. Therefore, in step 1320, the content of the corresponding audio data is replaced with predetermined audio data. Here, it is assumed that the replacement with the voice data makes the voice data file corresponding to the word at the top (left side) of $ F silence.

  At this stage, $ F is [http: // blog1. com / u1 / 10/03. wav, http: // blog1. com / u1 / 10/04. wav], the content of “http://blog1.com/u1/10/03.wav”, which is the audio data 1202 of FIG. 12, is not a sound corresponding to “ba” but a silence.

  Next, in step 1322, $ F and $ S are shifted to the left by one element, and $ F is changed to [http: // blog1. com / u1 / 10/04. wav] and $ S becomes [ka]. Since there is no pattern match from step 1314 to step 1318, the process proceeds to step 1322, where $ F and $ S are shifted to the left by one element, and again in step 1314, both $ F and $ S are left blank. Thus, the processing of the content examination unit A 1508 is terminated (step 1330).

  In the above state, the user of the personal computer 2 accesses the blog site C1500 using the Web browser 10 and browses the HTML text (FIG. 11) in the HTML text database 26 via the Web server 22. To do. At this time, the display 4 of the personal computer 2 is displayed as shown in FIG. 9 as in the first to third embodiments.

  Here, it is assumed that the user operating the personal computer 2 selects the underlines 906 and 908 in this order with the mouse.

  The links corresponding to the underlines 906 and 908 are the lines 11062 to 11064 and the lines 11072 to 11074 in FIG. 11, respectively, “http://blog1.com/u1/10/03.wav” and “http: //Blog1.com/u1/10/04.wav ". Since the extension is “.wav”, the sound reproduction unit 12 of the personal computer 2 can execute “http://blog1.com/u1/10/03.wav” and “http: //” as in the first embodiment. blog1.com/u1/10/04.wav "is attempted.

  Here, as described above, since the content of “http://blog1.com/u1/10/03.wav” is silent, the sound of “ba” is selected when the underline 906 is selected with the mouse. Is not reproduced and the speaker 14 remains silent, and the sound “ka” is reproduced when the underline 908 is selected with the mouse, so that the speaker 14 does not hear “baka”.

  In this embodiment, in the voice data replacement in step 1320, the voice data file corresponding to the word on the left is silenced. However, instead of silence, a sound fixed in advance may be used.

  In the fourth embodiment, the text data output from the receiving unit A1012 at the speech synthesis site A1010 is examined with the text data before being input to the speech synthesis unit 44, and the text data to be sent to the speech synthesis unit 44 is It may be changed.

  In this embodiment, the case of Japanese is used, but other languages such as English and Chinese can also be handled in the same manner because text data expressing utterances can be used.

1 is a configuration diagram of Example 1. FIG. It is an example of the input screen of a blog. It is a process flowchart of an edit part. It is the figure which showed the example of the stored HTML text. It is the figure which showed the example of the substituted HTML sentence. It is a figure which shows the example of the data stored in the audio | voice text database. It is a processing flowchart of a content examination part. It is a figure which shows the example of the term stored in the reading prohibition term database. It is a figure which shows the example of a display screen of a blog. FIG. 6 is a configuration diagram of Example 2. It is a figure which shows the example of the substituted HTML text. It is a figure which shows the example of the audio | voice data with a digital watermark stored in the audio | voice database. It is a processing flowchart of a content examination part. FIG. 6 is a configuration diagram of Example 3. FIG. 6 is a configuration diagram of Example 4.

Explanation of symbols

  1, 2, personal computer, 3, 4: display, 5, 6: keyboard, 7, 8: mouse, 9, 10: web browser, 11, 12 audio playback unit, 13, 14 speakers, 20, 1000, 1400, 1500: Blog site, 22: Web server, 24: Editing unit, 26: HTML text database, 28: Speech text database, 30: Conversion request unit, 32: Reading prohibited term database, 34: Content review unit, 40, 1010: Speech synthesis site, 42: reception unit, 44: speech synthesis unit, 46: transmission unit, 200: input screen, 202: title input part, 204: text input part, 206: voice-pictogram correspondence display part, 208: cancel button 210: Write button 400, 500: HTML text 602: Audio data 604: Text data, 606: voice data, 608: text data, 802, 804: reading banned words, 1402,1502: content review site.

Claims (5)

In response to receiving the text data, converting the text data into voice data and transmitting the voice data and text data representing the reading of the voice data;
A text database for storing text data received from a connected terminal, a plurality of partial text data is extracted from the text data in response to an instruction from the terminal, and the extracted plurality of partial text data is sent to the speech synthesis site A conversion request unit for transmitting and storing in the speech text database speech data corresponding to the plurality of partial text data transmitted from the speech synthesis site and text data representing reading of the speech data, and the plurality of partial texts When the text data representing the reading of the voice data corresponding to the data is continuous, and the text data representing the continuous reading corresponds to a preset reading prohibition term, the voice corresponding to the partial text data Includes a content review section that replaces the data with predetermined audio data. Text-to-speech conversion service system which is characterized by having a blog site.   2. The text-to-speech conversion service system according to claim 1, wherein text data representing the reading of the voice data is inserted into the voice data as a digital watermark. In response to receiving the text data, converting the text data into voice data and transmitting the voice data and text data representing the reading of the voice data;
A text database for storing text data received from a terminal to be connected, a plurality of partial text data is extracted from the text data in response to an instruction from the terminal, and the extracted plurality of partial text data is extracted from the speech synthesis site. A blog site including a conversion request unit that stores voice data corresponding to the plurality of partial text data transmitted from the voice synthesis site and text data representing the reading of the voice data in a voice text database;
When the text data representing the reading of the audio data corresponding to the plurality of partial text data is made continuous, the partial text data when the text data representing the continuous reading corresponds to a preset reading prohibition term A text-to-speech conversion service system comprising a content examination site for replacing the voice data corresponding to the above with predetermined voice data.   4. The text-to-speech conversion service system according to claim 3, wherein text data representing the reading of the voice data is inserted into the voice data as a digital watermark. Receive text data from the connected device,
Extracting a plurality of partial text data from the text data in response to an instruction from the terminal;
Converting the extracted partial text data into voice data;
Generating text data representing the reading of the converted voice data;
When the text data representing the reading of the audio data corresponding to the plurality of partial text data is made continuous, the partial text data when the text data representing the continuous reading corresponds to a preset reading prohibition term A text-to-speech conversion service method, wherein the speech data corresponding to the above is replaced with predetermined speech data.

Images