TW434492B - Hyper text-to-speech conversion method - Google Patents

Abstract

The present invention relates to a system that can convert hyper text into speech signals. The system consists of a hyper text markup language (HTML) analyzer, an articulation control command analyzer, a tag converter, a text converter, and a traditional text-to-speech converter. The HTML analyzer reads and analyzes the input hyper text and divides it into a text content, a HTML tag for marking up the text structure, and an articulation control command for controlling the way of articulation. The articulation control command analyzer analyzes the articulation control command and, in accordance with the command type, stores the command content into a tag corresponding table, a sound effect table, a parameter table, a heteronym table, or a proper noun table. In accordance with the heteronym table and the proper noun table, the text converter carries out text replacement or conversion to correct the text articulation. The tag converter not only analyzes the HTML tag and, in accordance with the tag corresponding table and the parameter table, controls the traditional text-to-speech converter's way of articulation but also, in accordance with the tag corresponding table and the sound effect table, synthesizes the sound effects to allow listeners to easily distinguish the text structure. Based on the way of articulation set by the tag converter, the traditional text-to-speech converter converts the text content converted by the text converter into speech signals.

Description

Printed by the Consumer Cooperatives of the Central Standards Bureau of the Ministry of Economic Affairs 434492 A7 ____ B7_ V. Description of the Invention (/) Background Description A text-to-speech converter is a device that converts text to speech. For people with visual impairments, this device helps them to hear information from the outside world. Under certain circumstances, this device is also an important device for ordinary people to obtain information, such as when driving a car or using a phone. The source of this information may be electronic documents or text information converted through optical scanners and text recognition devices. In everyday life, the sources of electronic information are becoming more numerous and growing in progression, such as e-mail, calendars, electronic news, stock information, and the high-profile global information network. To convert these electronic information into digital voice, if manual recording is used and then digitized, it will not only consume a huge amount of manpower and huge storage space, but also the manual recording method cannot be applied to computer systems to automatically integrate according to user needs. Electronic information generated. How to convert a variety of electronic information that was originally suitable for visual display to speech is a challenge for designers of text-to-speech converters. The most important reason is the presentation of electronic information, not only the presentation of its text content, but also the presentation of these text content, such as capitalization, bold, italics, paragraph patterns, and enumeration patterns in visual display. the way. When performing text-to-speech conversion, these were originally used to control the visual display format and font control. They cannot be directly converted to the punctuation marks in the text content. They are also directly converted to the speech H — the pronunciation of the string. Before and after the text, there are also pronunciation methods, such as the syllabary's broken sounds i_ i—-I— I--I— I --1--. I--t ^ i I- ^ n.: U3, Luo (please read the notes on the back before filling out this page) -2-Consumption Cooperation of Employees of the Central Bureau of Standards, Ministry of Economic Affairs, printed 434492 A7 ____ __ B7 V. Explanation of Invention (〆) The tone is a typical example. To solve these problems, the prior art inventions have proposed various solutions. U.S. Patent No. 5,555,343 discloses a text #tone conversion technology to solve such problems, which includes formatting and font control code processing methods, as well as punctuation marks and specific alphanumeric format processing methods. This method uses the first pre-built form to map the formatting and font control mark to the voice control mark, which is used to control the speed or volume of the pronunciation. This method uses the second pre-built form to change the specific The alphanumeric format corresponds to a spoken text string. These specific alphanumeric formats include numeric strings separated by colons to represent time, numeric strings separated by slashes to represent dates, and text strings separated by slashes to represent file directories. This method uses a third pre-built table, which corresponds to punctuation or mathematical operation symbols into spoken text strings or voice control codes. This method uses a pre-built form to determine whether the input character is articulate or non-pronounceable. When it encounters characters that cannot be pronounced, the first, second, and third pre-built forms are used to determine the proper pronunciation method. U.S. Patent No. 5,634,084 discloses another text-to-speech technology that solves this problem. This method first classifies the input text into categories such as numbers, units of measure, geographic terms, and time and date according to the context's relationship. This classified text is expanded according to one or more different types of abbreviated word lists to correspond to spoken words. For example, the abbreviation of a place name, SF, CA〃, this method can be converted to San Francisco California &quot; It can also convert 'MPEG〃 to spoken' &quot; mpeg &quot;. The size of this paper is applicable to China National Standard (CNS) a4 (uox ^ 7mm) ------- ^ ------ ir _-- ^ ---- β (Please read the back first Please pay attention to this page and fill in this page) 4344 92 Printed by the Consumer Standards Cooperative of the Central Standards Bureau of the Ministry of Economic Affairs-A7 _____B7_ V. Description of the Invention (Even) Due to the popularity of the Internet and the World Wide Web, 'Global Information The Internet has become one of the main sources of electronic information today. Most of the electronic information on the World Wide Web is in the format of Hyper Text Markup Language (HTML), which we call a hypermedia file. The difference between a hypermedia file and other electronic files is that in its original file order, in addition to the content of the file, it also contains a hypermedia tag (HTML tag). Hypermedia tags are text tags defined by the hypermedia markup language. They are used to indicate the content and structure of a document, or to control the display of a document. For example, the following example represents the original file of a hypermedia Zhao file: &lt;! BODY BGCOLOR = # DBFFFF &gt; &lt; body bgcolor = white &gt; <CENTER> &lt; map name = ,, Main1, &gt; &lt; area shape = ' 'rect ”coords =” 157,12,257,112 ”href = ,, Main.html ,, &gt; &lt; area shape = ,, rect,' coords = 5,293,141,393,24Γ 'href ^ VRML.html”> 〈area shape = ”Rect” coords = `` 18,141,118,241 ”href =, 'VRML ._” &gt; <area shape = ”rect” coords = ”] L57,226,257,366” href = ,, Main.htmr, &gt; &lt; / map &gt; &lt; img src = ”Images / Main.gif” usemap = ”# Main” -4-This paper iron scale is applicable to China National Standard (CNS) A4 specification (210X297 male | T) ^^ ~ '(Please first Read the notes on the back and fill in this page} 4344 9 2 A7 B7 Printed by the Consumer Cooperatives of the Central Standards Bureau of the Ministry of Economic Affairs 5. Description of the invention) border II 0 &gt; &lt; / img &gt; &lt; brxbrxbrxbr &gt; &lt; b &gt; &lt; font size = 3 color = black &gt; Welcome to the VR workshop of our company &lt; / font &gt; &lt; a href = "http://www.ccUtri.org.tw" xfont size = 3 color = blue &gt; ITRI &lt; / fontx / a &gt; &lt; font size = 3 color = black &gt; / &lt; / font &gt; &lt; a href = "http://www.ccl.itri.org.tw" &gt; &lt; font size = 3 color = blue &gt; CCL &lt; / fontx / a &gt; &lt; font size = 3 color = black &gt;. We have been &lt; br &gt; developing some advanced technologies as follows, &lt; br &gt; &lt; / b &gt; &lt; ul &gt; &lt; a href = ,, Main.html, s &gt; &lt; lixfont size = 3 color = blue &gt; PanoVR &lt; / font &gt; &lt; / a &gt; &lt; font size = 3 &gt; (A panoramic image-based VR) &lt; / f〇ntxbr &gt; &lt; a href = ”VRML, htmr '&gt; &lt; lixfont size = 3 color = blue &gt; CyberVR &lt; / font &gt; &lt; / a &gt; &lt; font size = 3 &gt; (A VRML 1.0 browser) &lt; / f〇 ntxbr &gt; &lt; / ul &gt; &lt; brxbr &gt; (Please read the notes on the back & fill out this f)} This paper music standard General Chinese National Standard (CNS) A4 specification (2 丨 0X 29? mm) Central Ministry of Economic Affairs Printed by the Consumer Cooperatives of the Bureau of Accreditation Service 434492 A7 _______ B7 ^ _ V. Description of Invention (7) &lt; a href = `` Winner'htmr &gt; &lt; iing src = ,, Images / Winner.gif, border = no &gt; &lt; / img &gt; &lt; / axbr &gt; &lt; / a &gt; &lt; brxbr &gt; &lt; font size = 3 color = black &gt; &lt; br &gt; You are the &lt; img src = 1, cgi-bin / Count.cgi? df = vvr.dat ”border = 0 align = middle: &gt; th visitor &lt; br &gt; &lt; / font &gt; &lt; HR size = 2 WIDTH = 480 ALIGN = CENTER &gt; (C) Copyright 1996 Computer and Communication Laboratory, &lt; br &gt;

Industrial Technology Research Institute, Taiwan, R.O.C, &lt; / BODY &gt; As can be seen from this example, in the original file of the hypermedia file, all are displayable characters, and there is no special control code that cannot be displayed. Hypermedia tags are marked with ', &lt; 〃 and * &gt; 〃, and are divided into a start tag and an end tag. The opening tag starts with ,, &lt; and the ending tag starts with, &lt; / &quot;. Therefore, '', &lt; font size = 3 col〇r = black &gt; 〃 is the start tag of '&quot; font々 hypermedia tag, and &lt; // font &gt; 〃 is the end tag. The hypermedia markup language gives special meaning to the text content marked by the start tag and end tag of the hypermedia tag to express the structure of the document, such as a title, paragraph, enumeration, and table. And the paper standard of these structural elements is in accordance with the Chinese National Standard (CNS) Eight Private Sights (2! 〇 × 297 public celebration) ---; --Beijing (Please read the precautions on the back before filling this page) Printed by the Consumer Cooperatives of the Central Standards Bureau of the Ministry of Economic Affairs 434492 A7 ________B7 5. The description of the invention (?) control. Therefore, the same-hypermedia file ′ will be displayed in different ways on different global information browsers or Gu display systems. In addition, hypermedia tags also allow nested arrangements. For example, in the upper shot, W and '&lt; f⑽ 加 = 3 color = black &gt; &quot; are nested and added to the' Wd_e t〇 workgroup of our company ^ ° Although the above_ is a hypermedia file written in English, the text content of the hypermedia file is allowed in other languages, such as Chinese, Japanese, and Korean. For readers of the Chinese web pages of the World Wide Web, they are no strangers to some English-language proper nouns, such as ^ rld Wide Webm. Therefore, in common Chinese webpages, English proper nouns are often mixed. In technical documents, meeting notices, and memos written in the body language, English proper nouns or abbreviations are often used. As a result, many languages are often mixed in the web-like World Wide Web pages. In addition, we also need to consider the problem that the same string has multiple pronunciations. For example, in the case of Chinese broken sounds, the same son may have different pronunciations in different words or in the context. The above-mentioned traditional text-to-speech converter is not suitable for converting hypermedia files into speech signals. During the conversion process, the first problem that must be solved is how to parse the hypermedia file to identify the hypermedia tag. Because the hypermedia tags are composed of displayable characters, not special control codes, and the hypermedia tags allow nested arrangements, the technical invention of the pre-rhyme-related text-to-speech converter is not applicable to Anatomy of a hypermedia file. Furthermore, the above-mentioned traditional text-to-speech converter cannot solve the problem of broken words. Although some Chinese text-to-speech converters can solve some of the paper's standards, such as 赖 (CNS 丨 ----------------- order --------) Read the notes on the back and fill in this page) 43 ^ 492 A7 B7 Printed by the Consumer Cooperatives of the Central Biaolu Bureau of the Ministry of Economy Problem 0 The present invention focuses on overcoming these problems. SUMMARY OF THE INVENTION According to the method proposed by the present invention, a hypermedia file can be converted into a voice signal, or other purposes can be achieved. The embodiment proposed here is a Computer system for converting hypermedia files into voice signals ^ This computer system includes a hypermedia markup language parser, a pronunciation control command parser, a tag converter, a text converter, and a traditional text-to-speech Is. Hypermedia Marker The language parser reads and analyzes the input hypermedia file, and separates it into text content, hypermedia markup that indicates the structure of the file, and pronunciation control instructions that control the way of pronunciation. Pronunciation control The parsing command parser analyzes the pronunciation control commands. 1 According to the type of the command, each of them is stored in a tag correspondence table, a sound effect table, a parameter table, a broken word table or a proper noun table. The text converter according to the broken word table, In the text content of the media file, 'everyone must modify the pronunciation string to replace the string to correct the pronunciation of the text. At the same time, the text converter also uses the proper noun table to convert all the text strings in the hypermedia file's text content. Replace the pronunciation of the original string with a text string. The tag converter analyzes the hypermedia tags 丄 and controls the pronunciation parameters of the traditional text-to-speech converter according to the tag correspondence table and parameters 纟 to change the text content marked by the hypermedia tag Volume, speed, and rhyme sounding parameters. The Tag Converter also performs sound synthesis according to the tag correspondence table and sound effect table, so that you can hear the structure of the file easily. The traditional text-to-speech converter is responsible for converting The text content is converted into a voice signal according to the pronunciation method set by the label converter. National Standard of the Country (CNS —) Α4 Entertainment Right f ^ Pack-(Please read the notes on the back before filling this page) Order-

. «-1 · 1- · 11 I

.I--t I t- -I, -f-t ', -f-t' Printed by A7, Shellfish Consumer Cooperative, Central Bureau of Standards of the Ministry of Economic Affairs 5. Description of the invention (y) This system is constructed in accordance with the present invention. , Can convert hypermedia files into voice signals, and also solve problems such as broken sounds and mixed languages. This system not only has a compact design, but also has great application and expansion capabilities. This device and method can not only provide personalized text-to-speech conversion, but also make it easier for hypermedia file providers to design hypermedia file speech expressions. 434492 A7 B7 丨 — ~.. V. Description of the invention (7) Brief illustration of the drawing Printed by the Consumer Cooperative of the Central Standards Bureau of the Ministry of Economic Affairs Figure 1 illustrates a specific embodiment of the present invention. Figure 2 illustrates the data flow between the various components of the invention. Figure 3 illustrates the sequence of pronunciation control instructions hidden in a hypermedia file. 4A '4B and 4C illustrate the parameter table, sound effect table, and label correspondence, respectively. Figs. 5A and 5B illustrate the broken word table and proper noun table, respectively. ^ Figure 6 illustrates the steps performed by the file read controller. FIG. 7 illustrates the execution steps of the pronunciation control instruction parser. Figure 8 illustrates the execution steps of a text converter. Figure 9 illustrates the execution steps of a standard converter. DETAILED DESCRIPTION OF THE INVENTION Fig. 1 illustrates a hypermedia speech conversion system 10 constructed in accordance with the present invention. This system is a computer system including a central processing unit U, a main memory 12, a network device 13, a telephone interface device 14, a mouse 15, a sound effect device 16, a display 17, and a storage device 18. This system uses a ribbon cable π to connect these devices 1M8 together. Through this line 19 ', these devices 11-18 can pass wheel instructions or information. The storage device 18 may be a disk drive for storing data and programs (prCeSS). The main memory 12 is also used to store data and programs, but is used to store the instructions currently being executed by the central processing unit 21; The central processing unit ϋ n is the command to handle the data and process the data. Network device 13 is used to connect to a computer network. For example, Ethernet (please read the precautions on the back before filling this page), -5 __________-10-G-law scale is suitable for Chinese national standards. Gs, M7 printed by the Central Standards Bureau, Ministry of Economic Affairs, Consumer Cooperatives -------- -B7 V. Description of the invention (丨 0) Road connector or other types of network cards. The telephone interface device 14 is used to connect with the telephone network. The keyboard and slide & 15 A are used to receive commands or feeds entered by the user. The display 丨 7 displays electronic information in text or graphic form. The sound effect device 16 receives a digital voice signal, and emits a sound or sound effect through a sound cable or a headset. . As shown in FIG. 1, the storage device 8 stores an operating system, an application program hypermedia file courtesy 23, a pronunciation control instruction module 21, and a document reading module 29. The operation line and the application technique are generally well-known techniques and will not be repeated here. File reading module 29 includes file reading, controller 28, text-to-speech converter 27, analysis of hypermedia markup language, 24, pronunciation control command parser 22, tag converter, tag correspondence table 41, parameter table 42 ' Sound effect table 43, text converter 26, broken sound table 31 and proper noun table%. Although the above procedures 22, 24-28 are executed by the central processing unit Π 贞 责 in a time-sharing manner, this is only for the convenience of stating the method of the present invention. The above procedures 彳 22, 24-28 can also use the well-known hardware technology to achieve the same function in a hardware implementation, which is not described in detail here. In addition, the text-to-speech converter 27 and the hypermedia markup language parser 24 are also well-known techniques and will not be described in detail here. The file reading controller 28 is responsible for controlling each program_data flow in the entire conversion process. FIG. 2 shows the data flow in the file reading controller 28. The hypermedia file file 23 may originate from the network device 13 or be read from the storage device 18. The pronunciation control command file 21 is also derived from the network device 13 or read from the storage device. __-π-National Standard (CNS) A4 is now available (------ —II— n HI Ά —I n .—I I ---] —. Ding — — _____ n ^-, T- . '(Read the precautions on the back before filling out this page) Printed by the Central Government Bureau of the Ministry of Economic Affairs, Employees' Cooperatives, 4344 92 A7 ________ ___B7 V. Invention Description ((1) Hypermedia Markup Language Rider 24 Analysis Hypermedia == 'in document mark 23 separates it into text content, a hypermedia label indicating the structure of the file, and a pronunciation command that controls the way of pronunciation ^ The hyper-markup language parser 24 sends the separated hypermedia label to the label converter 25. Hypermedia markup language parser 24 sends the separated text content to text converter 26. Hypermedia markup language parser 24 sends the separated pronunciation control command to pronunciation control command parser 22. Pronunciation control command parser 22 is responsible for analyzing the sound control instructions. The sound control instructions may be stored in a separate sound control instruction file 21 or hidden in the hypermedia file file 23. The sound control instructions are divided into the following four types: (1) Voice control instructions' Its format is : PARAM hypermedia tag attribute pronunciation parameters; (2) Audio control command 'in the format: AUDIO hypermedia tag attribute sound data; (3) Broken word control command, the format is: ALT Broken word _ alternative string before and after text string (4) Proper noun control instruction, its format is: TERM proper noun string translation string. Figure 3 shows a sequence of pronunciation control instructions no, 12〇, 13〇, 140, 150 '160' 170 and 180 These pronunciation control instructions are marked with a comment tag in the hypermedia markup language so that they can be hidden in the hypermedia file. The instruction U0 is a sound control instruction because it uses the ^ PARAIVT identification code 111. This The voice control instruction 11 defines all the text content marked with hypermedia label 113, Lr, -12-This paper size applies to China National Standard (CNS) A4 specification (210 X29 * / mm) ------- -II 1 (Please read the notes on the back before filling out this page) Order 4344 9 A7 B7 Printed by the Central Government Bureau of the Ministry of Economic Affairs, Employees' Cooperatives. 5. Description of the invention (丨 &gt;) Defined according to its pronunciation parameter 115 Ginseng For voice control. In this voice control command 110, the parameters defined by the pronunciation parameter 115 are: speed (speed) is 1.0, volume (volume) is 0.8, and rhythm (pitch) is 1.2. In the voice control command, you can Optionally, attribute tags are added between the media tag 113 and the pronunciation parameter 115. The attribute field is used to limit the scope to which the sound control instruction 110 is applicable. For example, the attribute field may be some attributes of the hypermedia tag, and the use of the attribute field may limit the sound control instruction 110 to only apply to the hypermedia tag 113 having a specific attribute. Command 120 is a sound control command because it uses &quot; AUDICT ID 121. This sound effect control instruction 120 defines that when encountering the hypermedia label 123 * LI〃, sound effect data 125 ~ beep.au〃 must be inserted. In this example, the audio data 125 is an audio data file named beep.au. In the sound effect control instruction 120, an attribute block may be optionally added to limit the scope to which the sound effect control instruction 120 is applicable. When the sound control instruction parser 22 analyzes the sound control instruction and the sound effect control instruction, it will modify the parameter table 42 shown in FIG. 4A or the sound effect table 43 shown in FIG. 4B according to the content of the instruction. The pronunciation control command parser 22 then modifies the tag correspondence table 41 shown in FIG. 4C. As shown in FIG. 4A, when the sound control command parser 22 analyzes the sound control command 110, it adds or modifies the item .42-1 in the parameter table 42. First, the pronunciation control command parser 22 must obtain the usable items 42-1 in the parameter table 42. The item 42-1 can be obtained by inserting a new item in the parameter table 42 or repeatedly using an item that is no longer used in the parameter table 42. After obtaining item 42-1, the pronunciation control instruction -13-this paper size applies the Chinese national standard (CNS 丨 A4 specification (210X297 mm) 1-I-—-H --- r-»——-ί! I β ^-_ I-. »I ί —--- ΐ _n J. (Please read the precautions on the back before filling out this page) 4344 92 Printed by the Shell Sample Consumer Cooperative of the Central Sample Bureau of the Ministry of Economy A7 —______ B7 Description of the Invention (丨 勹) ^ ~~~ 'The parser 22 stores the parameters defined in the sound control command 110 丨 丨 5 into the fields 42-12, 4243, and 42-U of item 42-1, as shown in the figure. The PID field 4241 in 4A is a parameter identification code. This field is for convenience of explanation. 'It can be ignored in practical use. As shown in FIG. 4B, the pronunciation control command parser 22 analyzes the sound control command 120.' Will add or modify the second item in the sound effect table 43. "Item 43-1." First, the sound control command parser 22 must obtain the usable item 43-1 in the sound effect table 43. The way to obtain the item 43-1 can be in the sound effect. Insert a new item in Table 43, or repeat the use of an item that is no longer used in Sound Table 43. After getting item 43_ After that, the pronunciation control command parser 22 stores the sound effect data file name 125 and the sound effect data content in the sound effect control command 120 into the fields 43-12 and M-ΐ 3 of the item 43-1. AID in FIG. 4B Block 43-11 is the sound data identification code. This block is for convenience of explanation and can be ignored in actual use. In addition, in order to save memory space, the use of the pronunciation control command parser 22 before modifying the sound table 43 You can use the sound effect data file name 125 to search the sound effect table 43. If you find that the same file name already exists, you will not modify it. After the pronunciation control command parser 22 modifies the parameter table 42 or sound effect table 43, Then modify the contents of the unmarked correspondence table 首先 I. First, the pronunciation control command parser 22 recognizes the pronunciation control command 丨 or 120, 111 or 121, the hypermedia tag 113 or i23, and the attribute stop retrieval tag correspondence table 41 'and obtains Item 4M or 41-2. In the search, if item 41- 丨 or 41-2 does not exist, a new item or 41-2 is created. Then the sound control command parser 22 will control the sound control command 11,- 14-This paper is a paper size (CNS) A4 specification (21QX 297). F Please read the notes on the back ¾ before filling in this page.) Binding. Order-t ------- 1 1 II [. .1 4344 y? Α7 Β7 V. Description of the invention (丨 仏) Printed by the Consumer Cooperatives of the Central Standards Bureau of the Ministry of Economic Affairs, set the type field 4M3 of head 4i-i to tender, to indicate that it is bad Bit 4H4 points to the parameter table 42t; ㈣ the pronunciation control command parser 22 also sets the index booth 41_14 as the sound control command ι〇 The corresponding item 42 in the parameter table 42-For the sound control command 12, the sound control command The parser 22 sets the pattern bit positions 41-23 of the item to AUDIO 'to indicate that the index position 4 丨 _24 is pointing to the sound effect table 43. At the same time, the analysis of the sound control command H 2 2 will also set the finger position 4 丨 _ 2 4 as the sound control command 12. Corresponding item 43-1 in the sound effect table 43. »In Figure 3, the commands 13, 14Q, and 150 It is a cracked word control instruction. The identification codes 131, 141, and 15 of these instructions are all “, AIjr ^ Each cracked sub-control instruction 30 ′, 140, and 150 define the cracked word strings 133, 143, and 153, and replace the broken word. Phonetic strings 133, 143, and 153 substitute strings 135'145 and 155. The purpose is to use the alternative strings 135, 145 and 155 to make the text-to-speech converter 27 produce the correct pronunciation. The instructions 14 and 15 also define the front and back text strings 147 and 157, which are used to limit the scope to which the instructions 140 and 150 are applicable. As shown in FIG. 5A, when the pronunciation control command parser 22 analyzes the broken word control commands 130, 140, and 150, it will sequentially add or modify items 31-1 '31-2 and 31-3 in the broken word table 31. . The pronunciation control command parser 22 reads the broken strings 133, 143, and 153 and converts them appropriately, and then saves the broken strings 31-11, 31-1, 31-2, and 31-3, 31-21 and 31-31. The pronunciation control command parser 22 reads in the substitute strings 135, 145, and 155 and converts them appropriately, and then stores them in the substitute string fields 3μ of the items 31-1 '31-2 and 31-3 respectively Π' 31 -22 and 31-32. The pronunciation control command parser 22 will read the context I. ---- ^ -----. 1 &quot; i. (Please read the precautions on the back before filling this page) -15-This paper applies the Chinese national standard (CN'S) A4 specification (210 · X 297 mm) 434492 Printed by the Consumer Cooperative of the Central Standards Bureau of the Ministry of Economic Affairs A7 B7 V. Description of the invention (丨 S) Strings 147 and 157 are read and converted into appropriate text, then Deposit the text strings 31-23 and 31-33 before and after items 31-2 and 31-3. In FIG. 3, the instructions 160, 170, and 180 are proper noun control instructions, and the identification codes 161, 171, and 181 of these instructions are &quot; TERM &quot;. Each proper noun control instruction 160, 170, and 180 defines proper noun strings 163, 173, and 183, and translation strings 165, Π5, and 185 that are used in place of proper noun strings 163, 173, and 183. The purpose is to use the translated strings 165, 175, and 185, so that the text-to-speech converter 27 can produce proper pronunciation for proper noun strings 163, 173, and 183. For example, when the text-to-speech converter 27 can only convert Chinese characters, proper noun control instructions 160, 170, and 180 can be used to convert English or _English mixed proper nouns into Chinese voice signals. As shown in FIG. 5B, when the pronunciation control instruction parser 22 analyzes the proper noun control instructions 160, 170, and 180, it will sequentially add or modify items 32-1, 32-2, and 32 in the proper noun table 32. -3. The pronunciation control command parser 22 reads proper noun strings 163, 173, and 183 and converts them as appropriate (detailed later), and then saves them into the specialists of items 32-1, 32-2, and 32-3, respectively. There are noun fields 32-11, 32-21 and 32-31. The pronunciation control command parser 22 reads the translated strings 165, 175, and 185 and converts them as appropriate (detailed later), and stores them into the translation fields of items 32-1, 32-2, and 32-3, respectively. 32-12, 32-22 and 32-32. In Fig. 2, the text converter 26 receives and processes the text content generated by the hypermedia markup language parser 24. The text converter 26 searches for the content of the text to determine whether there is a broken string in the broken word table 31 and a proper noun in the proper name table 32. If you find a broken sound ---------- ^ ------- iT --- ^ ---- 1 -------(Please read the precautions on the back before filling this page) -16-This paper size is applicable to Chinese National Standard (CMS) A4 (2iOx 297 mm) 4344 9 2 A7 B7 printed by the Consumer Cooperatives of the Central Bureau of Standards of the Ministry of Economic Affairs 5. Description of the invention (丨 ^) If there is a noun, the text converter 26 will perform string substitution according to the contents of the phonetic word table 31 or the proper noun table 32. The text converter 26 sends the processed text content to the text-to-speech converter 27. The label converter 25 receives and processes the hypermedia labels generated by the hypermedia markup language parser 24. The tag converter 25 uses this hypermedia tag to search the tag correspondence table 41 to determine whether the hypermedia tag needs to be controlled by sound or effect. If it is found that the hypermedia tag needs to be controlled by voice, the tag converter 25 obtains corresponding parameters from the parameter table 42 according to the index stop of the retrieved items, and sends these pronunciation parameters to the text-to-speech converter 27. If it is found that the hypermedia tag needs sound control, the tag converter 25 obtains the corresponding sound effect data from the sound effect table 43 according to the index field of the retrieved item, and sends it to the sound effect device 16 or the telephone interface device 14. The text-to-speech converter 27 receives and processes the text content from the text converter 26 and the pronunciation parameters from the tag converter 25. The text-to-speech converter 27 changes its parameter settings according to the newly received pronunciation parameters, such as changing the settings of parameters such as the speed, volume, and rhythm of the sound. When the text-to-speech converter 27 receives the text content, it will convert the text content into a voice signal according to the setting value of the pronunciation parameter at that time, and send the result to the sound effect device 16 or the telephone interface device 14. FIG. 6 illustrates the execution steps of the file reading controller 28. In step S1, the hypermedia speech conversion system 10 (the CPU 11 executes the operating system or application program) determines whether it is necessary to read the independent pronunciation control instruction file 21. If yes, the document reading controller 28 will perform the steps (please read the precautions on the back before filling this page) Order-17-This paper size is applicable to China National Standard (CNS) M specifications (2 [0 × 297mm) 4344 92 Printed by A7, B7, Consumer Cooperatives of the Central Standards Bureau of the Ministry of Economic Affairs 5. Description of the invention (丨 1) S2, read the content of the file, and send the content of the file to the pronunciation control instruction parser 22 to analyze the pronunciation control instruction in step S6. After executing step S6, or if it is not necessary to read the independent pronunciation control instruction file 21, the file reading controller 28 executes step S3 to read the content of the hypermedia file file 23. In step S4, the file reading controller 28 passes the contents of the hypermedia file archive 23 to the hypermedia markup language parser 24 · and separates them into file elements. A file element can be a hypermedia tag, a text string of text content, or a pronunciation control instruction. The file reading controller 28 sequentially reads and processes the file components separated by the hypermedia markup language parser 24. In step S5, the file reading controller 28 determines whether the file element separated by the hypermedia markup language parser 24 is a pronunciation control command. If so, the file reading controller 28 executes step S6 and passes the sound control command analysis to the sound control command parser 22. After executing step S6, the file reading controller 28 returns to step S4 to read and process the next file element separated by the hypermedia markup language parser 24. In step S5, if the read file element is not a sound control command, the file reading controller 28 executes step S7. If the file reading controller 28 finds that the file element is a hypermedia tag in step S7, the file reading controller 28 hands the hypermedia tag to the tag converter 25 for sound and sound control in step S8. The file reading controller 28 then returns to step S4 to read and process the next file element. In step S7, if the read file element is not a hypermedia tag, the file reading controller 28 executes step S9. In step S9, the file reading controller 28 regards the read file element as a text content of a text content, and passes it to a text converter (please read the precautions on the back before filling this page). Line- 18-This paper is compliant with China National Standard (CNS) A4 (210X297 mm), and printed by A7 __________ B7 in the consumer cooperation of the Card Central Standards Bureau of the Ministry of Economic Affairs. In step S10, the file reading controller 28 passes the conversion result to the text-to-speech converter 27 for processing. The sound converter 27 will convert it into a voice signal and pass the sound effect device 16 or the telephone interface device 14 Play it out. The file reading controller 28 then returns to step S4 'to read and process the file element. The file reading controller 28 executes these steps S4_Si repeatedly until all the file elements in the hypermedia file file 23 have been processed. FIG. 7 illustrates an execution flow of the pronunciation control command parser 22. In step S11, the pronunciation control command parser 22 reads the pronunciation control command. In step S12, the pronunciation control command parser 22 determines whether the pronunciation control command is a voice control command according to the command ^ identification code. If so, the 'pronunciation control command parser 22 executes step sn, adds or modifies an item in the tag correspondence table 41, and stores the super tag, the command type (here, PARAM) of the command, and the attribute and parameter indicators. Enter the corresponding fields. The pronunciation control instruction parser 22 then executes step S14, and stores the pronunciation parameters defined by the instruction in the parameter table 42. After executing step S14, the pronunciation control command parser 22 returns to step su to read and process the next pronunciation control command. In step S12, if the command is not a voice control command, the pronunciation control command parser 22 executes step S15 to determine whether the command is a valid control command. If yes, the pronunciation control command parser 22 executes step S16, adds or modifies an item in the tag correspondence table 41, and sets the hypermedia tag, command type (here, Aum〇), attributes, and sound data indicators of the command. Save the corresponding field. Then the pronunciation control finger analyzer 22 executes step S25, and the sound effect defined by the instruction ___- 19-CNS) A4 specification (2 丨 0x297 mm) (Please read the precautions on the back before filling this page) I- I —-I r 1— —κ τ »-6 domain

Oi Μ printed by the Central Bureau of Standards Consumer Cooperatives printed A7 B7 V. Description of the invention (丨 and its file contents are stored in sound effects # μ pronunciation control instructions. 1 step Yang, please take and deal with-offender if the instruction is not sound effects Control instruction, pronunciation control command analysis n 22 execute step S17, · order whether ^ ί = Γ J Yes, 'pronunciation control command analysis ... the existing contents of the phonetic word table 31, the finger ^ conversion. That is,' target For each m of the broken sound table 31, is there a broken sound string that needs to be converted for this instruction. On the right, a broken sound string that needs to be converted is found in the correction order, and the command parser 22 is sent. … OK; ^ According to the existing contents of the distinguished name list 32, the command pairs "nouns ..." for each item: check α! Φ I I 疋 there are * proper noun strings to be converted. If you refer to: It found that the proper noun string that needs to be converted 'pronunciation control; 22 then convert it into the corresponding lesson text string. After executing step 519, = analysis of the control instruction H 22 execute step S2Q to save the converted special instruction People proper nouns table 32. Ran After that, the pronunciation control finger analysis 22 returns to step S11, and reads and processes-the pronunciation control command is only β. In step SH ', if the instruction is not a proper noun control instruction, the pronunciation control instruction parser 22 executes step S2 to determine the Whether the command is a cracked word control command. If it is, the 'pronunciation control command parser M executes step S22, according to the existing content of the cracked word table μ, the

The Paper Size Office (CNS) Gift Box (21C) X (Please read the notes on the back before filling this page)

Α7 Β7 Printed by the Consumer Cooperatives of the Central Standards Bureau of the Ministry of Economic Affairs 5. Description of Invention (&gt; 0) Order text conversion. That is, for each item of the broken word table 31, it is sequentially checked whether there is a broken string in the instruction that needs to be converted. If a broken string to be converted is found in the command, the pronunciation control command parser 22 converts it into a corresponding substitute string. After executing step S22, the pronunciation control instruction parser 22 executes step S23 to store the converted broken word control instruction in the broken word table 3 丨. The pronunciation control command parser 22 then returns to step SU, and reads and processes the next sound control command. In step S21, if the instruction is not a broken word control instruction, the pronunciation control instruction parser 22 executes step S24. In step S24, the utterance control command parser 22 treats the read data as a comment, and therefore ignores it. The pronunciation control command parser 22 then returns to step S11 to read and process the next pronunciation control command. The pronunciation control command parser 22 will repeatedly execute steps S11-S24 until all the pronunciation control commands have been processed. FIG. 8 illustrates an execution flow of the character converter 26. In step S31, the text converter 26 reads the text content of the hypermedia file file 23. Next, the text converter 26 executes step S32, and performs text conversion on the text content according to the existing content of the broken sound word table 31. In other words, the text converter 26 sequentially checks each item of the broken character table 31 for broken characters in the text content to be converted. If a broken string to be converted is found in the text content, the text converter 26 converts it into a corresponding substitute string. In order to perform efficiently, when the text converter 26 performs text conversion on an item of the broken character table 31, it first searches the text content to find the bit of the broken character string of the item nil —-I n 1 —-1 nnnnn 1— T nt— ί I. _ 1% (Please read the notes on the back before filling in this page) This paper size applies to China National Standard (CNS) A4 (210X297 mm) 434492 Employees of the Central Bureau of Standards of the Ministry of Economic Affairs The cooperative prints A7 B7 and the description of the invention (called 2 and then placing the last character or string, plus the broken string, and the purpose of the text string before and after to determine whether the ^ broken 9 string content Replace with the item's alternative string. If = ', the text converter 26 will replace the = string with the item's alternative string! If the item does not define a text string before and after, the text to 26 will be replaced directly取代 to replace this broken sound string. Afterwards, the post sub-converter 26 continues to search for the text content in this way, and ======================================================================================================== This is the end of the search. 26 Then execute the steps such as text conversion of the text content converted in step S32 according to the existing contents of the proper noun table 32. That is to say, the text converter% will aim at each item of the proprietary name 32. Instruction whether there is a proper noun word that needs to be converted. If a proper noun string that needs to be converted is found in the text content, the text converter% will replace it with the corresponding translation string. The text converter 26 is When an item of the proper noun table 32 is used for text conversion, first search the text content to find the proper noun string of the item, and directly replace the proper noun string with the translated string. After processing, the text The converter 26 continues to search for the text content, and finds and processes the proper noun string of the item until the text content is searched. After executing step S33, the text converter 26 returns to step S31, reads Take and process the next string of text generated by the Hypermedia Markup Language Parser 24. ______- 22-This paper is approved by the GF countries (CNS) 7 ^ 77-297 public meals_) (Please first listen Note to fill out the back of this X)

434 ^^ 434 ^^ Printed by the Consumer Cooperatives of the Central Standards Bureau of the Ministry of Economic Affairs-23-A7 B7 V. Description of the invention (Figure 9 illustrates; ^ sign the converter 25 execution steps. In order to handle the nested arrangement of hypermedia labels The label conversion stomach 25 uses a stack stored in the main memory hip i2 or the central processing unit u (likely, in order to facilitate the conversion process of the hypermedia tag. In step, for a hypermedia, markup language parser The hypermedia label generated by 24, and the label conversion benefit 25 first determines whether it is a start label. If the hypermedia label is a start label, the label converter 25 executes step s42 to push it into a stack. Otherwise, the tag converter executes step S43 and pops (p0p) a hypermedia tag from the stack. In step S44, the tag converter 25 performs label conversion on the hypermedia tag at the top of the stack, and uses the hypermedia Tag money tag label on the shore 41 ° At step S45, the tag converter 25 determines the parameter setting item corresponding to the hypermedia tag material (its type is PARAM). If there is a tag conversion, Device Step 25: Use the parameter index of this item to read the corresponding parameters from the parameter table 42. Then, the label is transferred to the text-to-speech converter 27 and transferred to the text-to-speech converter 27 to change the pronunciation of the text content in the future. After executing step S46 'or the hypermedia tag has no corresponding parameter setting item, the standard converter 25 executes step S47, and determines whether the hypermedia tag has a corresponding sound control item (its type of booth) according to the search result. It is AUDIC. If there is, the tag converter 25 executes Buma S48 'to use the item's sound effect f material index, and reads the corresponding sound effect data from the tone 43. Weaving standard __ 25 sends this sound effect data Broadcast to sound effects device 16 or telephone interface device 丨 4. This paper New Zealand Financial (CNS) Mim (210 &gt; ^ 97 ^ (Please read the notes on the back before filling in this card-Binding 1. 434492 A7 B7 Five 2. Description of the invention (In step S47, if the hypermedia tag does not have a corresponding sound effect control item, the tag converter 25 executes step S49, and ignores the hypermedia tag and does not process it. After performing step S48 or step S49, the tag converter 25 Return to step S41 and wait to process the next hypermedia tag generated by the hypermedia markup language parser 24. The stack used by the tag converter 25 can ensure that the inner hypermedia element (HTML dement) can use its own voice And sound effect control. At the same time, when returning to the upper-level hypermedia element, the sound and sound effect control used by the element can still be restored. The above-mentioned embodiments are only for explaining the method proposed by the present invention. Different embodiments can still be derived without the spirit and scope disclosed in the following patent application scope. --------- ^ ------- 丨 Order --- ^ ----- 竦 •, (Please read the notes on the back before filling this page) The paper size printed by the consumer cooperative is applicable to the Chinese National Standard (CNS) A4 specification (210X 297 mm)

Claims (1)

^ 4 3 44 92 ABCD VI. Scope of Patent Application 1. A computer system for converting hypermedia files into voice signals, including: a hypermedia markup language parser, separating a hypermedia markup language format file into text content and markup file structure Hypermedia tags and pronunciation control instructions that control the way of pronunciation; a pronunciation control instruction parser, analyzes the pronunciation control instruction, and modifies the tag correspondence table, sound effect table, parameter table, broken word table and proprietary according to the content of the pronunciation control instruction Noun list; a text converter, which specifies the pronunciation string that must be modified in the broken pronunciation list. When it appears in the text, the pronunciation is modified according to the way specified in the broken pronunciation list. The string specified in the table must be translated. When it appears in the text, it will be translated according to the method specified in the proper noun table. A label converter, according to the label corresponding table, the pronunciation parameters must be modified or inserted. The hypermedia label of the sound effect, when it appears in the hypermedia file, The sound effect table items specified in the table are inserted into the sound effect data, and the corresponding items of the parameter table specified in the table corresponding to the label are used to modify the pronunciation parameters of the text content marked by the hypermedia label; Printed by the Standards Bureau Consumer Cooperative (please read the notes on the back before filling out this page) A text-to-speech converter converts the text content to the voice signal through the modified result of the text converter and the label converter. 2. —A method for converting hypermedia files into speech signals in a pronunciation control instruction parser, including the following steps: analyzing sound control instructions specifying pronunciation parameters such as volume, speed, and prosody that should be used for a particular hypermedia tag, and Analyze that the specified specific ultra-paper size is applicable to the Chinese National Standard (CNS) A4 specifications (2x297 mm) 434492 A8 B8 C8 DS 3. Printed by the Consumer Cooperative of the Central Standards Bureau of the Ministry of Economic Affairs The sound effect control instruction of the sound effect data, and according to the content of the sound control instruction, the pronunciation parameters are stored in the parameter table item, and the corresponding relationship between the hypermedia standard and the parameter table item is set in the tag correspondence table, and according to The content of the sound effect control instruction stores sound effect data into the sound effect table item, and sets the corresponding relationship between the hypermedia label and the sound effect table item in the tag correspondence table; and analyzes the specific content of the text in the specified hypermedia file. The way to change the pronunciation of Zi Zai and the broken word control instructions of the text string before and after, Analyze a proper noun control instruction that specifies a translation method for a specific string in the text content. The pronunciation modification method and the translation method are text strings that can be converted into a speech signal by a text-to-speech converter. The content of the broken word control instruction generates a broken word table, which stores the specific string and the corresponding pronunciation modification method and the before and after text strings, and generates a proper noun table according to the content of the proper noun control instruction, and stores it. The specific string and the corresponding translation method. Steps for converting a hypermedia file into a voice signal in a text converter: The special character string that must be modified according to each item of the word list, and the specific string in the text content of the hypermedia file is used for its specified purpose. For the replacement of the pronunciation of the alternative string, the alternative string can specify multiple pronunciations of the specific word _ as a specific pronunciation; and, this paper must be translated for each item of the proper noun list Ruler, .I ί, π ..-I--J-I s _ I__I (Please read the precautions on the back before filling in this tile) 4344 9 VI. Application for patent scope A8 B8 C8 D8 Central Shake Office of the Ministry of Economic Affairs Printed by the Shellfish Consumer Cooperative The following steps-28-= substring, replace the specific word in the text content of the hypermedia file = with the specified translation string for translation instead of 'the translation word ...: to make the special ^ string The middle part that cannot be converted into a voice signal is converted into a specified voice signal by a text-to-speech converter. 4 .: The following steps are to convert tags into ^ to change the super ^ into a voice signal to modify the pronunciation parameters and insert sound effects: Establish a hypermedia tag for each item in the tag correspondence table. , Modify the content of the hypermedia label: hypertext label in the hypermedia file according to the number table item specified by the corresponding item of the label to modify the volume, speed, and foot special parameters; and for each label in the corresponding table of the label The project requires that a hypermedia tag of the f effect be inserted, and the hypermedia tag in the hypermedia file is generated in accordance with the sound effect table item specified by the tag correspondence table item. 5. — A step of converting a hypermedia file into a voice signal: W Analyze the pronunciation control instruction. This step includes the following steps. ^ — For the sound control instruction, generate a tag correspondence retrieved by the hypermedia label specified by the sound control instruction. A table item, and generating a parameter table item, storing the volume, tempo, and prosody parameters specified by the sound control instruction; and in the label corresponding table item, setting an index in its index field to point to the parameter table item ; National Standard for Good Paper Standards (CNS) Α4ίι ^ ~ ^ 0χ297 Public Luxury --------- ^ ---- „—., 玎 (Please read the notes on the back before filling in this page) The scope of the patent application for the sound effect control instruction generates a tag correspondence table item retrieved by the hypermedia tag specified by the sound effect control instruction, and generates a sound effect table item, stores the sound effect data specified by the sound effect control instruction, and stores the sound effect data specified in the sound effect control instruction. In the corresponding table item, set the index in the index block to point to the sound table item; For the broken word control command, generate a broken word table entry that is specified by the broken word control command to specify a specific string that must be modified to be retrieved and stored. An alternative string specified by the broken word control instruction, which can convert the specific string with multiple pronunciations into one of the specific pronunciations; and for a proper noun control instruction, a proper noun is generated The control command specifies a proper noun table entry that must be retrieved by translating a specific string, and stores the translation string specified by the proper noun control command. This translation string can make it impossible to convert from a text-to-speech converter to a voice signal. To convert that specific string to a specific voice signal. β 1 -------- Order (Please read the precautions on the back before filling out this page) Printed by the Consumers' Cooperative of the Central Government Bureau of the Ministry of Economic Affairs 6 · As described in item 5 of the scope of patent application, enter The step includes the following steps: the method described in item 5 of the scope of patent application from the -transfer + transfer of job control instructions of the super file, the step further includes the step of reading each of the pronunciation control instructions. -29 Specifications of this paper · Regulation 43449P ABCD VI. Application scope of patent 8. The method described in item 5 of the scope of patent application further includes the following steps: Analyze the data of a hypermedia file; Retrieve the wearing correspondence table with the hypermedia tag. For the tag correspondence table item obtained through the retrieval, use the index in its index field to obtain the parameter table item and the sound effect table item; use the parameter table item. A set of pronunciation parameters, modify the pronunciation parameters used by the text-to-speech converter in the future; and insert the sound effect data stored in the sound effect table item. 9. The method according to item 8 of the scope of patent application, further comprising the steps of: when encountering a start tag in a hypermedia tag, pushing the start tag into a stack; and encountering a hypermedia tag When the end tag of the tag is popped, a hypermedia tag is popped out of the stack, and the specific hypermedia tag used for retrieval is a hypermedia tag located at the top of the stack. (Please read the precautions on the back before filling this page) Printed by the Central Consumers Bureau of the Ministry of Economic Affairs, Consumer Cooperative 10. The method described in item 5 of the scope of patent application, further includes the following steps: Analyze the data of a hypermedia file Search for the text content of this hypermedia file; -30-This paper size uses the Chinese National Standard (CNS) and private opinion (210 X 297 mm) 434492 A8 BS C8 D8 Printed by the Consumer Standards Cooperative of the Central Standards Bureau of the Ministry of Economic Affairs System 6. The scope of the patent application shall replace all the strings in the text content with the pronunciation string that must be modified in accordance with the provisions of the broken character list, and replace the specified strings in the broken character list item; and in the text content, Any string that must be translated in accordance with the provisions of the proper noun list must be replaced by the translation string specified in the proper noun list item. 11 'The method as described in item 10 of the scope of patent application, wherein the broken word list item also contains a context block' The text content meets the rules of the broken word list item and the pronunciation string must be modified. It must also conform to the pre- and post-text strings specified in the pre- and post-fields of the item. In this case, 'the string will be replaced with the alternative string specified in the broken-character table entry. 12. The method according to item 11 of the scope of patent application, further comprising the following steps: generating a voice signal, the content of which includes the voice signal generated by the sound effect data, the text content of the hypermedia file, the substitute word String and the translated string, the speech signal converted according to the pronunciation parameter. -------—____ '31-This paper size applies to China National Standard (CNS) A4 (210X297 mm) -------------- ft ----- · —— Order (Please read the notes on the back before filling this page)