JP2009186498A - Speech synthesis device and speech synthesis program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a speech synthesis device and a speech synthesis program, for changing sound quality, when reading text with a tag structure. <P>SOLUTION: When reading of a document which is displayed in a browser (S1) is requested (S5:YES), a parse tree is obtained from an XML parser (S10). The structure of a tag is checked for each node according to the parse tree, and if a condition is set in a parameter change table, a change value thereof is obtained, and a read character string and the parameter after changing are set in a buffer (S35). Then, the type of the tag is checked, and if a condition is set in the parameter change table for the tag type, a change value thereof is obtained, and the read character string and the parameter after changing are set in the buffer (S40). The set read character string is speech synthesized by using the specified parameter (S45), and output (S50). <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a speech synthesizer and a speech synthesis program, and more specifically to a speech synthesizer and a speech synthesis program that switch voice quality of synthesized speech in accordance with a tag attached to text.

  In general, when generating synthesized speech from text, first, morphological analysis of the text is performed, and phoneme data and parameters (phoneme duration, pitch pattern, power) are generated using the analysis result. The generated phonological data is extracted from the phonological database, and synthesized speech is generated according to the phonological data and parameters. The voice quality of the generated synthesized speech is mainly determined by parameters. For this reason, when it is desired to use a plurality of voice qualities, the parameters are switched and generated.

Such parameter switching may be performed according to user's preference, but a method for outputting different voice qualities according to text attributes (font, type, content, importance, source, etc.) and display form. Has also been proposed (see Patent Document 1).
JP-A-10-260815

  However, there are not so many kinds of attributes attached to text. On the other hand, in the markup language, description is made using various tags for display. In order to obtain an effect close to display even when reading aloud, there has been a demand to read the text by changing the voice quality according to the tag.

  The present invention has been made to solve the above problem, and an object of the present invention is to provide a speech synthesizer and a speech synthesis program for switching voice qualities when reading text having a tag structure.

  In order to achieve the above object, a speech synthesizer according to claim 1 of the present invention is a speech synthesizer that generates synthesized speech from tagged text, and linguistically analyzes the text to give a reading character string. A reading character string assigning means, a phoneme generating means for generating phoneme data and phoneme parameters corresponding to the reading character string given by the reading character string giving means, and a result obtained by parsing the tagged text according to the tag And a change means for changing a parameter generated by the phonological generation means for the text with a predetermined tag based on a result of the parsing acquired by the acquisition means, and a change by the change means Speech synthesis means for generating synthesized speech from the parameters and the phoneme data generated by the phoneme generation means. To.

  Further, in the speech synthesizer according to claim 2 of the present invention, in addition to the configuration of the invention according to claim 1, the changing means changes the parameter for the text with a tag of a predetermined name. It is characterized by.

  According to a third aspect of the present invention, in addition to the configuration of the second aspect of the present invention, the speech synthesizer includes a designation unit that designates the name of a tag whose parameter is changed by the changing unit. Features.

  According to a fourth aspect of the present invention, in the speech synthesizer according to the fourth aspect of the present invention, in addition to the configuration of the first aspect of the present invention, the changing unit is configured to process a text with a tag having a predetermined attribute. The parameter is changed.

  According to a fifth aspect of the present invention, in the speech synthesizer according to the fifth aspect of the present invention, in addition to the configuration of the first aspect, the changing unit changes the parameter for text having a predetermined tag structure. It is characterized by doing.

  According to a sixth aspect of the present invention, in the speech synthesizer according to the sixth aspect of the present invention, in addition to the configuration of the first aspect of the present invention, the change unit is configured to add a text with a tag designating a display format. The parameter is changed.

  A speech synthesis program according to claim 7 of the present invention causes a computer to function as various processing means of the invention according to any of claims 1 to 6.

  The speech synthesizer according to claim 1 of the present invention acquires the result of the syntax analysis according to the tag attached to the text, and changes the parameter of the synthesized speech generated from the text according to the attached tag. Therefore, the voice quality of the synthesized speech can be switched according to the tag, and the display characteristics by the tag configuration can be conveyed by the difference in speech.

  Moreover, since the speech synthesizer according to claim 2 of the present invention can change the parameter according to the tag name in addition to the effect of the invention according to claim 1, the voice quality can be associated with the tag name. For example, only certain columns in the table can be characterized and heard.

  In addition to the effect of the invention described in claim 2, the speech synthesizer described in claim 3 of the present invention can specify the name of the tag whose parameter is to be changed. You can create tagged text to read.

  Further, in addition to the effect of the invention according to any one of claims 1 to 3, the speech synthesizer according to claim 4 of the present invention can change parameters according to tag attributes. For example, the voice quality can be changed.

  In addition to the effect of the invention according to any one of claims 1 to 4, the speech synthesizer according to claim 5 of the present invention can change parameters depending on the tag structure, so that the tag hierarchy, tag repetition, etc. The characteristic of the display expressed by can be made to correspond to the voice quality that can be read out.

  In addition to the effect of the invention according to any one of claims 1 to 5, the speech synthesizer according to claim 6 of the present invention can change a parameter by a tag designating a display format. When is specified, the voice quality can be changed, and the characteristics of the display can be conveyed more clearly by the difference in voice.

  Moreover, the speech synthesis program according to claim 7 of the present invention can exhibit the operational effects of the invention according to any of claims 1 to 6 by being executed by a computer.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a hardware block diagram of the speech synthesizer 100. The speech synthesizer 100 is a so-called personal computer. As shown in FIG. 1, the speech synthesizer 100 is provided with a CPU 10 that controls the speech synthesizer 100. The CPU 10 includes a RAM 11 that temporarily stores various data, and a ROM 12 that stores BIOS and the like. Is connected. A hard disk device 13, an output control unit 14, an input control unit 15, and an audio output control unit 16 are connected to the CPU 10 via a bus. An output device 24 is connected to the output control unit 14, and an input device 25 is connected to the input control unit 15. The output device 24 is, for example, a display, and the input device 25 is, for example, a mouse or a keyboard. A speaker 26 is connected to the audio output control unit 16.

  The hard disk device (HDD) 13 is provided with at least a change parameter storage area 131, an acoustic model storage area 133, a program storage area 134, and other information storage areas 135. The change parameter storage area 131 stores a parameter change table that is information for changing a voice parameter when a tag that satisfies a predetermined condition exists in a sentence to be read out. The acoustic model storage area 133 stores a plurality of acoustic models for outputting sound from the speaker 26. In the program storage area 134, various programs executed by the CPU 10 are stored. In the other information storage area 135, other information used in the speech synthesizer 100 is stored.

  Next, a parameter change table provided in the change parameter storage area of the HDD 13 will be described with reference to FIG. FIG. 2 is a schematic diagram showing the configuration of the parameter change table. The parameter change table stores information for changing the voice parameter according to the tag or structure specified in the target sentence. As shown in FIG. 2, the parameter change table includes “condition”, “parameter”, and “supplementary character string” as data items. “Parameter” includes “acoustic model”, “power”, “pitch”, “speed”, and “sound quality”. In this embodiment, the conditions for changing the voice parameter are “user designation”, “DTD designation”, “style sheet designation”, “attribute designation”, “supplement designation”, “condition designation”, “between tags”. "Relative position", "relationship between tags", and "tag depth". Among these, “user designation”, “DTD designation”, “style sheet designation”, “attribute designation”, “supplement designation”, and “condition designation” are conditions depending on the type of tag. “Relative position between tags”, “Relationship between tags”, and “Tag depth” are conditions depending on the tag structure. In the present embodiment, the analysis result of the XML document to be read out is examined in the order of the tag structure and the tag type, and when the corresponding condition is met, the parameter is changed to the corresponding parameter and the parameter is read together with the read character string. Is set in the RAM 11.

  Hereinafter, each condition will be described. First, “user designation” is a reading condition character string arbitrarily designated by the user. A specification is received as a character string, a tag name is acquired based on the result of analyzing the input sentence, and if the tag is applicable, the acoustic model is changed.

  “Specify DTD” is a condition as to whether the tag is specified in the DTD declaration. DTD (Document Type Definition) is a "document type definition" that defines what tags and attributes are used in a document when it is described in a markup language such as XML. is there. For example, the number of times of appearance immediately below a certain element is always one element, the appearance immediately below a certain element is optional (?? *), the appearance appears more than once (+) directly below a certain element, and the appearance order immediately below a certain element is Arbitrary elements (|), whether there are child elements under a certain element, etc. can be defined. Therefore, in the present embodiment, as shown in FIG. 2, when an appearance optional element, an element that appears once or more, and an appearance order arbitrary element are used as DTD designation corresponding tags, Is going to change.

  “Style sheet designation” is a condition regarding whether or not the tag is designated in the style sheet. A style sheet is a template that gathers together information about the appearance of a document. By applying a style sheet, it is possible to unify the appearance of a plurality of documents, or to separate only layout information from an XML document and describe only the document structure. Typical style sheet languages include CSS (Cascading Style Sheets) and XSL (Extensible Stylesheet Language). The style sheet shown in FIG. 7 is an example of XSL. In this embodiment, as shown in FIG. 2, when the <th> tag, the <td> tag, and the <span> tag are style sheet designation tags and the target node is these tags, the parameters are Is going to change.

  “Attribute designation” is a condition as to whether or not the tag has a specific attribute. Examples of tag attributes include a link tag attribute, an image tag attribute, and a font designation attribute. As shown in FIG. 2, in this embodiment, the link tag and the image tag are attribute designation tags. When the tag of the target node has these attributes, the parameter of the power is set to 0 by referring to the parameter change table. .6 is specified.

  “Supplementary designation” is a condition as to whether or not the tag is designated as being preferably read out by supplementing the tag type when reading out. For example, in the case of a link tag, it is easier to understand if it is supplemented with “It is a link of ~” rather than reading it as it is. As shown in FIG. 2, in this embodiment, the <a> tag and the <img> tag are supplementary designation tags. In the case of the <a> tag, the power parameter is set to 0.6, and the character string “is a link” is supplemented. In the case of the <img> tag, the power parameter is set to 0.6 and the character string “is an image of” is supplemented.

  “Condition designation” is a condition as to whether or not the tag is designated in advance with other tags. For example, a setting can be made such that the numerical value enclosed by the <rank> tag is changed if it is smaller than a specified value, for example, “3”. In the present embodiment, as a condition, when a numerical value is included in a character string surrounded by tags, the parameter is changed depending on whether the numerical value is a positive value or a negative value. As shown in FIG. 2, if the tag surrounds a positive numerical value, the pitch is 1.2 times, and if the tag surrounds a negative numerical value, the pitch is 0.8 times.

  The “relative position between tags” is a condition depending on the depth of the tag. An XML document can be described hierarchically by tags. Hierarchization based on the depth of the tag can be realized by an expression such as indentation in the case of display or in the form of a table. However, since the hierarchy is difficult to understand in the case of reading, the voice quality is changed depending on the depth of the tag. As shown in FIG. 2, depending on the depth of the tag, the power, pitch, speed, and sound quality parameters are 1.1 times if the tag depth is “1” and the tag depth is “1”. If it is “2”, it will be 1.2 times, and if the depth of the tag is “3”, it will be 1.3 times.

  The “relationship between tags” is a condition that defines a tag relationship such that the same tag is next to each other as in a table structure tag, or there is a tag with a different name in the same hierarchy. For example, in the case of a table-structured row, adjacent tags are the same. In the case of a table-structured column, the tags have different names in the same hierarchy. As described above, in the case of the table, the pitch parameter is changed so that the change of the column or row can be understood by reading.

  The “tag depth” is a condition depending on the relative position of the tag. In the case of a tag that designates a relative position, the highest value is the default value, and the power is 0.9 two levels below the highest value. In the deepest tag, the power is 0.6.

  Next, operations in the speech synthesizer 100 will be described with reference to FIGS. FIG. 3 is a flowchart of the main process of the speech synthesis process. FIG. 4 is a flowchart of the tag structure check process executed in the speech synthesis process. FIG. 5 is a flowchart of the tag type check process executed in the speech synthesis process. FIG. 6 is a schematic diagram illustrating an example of a display screen of a browser. FIG. 7 is an explanatory diagram illustrating an example of an XML document. FIG. 8 is an explanatory diagram illustrating an example of a style sheet. FIG. 9 is a schematic diagram illustrating an example of an analysis tree.

  In this embodiment, a document described in the XML language is displayed on the browser, and when the user is instructed to read it out, the parsing result is obtained from the XML parser, and the speech parameters are appropriately changed according to the obtained parsing result. Then, synthesize the speech and read it out.

  When the main process is started, as shown in FIG. 3, first, the designated XML document is displayed on the browser (S1). For example, when the XML document shown in FIG. 7 is displayed, the style sheet shown in FIG. 8 is referred to and the screen shown in FIG. 6 is displayed.

  Next, it is determined whether or not a reading request has been input from the user (S5). Input from the user may be accepted by providing buttons and menus on the screen. If no reading request is input (S5: NO), the process stands by with the screen displayed as it is. If a reading request is input (S5: YES), an parse tree is acquired from the XML parser (S10).

  An XML parser (also called an XML processor) is software for extracting only text data from an XML document or file and converting it into a form that can be easily used by programs and application software. An XML document is described using a tag uniquely defined in XML, but an XML parser can convert this into a specific format and output it. The XML parser performs syntax analysis in the process of conversion, and outputs an error if it does not conform to the XML specification. Depending on the XML parser, the analysis result can be output in a tree structure (analysis tree). In this embodiment, the parse tree output from the XML parser is acquired and used in this way. For example, in this embodiment, it is assumed that an analysis tree as shown in FIG. 9 is acquired.

  Next, it is determined whether or not there is a designation related to reading from the user (S15). As the user designation, for example, the designation can be accepted by a character string such as “emphasize the“ number ”column” in the display screen of FIG. When such designation is made (S15: YES), the input sentence is analyzed (S20), and the corresponding tag name is acquired based on the analyzed result (S25). For example, in the above example, a “pieces” tag that is a tag corresponding to the “number” column is acquired. If there is no user designation (S15: NO), the process proceeds to S30 as it is.

  Next, the parameter change table stored in the change parameter storage area 131 is acquired and placed in the work area in the RAM 11 (S30). Next, the tag structure of the target XML document is checked according to the parse tree acquired in S10 (S35).

  Here, details of the tag structure check process will be described with reference to FIG. As shown in FIG. 4, first, the number i of nodes in the analysis tree is initialized to 0 (S101). Next, it is determined whether or not the value of i is smaller than the total number of nodes (S105). If the value of i is smaller than the total number of nodes (S105: YES), a parameter change value corresponding to the tag depth of the target node is obtained from the parameter change table (S110). As shown in FIG. 2, depending on the depth of the tag, the power, pitch, speed, and sound quality parameters are 1.1 times if the tag depth is “1” and the tag depth is “1”. If it is “2”, it will be 1.2 times, and if the depth of the tag is “3”, it will be 1.3 times.

  Next, it is determined whether or not the tag of the target node specifies a relative position (configured in a hierarchy) (S115). If the tag has a hierarchy (S115: YES), the corresponding parameter change value is acquired from the parameter change table according to the hierarchy (S120). As shown in FIG. 2, in the case of a tag designating a relative position, the highest value is the default value, and the power is 0.9 two levels below the highest value. In the deepest tag, the power is 0.6.

  After acquisition of the parameter change value or when the tag is not a tag for specifying the relative position (S115: NO), it is determined whether or not the same tag is adjacent (S125). For example, in the analysis tree of FIG. 9, the “pan” tag is the same tag next to each other. Such a structure occurs in the case of a table-structured row. For this reason, the parameter is changed so that the line is also expressed in the case of reading out. If the same tag is adjacent (S125: YES), the corresponding parameter change value is acquired from the parameter change table (S130). That is, as shown in FIG. 2, it is regarded as a row change and the pitch is multiplied by 1.2.

  After obtaining the parameter change value or when the same tag is not next (S125: YES), it is determined whether there is a sibling tag (S135). Sibling tags, ie, “name” tag, “pieces” tag, “size” tag, “cc” tag, “rank” tag, “memo” tag, etc. If there is a tag (S135: YES), it becomes a table-structured column, so the parameter is changed so that this is also expressed in the case of reading (S140). That is, the corresponding parameter change value is acquired from the parameter change table, and the pitch is set to 0.9 times as shown in FIG.

  After obtaining the parameter change value or when there is no sibling tag (S135: NO), it is determined whether the tag of the target node has a child tag (S145). For example, the “memo” tag of the analysis tree in FIG. 9 includes a “keyword” tag and a “ref” tag as child tags. If there is a child tag (S145: YES), the corresponding parameter change value is acquired from the parameter change table (S150). That is, as shown in FIG. 2, the sound quality is set to 0.3 times.

  After acquiring the parameter change value or when there is no child tag (S145: NO), the parameter change value acquired in S120, S130, S140, and S150 is associated with the character string to be read out and read out in the RAM 11 The buffer is set (S155).

  Then, 1 is added to i (S160), and the above processing is repeated until the value of i reaches the total number of nodes. When the processing of all nodes is completed (S105: NO), the speech synthesis processing (FIG. 3) is performed. Return to).

  After the tag structure check process (S35) is completed, next, the type of each tag included in the target XML document is checked according to the parse tree acquired in S10 (S40).

  Here, details of the tag type check process will be described with reference to FIG. As shown in FIG. 5, first, the number i of nodes in the analysis tree is initialized to 0 (S201). Next, it is determined whether or not the value of i is smaller than the total number of nodes (S205). If the value of i is smaller than the total number of nodes (S205: YES), it is determined whether or not the tag of the target node is of the type specified by the user. If it is a user designation tag (S210: YES), the parameter change value corresponding to the user designation is acquired with reference to the parameter change table (S215). As shown in FIG. 2, in the case of a user-specified tag, the acoustic model is changed from the initial value “model A” to “model B”. For example, if the user specifies “emphasize the“ number ”column”, the “pieces” tag corresponding to the “number” column is acquired as the user-specified tag in S25 of FIG. Yes. Therefore, if the tag of the target node is a “pieces” tag, the acoustic model is changed to “model B”.

  After obtaining the parameter change value or if the tag of the target node is not a user-specified tag (S210: NO), it is next determined whether the tag is a tag corresponding to the DTD specification (S220). In the present embodiment, as described above, when an appearance optional element, an element that appears once or more, and an appearance order arbitrary element are used as DTD designation corresponding tags, and these elements appear as tags (S220: YES), The parameter is changed (S225) (see FIG. 2).

  For example, in the DTD declaration at the beginning of the XML document in FIG. 7, <! ELEMENT pans (pan +)> is used, so the “pan” element appears to appear one or more times in the “pans” element. . Therefore, when the target node is a “pan” tag, the pitch is 1.2 times when the parameter change table is referred to. Also, because there is <! ELEMENT pan (name, pieces, size, cc, rank, memo?)>, "Name", "pieces", "size", "cc", " Each element of “rank” and “memo” appears arbitrarily. Therefore, when the target node is these tags, the sound quality becomes 0.5 when referring to the parameter change table.

  After acquiring the parameter change value or when the tag of the target node is not a tag corresponding to the DTD designation (S220: NO), it is next determined whether or not the tag is a tag corresponding to the style sheet designation (S230). In the present embodiment, as described above and as shown in FIG. 2, the <th> tag, the <td> tag, and the <span> tag are set as style sheet designation tags, and the target node is the tag. If there is (S230: YES), the parameter is changed (S235).

  For example, in the XML document of FIG. 7, after the DTD declaration, <? Xml-stylesheet type = "text / xsl" href = "pans.xsl"?>, And the style sheet "pans.xsl" (see FIG. 8) ) Is specified. Also, as shown in FIG. 8, in the style sheet, the tags “name”, “pieces”, “size”, “cc”, “rank”, and “memo” are applied to the <td> tag. When the target node is these tags, the pitch becomes 1.2 times by referring to the parameter change table.

  After obtaining the parameter change value or when the tag of the target node is not a style sheet designation tag (S230: NO), it is next determined whether or not it is an attribute designation tag (S240). In the present embodiment, as described above and as shown in FIG. 2, when the link tag and the image tag are attribute designation tags, and the tag of the target node has these attributes (S240: YES), the parameter With reference to the change table, the power parameter is set to 0.6 (S245).

  After obtaining the parameter change value or when the tag of the target node is not an attribute designation tag (S240: NO), it is next determined whether or not it is a supplementary designation tag (S250). The supplementary designation tag is preferably read out by narrowing the tag type when reading out. In the present embodiment, the <a> tag and the <img> tag are supplementary designation tags. When the tags of the target node are these tags (S250: YES), the change value is acquired with reference to the parameter change table (FIG. 2). As shown in FIG. 2, in the case of the <a> tag, the power parameter is set to 0.6, and the character string “is a link” is supplemented. In the case of the <img> tag, the power parameter is set to 0.6 and the character string “is an image of” is supplemented.

  After acquiring the parameter change value or when the tag of the target node is not a supplementary designation tag (S250: NO), it is next determined whether or not it is a condition designation tag (S260). In this embodiment, the condition designation tag changes the parameter depending on whether the numeric value is a positive value or a negative value when the numeric value is included in the character string enclosed by the tag as a condition. I have decided. In the case of a tag that matches such a condition specification (S260: YES), a change value is acquired with reference to the parameter change table (S265). As shown in FIG. 2, if the tag surrounds a positive numerical value, the pitch is 1.2 times, and if the tag surrounds a negative numerical value, the pitch is 0.8 times.

  After acquiring the parameter change value or when the tag of the target node is not a condition designation tag (S260: NO), it is next determined whether or not text is included between the tags (S270). If no text is included (S270: NO), “None” is prepared as an alternative character string (S275), and this is added to the reading character string. When the text is included (S270: YES), the process proceeds to S280 as it is.

  As described above, the parameter change value related to the tag type is acquired, and this is set in the reading buffer in association with each character string (S280).

  Then, 1 is added to i (S285), and the above processing is repeated until the value of i reaches the number of all nodes. When the processing of all the nodes is completed (S205: NO), the speech synthesis processing (FIG. 3). Return to).

  After completion of the tag type check process (S40), speech synthesis is performed by a well-known method using the acoustic model stored in the acoustic model storage area 133 according to the reading character string and the speech parameter stored in the reading buffer. Perform (S45). Then, the synthesized voice is output from the speaker 26 (S50). This completes the reading of the XML document displayed on the browser.

  As described above, according to the speech synthesizer 100 of the present embodiment, an XML document with a tag is displayed, and when the reading is instructed, a speech is generated according to the tag structure and the tag type. The parameters are changed, and speech synthesis is performed according to the changed speech parameters, and reading is performed. Therefore, even in the case of speech reading that tends to be poor in expression as compared with display, it is possible to perform read-out that is devised so that the display state can be understood clearly. Therefore, it is possible to provide convenience to visually impaired users and users who are not able to watch the display screen now.

  In the above embodiment, the CPU 10 that acquires the analysis tree in S10 corresponds to the acquisition unit of the present invention. Moreover, CPU10 which performs a tag structure check process and a tag kind check process by S35 and S40 is equivalent to the change means of this invention. The CPU 10 that executes the speech synthesis process in S45 corresponds to the speech synthesis means of the present invention.

2 is a hardware block diagram of the speech synthesizer 100. FIG. It is a schematic diagram which shows the structure of a parameter change table. It is a flowchart of the main process of a speech synthesis process. It is a flowchart of the tag structure check process performed in a speech synthesis process. It is a flowchart of the tag type check process performed in the speech synthesis process. It is a schematic diagram which shows the example of the display screen of a browser. It is explanatory drawing which shows the example of an XML document. It is explanatory drawing which shows the example of a style sheet. It is a schematic diagram which shows the example of an analysis tree.

Explanation of symbols

10 CPU
11 RAM
13 Hard Disk Device 14 Output Control Unit 100 Speech Synthesizer 131 Change Parameter Storage Area 133 Acoustic Model Storage Area 134 Program Storage Area

Claims (7)

A speech synthesizer that generates synthesized speech from tagged text,
A reading character string assigning unit that linguistically analyzes the text and assigns a reading character string;
Phoneme generation means for generating phoneme data and phoneme parameters corresponding to the reading character string assigned by the reading character string giving means;
Obtaining means for obtaining a result of parsing the tagged text according to the tag;
Based on the parsing result obtained by the obtaining means, for the text with a predetermined tag, a changing means for changing the parameter generated by the phonological generation means;
A speech synthesizer comprising: a speech synthesizer configured to generate a synthesized speech from the parameter changed by the change unit and the phoneme data generated by the phonological generation unit.   2. The speech synthesizer according to claim 1, wherein the changing unit changes the parameter for text with a tag with a predetermined name.   The speech synthesizer according to claim 2, further comprising designation means for designating a name of a tag whose parameter is changed by the changing means.   The speech synthesizer according to any one of claims 1 to 3, wherein the changing unit changes the parameter for text with a tag having a predetermined attribute.   The speech synthesizer according to claim 1, wherein the changing unit changes the parameter for text having a predetermined tag structure.   The speech synthesizer according to any one of claims 1 to 5, wherein the changing unit changes the parameter for a text with a tag for designating a display format.   A speech synthesis program for causing a computer to function as various processing means of the speech synthesis apparatus according to claim 1.

Images