JP3707872B2 - Audio output apparatus and method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To output test information corresponding to a reference level by voice, by switching the read-aloud style matching with the state of a careful read or a skip read according to the reference level for document information, and varying the vocalizing speed of a synthesized voice at skip read time, or the like. SOLUTION: A read-aloud style management part 44 of a voice synthesis part 40 generates a control parameter series by use of control rule sets corresponding to respective document reference levels while reflecting the document reference levels to be judged at a user request from a user input part 20. For example, when a document reference level is judged to be a skip read, a control rule set which increases the vocalization speed is used and a control rule set which reduces the width of pitch fluctuations is used to generate a synthesized voice. When it is judged that a careful read is made to the contrary, a control rule set which increases the width of pitch and power fluctuations is used to generate a synthesized voice.

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an audio output apparatus and method for converting at least text information in document information into audio and outputting the audio.
[0002]
[Prior art]
Research on TTS (Text-To-Speech) conversion that obtains synthesized speech from input text has been conducted for a long time, for example: Literature: Iwata et al .: “Software Japanese Text-to-Speech Synthesis for PCs,” Proc. , 2-8-13, pp. Many TTS systems such as those in 245-246 (October 1993) have been realized so far. These TTS systems are implemented as a single application or hardware that receives text from various other applications and reads it out with synthesized speech.
[0003]
On the other hand, in order to obtain knowledge for generating synthesized speech that is not monotonous, various reading styles, that is, announcement tone, reading tone, dialogue tone, fast tone, slow tone, high voice, low voice, etc. Research has been reported that analyzes the characteristics when changing the pitch and speaking speed of the voice. For example, Abe et al .: “Characteristic analysis of different first story styles,” IEICE technical report, SP93-7, pp. 37-42 (May 1993), Sakata et al .: “Prosodic features and synthesis of conversational speech,” IEICE Research Report, SP95-17, pp. 37-42. 55-62 (May 1995) and Arai et al .: “Use of Pitch Pattern Generation Rules Considering Generation Speed,” Proc. Of the Acoustical Society of Japan, 2-0-10, pp. 238-284 (March 1994) is a research report on the analysis.
[0004]
The reading style is closely related to the prosodic parameters such as pitch, speech speed, and power, changes over time, the segments themselves in the database for speech synthesis, and their selection rules. Is equivalent to controlling the value of the ancestor and its temporal change so that these control parameters correspond to the respective styles.
[0005]
By the way, several application systems have been developed as speech synthesis technology as screen reading software for visually impaired persons. For example, literature, Saito: “Production of software for visually impaired persons,” Information processing, Vol. 36, no. 12, pp. As described in 1116-1121 (Dec. 1995), a device has been devised so that visually handicapped persons who have difficulty in visual feedback at the time of operation can read a document only by operating a key. Specifically, it changes the reading method such as read skip repeated playback, monitors the screen display memory, and reads the text that is already displayed by reading the text being displayed on the screen or the user's cursor operation, etc. A function has been developed to automatically change reading aloud by key operation, sentence reading according to character type or voice reading operation key type, sound reading, and detailed (for example, “Tanbo no Tada”) reading. This instruction regarding how to read out, such as skipping and repeated reproduction, is realized by the user operating keys that are directly associated with these instructions in advance so that visually impaired persons can immediately issue instructions. In addition, the read-out function is basically used to grasp the display location of text on the screen that cannot be visually recognized by the visually impaired. The read-out location is displayed on the screen or is currently displayed on the screen. Is text.
[0006]
[Problems to be solved by the invention]
In general, when reading a document that you want to browse carefully, it is hard to feel stress even if you read it at the same speed as reading aloud aloud, but when you read diagonally or visually, it is much faster than the speed of reading aloud It is normal to run your eyes early. Therefore, when reading text in a document with synthesized speech, it is more natural to read a text in a reading style that is more crisp than a document that is read obliquely than a document that is to be referred to carefully.
[0007]
In addition, in the case of reading carefully and in the case of reading obliquely, the place where attention is paid to or the place where attention is read in the text in the document changes. For example, when reading carefully, there is a tendency to read a document from corner to corner, but when reading a long document diagonally, pay attention to the rough structure of the document, or when searching many documents visually Focus on specific parts such as document styles.
[0008]
In the present invention, which part of the text of the presented document is read out is switched according to the document reference status. For example, when reading carefully, all the text in the document is read out sequentially, and when reading diagonally, the text that is the key in the document structure is read out first. .
[0009]
By the way, the use of a speech reading function that determines a reading style and a reading position of a synthesized voice according to a document reference situation such as oblique reading or careful reading has not been performed so far.
[0010]
In the conventional TTS system, the synthesized speech reading style is uniformly given to the entire text designated by the user as a reading target in accordance with settings such as speech speed and pitch height that have been performed in advance. Further, until the user explicitly changes the setting of the reading style and thereafter the user changes the setting, the reading style is continued in the same reading style. This simply treats speech synthesis as a media-to-speech media conversion function, and leaves the variable parameters to the user when generating synthesized speech to be read out.
[0011]
In the case of screen reading for the visually impaired, the reading style that is associated with the operation key in advance is associated, and the user specifies the reading position with the cursor and selects the reading style with the reading operation key. Not too much.
[0012]
[Means for Solving the Problems]
User information for document information Presentation request time interval Or Presentation request input frequency Detecting at least one of the above Presentation request time interval Or Presentation request input frequency The reference level determination means for determining the reference level so that the reference level for the document information of the user is lower as at least one of the reference level is shorter, and the reference level determined by the reference level determination means Using the control rule set corresponding to this determined reading style The text information in the document information has at least speech synthesis means for converting the text information into speech and outputting the speech.
[0013]
Next, there is a display means for displaying the document information to be referred to, and the number of characters or lines or display area that can be displayed simultaneously on the display means and the number of characters or lines of the text included in the document information to be referred to or the document information Reference level determination means for obtaining a reference level so that the reference level for the document information of the user is lower as the ratio of display area or the size of display characters is larger, and the reference obtained by the reference level determination means level Using the control rule set corresponding to this determined reading style The text information in the document information has at least speech synthesis means for converting the text information into speech and outputting the speech.
[0014]
It has a display means for displaying the document information to be referenced, and the number of characters or lines or display area that can be displayed simultaneously on the display means and the number of characters or lines of text information contained in the document information to be referred to or the display area of the document information. The reference level determination means for determining the reference level so that the reference level for the document information of the user is lower as the ratio or the size of the display character is smaller, and the reference level determined by the reference level determination means Using the control rule set corresponding to this determined reading style The text information in the document information has at least speech synthesis means for converting the text information into speech and outputting the speech.
[0018]
Next, a user input unit that receives a character string input from the user, a character string search unit that searches the document information that refers to the character string input from the user input unit as a search target character string, and the character string search unit The reference level determination means for determining the reference level so that the reference level is higher as the number of locations where the character string to be searched is detected in the text information to be referenced is larger, and the reference level determined by the reference level determination means Using the control rule set corresponding to this determined reading style The text information in the document information has at least speech synthesis means for converting the text information into speech and outputting the speech.
[0019]
Further, user input means for receiving word input from the user, a synonym dictionary for obtaining synonyms of words input from the user input means, and characters for searching in document information referring to the synonyms as search target character strings A column search means, a reference level determination means for obtaining a reference level so that the reference level is higher as the number of locations where the character string search means detects the search target character string in the text information to be referred to is larger. Reference level obtained by reference level judgment means Using the control rule set corresponding to this determined reading style The text information in the document information has at least speech synthesis means for converting the text information into speech and outputting the speech.
[0020]
Further, the step of detecting at least one of the reference time or reference speed of the user with respect to the document information and the reference level with respect to the document information of the user are lowered as at least one of the reference time or the reference speed is shorter. Obtaining the reference level from the obtained reference level Determine the reading style and use the control rule set corresponding to the determined reading style. The text information in the document information includes at least a step of converting the text information into a voice and outputting it.
[0021]
DETAILED DESCRIPTION OF THE INVENTION
FIG. 1 shows a basic configuration of an audio output device 10 of the present embodiment.
The voice output device 10 includes a user input unit 20, a document input unit 30, a voice synthesis unit 40, and a document reference level management unit 50.
[0022]
FIG. 2 shows a more detailed configuration diagram. Each device will be described below.
Processing in the document input part
The document input unit 30 includes a document extraction unit 32 and a document storage unit 34.
The document handled by the document input unit 30 includes text, and the speech synthesis unit 40 reads the text in the document.
[0023]
“Text” is a character or symbol. Includes numbers. The “document” is assumed to include text, but it may be a multimedia document including not only text but also images and audio data. Alternatively, it may be a hammer document such as an HTML (Hyper Text Markup Language) document or an online document (online help).
[0024]
The document extraction unit 32 extracts a document from the document storage unit 34.
The document storage unit 34 is a storage device such as a disk or a memory, stores documents therein, and is provided inside the document input unit 30. In addition, the document is managed by a network address or a document search path that is placed outside connected via a network.
[0025]
When the user input requesting presentation of a new character is transmitted from the user input unit 20, the document input unit 30 extracts the corresponding document from the document storage unit 34.
[0026]
It is assumed that a method for accessing a document is specified as an instruction content from the user in the form of an address, a file search path, or the like. As such a document management method, a method already realized by a network file system or WWW (World Wide Web) can be used.
[0027]
Normal processing in the speech synthesizer
The speech synthesis unit 40 includes a text analysis unit 42, a reading style management unit 44, and a speech synthesizer 46.
[0028]
When the document extraction unit 32 extracts a new document, the speech synthesis unit 40 sends the document or the text content included in the document to the text analysis unit 42, and the analysis result is Reading The text is sent to the speech style synthesizer 46 sequentially to the speech style manager 44, and the text is read out with synthesized speech. In addition, " Reading The “upward portion” means a portion of the text in the document read out by the synthesized speech in the speech synthesizer 46.
[0029]
For a series of processing from the text analysis unit 42 to the speech synthesizer 46, an existing TTS (speech synthesis from text-to-speech text) technology can be used.
[0030]
For example, Sato et al. "Speech synthesis from Japanese text" research practical application report, Vol. 32, no. 11, pp. 2243-2252 (1983) is an example, and specifically, as follows.
[0031]
Performs text analysis, performs prosodic control such as pitch, power, speech rate, etc., selection and connection of speech units, and further performs phonological control such as sound quality control, and generates speech waveforms, Prosodic control and phonological control correspond to processing of the reading style management unit 44, and waveform generation processing corresponds to processing performed by the speech synthesizer 46.
[0032]
Furthermore, it demonstrates in detail.
(Text analysis unit 42)
The text analysis unit 42 analyzes the text in the document, and determines the prosodic control unit that gives the reading, accent, and intonation. The units of prosodic control that give readings, accents and intonations are regularly determined from information on readings and accents for each vocabulary in the dictionary used for language analysis, parts of speech and grammatical information on utilization.
[0033]
(Reading style manager 44)
The reading style management unit 44 can provide information such as the pitch of the synthesized speech, speech speed, and power so that the tone of the announcement, reading, dialogue, or tone such as fast, slow, high or low Prosodic feature parameters, selection of speech units that influence sound quality, and control of phonological feature parameters such as spectral deformation are performed.
[0034]
“Reading style” refers to “annual tone” such as announcement, reading, dialogue, or fast / slow tone, high / low voices. This is achieved by changing the control of prosodic feature parameters such as voice pitch, speech rate, and power, and spectral feature parameters of voice quality.
[0035]
The reading style management unit 44 controls prosodic or phonological feature parameters using a control rule set corresponding to the designated reading style. These control rules collect a lot of utterance data of natural speech of a desired reading style, analyze prosodic or phonological feature parameters, and derive parameter control rules.
[0036]
As an example of such rule derivation, Sakata et al. “Prosodic Features and Synthesis of Dialogue Speech”, IEICE Technical Report, SP95-17, pp. 199-101. 55-62 (May 1995).
[0037]
Specifically, a lot of dialogue voice utterance data is collected, a control parameter of the voice pitch is analyzed, and a control rule is derived by determining an appropriate value as a parameter to be given to the pitch control model at the time of synthesis. These parameter values are determined appropriately for the same reading style depending on the position in the sentence and the difference in accent type. This correspondence is called a pitch control rule.
[0038]
The rules derived for each of these reading styles are as shown in FIG. Reading It is stored in the reading style management unit 44 in the form of a table combined with the reading style, and the corresponding rule is applied at the time of synthesis by designating the reading style. At the time of synthesis, it is possible to obtain the position and accent type in the sentence by text analysis of the input text, obtain the parameters corresponding to each from the above restrictions, and generate synthesized speech.
[0039]
Although the pitch control has been described here, the same method can be applied to other parameters.
Processing in document reference level management
The document reference level management unit 50 includes a correspondence table 52 and a level determination unit 54.
[0040]
In the document reference level management unit 50, as shown in FIG. 2, the level determination unit 54 monitors the input status of a request for presentation of a new document, and determines the reference level of the user's document.
[0041]
Note that the “reference level” is a situation indicating how a user is reading a document. For example, the reference level is a stepwise representation of a state of reading obliquely or carefully. And the reference level is not always determined only at the present moment, but rather, it is a state of reading a document that has been captured with a slight width in time, for example, changing the presentation state of the same document It is assumed that the user is in the middle (no page turning, scrolling, or enlargement) or frequently requests new documents to be presented one after another.
[0042]
There are the following methods for determining the reference level.
(1) First method
As shown in FIG. 4 (example 1), the time T0 at which the request is input is stored in the correspondence table 52 every time a request for presenting a new document is input.
[0043]
When a request for presentation of a new document is input, the level determination unit 54 checks the difference from the time T1 when the presentation of the document that was presented immediately before is requested, and if the time difference T0−T1 is longer than a certain threshold Tth. Judge that it is referring carefully, and conversely, if it is short, it is judged that it is reading diagonally.
[0044]
The flow of this process is shown in FIG. In FIG. 5, the flow from step 11 to step 14 is a process of performing speech synthesis from text, and the flow from step 21 to step 24 is a process of determining the document reference level. In this case, the document reference level determination process in step 231 or step 232 corresponds to the document reference level in step 13. Reading Although it is necessary to perform it before the raising style change process, the process flow from step 11 to step 14 and the process flow from step 21 to step 24 may be performed in parallel.
[0045]
(2) Second method
Since speech synthesis often takes a relatively long processing time, the entire text to be controlled is not processed at once, but a break is set by punctuation or syntactic boundaries, and the text up to the break is processed once. As a voice synthesis. FIG. 6 shows the processing in that case.
[0046]
In the text analysis process (step 12) in FIG. 6, a process (step 122) for finding a break at the final stage of the analysis is performed. The process from step 122 to the control parameter generation process (step 13) according to the reading style corresponding to the reference level and the speech synthesis process according to the control parameter (step 14) are repeated until the entire text process is completed.
[0047]
(3) Third method
In addition to the method of paying attention to the time difference from the previous document presentation request, a threshold value is similarly set for the input frequency obtained by the number of input per unit time of the document presentation request. You may judge whether you are reading diagonally or carefully.
[0048]
(4) Fourth method
As shown in (Example 2) of FIG. 4, the parameter p obtained by weighting and adding to the passage of time may be compared with a certain threshold value Pth.
[0049]
In this case, the user's operation start time and operation end time for the interface of the user input unit 20 are detected, and the user's operation duration is obtained from the time difference. When the value exceeds the threshold, the reference level for the document May be determined to be low. This corresponds to lowering the reference level by determining that a continuous operation on the interface of the user input unit 20 is a user input requesting browsing or quick reading of a document.
[0050]
Conversely, the operation duration time of the user may be obtained, and if the value exceeds a threshold value, it may be determined to increase the reference level for the document. This is equivalent to increasing the reference level by determining that a continuous operation on the interface of the user input unit 20 is a user input requesting a detailed reading of the document.
[0051]
Furthermore, in addition to the user operation duration, a threshold is set for the frequency per unit time of the user operation on the interface of the user input unit 20, and a reference level is set according to whether the frequency of the user operation exceeds the threshold May be.
[0052]
Further, when the user input unit 20 has a plurality of interfaces, it may be determined that each of the user requests “fine reading”, “browsing”, and “fast reading”, and a corresponding reference level may be set.
[0053]
(5) Fifth method
The document reference level may be changed by direct input by the user in addition to the automatic determination as described above. For example, a document reference level is switched by a user operation using a button or a switch for switching such as oblique reading or careful reading as an interface of the user input unit 20.
[0054]
Alternatively, an intermediate level situation may be provided between the situation where the user is reading carefully as in the interface shown in FIG. 7 and the situation where the user is reading obliquely, and the user may adjust the level with a slider. 7A is an explanatory diagram of an interface for setting a document reference level by a slider, and FIG. 7B is an explanatory diagram of an interface for setting a document reference level by a button.
[0055]
In addition, a speech recognition unit is provided as an interface of the user input unit 20 and, for example, when a speech recognition result such as “details”, “details”, “e?” Is obtained, a determination is made to increase the reference level for the document. When a speech recognition result such as “skip” or “skip” is obtained, it may be determined to lower the reference level for the document.
[0056]
Processing by the document reference level in the speech synthesizer
Reflecting such a document reference level, the reading style management unit 44 in the speech synthesizer 40 generates a control parameter series using a control rule set corresponding to each document reference level.
[0057]
For example, when the document reference level is determined to be oblique reading, synthesized speech is generated using a control restriction set that increases the speech rate or a control restriction set that reduces the pitch increase / decrease width. On the other hand, if it is determined that the user is reading carefully, synthesized speech is generated using a control restriction set that increases the range of pitch and power increase / decrease. Of course, not only as in this example, the pitch and power increase / decrease are increased when it is determined that the reading is oblique, and the pitch and power increase / decrease is decreased when it is determined that the reading is performed carefully. The synthesized speech may be generated using the control restriction set.
[0058]
FIG. 3 shows an example in which the processing in the reading style management unit 44 is realized by a correspondence table between the document reference level and the reading style and a correspondence table between the reading style and the control restriction set.
[0059]
As shown in the flowchart of FIG. 8, the text analysis unit 42 in the speech synthesizer 40 determines whether or not the word has a small amount of information by using the part of speech of the dictionary used at the time of language analysis, grammatical information such as utilization, and the like. Can do. In this example, whether an independent word or an attached word is classified, it is determined that the word has a large amount of information if it is an independent word, and the word has a small amount of information if it is an auxiliary word. Further, if the meaning of the word itself is weak although it is an independent word such as a noun or a pronoun used as an adverb, it may be determined that the word has a small amount of information.
[0060]
If it is determined that the word has a small amount of information, as shown in the flowchart of FIG. 9, the sales style management unit 44 changes the sales style so as to increase the utterance speed, and generates synthesized speech according to the change. . Alternatively, the text analysis unit 42 may determine a prosodic control unit that gives a reading, accent, and intonation to a phrase chain excluding unimportant words, and may generate this synthesized speech.
[0061]
The structure of the text in the document can be obtained by performing the structure analysis of the document in which the structured tag such as HTML, SGML, TeX is embedded in the text analysis unit 42 of the speech synthesis unit 40. In this case, a structure can be easily obtained by searching a tag for structuring. In this case, it is possible to change the text reading location according to the document reference level managed by the document reference level management unit 50.
[0062]
For example, as shown in the flowchart of FIG. 10, when the document reference level is oblique reading, it is possible to sequentially read the text from the upper text on the structure and present the sentence structure by voice.
[0063]
Further, when reading such structured text, for example, when reading a chapter title, the chapter titles in the scroll direction are read sequentially, and instructions are given from the user input unit 20 when reading the chapter title. When there is an input, it is possible to jump on the display screen so that the text of the chapter can be seen, or to read the text of the section corresponding to the substructure of the chapter.
[0064]
The above process may be stored in a recording medium such as an FD or a CD-ROM, installed in a computer having a speech synthesizer, and the computer may execute the above process.
[0065]
Second embodiment
A sound output device 100 of the second embodiment is shown in FIG.
The configuration different from the audio output device 10 of the first embodiment is to add a document presentation unit 60 to be presented to the user. The document presentation unit 60 presents the document received from the document withdrawal unit 32.
[0066]
“Presentation” is based on visual display using a display, but is not limited to visual display. For example, when audio data is included, the reproduction output is also regarded as a kind of presentation.
[0067]
“Presentation state” is a state in which the document content is being output to the user. A display state (at the moment) that the size is visually displayed, which part of the entire document is displayed, and which part is hidden without being displayed. This is the playback state (at the moment) of how much audio data is being played and how much of the audio data is being played.
[0068]
When the document presentation unit 60 has a visual output unit such as a display, the document presentation unit 60 can change the presentation state such as enlargement or reduction in accordance with a user instruction. Further, when the document size does not fit in the display size, page turning or scrolling can be performed. These presentation states are changed via the user input unit 20.
[0069]
When the document presentation unit 60 reproduces audio data of a multimedia document, such a change in the presentation state corresponds to volume adjustment, rewinding, fast-forwarding, interruption, or the like.
[0070]
For example, the user input unit 20 may be a device such as a tablet in which a key or a device having a pointing function such as a mouse, a pen, or a touch panel is combined with a display. Alternatively, it may be implemented as a device whose level can be set, such as a switch board or a dial.
[0071]
According to the present invention, it is possible to output by changing the reading style of the synthesized speech in accordance with the degree of interest of the user such as whether the user is reading the document obliquely or reading it carefully.
[0072]
When the document presentation unit 60 is provided as described above, the frequency of input from the user who requests a change in the presentation state of the document such as enlargement, reduction, scrolling, etc. is checked, and if the input frequency is high, the same document is referred carefully. It can be determined that you are reading carefully. Alternatively, the frequency of input from a user requesting a change to the presentation state of a document such as scrolling or page turning is similarly examined, and if the input frequency is high, it can be determined that the reading is performed obliquely. Like the method shown in FIG. 2, the input frequency and time are checked as shown in FIG. 11 using a correspondence table that records the time input from the user to the user input unit 20. Can be done.
[0073]
The above process may be stored in a storage medium such as an FD or a CD-ROM, installed in a computer having a speech synthesizer, and the computer may execute the above process.
Third embodiment
An audio output device 200 of the third embodiment is shown in FIG.
[0074]
In this audio output device 200, the document reference level management unit 50 checks the flow rate of document data when a document is being drawn for the document drawing unit 32 to newly present, and if the flow rate per unit time is large. It can also be determined that you are skipping.
[0075]
The above process may be stored in a storage medium such as an FD or a CD-ROM, installed in a computer having a speech synthesizer, and the computer may execute the above process.
[0076]
【The invention's effect】
According to the present invention, a reading style suitable for a situation in which reading is performed carefully or obliquely is automatically switched from a user operation on the document browser. In addition, the oblique reading can be performed speedily by changing the utterance speed of the synthesized speech during the oblique reading. In addition, when reading diagonally, the speech control can be read aloud by increasing the tone component of the pitch control or by taking a high baseline so that it is possible to check whether the document is in the browsing mode without having to visually check the document. Can be judged by ear.
[0077]
Furthermore, by analyzing the structure of the document and examining the structure of the text in the document, when it is determined that the reading is oblique, the higher-level hierarchy such as the chapter title or section title of the text in the document that could not be displayed on the screen. Sentences are read aloud sequentially and the sentence structure can be shown by voice without checking the screen.
[Brief description of the drawings]
FIG. 1 is a block diagram of an audio output device according to a first embodiment.
FIG. 2 is a detailed block diagram of the audio output device according to the first embodiment;
FIG. 3 is a processing diagram in a reading style management unit;
FIG. 4 is a diagram illustrating determination of a document reference level in a level determination unit.
FIG. 5 is a flowchart of determination of a document reference level from an input time interval of a new document presentation request by a user.
FIG. 6 is a flowchart of determination of a document reference level from an input time interval of a new document presentation request by a user.
FIG. 7A is an explanatory diagram of an interface for setting a document reference level with a slider;
(B) is an explanatory diagram of an interface for setting a document reference level with a button;
FIG. 8 is a diagram showing information amount determination of words in sales text using a part-of-speech and information amount correspondence table;
FIG. 9 is a flowchart of reading using the determination result of the information amount of words in the sales text.
FIG. 10 is a flowchart of a reading portion for a structured document according to a document reference level.
FIG. 11 is a detailed block diagram of an audio output device according to a second embodiment.
FIG. 12 is a detailed block diagram of an audio output device according to a third embodiment.
[Explanation of symbols]
20: User input section
30 ... Document input section
32 ... Document drawer
34 ... Document storage unit
42 ... Text analysis section
44 ... Sales Style Management Department
46 ... Synthesizer
50. Document reference level management section
52 ... correspondence table
54 ... Level judgment part

Claims (6)

  By detecting at least one of the time interval of the user's presentation request or the input frequency of the presentation request for the document information, and the at least one of the time interval of the presentation request or the input frequency of the presentation request is shorter, the document of the user A reference level determining means for obtaining the reference level so that a reference level for information is low, a reading style is determined from the reference level obtained by the reference level determining means, and a control rule set corresponding to the determined reading style is obtained. A speech synthesizing apparatus comprising speech synthesizing means for converting at least text information in the document information into speech and outputting the speech.   It has a display means for displaying the document information to be referred to, and the number of characters or lines or display area that can be displayed simultaneously on the display means and the number of characters or lines of text included in the document information to be referred to or the display area of the document information. The reference level determination means for obtaining the reference level so that the reference level for the document information of the user is lower as the ratio or the size of the display character is larger, and reading from the reference level obtained by the reference level determination means A speech synthesizer comprising speech synthesis means for determining a style and converting at least text information in the document information into speech using a control rule set corresponding to the determined reading style.   It has a display means for displaying the document information to be referenced, and the number of characters or lines or display area that can be displayed simultaneously on the display means and the number of characters or lines of text information contained in the document information to be referred to or the display area of the document information. The reference level determination means for obtaining the reference level so that the reference level for the document information of the user is lower as the ratio or the size of the displayed character is smaller, and the reading style is determined from the reference level obtained by the reference level determination means. A speech synthesizer comprising speech synthesis means for determining and converting at least text information in the document information into speech using a control rule set corresponding to the determined reading style.   User input means for receiving a character string input from a user, character string search means for searching in document information referring to a character string input from the user input means as a search target character string, and the character string search means Reference level determination means for obtaining a reference level so that the reference level becomes higher as the number of locations where the target character string is detected in the text information to be referred to is larger, and a reading style is determined from the reference level obtained by the reference level determination means. A speech synthesizer comprising speech synthesis means for determining and converting at least text information in the document information into speech using a control rule set corresponding to the determined reading style.   User input means for receiving word input from a user, synonym dictionary for obtaining synonyms of words input from user input means, and character string search for searching in document information referring to the synonyms as search target character strings Means, a reference level determination means for obtaining a reference level so that the reference level increases as the number of parts detected in the text information to which the character string search means refers to the search target character string increases, and the reference level A speech style is determined by determining a reading style from the reference level obtained by the determining means, and converting at least text information in the document information into speech using a control rule set corresponding to the determined reading style. A speech synthesizer characterized by the above.   The step of detecting at least one of a reference time or a reference speed of the user with respect to the document information, and the reference so that the reference level of the user with respect to the document information becomes lower as at least one of the reference time or the reference speed is shorter. A level is determined, and a reading style is determined from the determined reference level, and text information in the document information is converted to at least speech and output using a control rule set corresponding to the determined reading style. A speech synthesis method comprising steps.

Images