JP4319851B2 - Reading apparatus, reading method and reading processing program - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a reading device that reads a text such as an email, a reading method, and a reading processing program.
[0002]
[Prior art]
A conventionally used TTS (Text To Speech) device (hereinafter referred to as a reading device) is configured as shown in FIG. 5, for example. That is, a pronunciation definition database 1 in which a word character notation and its pronunciation data are defined one-to-one, an analysis unit 2 that analyzes a character string to be read out, and a pronunciation data conversion unit that outputs the pronunciation data for each word 3 and a speech synthesizer 4 for outputting synthesized speech based on the pronunciation data.
[0003]
The conventional reading apparatus having such a configuration operates according to the flowchart shown in FIG. That is, text character string data in which a sentence composed of words in which kana / kanji / alphanumeric characters are mixed is input to the analysis unit 2 and divided into each word constituting the sentence (step 601). Subsequently, the pronunciation data conversion unit 3 refers to the pronunciation definition database 1 to search for pronunciation data corresponding to each word constituting the sentence, and constructs and outputs a pronunciation data string for the input text sentence ( Step 602). The output pronunciation data string is input to the speech synthesizer 4, and the text character string data is read out based on the pronunciation data string (step 604).
[0004]
[Patent Document 1]
Japanese Patent Laid-Open No. 2003-14485
[Problems to be solved by the invention]
However, the conventional reading apparatus as described above has the following problems.
That is, if the character string data input to the reading device is, for example, a character string as shown in FIG. 7, the underlined characters “A” and “AM” may have the same notation. From the contextual structure, A (1) is pronounced as “A” as an alphabet, and A (2) is pronounced as “A” as an English prefix. AM (1) is pronounced “am” as an English verb, and AM (2) is pronounced “AM” as a radio modulation method. Therefore, the string shown in FIG. 7, "er, I Am A Member of AMR Tokyo fan club." Is to read aloud and correct.
[0006]
However, in the conventional reading device, since the character string and the corresponding pronunciation data can be defined only one-on-one in the pronunciation definition database 1 of FIG. 5, only one of the readings can be realized. As a result, the character string shown in FIG. 7, and "er, I Am er members of Am Tokyo fan club.", Apparently strange reading has been made, there is a case that would provide a meaning unknown voice information.
[0007]
Patent Document 1 is an example of such a conventional technique, and when there are a plurality of readings such as a native language and a local language, the course is determined by using each of the readings (both the native language and the local language). The voice of the guidance sentence is output. However, when only one reading method is required, when there are a plurality of reading methods in the dictionary, which one is to be uttered is determined in advance according to a rule, and therefore, appropriate utterance may not be made.
[0008]
The present invention has been proposed in order to solve the above-described problems of the prior art, and the object of the present invention is to reduce the occurrence of erroneous reading in a sentence having the same notation different pronunciation phrase. Another object of the present invention is to provide a reading apparatus, a reading method, and a reading processing program with high accuracy.
[0009]
[Means for Solving the Problems]
To achieve the above object, the reading device according to claim 1 refers to a pronunciation definition database storing one or a plurality of readings for each word constituting a character string to be read out, and the pronunciation definition database. A pronunciation data detection unit that detects the pronunciation data of each word, a determination unit that determines whether or not there is a plurality of pronunciation data detected for each word, and when there are a plurality of detected pronunciation data, An intermediate pronunciation data creation unit that synthesizes pronunciation data to create intermediate pronunciation data, and pronunciation data determined as a single reading by the determination unit and / or created by the intermediate pronunciation data creation unit And a speech synthesizer that outputs synthesized speech based on the intermediate pronunciation data.
[0010]
Further, the reading method according to claim 3 captures the invention according to claim 1 from the viewpoint of the method, and reads one or a plurality of readings for each word constituting the character string to be read. Referring to the stored pronunciation definition database, pronunciation data detection processing for detecting pronunciation data for each word, determination processing for determining whether or not there is a plurality of pronunciation data detected for each word, and detected pronunciation data In a plurality of cases, the intermediate pronunciation data creating process for synthesizing those pronunciation data to create intermediate pronunciation data, the pronunciation data determined as a single reading in the determination process, and / or the intermediate pronunciation data And voice synthesis processing for outputting synthesized speech based on the intermediate pronunciation data created by the pronunciation data creation processing.
[0011]
The invention according to claim 5 captures the invention according to claim 3 from the viewpoint of a computer program, and reads out a character string to be read out by controlling the computer. In the computer, the program refers to the pronunciation definition database storing one or more readings for each word constituting the character string to be read out, and detects pronunciation data of each word. A determination step for determining whether or not there is a plurality of pronunciation data detected for each word; and if there are a plurality of detected pronunciation data, the pronunciation data is synthesized to create intermediate pronunciation data Intermediate pronunciation data creation step, pronunciation data determined as a single reading in the determination step, and Or, on the basis of the intermediate sound data generated in the intermediate sound data creation step, characterized in that to execute a speech synthesis step of outputting synthetic speech.
[0012]
According to the inventions of claim 1, claim 3 and claim 5 having the above-described configuration, for a word having a plurality of readings, intermediate pronunciation data obtained by synthesizing the plurality of readings is created. Can be output as a voice that can be heard by both. As a result, it is possible to greatly reduce the occurrence of erroneous reading out in a sentence in which the same notation different pronunciation phrase exists.
[0013]
The reading device according to claim 2 includes an extraction unit that extracts the same notation different pronunciation word from the character string to be read, an intermediate pronunciation data definition database that defines intermediate pronunciation data for the same notation different pronunciation word, A first pronunciation data conversion unit that converts the same pronounced different pronunciation word into intermediate pronunciation data with reference to the intermediate pronunciation data definition database, and a single pronunciation that defines pronunciation data for a word having a single reading A data definition database, and a second pronunciation data conversion unit that converts character string data that has not been converted by the first pronunciation data conversion unit into predetermined pronunciation data with reference to the single pronunciation data definition database And outputting synthesized speech based on the pronunciation data converted by each of the first pronunciation data conversion unit and the second pronunciation data conversion unit. And having a speech synthesizer.
[0014]
Further, the reading method according to claim 4 captures the invention according to claim 2 from the viewpoint of the method, and includes an extraction process for extracting the same notation different pronunciation word from the character string to be read. A first pronunciation data conversion process for converting the same notation different pronunciation words into intermediate pronunciation data with reference to an intermediate pronunciation data definition database defining intermediate pronunciation data for the same notation different pronunciation words; A second pronunciation that converts character string data that has not been converted by the first pronunciation data conversion processing into predetermined pronunciation data, with reference to a single pronunciation data definition database that defines pronunciation data for words having Based on the pronunciation data converted by each of the pronunciation data conversion processing and the first pronunciation data conversion processing and the second pronunciation data conversion processing, Characterized in that it comprises a speech synthesis process of force.
[0015]
The invention according to claim 6 is the computer program for reading out a character string to be read out by controlling the computer by capturing the invention according to claim 4 from the viewpoint of a computer program. In the computer, the program refers to the extraction step of extracting the same notation different pronunciation word from the character string to be read out, and the intermediate pronunciation data definition database defining the intermediate pronunciation data for the same notation different pronunciation word. A first pronunciation data conversion step for converting the same notation different pronunciation words into intermediate pronunciation data, and a single pronunciation data definition database that defines pronunciation data for words having a single reading; Character string data that has not yet been converted by the first pronunciation data conversion step Speech synthesis for outputting synthesized speech based on the second pronunciation data conversion step for converting to the second pronunciation data, and the pronunciation data converted by each of the first pronunciation data conversion step and the second pronunciation data conversion step Steps are executed.
[0016]
According to the invention of claim 2, claim 4 and claim 6 having the above-mentioned configuration, intermediate pronunciation data is defined in advance for words having a plurality of readings. By simply obtaining intermediate pronunciation data by referring to the definition database, it can be output as a voice that the user can hear. As a result, it is possible to greatly reduce the occurrence of erroneous reading out in a sentence in which the same notation different pronunciation phrase exists.
[0017]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention (hereinafter referred to as embodiments) will be specifically described with reference to the drawings.
Each function of the present invention is generally realized by controlling a computer with software. In this case, a storage device such as a register, a memory, or an external storage device included in the computer temporarily holds or permanently stores information in various formats. Then, the CPU adds processing such as processing and determination to these pieces of information according to the software, and further controls the order of processing.
[0018]
The software for controlling the computer is created by combining instructions corresponding to the processing described in each claim of the present application and the present specification, and the created software is in the form of assembled or compiled embedded software or the like. The above hardware resources are utilized by executing the above.
[0019]
However, the above-described aspects for realizing the present invention can be variously modified. For example, a recording medium such as a ROM chip or a CD-ROM storing software for realizing the present invention can be used alone. It is one embodiment of the invention. Also, some of the functions of the present invention can be realized by a physical electronic circuit such as an LSI.
[0020]
(1) First Embodiment (1-1) Configuration FIG. 1 is a block diagram showing the overall configuration of a reading apparatus according to this embodiment. That is, the reading device of the present embodiment includes a word extraction unit 10 that extracts a word to be processed from an input sentence (text data), and a pronunciation definition database 11 that stores one or more readings of each word. The pronunciation definition database 11 is referred to, a pronunciation data detection unit 12 that detects the pronunciation data of a word to be processed, a determination unit 13 that determines whether or not there is a plurality of pronunciation data detected for the word, When there is a plurality of detected pronunciation data, the pronunciation data are combined to create intermediate pronunciation data, and the intermediate pronunciation data creation section 14 that creates intermediate pronunciation data, and the determination section 13 reads a single reading for each word. A pronunciation data storage unit 15 for storing the determined pronunciation data or the intermediate pronunciation data created by the intermediate pronunciation data creation unit 14, and a pronunciation data for each word Based on the data, and a speech synthesis unit 16 for outputting a synthesized speech.
It is assumed that the method of synthesizing the plurality of pronunciation data in the intermediate pronunciation data creation unit 14 is set in advance.
[0021]
(1-2) Operation The flow of processing in the reading apparatus of the present embodiment having the above-described configuration will be described with reference to the flowchart shown in FIG.
First, the word extraction unit 10 acquires a character string (text data) to be read out, and extracts one word from the character string (steps 201 and 202). Then, for each extracted word, the pronunciation definition database 11 is referenced to detect the pronunciation data of the word (step S3).
[0022]
It is determined whether or not there is a plurality of pronunciation data detected for the word (step 204). If there are a plurality of pronunciation data, the plurality of pronunciation data are combined to create intermediate pronunciation data (step 205). The pronunciation data is stored (step 206). On the other hand, if there is only one pronunciation data detected for the word in step 204, the process proceeds to step 206 and the pronunciation data is stored.
[0023]
Subsequently, it is determined whether or not pronunciation data has been acquired for all the words included in the character string (text data) to be read out (step 207). If not yet completed, the process returns to step 202, The processing from step 202 to step 206 is repeated. On the other hand, if pronunciation data is acquired for all words, a character string (text data) is read out according to the pronunciation data (step 208).
[0024]
(1-3) Specific Example Next, the operation and effect of the present embodiment will be specifically described with reference to the sentence shown in FIG. In the character string shown in FIG. 7, for the same notation differently pronounced words indicated by the underline, all readings are detected in step 203 of FIG. 2, and a plurality of pronunciation data is detected in step 204. Judgment is made. In this case, the intermediate pronunciation data creation unit 14 synthesizes intermediate pronunciation data (intermediate pronunciation data) that can be heard by both users.
[0025]
In this specific example, the word extraction unit 10 acquires a character string (text data) to be read out, and extracts one word from the character string. For example, when the word “A” is extracted and the pronunciation data of the word is searched with reference to the pronunciation definition database 11, two ways of reading “A” and “A” are detected. In this way, when there are a plurality of readings, the intermediate pronunciation data creation unit 14 synthesizes the plurality of pronunciation data to create intermediate pronunciation data that sounds like “Ah”.
[0026]
Also, even when the word “AM” is extracted, when the pronunciation data of the word is searched, two ways of reading “am” and “AM” are detected. Intermediate pronunciation data that can be heard is created. On the other hand, the pronunciation data is detected from the pronunciation definition database 11 for words that have only a single reading.
[0027]
Accordingly, the character string shown in FIG. 7 is converted to “(** 1). I (** 2) (** 1) MENBER OF (** 2) TOKYO FUN CLUB.”. Note that (** 1) is intermediate sound data of “A” ≈ “Ah”, and (** 2) is intermediate sound data of “AM” ≈ “Ahm”.
[0028]
As a result, the output sound is the sound that can be heard as “Ah (*), I Ahm (*) Ah (*) Member of Ahm (*) Tokyo Fan Club”. Note that the sound with (*) means that it sounds like that as an example.
[0029]
Since the pronunciation data is acoustic data, it is very difficult to express it with characters. In the above example, they are represented as “Ah” and “Ahm”, respectively. Acoustic data adjusted to have a waveform-frequency characteristic that produces an intermediate sound is created.
[0030]
(1-4) Effect As described above, according to the reading device of the present embodiment, for a word having a plurality of readings, intermediate pronunciation data obtained by combining the plurality of readings is created. Both can be output as a sound that can be heard. As a result, it is possible to greatly reduce the occurrence of erroneous reading out in a sentence in which the same notation different pronunciation phrase exists.
[0031]
(2) Second Embodiment (2-1) Configuration FIG. 3 is a block diagram showing the overall configuration of the reading apparatus according to the present embodiment. That is, the reading device of the present embodiment includes the same notation different pronunciation word extraction unit 21 that extracts the same notation different pronunciation words such as “A” and “AM” from the input sentence (text data), The intermediate pronunciation data definition database 22 that defines intermediate pronunciation data for the same notation different pronunciation words, and the intermediate pronunciation data definition database 22 with reference to the intermediate pronunciation data definition database 22, which of the same notation different pronunciation words is used by the user A first pronunciation data conversion unit 23 for converting into intermediate pronunciation data that can be heard together, a single pronunciation data definition database 24 that defines pronunciation data for words having a single reading, and the same notation different pronunciation words only For the converted phonetic data converted for the text data and the text data mixed with the unconverted text character string, Referring to the definition database 24, the second pronunciation data conversion unit 25 that converts the remaining character string data into pronunciation data, and the first pronunciation data conversion unit 23 and the second pronunciation data conversion unit 25 perform conversion. And a speech synthesizer 26 for outputting synthesized speech based on the generated pronunciation data.
[0032]
(2-2) Operation The flow of processing in the reading apparatus of the present embodiment having the above-described configuration will be described with reference to the flowchart shown in FIG.
First, the same notation different pronunciation word extraction unit 21 acquires a sentence (text data) to be read out, and from this sentence, the same notation different pronunciation word (example) registered in the intermediate pronunciation data definition database 22 in advance. : “A”, “AM”) are extracted (step S401, step S402).
[0033]
Subsequently, the first pronunciation data conversion unit 23 converts each extracted word into preset intermediate pronunciation data with reference to the intermediate pronunciation data definition database 22 (step S403).
[0034]
It is determined whether or not conversion to intermediate pronunciation data has been completed for all of the extracted notation different pronunciation words (step S404), and then the second pronunciation data conversion unit 25 performs the remaining conversion. Is converted into a normal pronunciation data string with reference to the single pronunciation data definition database 24 (step S405).
A character string (text data) is read out based on the pronunciation data converted in step S403 and step S405 (step 406).
[0035]
(2-3) Specific Example Next, the operation and effect of the present invention will be specifically described using the sentence shown in FIG. 7 described in the section of the prior art as an example. In addition, in the character string shown in FIG. 7, words with the same notation and different pronunciation, which are indicated by underlining, are defined in the intermediate pronunciation data definition database 22 as words having a plurality of readings in advance, Intermediate pronunciation data that can be heard is also defined.
[0036]
The intermediate pronunciation data is defined as follows, for example, in the case of the above two readings “A” and “AM”. That is, for the notation “A”, if there are two readings “A” and “A”, for example, pronunciation data that sounds like “A” is defined. In addition, when the notation “AM” can be read in two ways, “AM” and “AM”, for example, pronunciation data that can be heard as “AM” is defined.
[0037]
Note that the pronunciation data is acoustic data, so it is very difficult to express it with characters. In the above example, they are represented as “Ah” and “Ahm”, respectively. Thus, acoustic data adjusted to have a waveform-frequency characteristic so as to be an intermediate voice is created.
[0038]
The same notation different pronunciation word extraction unit 21 refers to such an intermediate pronunciation data definition database 22 and selects a word character string to be intermediately pronounced defined in the database 22 from the input character string (in this example, “ A ”and“ AM ”) are extracted, and only the extracted character string portion is converted into preset pronunciation data (intermediate pronunciation data) as follows.
[0039]
That is, the character string shown in FIG. 7 is converted to “(** 1). I (** 2) (** 1) MENBER OF (** 2) TOKYO FUN CLUB.”. Note that (** 1) is “A” sounding data≈ “Ah”, and (** 2) is “AM” sounding data≈ “Ahm”.
[0040]
Subsequently, the second pronunciation data conversion unit 25 ignores the already-converted pronunciation data, and constructs a sentence with reference to the single pronunciation data definition database 24 for the remaining character strings that have not been converted. The phonetic data corresponding to each word is searched, a phonetic data string for the input text sentence is constructed and output, converted into utterance voice data by the voice synthesizer 26, and output as voice.
[0041]
As a result, the output sound is the sound that can be heard as “Ah (*), I Ahm (*) Ah (*) Member of Ahm (*) Tokyo Fan Club”. Note that the sound with (*) means that it sounds like that as an example.
[0042]
(2-4) Effects As described above, according to the reading device of the present embodiment, intermediate pronunciation data is defined in advance for words having a plurality of readings. By simply referring to the intermediate pronunciation data, it can be output as a sound that can be heard by both users. As a result, it is possible to greatly reduce the occurrence of erroneous reading out in a sentence in which the same notation different pronunciation phrase exists.
[0043]
(3) Other Embodiments The present invention is not limited to the above-described embodiments, and the following modifications are possible. In other words, the method of creating intermediate pronunciation data for a word with a plurality of readings is not particularly limited, a method of simply combining a plurality of readings, a method of emphasizing any of the readings, etc. Can be appropriately selected.
[0044]
【The invention's effect】
As described above, according to the present invention, it is possible to provide a high-accuracy reading device, a reading method, and a reading processing program capable of reducing the occurrence of erroneous reading in a sentence having the same notation different pronunciation phrase. be able to.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of a first embodiment of a reading apparatus according to the present invention. FIG. 2 is a flowchart showing a flow of processing in the reading apparatus according to the first embodiment shown in FIG. The block diagram which shows the structure of 2nd Embodiment of the reading apparatus which concerns on FIG. 4 is a flowchart which shows the flow of a process in the reading apparatus of 2nd Embodiment shown in FIG. 3 FIG. 5 shows the structure of the conventional reading apparatus Block diagram [FIG. 6] A flow chart showing a flow of processing in a conventional reading apparatus. [FIG. 7] An example of a sentence (text data) to be read out.
DESCRIPTION OF SYMBOLS 10 ... Word extraction part 11 ... Pronunciation definition database 12 ... Pronunciation data detection part 13 ... Determination part 14 ... Intermediate pronunciation data creation part 15 ... Pronunciation data storage part 16 ... Speech synthesis part 21 ... Same notation different pronunciation word extraction part 22 ... Intermediate Pronunciation data definition database 23 ... first pronunciation data conversion unit 24 ... single pronunciation data definition database 25 ... second pronunciation data conversion unit 26 ... speech synthesis unit

Claims (6)

A pronunciation definition database that stores one or more readings for each word constituting the character string to be read out;
With reference to this pronunciation definition database, a pronunciation data detection unit that detects pronunciation data of each word;
A determination unit for determining whether or not there is a plurality of pronunciation data detected for each word;
When the detected pronunciation data is plural, an intermediate pronunciation data creation unit that synthesizes the pronunciation data and creates intermediate pronunciation data;
A speech synthesizer that outputs synthesized speech based on the pronunciation data determined as a single reading by the determination unit and / or the intermediate pronunciation data created by the intermediate pronunciation data creation unit. Read-out device. An extraction unit that extracts the same notation different pronunciation words from the character string to be read out;
An intermediate pronunciation data definition database defining intermediate pronunciation data for the same notation different pronunciation words;
A first pronunciation data conversion unit for converting the same notation different pronunciation words into intermediate pronunciation data with reference to the intermediate pronunciation data definition database;
A single pronunciation data definition database that defines pronunciation data for words with a single reading;
A second phonetic data conversion unit that converts character string data that has not been converted by the first phonetic data conversion unit into predetermined phonetic data with reference to the single phonetic data definition database;
A speech reading apparatus comprising: a speech synthesis unit that outputs synthesized speech based on the pronunciation data converted by each of the first pronunciation data conversion unit and the second pronunciation data conversion unit. A pronunciation data detection process for detecting pronunciation data of each word with reference to a pronunciation definition database storing one or more readings for each word constituting a character string to be read;
A determination process for determining whether or not there is a plurality of pronunciation data detected for each word;
When the detected pronunciation data is plural, intermediate pronunciation data creation processing for synthesizing those pronunciation data to create intermediate pronunciation data;
A speech synthesis process for outputting synthesized speech based on the pronunciation data determined to be a single reading in the determination process and / or the intermediate pronunciation data created by the intermediate pronunciation data creation process. Read-out method. An extraction process for extracting the same notation different pronunciation word from the character string to be read out,
A first pronunciation data conversion process for converting the same notation different pronunciation words into intermediate pronunciation data with reference to an intermediate pronunciation data definition database that defines intermediate pronunciation data for the same notation different pronunciation words;
Refers to a single pronunciation data definition database that defines pronunciation data for a word having a single reading, and converts character string data that has not yet been converted by the first pronunciation data conversion processing into predetermined pronunciation data A second pronunciation data conversion process,
A speech reading method comprising: speech synthesis processing for outputting synthesized speech based on the pronunciation data converted by each of the first pronunciation data conversion processing and the second pronunciation data conversion processing. In the reading processing program that reads out the character string to be read out by controlling the computer,
The program is stored in the computer,
A pronunciation data detection step of detecting pronunciation data of each word with reference to a pronunciation definition database storing one or more readings for each word constituting a character string to be read;
A determination step of determining whether or not there is a plurality of pronunciation data detected for each word;
When the detected pronunciation data is plural, an intermediate pronunciation data creation step of synthesizing those pronunciation data to create intermediate pronunciation data;
A speech synthesis step of outputting synthesized speech based on the pronunciation data determined as a single reading in the determination step and / or the intermediate pronunciation data created in the intermediate pronunciation data creation step. A program for reading processing, characterized by being. In the reading processing program that reads out the character string to be read out by controlling the computer,
The program is stored in the computer,
An extraction step of extracting the same notation different pronunciation word from the character string to be read out;
Referring to an intermediate pronunciation data definition database defining intermediate pronunciation data for the same notation different pronunciation words, a first pronunciation data conversion step of converting the same notation different pronunciation words into intermediate pronunciation data;
Refers to a single pronunciation data definition database that defines pronunciation data for a word having a single reading, and converts character string data that has not been converted by the first pronunciation data conversion step into predetermined pronunciation data A second pronunciation data conversion step,
A speech processing step for executing a speech synthesis step for outputting synthesized speech based on the pronunciation data converted by each of the first pronunciation data conversion step and the second pronunciation data conversion step Program.

Images