JPH08129398A - Text analysis device - Google Patents

Abstract

PURPOSE: To provide a text analysis device in which two opposing requirements, i.e., the expansion of a word dictionary and the prevention of the speed reduction in a text analysis processing are satisfied. CONSTITUTION: A fixed dictionary storage section 23 and a user dictionary storage section 24 have plural word storage regions 23A, 23B, 23C,..., 24A, 24B, 24C,..., respectively corresponding to plural categories A, B,... A control section 12 of a text data generation device 10 transfers the commands, which specify the categories to be used, together with text data. A control section 22 of a text audio conversion device 20 extracts the commands from the received data. A category selection section 25 selects the category specified by the extracted command. The section 22 analyzes the text data employing the words which belong to the selected category.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a text analysis device for converting text data into voice waveform data, for analyzing text data by using a word dictionary.

[0002]

2. Description of the Related Art In recent years, a text-to-speech conversion system has been attracting attention as a speech synthesis system because it has no limitation on output vocabulary. This text-to-speech system
By converting text data, that is, a character code string into voice waveform data, a sentence can be read aloud.

Text-to-speech conversion processing for converting text data into speech waveform data is roughly classified into text analysis processing, synthesis parameter generation processing, and speech synthesis processing.

Here, the text analysis process is a process of analyzing text data using a word dictionary and generating a phonological prosodic symbol string in which reading, accent, intonation, etc. of the text data are described.

The synthesis parameter generation process is based on the phoneme prosodic symbol string obtained by the text analysis process, and includes speech units, durations and pitches (voice pitches), and amplitudes (voice volume) of each phoneme. This is a process of generating parameters necessary for speech synthesis, such as a temporal change pattern.

The voice synthesizing process is a process of generating voice waveform data from the synthesis parameter generated by the synthesis parameter generating process.

The word dictionary generally includes a fixed dictionary and a user dictionary. Here, the fixed dictionary is a dictionary in which standard words are registered in advance. On the other hand, the user dictionary is a dictionary in which the user can freely register words.

According to the structure in which the user dictionary is provided as described above, even if the text data includes a word that is not registered in the fixed dictionary, by registering this word in the user dictionary, the text data can be stored. Can be accurately analyzed.

[0009]

However, the above-described conventional text-to-speech conversion system has the following problems.

That is, the word dictionary is generally wide and shallowly edited in order to correspond to various text data. However, such a configuration may not be able to cover technical terms in a specific field, and thus text data in the specific field may not be accurately analyzed.

To solve this problem, the word dictionary may be edited wide and deep. That is, the word dictionary should be expanded. However, in this case, the word dictionary becomes too large and the matching time between the text data and the word dictionary becomes long. As a result, there is a time lag between the input of text data and the output of voice.

From the above, in the text-to-speech conversion system, a text analysis device capable of accurately analyzing text data in a specific field without causing a time lag is desired. In other words, there is a demand for a text analysis device capable of satisfying two contradictory requirements of expansion of a word dictionary and prevention of reduction in processing speed of text analysis processing.

[0013]

In order to solve the above problems, the present invention has a plurality of word storage areas respectively corresponding to a plurality of word classification categories and corresponds to each word storage area. A method for storing words belonging to a category, a method for selecting a category used for analyzing text data from a plurality of categories, and a word stored in a word storage area corresponding to the selected category And means for analyzing the text data.

[0014]

In the above structure, the category used for the analysis is selected prior to the execution of the text analysis process.
Next, the text data is analyzed using the words stored in the word storage area corresponding to the selected category.

With such a configuration, the text data can be analyzed using only the portion of the word dictionary in which the words required for analysis are registered. Therefore, even if the dictionary is edited wide and deep, it is possible to prevent the processing speed of the text analysis processing from decreasing. As a result, it is possible to satisfy two contradictory requirements of expansion of the word dictionary and prevention of reduction in the processing speed of the text analysis processing.

[0016]

Embodiments of the present invention will now be described in detail with reference to the drawings. In the following description, the case where the present invention is applied to a text analysis device of a text-to-speech conversion system will be described as a representative.

(1) First Embodiment First, a first embodiment of the present invention will be described in detail.

(1-1) Outline of First Embodiment In this embodiment, when words are registered in a word dictionary, they are classified and registered according to a certain category, and when text data is analyzed, a plurality of categories are selected. The category to be used is selected from, and analysis is performed using the words belonging to the selected category.

Further, in this embodiment, when selecting a category to be used, the information designating the category is output together with the text data, the category designating information is extracted from this output, and the extracted category designating information is added to the extracted category designating information. On the basis of,
The category to be used is selected.

Furthermore, in this embodiment, when the text data is analyzed using the words belonging to the selected category, the words belonging to the selected category are referred to in accordance with a predetermined priority order. is there.

(1-2) Configuration of the First Embodiment FIG. 1 is a block diagram showing the configuration of the first embodiment of the present invention. Note that FIG. 1 shows the overall configuration of a text-to-speech conversion system equipped with the text analysis device of the present invention.

In the figure, 10 is a text data generator for generating text data. The text data generator 10 is composed of, for example, a personal computer or a word processor. 20
Is a text-to-speech converter for converting the text data output from the text data generator 10 into speech waveform data.

In the text data generator 10, 1
Reference numeral 1 denotes a file storage unit that stores a plurality of text files. Reference numeral 12 denotes a control unit that executes interface control with the text-to-speech conversion device 20, read processing of text data, and the like. A data input / output unit 13 exchanges data and the like with the text-to-speech conversion device 20.

In the text-to-speech conversion device 20, 2
Reference numeral 1 is a data input / output unit for exchanging data and the like with the text data generator 10. A control unit 22 executes interface control with the text data generator 10, text analysis processing, synthesis parameter generation processing, and the like.

Reference numeral 23 is a fixed dictionary storage unit for storing a fixed dictionary. The fixed dictionary storage unit 23 is composed of, for example, a read-only memory. Reference numeral 24 is a user dictionary storage unit that stores a user dictionary. The user dictionary storage unit 24 is composed of, for example, a random access memory.

Reference numeral 25 is a category selection unit for selecting a category used for text analysis processing. Reference numeral 26 is a priority order register that holds information indicating the priority order of categories when referring to a fixed dictionary or a user dictionary. 27 is
This is an internal buffer for temporarily holding the processing result of the synthesis parameter generation processing.

Reference numeral 28 is a voice synthesizing section for generating voice waveform data based on the synthesis parameters held in the internal buffer 27. A voice output unit 29 performs, for example, digital-analog conversion on the voice waveform data generated by the voice synthesis unit 28, outputs voice based on the converted output, and transmits the converted output to another device. Is.

The fixed dictionary storage unit 23 has a plurality of word storage areas 23A, 23B, 23C, ... Corresponding to a plurality of categories A, B, C ,.
Have. Each word storage area 23A, 23B, 23C, ...
Stores the words belonging to the corresponding categories A, B, C, ....

The user dictionary storage section 24 also has a plurality of word storage areas 24A, 24B, 24C, ... Corresponding to a plurality of categories A, B, C ,. Each word storage area 24A, 24B, 24C ... Stores words belonging to the corresponding category. In this case, the categories corresponding to the word storage areas 24A, 24B, 24C, ... Are respectively word storage areas 2 of the fixed dictionary storage unit 23.
Matches the categories corresponding to 3A, 23B, 23C, ....

The categories A, B, C, ...
For example, there are “place name”, “person name”, “company name” and the like. Further, it is not limited to the category including only nouns, and there are categories including verbs such as “medicine”, “physics”, and “law”.

The category selection unit 25 selects a category used for the text analysis process from the above-mentioned plurality of categories A, B, C, .... This selection is made based on the category designation command sent from the text data generator 10.

Control unit 12 of text data generator 10
Sends a category specifying command for specifying the category used in the analysis of this text data together with this text data for each text data.

In this case, it is necessary to distinguish between the category designation command and the text data. As this method, for example, a method of transferring the position information of the category designation command and a method of making the code form of the command distinguishable from the text data can be considered. In this embodiment, the latter method is adopted, which will be described in detail later.

Control unit 22 of text-to-speech converter 20
Extracts a category designation command from the received data by utilizing the difference in code form. The extracted category designation command is supplied to the category selection unit 25.

As shown in FIG. 2, the priority register 26 stores a head address holding area 261 for holding head addresses of the categories A, B, ... And data indicating whether or not each category is selected. It has a selection data holding area 262 for holding.

In this case, the head addresses of the categories A, B, C, ... Are held in order of priority. In the figure,
The case where the category D has the highest priority is shown. The selection data is written by the control unit 22 based on the category designation command. In this case, for example, "1" indicates that it is selected and "0" indicates that it is not selected. In the figure, the case where categories D and C are selected and categories A and B are not selected is shown as a representative.

(1-3) Operation of the First Embodiment The operation of the above configuration will be described. In the following description, the case where the text data is Japanese text data, that is, a kanji-kana mixing sentence will be described as a representative.

First, a text-to-speech conversion process for converting text data into voice waveform data will be described.

The user selects the text data generator 10
When a conversion command for instructing the execution of the text-to-speech conversion process is input, the Japanese text data specified by this conversion command is transferred to the file storage unit 11 by the control unit 12.
Is read from. This text data is supplied to the text-to-speech conversion device 20 via the data input / output unit 13.

The text data supplied to the text-to-speech conversion device 20 is temporarily held in a text buffer provided inside the data input / output unit 21, and then converted into voice waveform data by a text-to-speech conversion process.

FIG. 3 is a diagram showing an example of the text-to-speech conversion processing. The illustrated text-to-speech conversion processing is roughly classified into text analysis processing, synthesis parameter generation processing, and speech synthesis processing.

The text analysis process uses the word concatenation information, the fixed dictionary stored in the fixed dictionary storage unit 23, and the user dictionary stored in the user dictionary storage unit 24,
A phonological prosodic symbol string (intermediate language) is generated by executing processes such as word and phrase identification, reading addition, phrase accent determination, rendaku, and phonophoric processing.

In the synthesis parameter generation process, parameters necessary for speech synthesis are generated from the phonological prosody symbol string obtained by the text analysis process by the prosody rule assignment process. Here, the prosody rule assigning process is a process of determining a synthesis unit, inserting a pause, controlling phoneme duration, controlling amplitude, controlling pitch, and the like.

Further, the voice synthesizing process generates the voice waveform data by the waveform synthesizing process while referring to the waveform dictionary.

As general text analysis processing items, word segmentation, word reading determination, accent determination, part-of-speech information determination, word combination, compound sound processing, accent combination, phrase reading setting, phrase combination, accent. The size of the pose, the intonation position and size, and the pose length.

FIG. 4 is a diagram showing a general text analysis process. The illustrated text analysis process includes a morpheme analysis process, a clause generation process, and a phrase generation process.

Here, the morphological analysis process is a process of matching the input text data with the fixed dictionary and the user dictionary and dividing the text data into words.
FIG. 5 is a diagram showing general registration contents of a fixed dictionary and a user dictionary. As shown in the figure, the notations, readings, parts of speech, and accent positions of words are registered in these dictionaries. As a result, by collating the input text data with the dictionary,
If the notations match, the reading, part of speech, and accent position are obtained.

The phrase generation processing includes word combination, compound sound processing, and accent combination based on the information indicating the word reading, the part of speech, and the accent position obtained by the morphological analysis processing, and determines the reading for each clause. It is a process to do. For this determination, an accent combination rule and a clause boundary rule based on part-of-speech information are used.

The phrase generation process is a process of determining the reading of the phrase and determining the temperament within the phrase according to the connection rule between the phrases. FIG. 6 shows an example of a combination rule based on the part-of-speech information.

The text analysis process and the synthesis parameter generation process are executed by the control unit 22. On the other hand, the voice synthesizing process is executed by the voice synthesizing unit 28. In this case, the synthesis parameter generated by the control unit 22 is stored in the internal buffer 27 and then supplied to the voice synthesis unit 28.

The voice waveform data generated by the voice synthesizer 28 is supplied to the voice output unit 30. The voice output unit 30 performs digital / analog conversion of voice waveform data, drives a speaker to output voice by the converted output, and transmits the converted output to another device via a communication line.

Next, the characteristic operation of the present invention will be described. When transferring the text data to the text-to-speech conversion device 20, the control unit 12 of the text data generation device 10 first transfers the above-mentioned category designation command, and then transfers the text data.

Control unit 22 of text-to-speech converter 20
Extracts a category designation command from the received data and supplies it to the category selection unit 25. Further, the control unit 42 writes the selection data in the selection data holding area 262 of the priority order register 26 based on the extracted category designation command. As a result, "1" is written in the position corresponding to the category specified by this command.

When the category selection unit 25 receives the category designation command, it selects the category designated by this command. As a result, the word storage area corresponding to the selected category is set to the enabled state. For example,
When the categories C and D are designated by the category designation command, the word storage area 2 corresponding to the categories C and D
3C, 23D, 24C and 24D are set to the enable state.

When the category selection processing is completed, the control unit 22 analyzes the received text data using the words stored in the word storage area corresponding to the selected category. In this case, for example, the control unit 22 first refers to the fixed dictionary, and when there is no target word in the fixed dictionary, the control unit 22 refers to the user dictionary.

When referring to each dictionary, the control unit 22
Refers to each category according to the priority of the selected category. That is, in the example of FIG. 2, the control unit 22 first sets the word storage area 23 corresponding to the category D.
D (or 24D) is referred to, and if the target word is not stored in this word storage area, the word storage area 23C (or 24C) corresponding to category C is referenced.

In this case, the priority register 26 stores the start addresses of the categories A, B, C, D, .... Accordingly, the control unit 22 can access the word storage area without knowing the position of the word storage area corresponding to the selected category.

The category designation command is defined as, for example, a hexadecimal code indicating 2 bytes enclosed in braces. Each bit of the 2-byte code corresponds to one of the above-mentioned plurality of categories A, B, C ,. As a result, the 2-byte code can support up to 16 categories A, B, C, ....

Here, each category A, B, ... Is selected when the corresponding bit is “1” and is not selected when the corresponding bit is “0”, all categories A, B ,. The command code is represented as {FFFF}.

When this category designation command is transferred from the text data generator 10 to the text-to-speech converter 20, the category designation command is temporarily held in the text data buffer of the data input / output unit 21.

Control unit 22 of text-to-speech converter 20
Reads out the data from the text data buffer every 2 bytes and judges whether or not this is "{". If “{”, the following data can be interpreted in two ways, that is, a category designation command or text data.

Therefore, the following 8 bytes ("F" is a 2-byte character code) are read. This is "FFF
If it is F ", this is also valid as a command. Therefore, in this case, the following 2 bytes are read and this is"} ".
If it is, it is determined to be a category designation command of "{FFFF}", and if it is not "}", it is determined to be text data.
In this way, the control unit 22 extracts the category designation command from the received data.

FIG. 7 is a block diagram showing an example of a concrete configuration of the category selecting section 25 described above.

The illustrated category selecting section 25 is composed of a latch circuit 251. The data input / output unit 21 is connected to the input terminal of the latch circuit 251 by the control unit 22.
2 bytes of data read in parallel from the text data buffer are supplied.

Further, as shown in FIG. 8, the output terminal of the latch circuit 251 has a corresponding word storage area 23A, 24A, 23B, 24B, 23C, 2 for each bit.
4C, ... Chip select terminals CE are connected.
For example, the output terminal of the first bit of the latch circuit 251 is
Chip select terminals CE of the word storage areas 23A and 24A
The second bit output terminal is connected to the word storage area 2
Connected to the chip select terminals CE of 3B and 24B,
The output terminal of the bit is connected to the chip select terminal CE of the word storage areas 23C and 24C.

In the above configuration, the latch circuit 251
Prior to the start of the text-to-speech conversion process, the control unit 2
It is reset by the reset signal output from 2.
As a result, the output of the latch circuit 251 is, for example, “F
FFF ". As a result, in this case, all word storage areas 23A, 24A, 23B, 24B, 23
C, 24C, ... Are set to the enabled state.

In this state, when the category designation command is extracted by the control unit 22, the control unit 22 outputs a latch signal. As a result, the category designation command is latched in the latch circuit 251. As a result, this time, the word storage area corresponding to the category designated by the category designation command is set to the enabled state.

(1-4) Effects of the First Embodiment According to this embodiment described in detail above, the following effects can be obtained.

First, according to this embodiment, when words are registered in the word dictionary, they are classified and registered according to a certain category, and when analyzing text data, a category to be used is selected from a plurality of categories, Since the analysis is performed based on the words belonging to the selected category, it is possible to analyze the text data using only the part necessary for the analysis of the fixed dictionary and the user dictionary.

As a result, even if the dictionary is wide and deeply edited, it is possible to prevent the processing speed of the text analysis processing from decreasing as much as possible. As a result, according to this embodiment, it is possible to satisfy two conflicting requirements of expanding the word dictionary and preventing the processing speed of the text analysis processing from decreasing.

Further, according to this embodiment, since the category is designated on the side of the text data generator 10, a plurality of text-to-speech converters 20 are provided to one text data generator 10 through a telephone line or radio waves. In a system that connects and broadcasts the same text data, that is, a sender-centered system, it is possible to select the optimum category for all text-to-speech conversion devices 20.

That is, in the system as described above, it is possible to specify the category to be used on the receiver side. However, in such a configuration, each recipient does not always know the content of the text data, and thus it is not always possible to specify the most suitable category for the received text data.

On the other hand, according to the configuration in which the sender specifies the category as in this embodiment, a person who is familiar with the contents of the text data can specify the category. The person can select the category most suitable for the received text data.

Furthermore, according to this embodiment, when analyzing text data, according to a predetermined priority order,
Since the words belonging to the selected category are referred to, the processing speed of the text analysis processing can be increased.

(2) Second Embodiment Next, a second embodiment of the present invention will be described in detail.

(2-1) Outline of Second Embodiment In the previous embodiment, the text data generating device transfers the category designation command together with the text data to the text-to-speech conversion device. The case where the category designation command is extracted from the received data and the category to be used is selected based on the extracted category designation command has been described.

On the other hand, in this embodiment, the text-to-speech converter is provided with a category designating switch for designating the category to be used, and the category designated by this switch is selected.

Further, in the above embodiment, the case where the text data is analyzed by referring to the selected category in accordance with the predetermined priority order has been described.

On the other hand, in this embodiment, a search table for searching a target word is created, and the text data is analyzed by referring to the category selected using this search table. It was done.

(2-1) Configuration of the Second Embodiment FIG. 9 is a block diagram showing the configuration of the second embodiment of the present invention. Note that FIG. 9 shows the overall configuration of the text-to-speech conversion system equipped with the text analysis device of the present invention.

In the figure, 40 is a text data generator, and 50 is a text-to-speech converter. In the text data generating device 40, 41 is a file storage unit, 42 is a control unit, and 43 is a data input / output unit.

In the text-to-speech conversion device 50, 5
1 is a data input / output unit, 52 is a control unit,
Reference numeral 53 is a fixed dictionary storage unit, 54 is a user dictionary storage unit, 55 is a category selection unit, and 56 is
A search table storage unit, 57 is an internal buffer, 58 is a voice synthesis unit, 59 is a voice output unit, and 60 is a category designation switch.

The fixed dictionary storage section 53 has a plurality of word storage areas 53A, 53B, 53C, ... Corresponding to a plurality of categories A, B, C ,. The user dictionary storage unit 54 also has a plurality of word storage areas 54A, 54B, corresponding to the plurality of categories A, B, C ,.
54C, ...

The main points of the characteristic structure of this embodiment will be described below, for example.

In this embodiment, the category to be used is designated by the user using the category designation switch 60. Therefore, in this embodiment, the control unit 42 of the text data generator 40 does not transfer the category designation command.

Further, in this embodiment, when analyzing text data, the search table stored in the search table storage section 60 is used to search for the target word.
When the category to be used is designated by the category designation switch 60, this search table is based on the designated category and the stored contents of the fixed dictionary and the user dictionary.
It is created by the control unit 52.

(2-3) Operation of the Second Embodiment The operation of the above configuration will be described. It should be noted that the following description focuses on the characteristic operations of this embodiment, that is, the category selecting operation, the search table creating operation, and the text analysis process using the search table.

First, the category selection operation will be described.
When the user uses the category designating switch 60 to designate the category to be used, the category designating data is supplied from the category designating switch 60 to the category selecting section 55. Upon receiving this category designation data, the category selection unit 55 selects the category designated by this data.

FIG. 10 is a diagram showing an example of a specific configuration of the category designation switch 60. The illustrated category designating switch 60 is configured by, for example, a 16-bit DIP switch (DIP switch). This gives 1
Six categories A, B, ..., P can be dealt with. The output of each bit switch 60A, 60B, ..., 60P of the category designating switch 60 is supplied to the category selecting section 55.

The category selecting section 55 is composed of a latch circuit 551, for example, as in the previous embodiment. The latch circuit 551 is temporarily reset by a reset signal supplied from the control unit 52 at the start of the text analysis process, and then similarly latches the input data according to the latch signal supplied from the control unit 52. As a result, the category specifying data output from the category specifying switch 60 is latched in the latch circuit 561.

Next, the search table creating operation will be described. Each time the category designation switch 60 designates a category to be used, the control unit 52 creates a search table for searching a word belonging to the designated category.

This search table is composed of a search table in which the storage address of a word is registered and a search table in which the top address of a link in which the storage address of a word starting with a certain character is stored is registered. Hereinafter, the former search table is referred to as a secondary search table, and the latter search table is referred to as a primary search table.

FIG. 11 is a diagram showing an example of registered contents of the fixed dictionary stored in the fixed dictionary storage unit 53 and the user dictionary stored in the user dictionary storage unit 54. In addition,
In the figure, for the sake of simplicity, the distinction between the fixed dictionary and the user dictionary is eliminated, and the registered contents of one word dictionary are shown.
FIG. 12 is a diagram showing an example of registered contents of the secondary search table. FIG. 13 is a diagram showing an example of registered contents of the primary search table.

In the dictionary shown in FIG. 11, an address of 10000h (h indicates hexadecimal) is given for each category. Each address contains notation, reading, and other information used for dictionary matching. Further, in the illustrated example, the case where category A shows a chemical term, category B shows a music term, category C shows a medical term, and category D shows a biological term is shown as a representative.

When creating the search table, the control unit 52 first sorts, for example, from the beginning of the dictionary shown in FIG.
The storage address of a word starting with the character code of "A" is extracted. This extraction is performed only for words belonging to the specified category.

If all categories A, B, C and D are designated, the storage address of "alkaloid" is 10000h and the storage address of "antimony" is 1000h.
1h, storage address of "Andante" 20000h,
The storage address 30000h of "Ampool" and the storage address 40001h of "Ancora" are extracted.

Next, all the extracted storage addresses 10
000h, 10001h, 20000h, 30,000
h, 40001h are sorted in order from the longest corresponding word. In the case of the present example, the notation is long in the order of "alkaloid", "andante", "ampur", "antimony", and "angler". Therefore, the extracted storage addresses are 10000h, 20000h, 30
They are rearranged in the order of 000h, 10001h, 40001h.

The lengths of the expressions "Andante" and "Ampool" are the same 6. In this case, for example, the items having the category first, that is, "Andante" are arranged first.

When this rearrangement is completed, each storage address 10000h, 20000h, 30000h, 1000
1h and 40001h are written in order from the address 1000h. When this writing is completed, a code for indicating the end of the link of the storage address of the word starting with "A", for example, "FFFFFFFFh", is added to the next address 1fff.
Store in h.

Next, similar processing is performed for "a". In this case, the storage address is written at the address 1f.
It starts from the next address 2000h after ffh. Hereinafter, the same processing is performed for “C”, “D”, .... When this process ends for all the characters that are the target of the text analysis process, the creation of the secondary search table ends.

When the creation of the secondary search table is completed,
The start address of the link starting with "a" and the character code of "a" are stored in address 0100h. Hereinafter, similarly, the same processing is performed for “a”, “c”, “d”, .... When this process ends for all the characters that are the target of the text analysis process, the creation of the primary search table ends.

Next, the text analysis process using the search table will be described.

Now, it is assumed that the sent text data is "Anchou ...". In this case, the control unit 52
Pays attention to the leading character "A" and searches for "A" in the secondary search table to obtain the address where the link "A" exists.

In the case of FIG. 13, "A" is stored in the address 0100h, and the address data following this "A" is 1000h, which indicates that the link of "A" is at the address 1000h.

Next, the controller 52 sends the address 1000h.
Is read out, the contents of the address indicated by this are compared with the text data, and it is determined whether the two match. In the case of the present example, 1000h is 10,000
Since h is stored, the content of the address 10000h is read, and this is compared with the text data to determine whether the two match. In the case of this example, address 10
Since "alkaloid" is stored in 000h, it is determined that they do not match.

Next, the controller 52 sends the address 1001h.
Is read out, the contents of the address indicated by this are compared with the text data, and it is determined whether the two match. In the case of the present example, the address 1001h has 2
Since "0000h" is stored and "Andante" is stored in the address 20000h, it is determined that they do not match.

Similarly, the above-mentioned processing is repeated while sequentially updating the addresses of the secondary search table.
As a result, the user finally reaches the “angler” stored at the address 40001h.

In such a structure, if categories A, B, and C shown in FIG. 11 are not designated, the search table shown in FIG. 12 becomes smaller. As a result, the number of steps to reach "angler" is reduced, and high-speed search is possible.

(2-4) Effects of the Second Embodiment Also in this embodiment described in detail above, it is possible to obtain almost the same effects as those of the first embodiment described above, and further, the following effects. Such an effect can be obtained.

First, according to this embodiment, the text-to-speech conversion device 50 side specifies the category. Therefore, there are a plurality of text data generation devices 40, and the text-speech conversion device 50 has a plurality of text data generation devices. Four
In a system that selects desired text data from a plurality of text data sent from 0, that is, in a receiver-centered system, a good category is selected for each text data without increasing communication cost. can do.

That is, also in this case, it is possible to specify the category on the sender side as in the first embodiment. However, in this system, there is a gap between the timing at which the sender sends the category designation command and the timing at which the receiver selects the text data, in other words, the timing at which the transmission station is selected.

To reduce this deviation, it is necessary to embed many category designation commands in the transmitted text data. However, in this case, the transmission data becomes large and the communication cost increases.

On the other hand, in this embodiment, since the category is specified on the receiver side, the timing adjustment as described above becomes unnecessary. As a result, a good category can be selected for each text data without incurring an increase in communication cost due to an increase in transmission data.

Further, according to this embodiment, every time the designated category is changed, a search table for searching the target word is created, and the text analysis process is executed using this search table. Therefore, the processing speed of the analysis processing can be increased.

(3) Other Embodiments Two embodiments of the present invention have been described in detail above, but the present invention is not limited to the above-mentioned embodiments.

For example, in the first embodiment described above, the case where the priority order of the categories is fixed has been described. However, according to the present invention, for example, the search frequency of each category is monitored, and when the analysis processing of some text files is completed, the priority order is automatically changed based on the monitoring result of the search frequency. May be.

Further, in the second embodiment described above, a case has been described in which the plurality of extracted storage addresses are sequentially rearranged and registered from the one having the longest corresponding character code description. However, according to the present invention, for example, priorities are assigned to the respective categories, different categories are sorted according to their priorities, and the same categories are sorted according to, for example, the notation length. You may make it rearrange according to.

Further, in the above embodiment, the case where the present invention is applied to the text analysis device of the text-to-speech conversion system has been described. However, the present invention can be applied to a text analysis device other than the text-to-speech conversion system.

In addition, in the previous embodiment, the present invention is
The case has been described where the present invention is applied to a text analysis device that generates a phonological prosodic symbol string from text data using a word dictionary.
However, the present invention can be applied to at least a text analysis device that does not generate a phonological prosodic symbol string as long as the text analysis device performs word division.

In addition to this, it goes without saying that the present invention can be variously modified and implemented without departing from the scope of the invention.

[0121]

As described above in detail, according to the present invention,
When registering words in the dictionary, they are classified and registered according to multiple categories, and when analyzing text data, the category to be used is selected and the words belonging to the selected category are used for analysis. You can parse text data using only the part of the dictionary that is really needed for the text data to be parsed. This makes it possible to satisfy two conflicting requirements of expanding the word dictionary and preventing the processing speed of text analysis processing from decreasing.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a first embodiment of the present invention.

FIG. 2 is a diagram showing a configuration of a priority register according to the first embodiment.

FIG. 3 is a diagram illustrating an example of text-to-speech conversion processing according to the first embodiment.

FIG. 4 is a diagram illustrating an example of text analysis processing according to the first embodiment.

FIG. 5 is a diagram showing an example of registered contents of a word dictionary according to the first embodiment.

FIG. 6 is a diagram showing an example of a part-of-speech combination rule according to the first embodiment.

FIG. 7 is a diagram showing an example of a specific configuration of a category selection unit of the first exemplary embodiment.

FIG. 8 is a block diagram showing a relationship between a category selection unit and a dictionary storage unit according to the first embodiment.

FIG. 9 is a block diagram showing a configuration of a second exemplary embodiment of the present invention.

FIG. 10 is a diagram showing an example of a specific configuration of a category designating switch of the second embodiment.

FIG. 11 is a diagram illustrating a search table creation process according to the second embodiment.

FIG. 12 is a diagram for explaining a search table creation process of the second embodiment.

FIG. 13 is a diagram for explaining a search table creation process of the second embodiment.

[Explanation of symbols]

10, 40 ... Text data generator 20, 50 ... Text-to-speech converter 11, 41 ... File storage unit 12, 22, 42, 52 ... Control unit 13, 21, 43, 51 ... Data input / output unit 23, 53 ... Fixed Dictionary storage unit 24, 54 ... User dictionary storage unit 25, 55 ... Category selection unit 26 ... Priority register 27, 57 ... Internal buffer 28, 58 ... Voice synthesis unit 29, 59 ... Voice output unit 56 ... Search table storage unit 23A , 24A, 23B, 24B, 23C, 24C,
..., 53A, 54A, 53B, 54B, 53C, 54
C, ... Word storage area 60 ... Category designation switch

Claims (5)

[Claims] 1. A text analysis device for analyzing text data using a word dictionary, comprising: a plurality of word storage areas corresponding to a plurality of categories for word classification, and a category corresponding to each word storage area. A word dictionary storing means for storing words belonging to, category selecting means for selecting a category to be used in the analysis of the text data from the plurality of categories, and a category selecting means corresponding to the category selected by the category selecting means. A text analysis device, comprising: a text analysis unit that analyzes the text data using a word stored in a word storage area. 2. The category selecting means outputs a category specifying information for specifying a category to be used together with text data, and a category for extracting the category specifying information from an output of the data outputting means. The text analysis device according to claim 1, further comprising: designated information extracting means; and selecting means for selecting a category to be used based on the category specifying information extracted by the category specifying information extracting means. 3. The category selecting means comprises: a category specifying means for a user to specify a use category; and a selecting means for selecting a category specified by the category specifying means. 1
The described text analysis device. 4. The text analysis means is configured to analyze the text data while referring to a word belonging to a category selected by the category selection means in accordance with a predetermined priority order. The text analysis device according to claim 1. 5. The text analysis means creates a search table for creating a search table for searching a target word based on the category selected by the category selection means and the stored contents of the dictionary storage means. 2. The text according to claim 1, further comprising: a means and an analyzing means for analyzing the text data by referring to the word dictionary using the search table generated by the search table generating means. Analyzer.