JPS6213692B2 - - Google Patents

Abstract

PURPOSE:To swiftly prepare a voice information index table by inputting a voice response writing whose editing is desired, by use of phonetic characters, and automatically extracting the voice element piece information conforming with a partial character-string which constitutes said voice response writing, out of the existing voice element pieces. CONSTITUTION:The voice element piece information 101, 102... which have been changed to existing digital signals, and the store position information 111, 112... are stored in the digital voice information area 1 in the memory device, and the character element piece information 201, 202... which have represented the information 101, 102 by ''kana'' (Japanese syllabary) characters are arrayed in the character element piece store position information area 2, and are stored together with the information 111, 112.... And, a desired voice response writing 301 is input to the input voice response statement buffer 4, the partial character-strings 311... are extracted by the partial character-string sampling circuit 5, are stored in the partial character-string temporary memory 6, and the information 202... from the area 2 are extracted by the character element piece information sampling circuit 7, and are stored in the character element piece information temporary memory 8. And, the information of the temporary memories 6, 8 is collated by the collating circuit 9, and the coincidence sound information is stored in the digital voice information index table 17 of th writing information store buffer 12.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a method for editing a voice response sentence, and particularly to a method for editing a response sentence for a voice response device.

As voice response services become widespread in automated telephone networks, the types of voice response texts are also increasing. It is effective to create a desired voice response sentence by editing existing voice segments that have been converted into digital signals using the recently advanced voice data compression technology.

Previously, there were only a few types of voice response sentences, so
Editing of voice response sentences was done manually.
However, as the types of voice response sentences increase, the editing effort also increases, and as the number of voice segments to be stored increases, manual editing is becoming difficult.

An object of the present invention is to break through the limitations of conventional manual editing of voice response sentences and to realize a voice response sentence collection system that can cope with the rapidly increasing number of voice response services.

The purpose of this is to convert a plurality of voice unit information converted into digital signals, storage position information in a storage device of the voice unit information, and convert the voice unit information into phonetic characters in editing a response sentence for a voice response device. a storage device for storing character segment information represented by , a means for extracting a partial character string from a voice response sentence input as a phonetic character string, and a means for storing the extracted partial character string and the character segment information. By providing means for collating and means for storing storage position information of voice segment information corresponding to character segment information that matches as a result of the verification in a storage device according to the position of the partial character string in the voice response sentence. achieved.

Hereinafter, one embodiment of the present invention will be described with reference to the drawings.
In the figure, digital audio information area 1 in the storage device
The existing digital signalized speech segment information 1
01, 102, ...... are respectively storage positions 11
1, 112, ...... are stored. In addition, each voice segment information 10 in the digital voice information area 1 is stored in the character segment storage position information area 2 in the storage device.
Character segment information 201, 202, . . . , in which 1, 102, . . . The information 101, 102, . . . is stored together with the accommodation position information 111, 112, . In the figure, accommodation position information 112
The character segment information 202 in which the speech segment information 102 of is expressed in kana characters is the character segment information 20 in which the speech segment information 101 of the housing position information 111 is expressed in kana characters.
1, it is stored at the front in character segment storage position information area 2 arranged in alphabetical order. The voice response sentence 301 desired to be edited is input in kana characters, for example, using the typewriter device 3, and is stored in the input voice response sentence buffer 4. A partial character string 311 consisting of a predetermined number of characters, for example 10 characters, from the beginning of the stored voice response sentence 301 is extracted by the partial character string extraction circuit 5 and stored in the partial character string temporary storage 6. Next, character segment storage position information area 2
First character segment information 2 arranged in alphabetical order from
02 is extracted by the character segment extraction circuit 7 and stored in the character segment information temporary storage 8. Next, the matching circuit 9 is activated and matches the partial character string 311 stored in both temporary memories 6 and 8 with the character segment information 202. If the two match as a result of the comparison, the accommodation position information 112 of the corresponding speech segment information 102 stored in the character segment information 202 in the character segment accommodation location information area 2 is extracted by the accommodation location information extraction circuit 10, It is stored at the beginning position of the text information storage buffer 12 in the storage device via the accommodation position information storage circuit 11 (immediately after the text number 120 of the voice response text 301 that you wish to edit now). As described above, the storage position information 112 of the speech unit information 102 corresponding to the partial character string 311 consisting of the first ten kana characters of the voice response sentence 301 is stored at the beginning position of the text information storage buffer 12. Subsequently, the accommodation position information storage circuit 11 causes the pointer update circuit 13 to update the partial character string extraction pointer by 10 characters. Therefore, the partial character string extraction circuit 5 extracts a partial character string 312 consisting of the next 10 characters of the voice response sentence 301 already stored in the input voice response sentence buffer 4, and stores it in the partial character string temporary storage 6 in the same manner as described above. do. Subsequently, as described above, the character segment extraction information extraction circuit 7 extracts the first character segment information 202 from the character segment storage position information area 2 and stores it in the character segment information temporary storage 8. Next, the matching circuit 9 uses the partial character string 312 and the character segment information 202 stored in both temporary memories 6 and 8.
This time, the two did not match. The matching circuit 9 causes the pointer updating circuit 14 to update the pointer for extracting character segment information by one piece of information. Therefore, the character segment information extraction circuit 7 uses the character segment storage position information area 2
The second character segment information 201 is extracted from the character segment information 201 and stored in the character segment information temporary storage 8. Thereafter, the matching operation is repeated in the same way, and the pointer update circuit 14 continues until the character segment information that matches the partial character string 312 stored in the partial character string temporary storage 6 is extracted from the character segment storage position information area 2. updates the pointer to one piece of information,
Repeat verification. Even if all the character segment information in the character segment storage position information area 2 is extracted, the partial character string 3
12, if there is no match, the pointer update circuit 14 activates the mismatch character output circuit 15, and outputs the first character of the partial character string 312 for which matching character segment information was not obtained to a typewriter device, for example. 16. At the same time, the mismatched character output circuit 15 causes the pointer update circuit 13 to update the partial character string extraction pointer by one character. Therefore, the partial character string extraction circuit 5 extracts 10 characters from the new pointer given from the pointer update circuit 13, and compares them with the character segment information as described above. The mismatched characters output to the typewriter device 16 are presented to the voice response sentence editing requester, and new voice segment information is placed in the digital voice information area 1, and corresponding character segment information is transferred to the character segment storage position information area. 2, or change the expression of the voice response sentence 301 and input it again.

In the above explanation, the partial character strings 311, 312 extracted from the input voice response sentence buffer 4 by the partial character string extraction circuit 5 are fixed to a predetermined fixed number of characters, which is 10 characters in this embodiment. However, the following methods are also effective for editing more efficiently. That is, when inputting the voice response sentence 301 desired to be edited from the typewriter device 3, a delimiter code, for example, a blank code, predetermined according to the words or phrases in the sentence is inserted. When extracting a partial character string from the input voice response sentence buffer 4, the partial character string extraction circuit 5 extracts partial character strings with a variable number of characters separated by the delimiter code (blank code), and stores character segments in the same way as described above. Location information area 2
A comparison is made with the character segments 201, 202, . . . In this case, all character segments 20
1, 202, . . . for a partial character string that does not match as a result of the comparison, the unmatched character output circuit 15 outputs all characters of the partial character string to the typewriter device 16, and the pointer update circuit 13 outputs one character. Do not update the address and use the next delimiter (blank code)
The partial character string extraction circuit 5 is instructed to extract the next partial character string surrounded by . Regarding the unmatched partial character strings output to the typewriter device 16, the person requesting the editing of the voice response text performs a pair search similar to the above-mentioned partial character string extraction consisting of a certain number of characters, as well as searching for the insertion position of the delimiter (blank code). I will also try making changes.

In any of the partial character string extraction methods described above, for the full text of the voice response sentence 301 that you wish to edit, the storage location information of the voice segment information corresponding to the character segment information that has found a matching match is stored in the text information storage buffer. 12, the typewriter device 3
A digital voice information index table 17 corresponding to the voice response sentence 301 input in kana characters is completed in the text information storage buffer 12. By registering the obtained digital voice information index table 17 in a voice response section (not shown), a desired voice response sentence can be activated.

As is clear from the above description, according to this embodiment, the voice response sentence 301 in kana characters inputted from the typewriter device 3 has a predetermined number of characters (10 characters) or a delimiter inserted in advance. It is decomposed into partial character strings extracted by (blank code), and the speech segment information that matches each partial character string is detected using the character segment information as a medium, and the desired digital speech information index table 17 can be obtained. .

Please note that this figure is only one embodiment of the present invention.
For example, the functional assignment of the various circuits shown in the figures may be modified in many ways, but the effects of the present invention will not change in any case. In addition, the number of characters of a partial string extracted from a voice response sentence indicated by a phonetic string is not limited to 10 characters, and the delimiter used to extract a partial string is a blank symbol. It is not limited to.

As described above, according to the present invention, by inputting a voice response sentence to be edited into phonetic characters, voice segment information matching a partial character string constituting the voice response sentence is automatically generated from among existing voice segments. A desired digital voice information index table can be quickly obtained, and a voice response sentence editing method that can cope with the rapid increase in voice response services is realized.

[Brief explanation of the drawing]

The figure is a diagram of a voice response sentence editing method according to an embodiment of the present invention. In the figure, 1...digital audio information area, 2
...Character segment storage position information area, 3, 16...Typewriter device, 4...Input voice response sentence buffer, 5...Partial character string extraction circuit, 6...Partial character string temporary storage, 7...Character Fragment information extraction circuit, 8...
...Temporary memory of character segment information, 9...Verification circuit, 10
...Accommodation position information extraction circuit, 11... Accommodation position information storage circuit, 12... Text information storage buffer, 1
3, 14...Pointer update circuit, 15...Unmatched character output circuit, 17...Digital speech information index table, 101, 102...Speech unit information, 11
1,112...accommodation position information, 201,202...
... Character element information, 301 ... Voice response sentence, 31
1,312...Partial character string, 120...Text number.

Claims (1)

[Scope of Claims] 1. In editing a response sentence for a voice response device, a plurality of voice unit information converted into digital signals, storage position information in a storage device of the voice unit information, and the voice unit information are a storage device for storing character segment information expressed in phonetic characters; a means for extracting a partial character string from a voice response sentence input as a phonetic character string; and the extracted partial character string and the character segment. and means for storing storage position information of voice segment information corresponding to character segment information that matches as a result of the verification in a storage device according to the position of the partial character string in the voice response sentence. A voice response sentence editing method characterized by providing. 2. The voice response sentence editing method according to claim 1, wherein the partial character string extracted from the voice response sentence consists of a predetermined number of phonetic characters. 3. The voice response sentence editing method according to claim 1, wherein the partial character strings extracted from the voice response sentence are separated by delimiters inserted in advance into the voice response sentence.