JPS5922240B2 - Voice response editing method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a voice response sentence editing method that allows voice response sentences to be easily changed and added.

In recent years, in response to voice output requests from outside, voice response sentences are created in which the digital information of the voice stored in units of words (hereinafter referred to as word format) in the voice memory is edited and the response sentence is output as voice. An editing method has been proposed and put into practical use.

FIG. 1 is a block diagram showing the configuration of a voice response device that implements a conventional voice response sentence editing method, and 1 is a block diagram showing the structure of a voice response device that implements a conventional voice response sentence editing method, where 1 gives the voice response device the text number of the sentence to be outputted and the line number for outputting the voice. 2 is a control device that edits voice response sentences using a microprogram; 3 is a control memory in which a microprogram for operating the control device 2 is stored; 4 is a PARCOR code, an LSP code, and a PARCOR code;
A voice memory in which voices digitized by CM code etc. are recorded in the form of words; 5 is a storage position information of a word series of a sentence to be outputted, that is, all the words constituting the sentence to be outputted in voice; 61 to 61 (subscripts 1 to i indicate line numbers) are synthesizers that convert digital voice information from the voice memory 4 into voice information; 7 is a control device; 2, control memory 3, audio memory 4, buffer memory 5, synthesizer 61
This is a voice response device consisting of 61.

As shown in Fig. 2, the voice memory 4 currently stores digital information of voices such as ``Please call me back'', ``Go home'', and ``Please call me'', and also includes voice responses. The device 7 selects the next two sentence numbers # based on the audio digital information stored in the audio memory 4.
Assume that it is possible to synthesize sentences #1 and #2.

In FIG. 2, A represents word delimiter information #1. Please call again. #2. Please call home.

The operation of the voice response device 7 in this case will be explained below.

That is, if the host device 1 outputs audio output request information that says, for example, "output the text with text number #2 to the line with line number 1," and the control device 2 receives this audio output request information, the control The device 2 branches according to the document number according to the microprogram having the logic shown in FIG. 3 stored in the control memory 3, and selects the document number #2.
The two sentences that make up the sentence are “home” and “please call me.”
The accommodation position information and sentence end information of one word in the voice memory 4 are set in the area allocated to the buffer memory 5 corresponding to the line number 1. FIG. 4 shows the contents of the buffer memory 5 at this point. As shown in FIG. 4, once the storage position information of the two words "home" and "please call" and the sentence end information are set in the area corresponding to line number 1 of the buffer memory 5, the control Based on the storage location information set in the buffer memory 5, the device 2 sequentially reads digital information of voices indicating "home" and "please call" from the voice memory 4, and sends the digital information to the synthesizer 61 corresponding to line number 1. In addition, the synthesizer 61 performs speech synthesis based on the provided digital information.

The conventional voice response text editing method edits the response text and performs speech synthesis as described above, but when adding new sentences or changing sentences, the disadvantages described below occur. It was hot.

Now, let us consider the case where the following sentence with sentence number #3 is to be added. #3 Please call the company.

In this case, if we look at the voice memory 4, the words for "Please call us" have already been stored, so
It is sufficient to newly store only the words "to the company" and make the contents as shown in Fig. 5, but at the same time, the microprogram stored in the control memory 3 is changed to one having the logic shown in Fig. 6. The need arises.

Note that in FIG. 5, A represents word delimiter information. As shown in Figures 5 and 6, in the conventional method,
When adding a new sentence, not only the voice memory 4 but also
Since the control memory 3, which is generally made up of read-only memory (ROM), also needs to be changed, adding text,
The drawback was that changes could not be easily made. The present invention has improved the above-mentioned drawbacks, and its purpose is to enable voice response sentences to be added or changed without adding or changing microprograms.

Examples will be described in detail below. FIG. 7 is a block diagram of a voice response device implementing the method of the present invention.
12 is a control device for editing the voice response sentences using a microprogram; 13 is for operating the control device 12; 's microphone. A control memory 14 stores a program, and a sentence pattern memory 14 stores storage position information of a word series of a response sentence (hereinafter referred to as a sentence pattern) for each sentence number.
Reference numeral 15 denotes a voice memory in which voices digitized by PCM codes, PARCOR codes, LSP codes, etc. are stored in the form of words, and 16 denotes a buffer memory in which storage position information of a word series of a sentence to be outputted in accordance with a line is stored. ,
171 to 171 (subscripts 1 to i indicate line numbers) are synthesizers that convert audio digital information from the audio memory 15 into audio information, 18 is a control device 12, control memo 1 river 3,
This voice response device consists of a sentence pattern memo 4, a voice memo 5, a buffer memory 16, and a synthesizer 17, to 171.

Now, for example, on May 5, a voice memo says, ``Please call me back.

”, “Go home”, and “Please call me” are stored as digital information as shown in FIG. Suppose that it is possible to synthesize the sentences of In FIG. 8, A represents word delimiter information. #1. “Please call back.

”#2. [Please call home.]

] In this case, in this embodiment, the newly created sentence pattern memo I month 4 contains the sentence pattern of the sentence number #1 in order from address O, as shown in FIG. , the delimiter information at the end of one sentence, the sentence pattern of the sentence with sentence number #2, and the delimiter information at the end of all sentences are stored.

Now, for example, if information is given from the host device 11 that says, "Output the text with text number #2 to the line with line number 1," the control device 12 outputs the microprogram stored in the control memory 13. Based on this, the sentence pattern memory 14 is sequentially searched from address O.

Then, according to the delimiter information at the end of one sentence, sentence number #2
When the control device 12 detects the first address of the area where the sentence pattern of the sentence is stored, the control device 12 transfers the storage location information of ``to my home'' and ``please call me'' stored in this area to the buffer memo 1. Set in the area assigned to line number 1 of month 6. FIG. 10 shows the contents of the buffer memory 16 at this point. Next, the control device 12 sends the message "Go to home" stored in the buffer memo I/6.
Based on the storage location information of "Please call me" and "Please call me", the digital information of the voice representing "Home Blade 2 "Please call me" is sequentially read from the voice memory 15 and added to the synthesizer 171 corresponding to line number 1. .

The synthesizer 171 performs voice synthesis based on the digital voice information provided from the voice memory 15, and outputs voice information. Furthermore, from the top table 11,
When there is a request to output audio for sentences other than sentence numbers #1 and #2, error processing is performed in the same way as in the conventional method, but when searching the sentence pattern memory 14 sequentially from address O, all sentences If the starting address of the area that stores the sentence pattern of the sentence requested for audio output cannot be detected before the end delimiter information is detected, the microprogram is configured to handle an error. Then, even when sentences are added, error handling can be performed without changing the microprogram. Next, consider the case where the following sentence is added as the sentence number #3.

#3. Please call the company.

In this case, digital information of the voice saying "to the office" is newly stored in the voice memo 1 and 5, and its contents are shown in FIG. 11 (A in the figure represents word delimiter information).

), and if the sentence pattern memory 14 is newly stored with storage position information for "to the office" and "please call me", and the contents are shown in FIG. 12, the control The output request for the text with text number #3 can be handled without changing the microprogram stored in the memory 13. That is, as described above, in the method of the present invention, the sentence pattern memory 14 is sequentially searched from address O, and the first address of the area in which the sentence pattern of the sentence of the sentence number for which the voice output request has been made is stored. The storage position information of the word detected and stored in this area is set in the buffer memory 16, and the voice memo 1 river 5 is read out based on the storage position information set in the buffer memory 16. As can be seen from Figure 12, the information regarding the added sentence number #3 is also stored in the voice memory 15 and sentence pattern memo January 4 in exactly the same recording format as the sentences #1 and #2. Since it is stored, voice response sentences can be added without changing the microprogram that is the editing logic for the voice response sentences stored in the control memo I river 3. In the embodiment, the sentence pattern memory 14, the voice memory 15, and the buffer memo January 6
Although these memories are provided separately, it is also possible to provide these memories in the same space to simplify the configuration.

Further, although the explanation has been omitted in the embodiment, when the voice response device 18 is started up, an index table is created that stores the beginning position information of the sentence pattern of each sentence number, as shown in FIG. 13, for example. Memory 13 that controls the logic that automatically generates
If you add this to the internal microprogram and search this index table to detect the first address of the area where the sentence pattern of the sentence whose sentence number is requested for voice output is stored, the number of sentences can be reduced. Even if there are a large number of search results, the search time can be shortened. As explained above, the present invention provides a sentence pattern memory that stores the sentence pattern for each sentence of each sentence number, searches the sentence pattern memory based on audio output request information, and responds to the audio output request. The area in which the sentence pattern of the response sentence is stored is detected, and the response sentence is edited based on the storage position information set in the area. It is possible to add or change a response sentence without changing the stored microprogram, and therefore there is an advantage that processing when adding or changing a voice response sentence is extremely easy. Further, according to the method of the present invention, there is no need to replace the control memory 13, which is generally made up of a read-only memory (ROM), when adding or changing a response sentence, so there is an advantage that it is economical.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of a voice response device implementing the conventional method, FIGS. 2 and 5 are diagrams showing the contents of the voice memory 4, and FIGS. 3 and 6 are diagrams showing the logic of the microprogram. 4 is a diagram showing the contents of the buffer memory 5, FIG. 7 is a block diagram of a voice response device implementing the method of the present invention, FIGS. 8 and 11 are diagrams showing the contents of the voice memory 15, and FIG. ,
Figure 12 is a diagram showing the contents of Sentence Pattern Memo January 4, and Figure 10 is Batsufua Memo! FIG. 13 is a diagram showing the contents of month 6, and FIG. 13 is a diagram showing the contents of the index table. 1 and 11 are host devices, 2 and 12 are control devices, 3 and 13 are control memories, 4 and 15 are voice memo 18, 5 and 16 are buffer memo I8 61 to 61, and 171 to 171 are synthesizers, 1
4 is a sentence pattern memory.

Claims (1)

[Claims] 1. In a voice response sentence editing method in which a response sentence is edited and output as voice information based on voice output request information from a host device, the voice output request information is edited according to a microprogram stored in a control memory. a control device that performs editing control according to the response; a voice memory that stores voice as digital information in the form of words; and a sentence that stores storage position information in the voice memory of the word series of each response sentence for each response sentence to be outputted as voice. a buffer memory for storing storage position information in the voice memory of a word sequence of a response sentence to be outputted in correspondence with an output line; and a synthesizer for converting digital information from the voice memory into voice information. In preparation, based on the audio output request information from the higher-level device,
The sentence pattern memory is searched to detect an area in which storage position information of a word series indicating the format of a response sentence for which a voice output request has been made is stored, and the storage position information stored in the area is stored. set in a predetermined area of the buffer memory, and read digital information stored in the audio memory based on accommodation position information set in the predetermined area;
A voice response sentence editing method characterized by being added to the synthesizer.