JPH0612520B2 - Voice response sentence output method - Google Patents

Description

Description: TECHNICAL FIELD OF THE INVENTION The present invention relates to a voice response device, and more particularly to a voice response sentence output system.

Background of technology In order to cope with the sophistication of communication networks accompanying the increase in information activities,
In addition to the digitization of communication networks, the introduction of various new services is planned. Among these new services, the present invention refers to a voice response sentence output method related to a voice response device using a voice processing technique which has been rapidly developed in recent years.

2. Related Art and Problems Conventionally, when outputting a voice response sentence, a method of editing the voice response sentence requested by the customer and then simply returning it once has been used. For this reason, if the customer misses the information he or she really wants to know, there is a problem in that the customer has to make another inquiry. Of course, such problems are related to the customer's attention.
From the viewpoint of improving the service, it is preferable to take some measures. Therefore, as a method of repeatedly outputting a part of a voice response sentence, by writing a group of speech pieces that need to be repeatedly output in the edit result area many times and extracting the edit results in order when outputting the voice, A method of repeatedly outputting voice may be considered. However, when assembling a voice response sentence by voice unit editing, it is common that the number of voice units has an upper limit, and if such a method is adopted as it is,
Since the number of speech pieces increases, the limit on the actual length of the voice response sentence becomes severe. On the other hand, it is possible to repeatedly output the entire voice response sentence by the above method, but the limitation on the length of the voice response sentence is more severe. For example, due to the limitation of the memory capacity, there is a problem that if the document is output twice, only a sentence with half the number of voices that can be output once can be composed.

OBJECT OF THE INVENTION It is an object of the present invention to propose a voice response sentence output system which can solve the above problems.

In order to achieve the above object, the present invention is prestored from an information string representing the structure of a sentence designated by a sentence designation code input from the outside and an information sequence designating a detailed form of the sentence. In a voice response device that edits a sentence using a plurality of types of voice information and outputs the edited sentence by voice, one or more parts of the sentence are repeatedly output in the information sequence of the edited sentence. And a means for holding the voice output state at the time of voice output, which is repeatedly designated from the control information and the voice output state holding information included in the information sequence of the edited sentence. The feature is that the voice output is repeated a number of times.

That is, by providing a function of partially outputting a repeated voice response sentence, when a predetermined voice response sentence is edited and output, an arbitrary phrase of the voice response sentence (a phrase including particularly important information) is repeated. The output is performed so that the user does not miss the information, and further, the entire sentence can be repeatedly output without the user's operation.

Embodiment of the Invention FIG. 1 is a block diagram showing a basic configuration example of a voice response system to which the system of the present invention is applied. In this figure, 11
Is a host device, more specifically, a digital exchange. The upper device 11 includes a digital communication path switch (SW) 13 and a central control unit (CC) 14. The voice response unit (ARU) 12 of the present invention is connected to the host device 11.

The voice response device 12 uses, for example, an LSP as voice data.
The speech data analyzed by the (line spectrum pair) method is accumulated, and sentences are synthesized by inputting the data to the speech synthesizer one after another at an appropriate timing. The digital speech path interface unit (HWI) 15 is used. A voice response sentence is output via the exchange control system interface unit (CCI) 16 to receive a text designation code and a text detail designation code. A voice synthesis unit 17 is connected to the HWI 15, and a voice editing unit 18 is connected to the CCI 16. The voice editing unit 18 mainly operates by the control device 181, and the control is performed by the block group denoted by reference numerals 182-186. A control memory 182 stores a microprogram for editing and synthesizing a voice to be responded to. Reference numeral 183 denotes a voice memory, which stores voice data obtained by LSP analysis in a single tone and word format. Reference numeral 184 denotes a sentence pattern memory that stores information representing the structure of a sentence. Specifically, it stores map data (see FIG. 2A) that indicates the accommodation position of the voice data in the voice memory 183. A buffer memory 185 stores the edited result of the voice response sentence to be output for each line (see line # 0 ... line # 31 shown in the voice synthesizer 17 in the figure) (see FIG. 3A). . A work area memory 186 is used as a work area by the control device 181, and is roughly divided into a data buffer area and a line control word (CLW) area. This data buffer is a reception buffer for temporarily storing the request from the host device 11. Also, line control word (CLW)
Is an area for holding the progress state of each line for synthesis when synthesizing the voice response sentence for each line (see FIG. 4).

In the embodiment shown in FIG. 1, 32 speech synthesizers 17
-0 to 17-31 can be operated in parallel.

FIG. 2A is a diagram showing an example of the sentence pattern map data stored in the sentence pattern memory 184 of FIG. 1 and is also a characteristic part of the present invention. The sentence pattern map is composed of, for example, M types of sentence patterns, and an example of the arbitrary Nth sentence pattern map N will be described in detail. As shown in the figure, the sentence pattern map includes, for example, fixed parts # 1, # 2, # 3 and # 4, variable parts # 1 and # 2, and an end mark indicating the end of the sentence pattern.

FIG. 2B is a diagram showing the format of the sentence pattern fixing portion shown in FIG. 2A, which has, for example, a 2-word (2W) structure of 16 bits per word. In the second word, the lower 16 bits of the address (speech address) of each sound piece of the sound data stored in the sound memory 183 are written, and the 0th to 3th bits of the first word are written. The upper 4 bits PA of the voice unit address are written in.

FIG. 2C is a diagram showing a format of the sentence pattern map variable portion shown in FIG. 2A, which has, for example, a 1-word (1 W) structure, and a mark (M) area is at the 7th bit and a flag is at the 0th to 2nd bits. Each (FLG) area is allocated. The remaining bits are empty. When the mark (M) has a logic "1", it represents a variable part. In addition, the flag (F
LG) indicates the processing contents of the variable part, particularly "1" and "1".
The pattern of "0" indicates the beginning of partial repetition, and the pattern of "1""1""1" indicates the end of partial repetition. The other patterns indicate the contents of various character processes.

FIG. 3A shows the buffer memory 18 of FIG. 1 as a result of the editing operation.
FIG. 3B is a diagram showing an example of the edited result stored in FIG.
The figure is a diagram showing an example of the voice unit address data in FIG. 3A. In particular, the configuration of Figure 3B is based on the present invention. In FIG. 3A, in the buffer memory 185, for example, in the case of 32 lines, the lines are divided into a line # 0 buffer region and a line # 31 buffer region, and an example of a detailed line #N buffer is shown in detail. . As described above, the buffer memory 185 stores the edited result of the voice response sentence to be output for each line.
It consists of 128 words per line. E in line #N buffer indicates an unused word. The line #N buffer comprises a plurality of voice unit addresses # 0, # 1 ... And an end mark. The basic structure of each voice unit address is almost the same as that shown in FIG. 2B, but will be described later. As shown in FIG. 3B, the region (indicated by m in the figure) of the partial repeat control information is assigned to the voice unit address after editing.

FIG. 3B is a diagram showing the format of the voice unit address shown in FIG. 3A, and the partial repetition control information m is assigned to, for example, the fourth bit. If the mark m is "1", it indicates that the speech piece requires partial repetition, and if it is "0", it indicates that repetition is unnecessary. In this case, at the time of synthesizing, the speech unit with the mark m of “1” is synthesized a predetermined number of times (for example, only twice), and the other portions are synthesized only once.

From the sentence pattern map 51 shown in FIG.
The editing operation of the control unit until the editing result shown in the buffer memory 52 of the figure is obtained will be described with reference to the flow of FIG.

FIG. 4 is a flow chart showing the operation at the time of voice editing including partial repetition.

(1) The first word (51 in FIG. 6) of the sentence pattern map regarding the sentence pattern instructed from the upper level device 11 in the sentence pattern memory 184 is loaded (step a).

(2) The mark M in the variable portion of the sentence pattern map (Fig. 2C) is "0".
(Step b) And since the partial repeat mark m1 indicating whether the edit processing target part held in the control unit is inside or outside the partial repeat range is "0" (step c), the route of the operation flowchart is taken. , Voice unit # 1 address,
It is loaded into the buffer memory 185 (step e) (see the voice unit # 1 address of 52 in FIG. 6).

(3) The third word (51 in FIG. 6) of the sentence pattern map is loaded from the sentence pattern memory 184.

(4) The mark M in the variable portion of the sentence pattern map is "1" and the flag F
LG is “beginning of partial repetition” (FLG =
Since it is "1""1""0") (step f) (sixth)
Starting from the partial repeat of 51 in the figure), the route of the operation flowchart is taken and the partial repeat mark m1 is set to "1" (step g).

(5) The fourth word (51 in FIG. 6) is loaded from the sentence pattern memory 184.

(6) Since the mark M in the variable part of the sentence pattern is "0" and the partial repeat mark m is "1", the repeat mark m is set to "1" for the voice unit address from the sentence map through the route. (Step d) and with the fifth word (speech unit # 2 address in 51 of FIG. 6), it is stored in the buffer memory 185 (see the speech unit # 2 address of 52 of FIG. 6).

(7) The sixth word (51 in FIG. 6) is loaded from the sentence pattern memory 184.

(8) The same processes as (5) and (6) above are performed and stored in the buffer memory 185 (refer to voice unit # 3 address 52 in FIG. 6).

(9) The 8th word (51 in FIG. 6) is loaded from the sentence pattern memory 184.

(10) Since the mark M of the variable portion is "1" and the flag FLG indicates "partial repetition end" (partial repetition of 51 in FIG. 6) (FLG = "1""1""in the example described above). 1 ") and the partial repeat mark m1 passing through the route is set to" 0 "(step h).

(11) The ninth word is loaded from the sentence pattern memory 184.

(12) The same processing as (1) and (2) above is performed.

(13) Since the 11th word is the end mark (the bottom row of 51 in FIG. 6), the editing is finished (step j), and the synthesis in the latter stage illustrated in FIG. 6 is started. In step i, the flag FLG (FIG. 2C) is FLG ≠ “1” “1” “0”,
In addition, various character processing is performed when FLG ≠ “1” “1” “1”.

FIG. 5 is a diagram showing the structure of a line control word (CLW) for one line. As described above, CLW is an area that constitutes a part of the work area memory 186 of FIG. 1, and holds the progress of synthesis when synthesizing the voice response sentence for each line. It will be updated with reference to. In the figure, as a general rule, the lower 16 bits of the voice data to be inputted to the synthesizer next are input to the second word, and the upper 4 bits PA 'of the same address are input to the first word.
When the composition of the partial repeat control information and the currently output voice unit is completed, the pointer is used to indicate the voice unit to be synthesized and output next.

In this embodiment, at the time of voice synthesis, the partial repeat mark m at the voice unit address in the buffer memory 185 is used.
Depending on "1" or "0" in (Fig. 3B above), CLW
The partial repeat mark m1 (Fig. 5) and the partial repeat second mark m2 (Fig. 5) in the above are set to "1" or "0" respectively, and these marks m1 and m2 are viewed, and any syllable of the voice response part is displayed. Is partially repeated and output.

FIG. 6 is a diagram schematically showing the process of editing a voice response sentence including partial repetition. Each voice unit # 1 to # 4 is specifically composed of the following voices, particularly voice unit # 2 and voice unit. It is assumed that the response sentence of piece # 3 is important and is repeated.

Voice unit # 1: "You have called" Voice unit # 2: "No. 123-4567" Voice unit # 3: "Not used" Voice unit # 3: "Please call again" Here Since voice unit # 2 and voice unit # 3 are within the range of partial repetition, in the end, the final voice response sentence is as follows: "You asked me, No. 123-4567 is not used, No. 123-4567 is not used. Please call again. " The part 51 in FIG. 6 belongs to the range of the sentence pattern memory 184 in FIG. 2A, and the part 52 is the third part.
A part 53 belongs to the range of the buffer memory 185 of FIG. A and is a final voice response output sentence.

The operation of the control unit for synthesizing the voice response sentence from the edited result shown in the buffer memory 52 of FIG. 6 will be described below with reference to FIG. FIG. 7 is a flow chart showing the operation of connecting the voice units during voice synthesis including partial repetition. The content of each step is as described in the figure. According to the specific example of the response sentence described above, (1) By the processing of the route, "you were the call" is synthesized. At this time, m1 = “0”, m2 =
It is "0". m1 and m2 are a partial repeat mark and a partial repeat second mark in CLW (FIG. 5), respectively.

(2) Depending on the processing of route and
No. 67 is not used. " At this time, m1 = "1" and m2 = "0".

(3) Depending on the route and
No. 67 is not used. " At this time, m1 = "1" and m2 = "0".

(4) "Recall" is synthesized by processing the route and. At this time, m1 = “0”, m2 =
It is "0".

The synthesis is completed here, and the speech synthesizers 17-0 to 17-31 shown in FIG.
A desired voice response sentence is sent to the customer via any of the above.

In the sentence pattern shown in FIG. 6 described above, only the fixed part is included for simplification of description, but the variable part is included in the repeating range.
It may be a sentence pattern included outside. For example, the telephone number portion of the above sentence is usually a portion to be newly designated every time there is an instruction to output a voice response sentence, and the sentence pattern should be a variable portion. Also, the partial repetition information m in FIG. 3B
In addition to the above-described embodiments, when the fifth and sixth bits are any of "01", "10", and "11", a sound piece that repeats 2, 3 and 4 times respectively. It is also possible to make an arrangement so as to indicate that it is the head of the group, and that when the fourth bit is "1", it indicates that the speech piece immediately before that is the tail of the partial repetition. In this case, it is possible to change the number of repetitions for each synthesized voice unit group. For example, it is possible to use a method of designating a repetition four times for a portion where the content is important and a repetition twice for a less important portion. The method of specifying the number of times may be a method of determining it when the sentence pattern map is determined, but it is also possible to allow it to be specified when the upper device instructs synthesis of the voice response sentence. In addition to the above, in carrying out the present invention, a method of constructing a sentence pattern, a format of an edited result, a method of constructing a line control word, an editing operation, a combining operation,
Alternatively, it goes without saying that various modifications can be made in the device configuration and the like.

Although the above description is for the case where only a part of one voice response sentence is repeatedly output by voice, the following description will be given of the entire repeated voice output of the voice response sentence. In the case of the entire repetition, by including the information indicating the presence or absence of the repetition or the number of times in the information string of the edited result in the editing operation,
At the time of the synthesizing operation, a repeating operation is performed a specified number of times based on this information and the information on the progress of synthesizing held in CLW. It is also possible to specify the number of total repetitions from the higher-level device in the same way as other codes when instructing the voice response sentence output from the higher-level device, and can be set in advance in the sentence pattern. is there. Even in the entire repetition, various modifications can be made in the method of implementation.

EFFECTS OF THE INVENTION As described above, according to the present invention, it is possible to repeatedly respond only to important parts, and it is possible to improve the service in voice response.

[Brief description of drawings]

FIG. 1 is a block diagram showing a basic configuration example of a voice response system to which the method of the present invention is applied, and FIG. 2A is a diagram showing an example of sentence pattern map data stored in the sentence pattern memory 184 of FIG. 2B is a diagram showing an example of the sentence pattern map fixing unit shown in FIG. 2A, FIG. 2C is a diagram showing the format of the sentence pattern changing unit shown in FIG. 2A, and FIG. 3A is the buffer of FIG. FIG. 3 is a diagram showing a voice response sentence stored in the memory 185, FIG. 3B is a diagram showing an example of the voice unit address shown in FIG. 3A, and FIG. 4 is an operation at the time of editing a voice response sentence including partial repetition. 5 is a flowchart showing an example of one line of a line control word (CLW), FIG. 6 is a diagram schematically showing a process of editing and synthesizing a voice response sentence including partial repetition, and FIG. The figure shows the operation when synthesizing a voice response sentence including partial repetition. Is a low chart. 11 ... Host device, 12 ... Voice response device, 17 ... Voice synthesis unit, 18 ... Voice editing unit, 181 ... Control device, 182 ... Control memory, 183 ... Voice memory, 184 ... Text pattern memory, 185 ... Buffer memory, 186 ... Work area memory.

Front page continuation (72) Inventor Shuichi Hashimoto 1015 Kamiodanaka, Nakahara-ku, Kawasaki-shi, Kanagawa Fujitsu Limited (72) Inventor Minoru Oyama 3-9-11 Midoricho, Musashino-shi, Tokyo Nippon Telegraph and Telephone Public Corporation Musashino Telecommunications Inside the research institute (72) Inventor, Naozumi Tsuchida 3-9-11 Midoricho, Musashino City, Tokyo Inside Musashino Telecommunications Research Laboratories, Nippon Telegraph and Telephone Public Corporation (56) Reference JP 58-62738 (JP, A) Actual development Sho 56- 64100 (JP, U)

Claims (1)

[Claims] 1. A voice response device (1) for generating and outputting a voice response sentence to be responded to according to an instruction from a higher-level device (11).
In 2), the voice response sentence is received as a sentence pattern map (51) from the higher-level device (11), and the sentence pattern map data stored in the sentence pattern map (51) is the voice response map. A plurality of sentence pattern map fixing units including a voice unit address for displaying an address corresponding to each voice unit in a voice memory (183) in which various voice units constituting a response sentence are stored in advance, and these sentence pattern map fixing units. And a sentence pattern map variable portion inserted between a position where the partial repetition is to be started and a position where the partial repetition is to be ended in the voice unit address. When the voice response sentence is edited with reference to the map (51), the voice unit addresses are sequentially written in the buffer memory (52), and each of the voices for which the repetition is designated by the sentence pattern changing unit. To one address For this purpose, a partial repeat mark is written, and further, an end mark is written for the final voice unit address, and when the voice response sentence is output, the voice memory (183) is sequentially read according to the contents of the buffer memory (52). A voice response sentence output method, wherein the voice is synthesized up to the end mark by accessing and by repeatedly accessing the partial repeat mark.