JP2000137496A - Scheme and system for retrieving telephone number employing speech recognition - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve the recognition rate in speech recognition while simplifying the constitution of the system by conducting a phoneme analysis twice for the name of the person being called told by a caller and the name repeated by an operator, conducting speech recognition that discriminates a matching portion as the name of the person being called and retrieving a telephone number of the person. SOLUTION: A caller A calls an operator, saying 'Would like to speak to Mr. B' requesting a telephone transfer to Mr. B (1). The voice pronounced by the caller A is transmitted to the operator and a speech recognition trunk (V-DET) ('B' for a first time). The operator says 'You wish to speak to Mr. B, do you? Please wait for a moment' to confirm the objective of the caller A (2). The voice 'Mr. B' pronounced by the operator is transmitted to the caller A and the trunk V-DET ('B' for a second time). At the trunk V-DET, the 'B' (the first time) from the caller A and the 'B' (the second time) from the operator are repeated. Thus, the system easily recognizes the fact that the caller A wishes to speak with Mr. B, the telephone number of Mr. B '3771' is retrieved and is displayed on a repeating exchange (3).

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a telephone number search by voice recognition, and more particularly to a telephone transfer in which an incoming telephone call from outside the company is transferred to a called party via an operator. In a preferred method or system to be applied, the telephone number of the target called party is searched by voice recognition during transfer, and in that case, the voice recognition is performed while simplifying the device configuration. It has made it possible to increase rates and, in turn, to make it more efficient in telephone number searches.

[0002]

BACKGROUND OF THE INVENTION Telephone transfer includes telephone transfer via an operator (exchanger). In a telephone transfer process involving such an operator, when a call is received, the operator talks with the caller,
It is common practice to ask the name of the called party (the name of the person to be called) before forwarding the call. For example, there is a case where a telephone response form is adopted in which an incoming telephone call (incoming call) from outside the company is transferred to the corresponding target person (called person) through an operator.

By the way, when a private telephone having more than 100 lines is installed, for example, as described above, an incoming call from an outside company (incoming call) is often transferred to the called party by the intermediary of an operator. However, in the case where the number of lines is large or a plurality of tenants are jointly allocating operators, it takes time to search for the telephone number of the called party. The longer it takes to retrieve the called party's phone number, the more the transfer process will be delayed, and such a delay will result in a decrease in the number of calls accepted and exchanges in the transfer scene, and the incoming call may continue. In terms of services for the caller of an unsuccessful incoming call (incoming call), it is difficult to promptly respond to the caller, for example, by making the caller wait.

[0004] Therefore, the introduction of a system for retrieving a telephone number by voice recognition when an operator mediates and transfers a call to a called party when an incoming call is received is effective in such a case. In other words, in order to shorten the telephone number search time, the voice recognition device linked to the private branch exchange extracts the name of the called party from the contents of the incoming call and displays the telephone number search and the list of candidate candidates. The performance is improved.

[0005] Therefore, it is desirable to introduce such a system, whereby the time for searching for the telephone number of the called party can be reduced, and a private telephone having a considerably large number of lines is installed. Even in such a case, or in a case where a plurality of tenants are to arrange an operator jointly, it is advantageous to solve the above disadvantages and inconveniences.

In FIG. 6, a private branch exchange 300
(PBX) is connected to a relay console (switching console) 350 (ATT), and has a voice recognition trunk 304 (V-DET). In the figure, reference numeral 100 denotes a central office switching network 200 (CO).
The telephone (T) of the other party's caller (speaker A) connected via
EL) 360 is a telephone (T) of the called party (speaker B).
EL), 301 is an attendant console interface (ATT-IF), and 302 is a telephone interface (S-T).
LC) (single line), 303 is a central office trunk (C
OT) are shown.

Here, in the above, the telephone transfer via the exchange when the call is received can be performed in the order shown in FIG. 7 (i) to (iv). The voice (i) of "Please ask Mr. B" from speaker A (the caller) is the console 35
0 to the voice recognition trunk (V-DET) 304, based on which the operator responds (ii), while the voice recognition trunk (V-DET) 304 receives the "B" Search for phone number "3771"
After the display (iii), the operator sees the telephone number display "B: 3771" and transfers the incoming call from the speaker A to the speaker B (iv).

[0008] In this way, for example, for a system having a considerable number of private telephone lines and transferring incoming telephone calls from outside the company via an operator, a speech recognition device is connected to the private branch exchange, and the target person (the subject) is connected. A system for retrieving the corresponding telephone number of the caller can be constructed and introduced.

Here, generally, speech recognition can be performed in the steps of FIG. Steps S1 to S4 consist of acoustic processing, phonological analysis, word analysis (segment analysis), and sentence analysis (semantic analysis), and a recognition result is obtained by such a speech recognition procedure based on a speech signal.

However, when introducing the above-mentioned telephone number search, if it is desired to perform voice recognition of the name of the called party to be forwarded, the following points should be considered. The following points can be pointed out.

(A) Phonological analysis (S2 in FIG. 8) can be realized at a high recognition rate by a relatively simple method using a neural network or the like, but word analysis (phrase analysis) (S3) and sentence analysis (semantic analysis) (S4) needs to have high processing ability and a large amount of matching data (dictionary),
Expensive systems are required.

(B) If the voice from the caller as shown in FIG. 7 (i) is only targeted as a voice input to the voice recognition trunk (V-DET), "B" corresponding to the called party name portion in that case is used. Are difficult to extract, have low recognition rates,
Therefore, it is necessary to perform accurate phrase analysis in order to identify the name of the called party only from the contents of a telephone call (incoming call) from outside the company, but in some cases, it is necessary to perform semantic analysis as well. .

In the case of the method shown in FIG. 7, the transfer target B must be identified only by the voice from the speaker A (the caller), and which part of the voice from the speaker A is It is difficult to identify whether the name is "B".

This will be described more specifically with reference to FIG. 9 as well. In this case, for example, "Sumita" is a case where the callee is a target, and speaker A (the caller)
This is a scene in which the operator was told the meaning of "Please Sumita-san," as shown in the illustration, as shown in the voice of "... Sumita-san, please." To the voice recognition device. Here, in order to recognize the “Sumita” portion indicated by the reference numeral c in the figure as a subject name, a phrase analysis or a semantic analysis is necessary (even if a phoneme is simply recognized, Only meaningful as a string).

Therefore, an incoming call (an incoming call from outside the company)
If the callee's name is to be extracted by speech recognition only from the contents of the above, not only simple phonological analysis but also accurate phrase analysis is required, so the speech recognition device used as in (a) above is large. Scale and expensive.

(C) In addition to the above (a) and (b), for an incoming call, the communication level may fluctuate or the SN ratio may deteriorate depending on the line condition. Such a variation in the communication level and a deterioration in the SN ratio also result in a low recognition rate in voice recognition of only incoming calls. In particular, since the voice of the speaker A who is a caller outside the company passes through the network 200, the volume generally varies and the S / N ratio is poor, and thus the voice of the speaker A shown in FIGS.
This is a factor that lowers the recognition rate in the telephone number search method using voice recognition. If the recognition rate of the name of the person to be transferred is low, this may cause a search error and an incorrect transfer number.

Therefore, it is more desirable to incorporate the above-mentioned system so that the corresponding telephone number of the target called party can be searched by voice recognition when transferring a telephone call, and to properly and easily extract the name of the called party when introducing the system. The system can be introduced and constructed so that the recognition rate can be increased, the recognition rate can be increased, and the configuration of the voice recognition mechanism and the device can be simplified while improving the recognition rate. It is more desirable to reduce the recognition rate even if the caller's voice is transmitted through a network and there is a concern that the call level may vary or the SN ratio may be deteriorated, regardless of the factors involved. Or to improve the performance of the callee, or to perform speech recognition of the called party's name without performing phrase analysis or semantic analysis by simply performing phonological analysis. Analysis and the like can be omitted.

The present invention is based on the above considerations,
And based on the considerations described below, it is suitable to be applied to a telephone transfer for transferring a telephone call to a called party via an operator,
In addition to searching for the telephone number of the target called person by voice recognition, it is also possible to improve the recognition rate in voice recognition while simplifying the configuration, and to realize efficiency in the telephone number search. Things.

Further, it is possible to omit semantic analysis or the like that requires advanced functions, and to perform speech recognition of only the name of the called party without performing phrase analysis or semantic analysis only by performing phonological analysis. , With additional improvements,
The purpose of the present invention is to make it easier to extract the name portion from the voice for the purpose of identifying the called person's name, thereby improving the accuracy of the phoneme analysis of the called person's name.

[0020]

According to the present invention, there is provided a telephone number search system and system by voice recognition as described below. In other words, according to the present invention, when an incoming call is transferred to the transfer target by the operator, the caller's name announced by the caller and the called name repeated by the operator are two.
The phoneme analysis is performed a number of times, and the matched part is determined to be the called person's name, whereby the voice of the called person's name is recognized and the telephone number is searched.

Further, when the operator repeats the name of the called person, there is provided means for notifying the voice recognition device of a signal for identifying the name part.

Further, the present invention is a system for performing a telephone number search by voice recognition when a call is received and a transfer is made to the called party through the intermediary of an operator. Means for phonologically analyzing the call content of the operator, and the operator repeats the name of the called party reported by the caller, so that a common phonemic sequence appears in the caller-side phonological analysis result and the operator-side phonological analysis result. On the basis of this, there is provided a means for identifying a called party name area and performing a telephone number search.

Also, in the above (claim 3), while the operator pronounces the recalled name of the called person, the switch is operated to make it easy to extract a portion corresponding to the name from the voice. It is characterized by the following.

Further, the switch according to the fourth aspect is a predetermined switch provided in the exchange, and is configured to be provided in the exchange, or the switch according to the fourth aspect is provided with button information for the exchange. , Using a button of a dedicated telephone provided with means for notifying the user.

Further, in the case where a computer or a communication terminal for a computer is used as the switching board, the keyboard of the computer or the communication terminal for the computer is used as the switch according to claim 4. It is.

Further, the present invention is characterized in that it is provided with a hand throw for notifying the voice recognition device of keyboard press information of the computer or the computer terminal.

Also, whether the voice recognition function is in the exchange,
Alternatively, it is provided as an external device of the exchange.

Further, the voice recognition function is provided as an external device of the exchange, and is provided with a hand for notifying the operation information of the switch of claim 4 or 5 to the external device.
It is characterized by the following.

[0029]

According to the present invention, when an operator talks with a caller upon receiving an incoming call and listens to the name of the person to be transferred, which is to be transferred, the operator performs voice recognition in order to reduce the telephone number search time. The serviceability can be improved by extracting the name of the called party by the device and searching for the telephone number or displaying the list of candidate candidates. In this case, for example, the operator mediates incoming calls from outside the company (incoming calls). When the call is transferred to the target person (called person), the name of the called person announced by the caller and the name of the called person repeated by the operator
By performing phonological analysis a number of times, it is possible to determine a match (may not be a perfect match) as the name of the person to be called, thereby making it possible to omit semantic analysis or the like that requires advanced functions. .

Even when a private telephone with a considerably large number of lines is installed, or when an operator is arranged jointly by a plurality of tenants, it is advantageous for eliminating the disadvantages and inconveniences as described above, and furthermore, It is possible to appropriately configure a system that also considers the points (a) to (c) described above,
It is possible to improve the recognition rate while simplifying the configuration, and to realize more efficient telephone number search.

If the present invention is not adopted, it is difficult to identify which part of the voice from the caller is the name of the person to be transferred, and variations in the call level, deterioration of the S / N ratio, etc. According to the present invention, a transfer target person can be identified from a common recognition word of a voice from a caller and a voice from an operator, which causes a reduction in the recognition rate. The operator's voice pronounced by the operator in the repeat of the operator is also recognized, and since this is a voice of higher quality than the voice of the caller via the network, the recognition rate can be improved in this respect as well. In addition to improving the recognition rate of the called party name, the operator can repeat the called party's name and also recognize the operator's voice while avoiding a reduction in the recognition rate in voice recognition alone. Recognition It can simplify the configuration of the device. In this case, it is possible to perform speech recognition of the name of the called party without performing phrase analysis or semantic analysis only by performing only phonological analysis, and to avoid using a large-scale and expensive speech recognition device. Thus, the device configuration can be simplified. In addition, by increasing the recognition rate of the transfer target person as described above, the effect of preventing search errors can be expected, the transfer number error can be eliminated, and with the low recognition rate, the potential target person can be identified. When displaying in a list, many subjects are displayed in a list, but by improving the recognition rate, the subjects can be narrowed down and the list display can be made smaller, and in this regard, telephone number search is also more efficient .

According to a preferred embodiment of the present invention, it is possible to use a means capable of notifying the voice recognition device of a signal identifying the name part when the operator repeats the name of the called party. In addition, the accuracy of phonological analysis of the name of the called party can be improved.This is useful when it is difficult to judge a match because, for example, the quality of incoming telephone calls from outside the company is far inferior to the voice quality of operators in the company. It is particularly effective. This means that, for example, only the name part is extracted from the voice by pressing a switch provided on the exchange board only while the operator pronounces the recalled name for the purpose of identifying the called name area. The switch may be provided in a switchboard (intermediary console) or may be provided by using a button of a dedicated telephone provided with a means for notifying the switchboard of button information. Alternatively, when a computer or a communication terminal for a computer is used as the exchange, the keyboard of the computer or the terminal for the computer may be used as the switch.

[0033]

Embodiments of the present invention will be described below with reference to the drawings. 1 to 3 show one embodiment of the present invention. FIG. 1 shows an example of the configuration of a telephone number search by voice recognition according to the present invention. In this embodiment, a private branch exchange has a built-in voice recognition mechanism (and a case where the private branch exchange has an external device). ) Operator (exchanger)
Indicate call forwarding via FIG. 2 shows a specific example of the transfer in the system for searching for the telephone number, and FIG. 3 shows a specific example of the telephone number search target determination.

As shown in FIG. 1, a private branch exchange (PBX)
A relay console (exchange console) 350 (ATT) handled by an operator (exchanger) is connected to 300, and a telephone (TEL) 360 as a local telephone is connected to the 300.
As in FIG. 6, a telephone (TEL) 100 on the other party's caller (speaker A) side is connected through the central office 200.

In this embodiment, a number of private telephone lines, for example, 100 or more private telephone lines, are installed. It is assumed that the incoming call (the incoming call) is mediated by the exchange and the incoming call is transferred to the called party who is the transfer target. Accordingly, in the figure, the telephone 360 is one of many such private telephones, and represents a telephone of a called party (speaker B) which is a person to be called in our company. Illustrated. The relay console (switching console) 350 may use a computer or a computer communication terminal.

The private branch exchange 300 includes a relay station (exchange table) 3
Console interface (ATT-I
F) 301, a telephone interface (SLC) 302 to which the telephone 360 is connected, and a central office trunk (COT) 303 to which the network 200 is connected. When a transfer is made to the target person (called person) by mediation, a system for extracting the name of the target person (called person) by using a voice recognition device connected to the private branch exchange 300 and searching for the corresponding telephone number is constructed. Here, it is assumed that a voice recognition mechanism is incorporated in the exchange. As a result, the voice recognition trunk (V
-DET) 310 is provided in the private branch exchange 300.

As described above, the voice recognition function may be provided in the exchange, or may be implemented, for example, as an external device of the exchange as a voice recognition device 311 as illustrated by a broken line in the figure. When this is installed as an external device, the introduction of this system can be easily implemented by making minimum modifications and configuration changes to the exchanges to which the system is applied, and as will be described later. According to the present embodiment, at the time of telephone number search by voice recognition, only a phonological analysis can be performed, and a phrase analysis or a semantic analysis is not required. Since this can be simplified, it becomes more effective. However, even in the case where the voice recognition device is incorporated in the exchange as shown by the solid line in the figure, the advantage obtained by the simplification of the configuration is not lost.

Here, in the system of the present embodiment, when a telephone call (incoming call) is transferred to the target person (called person) through the mediation of the exchange, the exchange responds to the caller. Paying attention to the fact that if the subject's name is read back, the name will appear twice, and the caller's name reported by the caller and the callee's name when the operator repeats this "Name" that appears twice in the call content
Based on the idea that the part can be advantageously used for speech recognition, this is embodied, for example, when a call is received from the telephone 100, the exchange talks with the caller speaker A, When a call is forwarded after hearing the name of the called party (for example, speaker B on the local telephone 360), only the phoneme analysis is performed by the following method, and the called party who does not need phrase analysis or semantic analysis is required. In this embodiment, in order to enable voice recognition of the name, a means for phoneme-analyzing the call content (for example, the upper part of FIGS. 2 and 3) of the caller A and a call content of the exchanger (for example, FIGS. A means for phonologically analyzing the lower part of FIG. 3 is provided, and the operator repeats the name of the called party (for example, “B” in FIG. 2 and “Sumita” in FIG. 3) given by the speaker A of the caller. , The common phoneme sequence becomes In the system of the present embodiment, these means are used for speech recognition incorporated in the private branch exchange 300 because the name of the called party is identified and the telephone number is searched for. It includes a trunk (V-DET) 310. Regarding phonological analysis, as described in step S2 of the example of general speech recognition in FIG. It may be due to.

More specifically, referring to FIGS. 2 and 3, in this embodiment, when a call is received from the speaker A of the telephone set 100, the telephone transfer via the exchange is performed as described below. The processing is performed as shown in FIG.

[1] The caller A calls the exchange and asks "Please give me B", and requests a call transfer to Mr. B (). The voice that "Please ask Mr. B" pronounced by speaker A is exchanged with a voice recognition trunk (V-DET) 3
10 (“B” (first time)).

[2] The exchange confirms to the speaker A that it is "B" (). In this case, the operator is listening to "B, please wait a moment.", And the voice pronounced by the operator, "B is," is sent to the speaker A and the voice recognition trunk (V-DET) 310. Is transmitted ("B"
(2nd)).

[3] Voice recognition trunk (V-DET) 3
In "10", since "B" (first time) from speaker A and "B" (first time) from the operator overlap, it is easily recognized as "B", and "B" The telephone number "3771" is retrieved and displayed on the console 350 (). Here, the recognition rate is improved by detecting the duplication of “B”.

This can be explained from the comparison and comparison between FIG. 3 and FIG. 9 as follows. In the technique according to FIG. 9, the subject name must be identified only from the voice from speaker A. As a result, in order to recognize the subject name, phrase analysis and semantic analysis are necessary, and phonemes are simply recognized. However, according to the telephone number search subject determination of FIG. 3, even if the voice is simply recognized, as shown by the reference symbol a in FIG. Since the same character sequence "Sumita" appears twice, the voice recognition character string of Speaker A (Fig. 3, "... Sumita-san, please")
The subject's name can be extracted by comparing the voice recognition character string of the exchange with the voice recognition character string of the exchanger ("... Sumita ..." at the top of FIG. 3).

Therefore, when an incoming call (incoming call) from outside the company is transferred to the target person (called person) by the intermediary of the caller, the name of the called person indicated by the caller and the called person to whom the caller repeats By performing phonological analysis of the name twice, and determining a match (which does not have to be a perfect match) as the name of the called person, it is possible to omit semantic analysis or the like that requires high functionality. If the name of the called party is to be extracted by voice recognition only from the contents of the incoming call, it is necessary to perform not only a simple phonological analysis but also an accurate phrase analysis (FIG. 8S3), and in some cases, a semantic analysis (FIG. 8S4). Therefore, it is necessary to provide a high processing capability and a large amount of matching data (dictionary), an expensive system is required, and a speech recognition device to be used is large-scale and expensive. On the other hand, in the present embodiment, such a situation can be avoided and a telephone number search system based on voice recognition can be introduced, and the telephone number of the called party can be searched for when the telephone is transferred via an operator. It does not take much time, and based on the extraction of the name of the called party by the voice recognition trunk (V-DET) 310 linked with the private branch exchange 300 in accordance with the above method, the telephone number search or the If it is possible to improve the serviceability by displaying the target Hou complement's list by. In addition, the process of duplicate detection by voice recognition of the switching voice accompanying the repetition of the switching voice as shown in the lower part of FIG. It will also help improve the recognition rate.

[4] Referring back to FIG. 2, the operator looks at B's telephone number (3771) display "B: 3771" and transfers the incoming call from speaker A to speaker B (). Thus, when the transfer is completed, the telephones of the speaker A and the speaker B are connected.

As described above, the telephone transfer can be performed by the telephone number search based on the voice recognition, the configuration can be simplified, the recognition rate can be improved, and the efficiency of the telephone number search can be realized. In the case of the conventional technique, the transfer target person B must be identified only by the voice from the speaker A, and which part of the voice from the speaker A is the transfer target name “B”. In contrast to this, it is difficult to identify the transfer target person “B” from the common recognition word of the voice from the speaker A and the voice from the exchange in this embodiment.

Also, since the voice of speaker A passes through the central office 200, the volume generally fluctuates or the S / N ratio is poor, which causes a reduction in the recognition rate. In this case, since the voice pronounced by the operator is also targeted for recognition, since the voice has higher quality than the speaker A, the recognition rate can be improved.

As described above, an increase in the recognition rate of the transfer target person "B" can be expected to be effective in preventing a search error, and can eliminate a transfer number error. In addition, in the low recognition rate, when displaying a list of potential subjects, many subjects are displayed in a list. However, as described above, by improving the recognition rate, the subjects are narrowed down,
As a result, even when the target candidate list is displayed, the list display can be reduced.

The following is another embodiment of the present invention. In the above embodiment (first embodiment), an incoming call from an outside company (an incoming call) is mediated by an exchange and the target person ( Callee)
When a call is forwarded to a private branch exchange, a voice recognition device connected to the private branch exchange extracts the name of the target person (called person) and searches for the corresponding telephone number. In voice recognition, the operator repeats the name of the target person (called person) and avoids a decrease in the recognition rate, and also recognizes the voice of the switching person by voice. In addition to improving the recognition rate, the configuration of the speech recognition apparatus can be simplified.
Basically, it is intended to make further improvements while utilizing such features and advantages.

That is, in this embodiment, in particular, when the operator repeats the name of the called party, a mechanism and means for notifying the voice recognition device of a signal identifying the name part are provided, so that the name of the called party is provided. To improve the accuracy of the phoneme analysis. This is particularly effective when it is difficult to determine a match because the quality of incoming telephone calls from outside the company is far inferior to the voice quality of the internal exchange. FIGS. 4 and 5 are diagrams respectively showing an example of a system configuration according to the present embodiment and a specific example of determination of a person to be searched for a telephone number. The same parts as those in the first embodiment are denoted by the same reference numerals. Since the basic configuration and the like are the same, the description thereof is omitted, and the main part will be described below.

As shown in FIG. 4, in the present embodiment, in addition to the configuration of the first embodiment, only while the operator pronounces the called party name repeated for the purpose of identifying the called party name area. A specific switch (SW) 351 provided on the exchange table 350 can be operated, for example, pressed. Here, the switch (SW) 351 is used for notifying a signal for identifying a name part, and as shown by a reference numeral c in FIG. , SW for specifying the recognition location (timing)
Can be set as an additional condition.

According to the present embodiment, the first
In addition to the same effects as described in the embodiment of FIG. 5, while the operator pronounces the recalled callee name, for example, in FIG. 5, as in FIG. Is pronounced, but the switch (S
W) Pressing 351 makes it easy to extract only the name part from the voice. Therefore,
When such a function is provided, even if the voice quality of the speaker A who has passed through the network 200 is much lower than the voice quality of the operator due to variations in volume, deterioration of the S / N ratio, etc. Even when it is difficult to determine the match by the method described in 3 above, the intended extraction can be reliably performed,
Such a case can be dealt with and is effective.

It should be noted that the present invention is not limited to the above, and the following changes,
Deformation is also possible. The above specific switch (SW) 35
1 may be provided in an exchange (relaying station) 350, or a button of a dedicated telephone equipped with a means for notifying button information to the exchange may be used.

As already mentioned in the first embodiment, in the case where a computer or a communication terminal for a computer is used as a switching board, in this case, the computer or the computer is used as the "specific switch". The terminal keyboard may be used. In this case, it is assumed that a hand-throw is provided for notifying the voice recognition device of the keyboard press information of the computer or the computer terminal.

Also, as already mentioned in the first embodiment, the voice recognition function may be provided in the exchange as shown in FIG. 4 or in the exchange as shown in FIG. When the aspect of the external device is adopted, the external device (311) may be provided with a means for notifying the external device of the "specific switch" pressing information.

The same effects as described above can be obtained with these modifications and variations, and the present invention may be implemented in this manner.

[Brief description of the drawings]

FIG. 1 is a diagram showing an example of a telephone number search method by voice recognition and a system configuration thereof according to an embodiment of the present invention.

FIG. 2 is a diagram showing an example of a sequence at the time of transfer for explanation of the embodiment.

FIG. 3 is a diagram showing a specific example of telephone number search target person determination.

FIG. 4 is a diagram showing another embodiment of the present invention.

FIG. 5 is a diagram showing a specific example of a telephone number search target person determination for explaining the embodiment.

FIG. 6 is a diagram showing a basic configuration example of a system for extracting a called party name from the contents of an incoming call by a voice recognition device linked to a private branch exchange and searching for a telephone number.

FIG. 7 is a diagram showing a basic mode of a transfer sequence shown in comparison with FIG. 2;

FIG. 8 is a diagram showing a general speech recognition procedure.

FIG. 9 is a diagram showing a telephone number search method by voice recognition in a case where the present invention is not adopted, in comparison with FIGS.

[Explanation of symbols]

Reference Signs List 100 telephone (TEL) 200 central office exchange network (CO) 300 private branch exchange (PBX) 301 attendant console interface (ATT-IF) 302 single line telephone interface (SL)
C) 303 Central office trunk (COT) 310 Voice recognition trunk (V-DET) 311 Voice recognition device (external device) 350 Attendant console (switchboard) (ATT) 351 Switch (SW) 360 Telephone (TEL)

──────────────────────────────────────────────────の Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) H04M 3/42 H04M 3/42 P 3/58 3/58 Z H04Q 3/58 101 H04Q 3/58 101F Terms (reference) 5D015 KK01 5K015 AA07 AA10 AF03 AF11 KA01 5K024 AA15 AA23 BB01 BB06 BB07 CC01 DD01 DD03 EE04 EE09 GG03 GG07 5K049 AA01 AA07 AA17 BB04 CC10 EE02 EE13 FF01 FF32 GG07

Claims (9)

[Claims] When an incoming call is transferred to a transfer target through an operator, phoneme analysis is performed twice for the called party name announced by the caller and the called party name read back by the operator, and the matching part is determined. A telephone number search method based on voice recognition characterized in that voice recognition of the name of the called person is performed and the telephone number is searched by determining the name of the called person. 2. The telephone number search by voice recognition according to claim 1, further comprising means for notifying the voice recognition device of a signal for identifying the name part when the operator repeats the name of the called party. method. 3. A system for performing a telephone number search by voice recognition when an operator mediates and transfers to a called party when a telephone call arrives, comprising: means for phoneme-analyzing call contents of a caller; Means for phonologically analyzing the content, based on the fact that the operator repeats the called party name reported by the caller, so that the common phonological sequence appears in the caller-side phonological analysis result and the operator-side phonological analysis result, A telephone number search system using voice recognition, comprising means for identifying a called party's name area and performing a telephone number search. 4. The method according to claim 3, wherein a portion corresponding to the name is easily extracted from the voice by operating a switch while the operator pronounces the recalled callee name. Phone number search system by voice recognition. 5. The switch according to claim 4, wherein the switch is a predetermined switch provided on a switchboard, and is configured to be provided on the switchboard. 5. The telephone number search system by voice recognition according to claim 4, wherein a button of a dedicated telephone equipped with a means for notifying is provided. 6. The computer according to claim 4, wherein a computer or a communication terminal for the computer is used as the switchboard, and a keyboard of the computer or the communication terminal for the computer is used as the switch of claim 4. Item 4. A telephone number search system using voice recognition according to item 4. 7. The telephone number search system by voice recognition according to claim 6, further comprising a hand-throwing operation for notifying a voice recognition device of keyboard press information of the computer or the computer terminal. 8. The telephone number search by voice recognition according to claim 3, wherein the voice recognition function is provided in the exchange or provided as an external device of the exchange. system. 9. The voice recognition function is provided as an external device of the exchange, and is provided with a hand throw for notifying the external device of the operation information of the switch according to claim 4 or 5. A telephone number search system using voice recognition according to claim 8.