JP4686160B2 - Conversation recording apparatus and conversation recording method - Google Patents

Description

  The present invention relates to a conversation recording apparatus and a conversation recording method for recording conversations exchanged in scenes such as merchandise transactions.

  In conversations at storefronts at financial institutions and in call centers that accept phone calls from customers, it is necessary to thoroughly prevent accountability by thoroughly explaining accountability when explaining financial products with investment risks. is there. As a trouble, there is a case where the customer loses after selling financial products, and there is a watering theory that says that there is such a risk. Some people complain that they didn't think so.

In order to prevent the trouble generated from the conversation content as described above, it can be dealt to some extent by recording the conversation content. However, at present, measures such as taking small notes are taken, and it is troublesome to record a long conversation. Even when taking notes, nuances can be a problem. Therefore, it is effective to record the conversation content. In the present invention, the content of the conversation is recorded using the voice recognition technology. As a conventional voice recognition technology, for example, Patent Document 1 that searches for a word registered in a dictionary and detects the word from the voice is disclosed. There are two. Further, Japanese Patent Application Laid-Open No. H10-228707 discloses a technique related to a time stamp pushed on an audio file used in the present invention.
Japanese Patent Application Laid-Open No. 2002-229584 JP 2004-012883 A JP 2004-194306 A

  However, recording all conversations between customers and salespeople is expensive and wasteful. Also, if you save the recorded content of all conversations, it takes time to search for important parts. Furthermore, it is not convenient to use the recorded contents for salesperson education and guidance after recording.

  In addition, simply recording a conversation cannot solve the problem of omission of explanation, in which all matters to be explained are not explained due to a salesperson's mistake. For example, although it was necessary to explain four items to the customer, it turned out that the sales staff inadvertently explained only up to three items after confirming the recorded content of the conversation. It can also happen.

  In addition, when customers ask critical questions about the risks and weaknesses of financial products, they sometimes complain. For example, a customer may say, “I don't have to worry about breaking the principal of this product,” and may be reassured, “I ’m fine, so I ’m not worried.” These prohibitions are likely to cause trouble later on.

  The object of the present invention is to eliminate such drawbacks of the prior art, to reduce the risk of dealing with customers, and to provide appropriate navigation for sales staff.

  In order to solve the above-described problems in the present invention, a conversation recording apparatus according to the present invention includes a keyword storage means for recording one or more keywords for instructing the start of recording and a keyword for instructing the end of recording, and a conversation, Voice recognition means for detecting a keyword for instructing start and end of recording, first recording means for recording, and conversation to the first recording means when a keyword for instructing start of recording is detected by the voice recognition means And recording control means for causing the first recording means to stop recording the conversation when the voice recognition means detects a keyword for instructing the end of the recording, so that only the necessary part of the conversation is recorded. Record.

  According to the present invention, keywords are skillfully interwoven in a natural conversation that is not understood by the customer, and this is voice-recognized, so that only the important part in the conversation is recorded. As a result, when the conversation contents become a problem in the future, it becomes clear whether there is a non-consideration between the salesperson and the customer. Further, since only the important part in the conversation is recorded, a large amount of recording memory is not required, and the cost can be reduced.

  According to the present invention, by navigating the salesperson during the conversation, the salesperson can explain the matters to be explained without omission, and accountability is surely fulfilled. In addition, even if a customer asks a critical question, the above-mentioned navigation alerts the sales staff not to say a forbidden phrase, and troubles caused by conversations are less likely to occur in the future.

  According to the present invention, the importance level of the recorded conversation can be determined, and the storage period of the recorded content according to the importance level can be determined.

  According to the present invention, by confirming the content of conversation recording, even if it does not develop into trouble, it can be used as a material for determining whether or not there is a problem. In other words, it can be used for salespersons' education and guidance, and future business improvements can be expected.

  Next, an embodiment of a conversation recording apparatus according to the present invention will be described in detail with reference to the accompanying drawings. In the text, signals are indicated by reference numerals of signal lines through which they pass, and similar elements are indicated by the same reference numerals. The description of elements not directly related to the present invention is omitted.

  FIG. 1 is a block diagram showing an embodiment of a conversation recording apparatus according to the present invention. The conversation recording device 10 may be a personal computer. The device 10 has a drive 12. The drive 12 is a device that detachably accommodates the storage medium 14 as indicated by an arrow 13, and reads the recorded content of the storage medium 14 and transmits it to the control unit 100 described later through the signal line 15, or the control unit 100 Information can be written to the storage medium 14 in accordance with the control signal 17 from. The storage medium 14 may be a readable / writable recording medium such as CD-RW (Compact Disk ReWritable).

  Customer information can be written to the storage medium 14. The customer information may be stored in the storage medium 14 as described above, or may be stored in the customer information database 16. Customer information may be given from the storage medium 14 to the conversation recording device 10 via the drive 12 as indicated by an arrow 13, or from the customer information database 16 shown in FIG. 1 to a signal line 18 such as a LAN (Local Area Network). It may be given to the conversation recording device 10 directly through.

  In the customer information, information for marketing, such as annual income and age, is recorded for each customer. Moreover, the history of the transaction so far is recorded. For example, a product sold in the past, a product that has not been sold but explained, and the date are recorded. It is also good to record that the sales clerk uttered words that are forbidden for financial products. Furthermore, the customer's reaction at the time of explanation and the impression held by the salesperson may be recorded.

  The dictionary 20 is a non-volatile storage device. (A) For each product, not only the contents of each product, but also a salesperson to issue when starting the description of the product (b) the start keyword and the description of the product (C) An end keyword is recorded which should be issued by the salesperson when ending. For example, the start keyword may be arbitrarily determined for a product to be explained, such as “breakdown of principal of financial product”. The end keyword may be arbitrarily determined, but it may be a word indicating that the explanation has been completed successfully, such as “I understand your understanding about the loss of the principal of financial products”. Note that the start keyword and the end keyword preferably correspond one-to-one with respect to a certain product as in the above example. This is because keywords can be incorporated into a natural conversation. However, these keywords only need to trigger the start and end of recording, which will be described later. Therefore, regardless of individual products, there are only two types: a keyword that instructs the start of recording and a keyword that instructs the stop of recording. May be prepared.

  The dictionary 20 also records (d) average time predicted for each product to elapse from the start to the end of the description of the product. Furthermore, words that are forbidden in the product description, such as “There will be no danger of breaking the principal”, are also recorded as (e) forbidden keywords. In addition, words such as “I apologize” used when a salesperson apologizes for misrepresentation or omissions are also registered as (f) apology keywords.

  The voice recognition unit 22 is a device that receives and recognizes a conversation from the microphone 24 via the signal line 26. As the speech recognition unit 22, for example, a speech recognition device described in Patent Document 1 or 2 may be used. Note that since the microphone 24 is exclusively for the salesperson, the conversation received by the voice recognition unit 22 is only the salesperson's conversation. When the speech recognition unit 22 receives the conversation, the speech recognition unit 22 refers to the dictionary 20 via the signal line 28 and detects various keywords recorded in the dictionary 20 from the conversation. When the keyword is detected, the voice recognition unit 22 sends a detection signal 30 that can be identified as follows to the control unit 100 described later. That is, (b) a start keyword detection signal, (c) an end keyword detection signal, (e) a prohibited phrase keyword detection signal, and (f) an apology keyword detection signal are sent.

  The recording control unit 32 is a device for operating the two recorders 34 and 36 with the control lines 39 and 40 and recording the sound 44 sent from the microphone 38 to these recorders 34 and 36. Note that the microphone 38 is a microphone that receives conversations between the salesperson and the customer. The recording control unit 32 can cause the recorders 34 and 36 to start recording the audio 44 from a desired point in time, and can also pause recording. The recording control unit 32 can also reproduce the content recorded by the two recorders 34 and 36 from an arbitrary point of time and output the content from the speaker 50 through the signal line 52.

  In this embodiment, the recording control unit 32 controls the recording of the two recorders 34 and 36 as follows. That is, when (b) the start keyword detection signal is received from the control unit 100 through the signal line 46, the recorder 34 starts recording. When (c) the end keyword detection signal is received through the signal line 46, the recording of the recorder 34 is temporarily stopped. On the other hand, the recorder 36 does not receive an instruction from the recording control unit 32 for the start of recording, and automatically starts recording from the time when a salesperson or customer conversation is detected from the microphone 38 for the first time. However, the end of recording is performed according to an instruction 40 of the recording control unit 32.

  FIG. 2 is a diagram conceptually showing the recording contents of the recorders 34 and 36. As shown in FIG. 1, the recorders 34 and 36 receive the same sound 44 from the microphone 38, but the recording methods are different. That is, the recorder 34 plays the role of recording the sound 44 intermittently in accordance with the instruction of the recording control unit 32. This is because only a necessary part of the audio 44 is recorded. In other words, as shown in FIG. 2, the recorder 34 records the conversation bracket 45 from the time when the start keyword registered in the dictionary 20 is issued to the time when the end keyword is issued in the voice 44. . Thus, hereinafter, each conversation portion that the recorder 34 intermittently records is referred to as a bracket. On the other hand, the recorder 36 plays a role of recording all the audio 44 as shown in FIG. 2 until a recording stop command 40 is issued from the recording control unit 32. This is to record a backup of the entire audio 44. The automatic correction S47 by the recording control unit and the correction S49 by manual means will be described later.

  The recorders 34 and 36 shown in FIG. 1 may be recording devices having a recording device such as a flash memory. Audio is recorded as a digital signal, and is recorded as an audio file for each conversation recorded over a certain period of time. This audio file may be, for example, an MP3 format file. In this audio file, the recording start time and recording end time of the recorded conversation are recorded as time stamps.

  The memory 48 shown in FIG. 1 may be a recording device such as a flash memory. In the memory 48, information corresponding to a time stamp indicating which time zone conversation should be recorded finally is written by the recording control unit 32 through the signal line 54. Information stored in the memory 48 corresponding to the time stamp is read out to the recording control unit 32 through the signal line 56.

  The navigation support unit 60 is a device that determines display contents of the salesperson screen 62 and the customer screen 64 through the control lines 61 and 67. Hereinafter, the navigation support unit 60 is abbreviated as “navigation support unit 60”. The salesperson screen 62 and the customer screen 64 are display devices to which the salesperson and the customer refer, respectively, at the time of negotiation. The contents displayed on the screens 62 and 64 by the navigation support unit 60 are listed below. First, customer information input via the signal line 65 from the customer information database 16 or the storage medium 14 via the control unit 100 can be displayed on the salesperson screen 62. If such customer information is displayed on the salesperson screen 62, the salesperson refers to this prior to the consultation with the customer, and from the past transaction history, what kind of product will be given to the customer in the next consultation. Can explain or make a plan. In other words, the salesperson can select one or more products as a topic constituting a conversation exchanged with the customer.

  The navigation support unit 60 can display all the product names acquired from the dictionary by the control unit 100 through the signal line 72 along with the customer information described above on the salesperson screen 62. Thereby, the salesperson can select some products to be explained to the customer at the next round of conversation from all displayed products. The product name is selected by the salesperson using the input unit 66 such as a keyboard, is input to the control unit 100 via the signal line 68, and is further sent to the navigation support unit 60.

  The navigation support unit 60 includes a navigation information unit 70 that is a volatile memory. The navigation information unit 70 is also abbreviated as “navigation information unit 70” below. The navigation information unit 70 stores the contents of the product selected by the salesperson as described above. That is, each time the salesperson selects a product name using the input unit 66 from among all the product names displayed on the salesperson screen 62, the navigation support unit 60 displays that the product has been selected. The control unit 100 that is displayed on 62 and receives the product name accesses the dictionary 20 through the signal line 72 and provides the navigation information unit 70 with information on the selected product. That is, (a) the contents of the product, (b) the start keyword, (c) the end keyword, and (e) the forbidden keyword registered in advance in the dictionary 20 are recorded in the navigation information unit 70 in association with the product name. .

  When information about the selected product is recorded in the navigation information unit 70, the navigation support unit 60 further registers, in the navigation information unit 70, information indicating that the description of the product has not been completed for each selected product. Keep it. The information indicating that the explanation has not been completed is of course “uncompleted” at the beginning, and “completed” when the discussion is completed after the discussion.

  As described above, when the information about the selected product and the unfinished information are registered in the navigation information unit 70, the navigation support unit 60 displays the information on the salesperson screen 62. That is, for each selected product, (a) the content of the product, (b) the start keyword, (c) the end keyword, and (e) the forbidden keyword are displayed on the screen 62 in correspondence with the product name. (X) Information that has not been explained is also displayed. It is not necessary to display these pieces of information on the screen 62 at the same time. For example, when a product name is clicked, information related to the product name may be displayed in detail. Information indicating any incomplete may be displayed, for example, by adding a check mark to an unexplained product name or changing the display color between the product name for which the explanation has been completed and the unexplained product name. At first, for all products, checkmarks may be added, or product names may be displayed in a color corresponding to “not yet”.

  When the navigation support unit 60 receives a start keyword of a product from the control unit 100, the navigation support unit 60 records in the navigation information unit 70 that the description of the product has been started, and the product is being described on the salesperson screen 62. Information indicating the presence may also be displayed. For example, the product name may be blinked. This flashing also means that the conversation is being recorded by the recorder 34. When the navigation support unit 60 receives the end keyword of the product, the navigation support unit 60 records information indicating that the explanation has not been completed in the navigation information unit 70 as “end”, and displays the product name on the salesperson screen 62. Then, the check mark indicating unexplained may be deleted.

  Although not shown in FIG. 1, the navigation support unit 60 may be provided with a recording support function on the salesperson screen 62. Specifically, a button for forcibly starting and stopping recording is displayed on the screen 62, and when a salesperson presses the button on the screen 62 using the input unit 66, the recording starts and stops. It is sufficient that the configuration can be freely set. A command using the recording support function is sent from the control unit 100 to the recording control unit 32, and the recorder 34 is operated. As described above, recording by the recorder 34 for partial recording is started and stopped in principle by the start and end keywords issued by the salesperson, but the salesperson wants to record and stop by the recorder 34 independently. Because there is also.

  When the navigation support unit 60 records and displays information as described above, the salesperson can check the product for which the explanation has not ended by looking at the salesperson screen 62. You can see if you should use it, and you can see forbidden keywords that should not be used in the description. When explaining the product, the salesperson explains the product by skillfully interweaving the start / end keywords in the conversation while referring to the screen 62 for the salesperson, and paying attention not to give a forbidden phrase. Can explain.

  On the other hand, the navigation support unit 60 may display the product name, information indicating that the product is being explained, and information that has not been explained on the customer screen 64. Thereby, the customer can list the products for which explanation is given, and can also confirm that the explanation of those products has not been completed.

  When the salesperson confirms that the explanation of all the products is completed on the salesperson screen 62, the salesperson can instruct the end of the talk from the input unit 66. This instruction is transmitted to the recording control unit 32 via the control unit 100, and the recording of the backup recording recorder 36 is stopped.

  Note that the recording stop of all recording recorders 36 may be automatically performed when silence continues for a certain period of time. For example, if silence continues for 10 minutes, the control unit 100 accesses the navigation support unit 60 to check whether all the explanations have been completed. Good. If not finished, it may be displayed on the screen 62 as a warning that there is an unexplained product.

  As already described, since the recorder 34 records the sound 44 intermittently, one or more sound files are stored in a built-in recording device. On the other hand, the recorder 36 stores a backup audio file of the entire conversation. One of the objects of the present invention is to save a storage area by recording only a necessary part of a conversation. However, the audio file recorded in the recorder 34 does not always record only the really necessary part of the conversation, and may be excessive or insufficient. Therefore, in the embodiment of the present invention, the backup audio file saved in the recorder 36 can be used to correct the audio file to be finally recorded.

  In order to correct the audio file, the navigation support unit 60 and the recording control unit 32 may be used in combination. The recording control unit 32 not only controls the recording performed by the recorders 34 and 36, but also can correct the audio file recorded in the recording device built in these recorders as described below.

  FIG. 3 is a time chart conceptually showing the automatic correction S47 by the recording control section and the correction S49 by manual means shown in FIG. If the start keyword 51 appears at time t0 during the conversation, the speech recognition unit 22 recognizes the start keyword 51 between times t1 and t2, and the recording start command 53 is recorded via the control unit 100 and the recording control unit 32. Given to 34. Accordingly, since recording starts at time t2, the start keyword 51 and the beginning of the product description 55 are not recorded in the audio file 53 created by the recorder 34. However, it is ideal to record including the start keyword 51. On the other hand, when the end keyword 57 appears during the conversation, a recording stop command 59 is given to the recorder 34 between the times t3 and t4 through the same route as when recording is started. Accordingly, the voice file 53 created by the recorder 34 also records conversations at times t3 to t4 after the end keyword 57. However, the recorded conversations at times t3 to t4 are extra.

  Therefore, as shown in FIG. 3, the purpose of automatic correction S47 by the recording control unit and correction S49 by the manual means shown in FIG. 2 is to obtain a corrected audio file from time t0 to t3. In the automatic correction S47 by the recording control unit 32, an audio file having a time stamp obtained by shifting the time stamp of the audio file 53 may be acquired from all the recording recorders 36. Further, the salesperson corrects the time stamp of the audio file by the correction S49 by the manual means. As a result, the correction that has not been properly performed in the automatic correction S47 can be further corrected.

  Hereinafter, the corrections S47 and S49 will be described in detail. When it is confirmed that the explanation of all the products is completed on the salesperson screen 62, the salesperson instructs the input unit 66 to end the talk. This instruction is transmitted to the recording control unit 32 via the control unit 100, and the recording of the backup recording recorder 36 is stopped. Then, the recording control unit 32 refers to the time stamp of the audio file of the partial recording recorder 34. Then, the time zone shifted from the time zones t2 to t4 recorded in the time stamp is stored in the memory 48. That is, the start time is shifted to a time preceding the time t2 shown in FIG. 3, and the end time is shifted to a time preceding the time t4. The shifting length may be arbitrarily determined. On the other hand, the control unit 100 gives instructions to the navigation support unit 60 to display on the salesperson screen 62 that the product description has already been completed. Display.

  The salesperson uses the input unit 66 to select any product from the salesperson screen 62. Then, the recording control unit 32 accesses the memory 48, acquires the time zone recorded by shifting the description of the product, and talks to all the recording recorders 36 for the conversation corresponding to the shifted time zone. The data is reproduced and output from the speaker 50. That is, the recorded content in the time zone in which the start keyword 51 to the end keyword 57 are considered to be recorded without excess or deficiency is reproduced. If there is any excess or deficiency, the sales clerk can listen to the recorded contents of the explanation of each product and further change the time zone. In order to change the time zone, the time zone may be changed by directly inputting the time zone from the input unit 66, or the time stamp may be visually changed by a GUI (Graphical User Interface).

  When the most appropriate time zone, that is, the times t0 to t3 shown in FIG. 3 is finally determined, the recording control unit 32 cuts out the recording content corresponding to the time zone from all the recording recorders 36, and displays it in FIG. A corrected audio file 63 is created as shown. Then, the audio file 63 is passed to the control unit 100. The control unit 100 adds the audio file 63 to the storage medium 14 or directly to the customer information database 16, thereby adding it to the customer information as a record of the product description.

  The bias calculation unit 80 is a device that calculates a value indicating the importance of the conversation between the salesperson and the customer and records it in the internal memory 82. This value is called a bias in this embodiment. The greater the bias, the more important the conversation, and the longer it is necessary to preserve the recorded conversation. In calculating the conversation bias, the present embodiment uses the following information. That is, the customer information, the magnitude of the risk of the product explained to the customer during the conversation, the length of the conversation actually required to explain the product, and the (f) apology keyword that appeared during the conversation. In this embodiment, the rules for determining and changing the bias are only described qualitatively, but a quantitative value for increasing or decreasing the bias may be freely determined.

  When the bias calculation unit 80 obtains customer information via the control unit 100 through the signal line 79, the bias calculation unit 80 determines a pre-bias that is an initial value of the bias according to the length of association with the customer. In other words, the longer the relationship with the customer, the more enthusiastic the customer is and the lower the possibility that the conversation will be dangerous, so the bias may be set small. On the other hand, if the relationship with the customer is short or if the customer is a first-time customer, the content of the conversation cannot be predicted and is likely to be dangerous, so the bias needs to be set high.

  In addition, the bias calculation unit 80 can obtain information about a product for which explanation is planned via the control unit 100. Here, the information about the product is the magnitude of the risk of the product and the average time expected to be necessary for the description of the product. Of these pieces of information, the bias calculator 80 adds a pre-bias according to the risk level of the product. The magnitude of the risk of any product may be registered in advance along with information about the product in the dictionary 20, and the control unit may obtain it and send it to the bias calculation unit 80, or the memory in the bias calculation unit 80 It may be registered in 82.

  FIG. 4 is a diagram conceptually illustrating a bias calculation process performed by the bias calculation unit 80. In the figure, as shown in step S81, first, the above-described pre-bias is determined. After the start of recording shown in step S82, the control unit 100 sends an apology keyword detection signal to the bias calculation unit 80 when the voice recognition unit 22 detects (f) an apology keyword for bias calculation. These apology keywords are, for example, words of a salesperson's apology such as “I'm sorry” and “Not so”, and are registered in the dictionary 20.

  Each time the bias calculation unit 80 receives an apology keyword detection signal, the bias calculation unit 80 adds a certain amount of bias as shown in steps S83, S87, and S89. That is, the bias is incremented. As the number of times the salesperson apologized to the customer due to incomplete explanations, etc., it can be inferred that the conversation was not smooth, and accordingly, it can be inferred that the conversation easily developed into a trouble later. is there.

  Thus, in this embodiment, there is one bias common to all conversation brackets recorded in the recorder 34, and a certain amount of bias is added each time an apology keyword is detected. However, the bias calculation method is not limited to this method and may be freely determined. For example, a difference may be provided in the bias addition value depending on whether the place where the apology keyword appears is inside or outside the bracket. Further, the bias value to be raised may be made different depending on which bracket the keyword has appeared. As explained for the pre-bias setting, the magnitude of risk is different for each product, that is, for each bracket, and the apology keyword that appeared in the description of the high-risk product is more than the apology keyword that appeared in the description of the low-risk product. Because there is a weight.

  When the conversation bracket 85 or 87 for explaining a certain product is completed, the control unit 100 receives the end keyword detection signal 30 from the voice recognition unit 22 and transfers it to the bias calculation unit 80. In addition, the control unit 100 can acquire, from the recorder 34, the time stamp pressed by the audio file that has recorded the conversation related to the description of the product via the recording control unit 32. Then, the time stamp information is sent to the bias calculator 80. Then, the bias calculation unit 80 compares the average time estimated in advance for the description of the product previously obtained from the dictionary 20 with the time actually required for the description of the product obtained from the time stamp information. The bias calculator 80 increases or decreases the bias according to the time difference as shown in steps S85 and S91. That is, when the time required for the description of the product is actually longer, it can be estimated that the conversation has not progressed smoothly, and the possibility of developing into a trouble later is high, so the bias is increased. On the other hand, if the actual explanation time is shorter, the bias is decreased.

  As described above, the bias calculation unit 80 sets the pre-bias based on the customer information and the magnitude of the risk of the product to be explained, and every time an apology keyword such as an apology is detected by the speech recognition unit 22. Each time an explanation about a product is completed with a bias increase, the actual time required to describe the product is compared with the average time expected to explain the product, and the bias is determined according to the difference. Increase or decrease.

  After the recording is completed as shown in step S92, the bias adjustment S93 shown in FIG. 4 can be performed by using the bias calculation unit 80 and the navigation support unit 60 in combination. The bias adjustment S93 may be executed by sending the bias calculated by the bias calculation unit 80 to the navigation support unit 60 via the control unit 100 and displaying it on the salesperson screen 62. The salesperson can freely adjust the automatically calculated bias by manual operation by using the input unit 66. The adjusted bias input from the input unit 66 is displayed on the screen 62 and recorded in the memory 82 in the bias calculation unit 80 via the control unit 100.

  Next, as shown in step S95, the bias calculation unit 80 determines whether to save the conversation brackets 85 and 87 of the recorded product description in the conversation 44 from the bias recorded in the memory 82. . For example, if the bias value is lower than a certain value, the brackets 85 and 87 may be immediately deleted as shown in step S99. In this case, the bias calculation unit 80 may set the storage period to zero and record it in the memory 82.

  As will be described later, the audio file in which the brackets 85 and 87 are recorded is finally added to the customer information. At this time, the storage period is also added to the customer information. This storage period is a kind of timed device, and for example, a server having the customer database 16 may sequentially delete the audio files from the customer database 16 according to the storage period. As a result, the audio file whose storage period has passed is automatically deleted, saving the storage area.

  On the other hand, when it is determined in step S95 that the recorded brackets 85 and 87 are stored, the bias calculation unit 80 determines a storage period corresponding to the bias and stores it in the memory 82, as shown in step S97. As described above, this storage period is stored in the customer database 16, and the storage period of the audio file recording the brackets 85 and 87 is determined.

  The monitoring unit 90 is a device that records (e) prohibited phrases keywords that appear in the product description. For example, when the speech recognition unit 22 detects a forbidden keyword such as “I'll make money first” and sends a forbidden keyword detection signal 30 to the control unit 100, the control unit 100 sends the detected forbidden keyword through the signal line 104 to the monitoring unit 90. Send to. Although not shown in FIG. 1, when a forbidden keyword arrives at the monitoring unit 90, the content may be automatically printed out and used later for the education and guidance of the salesperson. In addition, the control unit 100 sends the detected banned keyword to the navigation support unit 60 and displays it as a warning on the salesperson screen 62 so that the salesperson can immediately withdraw and apologize at the meeting with the customer. It is good to make it.

  The control unit 100 is connected to the various elements already described above, and is a control unit for the entire conversation recording apparatus that exchanges information with the various elements and operates the various elements. As described above, the control unit 100 mainly acquires customer information obtained from the storage medium 14 and the like and information related to products obtained from the dictionary 20, and supplies these information to necessary elements. The control unit 100 also receives various keyword detection signals 30 from the voice recognition unit 22 and sends appropriate commands to the recording control unit 32, the navigation support unit 60, and the bias calculation unit 80.

  Further, the control unit 100 adds the audio file finally determined to be saved to the customer information. Specifically, the control unit 100 refers to the time stamp recorded by the recording control unit 32 in the memory 48, and the audio file finally cut out from the recording recorder 36 is stored in the storage medium 14 via the drive 12. It is added to the customer information or added to the customer information in the customer information database 16 through the LAN 18. In addition, the control unit 100 receives the period corresponding to the bias calculated by the bias calculation unit 80 through the signal line 102 and adds it to the customer information as the storage period of the audio file. Further, the forbidden phrase recorded in the monitoring unit 90 is also added to the customer information.

  The operation of the embodiment of the present invention configured as described above will be described below.

  When a customer comes to the store and talks with a salesperson, the salesperson first checks the customer's information. 1 is inserted into the drive 12 or the customer information is directly acquired from the customer information database 16. The customer information is sent to the navigation support unit 60 via the control unit 100 and displayed on the salesperson screen 62. At this time, all product names are displayed on the salesperson screen 62 through the control unit 100 and the navigation support unit 60 together with the customer information. The salesperson selects several products to be explained to the customer at the next round of conversation using the input unit 66 from all product names while referring to the customer information. The selection of the product name is input to the control unit 100 and further sent to the navigation support unit 60.

  The navigation support unit 60 displays the product name selected by the salesperson on the screen 62. The control unit 100 that has received the product name stores the (a) product content, (b) start keyword, (c) end keyword, and (e) prohibited phrase keyword from the dictionary 20 in correspondence with the product name in the navigation information unit 70. Record. In addition to these pieces of information, the navigation support unit 60 registers (x) information indicating that the description of the product is incomplete in the navigation information unit for each selected product. Naturally, the initial state of the information indicating incomplete is “uncompleted”.

  On the other hand, the customer screen 64 may be displayed with the product name to be explained and any information that has not been explained. Thereby, when the customer comes to the store, the customer can confirm the product scheduled to be explained and which explanation has not been completed.

  The customer information is also sent from the control unit 100 to the bias calculation unit 80. The bias calculation unit 80 sets a pre-bias based on the length and frequency of the transaction history with the customer. In addition, the control unit 100 sends information on the product selected by the salesperson, that is, the magnitude of the risk of the product, and the average time required for explaining the product to the bias calculation unit 80. The bias calculation unit 80 records it in the memory 82. Of these pieces of information, the bias calculator 80 adds a pre-bias according to the risk level of the product.

  Now that the customer is ready to greet, the salesperson and the customer can actually enter into talks.

  When the customer visits the store, the salesperson / customer microphone 38 picks up the conversation of either the salesperson or the customer, and the recording recorder 36 automatically starts recording the conversation. As shown in FIG. 2, the all recording recorder 36 records all conversations (voices 44) between the customer and the salesperson.

  The salesperson refers to the salesperson screen 62, confirms the product to be explained and the start / end keywords of the product, and sets the product description start keyword in a natural conversation to explain the product to the customer. To start. When the voice recognition unit 22 detects the start keyword of any product via the salesperson's microphone 24, the voice recognition unit 22 notifies the control unit 100 of the product name. The control unit 100 instructs the recording control unit 32 to cause the partial recording recorder 34 to start recording.

  Further, the control unit 100 informs the navigation support unit 60 that the explanation of the product has been started. The navigation support unit 60 accesses the navigation information unit 70 and writes information that the product is being explained. As a result, the product names displayed on the salesperson screen 62 and the customer screen 64 blink. That is, it is displayed that the product is being explained.

  During the conversation, the control unit 100 sends an apology keyword detection signal to the bias calculation unit 80 when the voice recognition unit 22 detects (f) an apology keyword such as “I'm sorry”. The bias calculation unit adds the bias by a predetermined value. This is because it is thought that the risk of this conversation developing into trouble later increased.

  In addition, during a conversation, when a phrase-forbidden keyword such as “I'll make money first” is detected by the voice recognition unit 22, the control unit 100 records the contents of the phrase-forbidden keyword in the monitoring unit 90.

  The sales clerk explains a certain product, and when the customer says that he / she understands the content, issues a closing keyword for the product. This end keyword is detected by the voice recognition unit 22 and sent to the control unit 100. The control unit 100 notifies the recording control unit 32 and the navigation support unit 60 that the description of the product has been completed. Then, the recording control unit 32 stops recording by the recorder 34. This prevents unnecessary conversations from being recorded. On the other hand, the navigation support unit 60 sets the information indicating that the description of the product has not been completed to “END”. At the same time, the product names on the screens 62 and 64 stop blinking, the check mark disappears, and the explanation is finished. As a result, the salesperson can check the salesperson screen 62 for products that have not been explained.

  In the present embodiment, as described above, the recorder 34 records from when the start keyword of a certain product is issued until the end keyword is issued, but there may be cases where the start keyword continues due to a salesperson's mistake. In that case, for example, recording may be continued until end keywords corresponding to all of the detected start keywords are detected.

  Next, the control unit 100 instructs the recording control unit 32 to acquire the time taken for recording the product description from the time stamp of the file recorded in the partial recording recorder 34 and informs the bias calculation unit 80 of the time. . The bias calculation unit 80 increases or decreases the bias by comparing the average time estimated to be required for the description of the product with the actual recording time. In fact, the longer the time required to explain a product, the more time it takes to explain the product, so it can be inferred that this is a topic that is likely to develop into trouble later, and the bias increases. On the other hand, if the explanation is completed in a short time, the bias may be reduced.

  Even if the explanation is finished for all the products and the fact is displayed on the salesperson screen 62, the conversation may still be continued, so the salesperson does not take any action immediately. Eventually, after the customer returns, the salesperson inputs a talk end instruction from the input unit 66. The control unit 100 confirms whether or not all the explanations have been completed with reference to the navigation information unit 70 of the navigation support unit 60, and stops the recording of all the recording recorders 36. If an explanation for all the products has not been completed due to a salesperson's mistake and an end-of-concept instruction has been entered, a warning is issued and a warning that all explanations have not been completed is displayed on screen 62. To do.

  Subsequently, automatic correction S47 by the recording control unit 32 is performed. The time stamp obtained by shifting the time stamp of the voice file 53 in which the conversation is partially recorded as shown in FIG. 3 is temporarily recorded in the memory 48. Further, the salesperson corrects the time stamp of the audio file by the correction S49 by the manual means. As a result, when the time zone to be finally saved, that is, the times t0 to t3 shown in FIG. 3, is determined, the recording control unit 32 cuts out the recording content corresponding to the time zone from the recording recorder 36 for all recordings. A corrected audio file 63 shown in FIG. Then, the audio file 63 is passed to the control unit 100. The control unit 100 adds the audio file 63 to the storage medium 14 or directly to the customer information database 16, thereby adding it to the customer information as a record of the product description.

  The control unit 100 also adds to the customer information the storage period of the audio file determined by the bias calculation unit 80 based on the bias calculated by the bias calculation unit 80 and corrected by the salesman by manual means as necessary.

  Furthermore, the control unit 100 adds the prohibition phrase recorded in the monitoring unit 90 to the customer information.

  FIG. 5 shows an embodiment of a remote conversation recording system using the conversation recording apparatus described in FIG. The remote conversation recording system 200 includes a customer telephone 204 that is a subscriber telephone connected to a Public Switched Telephone Network (PSTN) 202. The telephone 204 is used by a customer from home or the like. The system 200 includes a salesperson telephone 208 that is a subscriber telephone provided in a store 206 such as a bank and connected to a public switched telephone network 202. Further, what is connected to the public switched telephone network 202 in the store 206 may be a computer telephony system 210 provided with a customer screen.

  Further, the system 200 includes a conversation recording server 214 provided by a business operator 212 which is a third party, and the conversation recording apparatus 10 shown in FIG. The conversation recording server 214 is also connected to the public telephone exchange network 202, and mediates a call between the customer telephone 204 and the salesperson telephone 208 in the store 206. That is, the customer does not directly call the salesperson's telephone 208 using the customer telephone 204, but first connects the call to the conversation recording server 214, and the conversation recording server 214 further connects the call connection line. Is used to connect the call to the salesperson telephone 208. As a result, all conversations between the customer and the salesperson go through the server 214.

  The conversation recording server 214 is connected to the conversation recording apparatus 10, converts the conversation exchanged between the customer telephone 204 and the salesperson telephone 208 into a packet, and transmits the packet to the conversation recording apparatus 10. The conversation recording device 10 reproduces the received packet as a voice and receives it from a portion corresponding to the microphones 24 and 38. Thereafter, as already described, the necessary part of the conversation is recorded based on the keywords registered in the dictionary 20 in advance.

  When using the system 200 shown in FIG. 5, the customer or salesperson (store) needs to make a contract with the business operator 212 and inform the business operator in advance of keywords and the like necessary for the start and end of recording. . According to the embodiment of FIG. 5, the customer can record the conversation even when consulting with the salesperson of the store by telephone from his home at a location remote from the store. In the present embodiment, the present invention is used from the customer side, but it may be the salesperson side who makes a contract with the business operator 212 and requests the recording of the conversation.

  FIG. 6 shows another embodiment of the remote conversation recording system using the conversation recording apparatus described in FIG. 1, and is a modification of the embodiment shown in FIG. The remote conversation recording system 300 shown in FIG. 6 accepts customer consultations and complaints over the telephone, accurately and promptly communicates the customer consultation details to the person in charge according to the customer consultation content, and the person in charge is appropriate for the customer. It aims at responding to.

  In the following, the present embodiment will be described in particular on the point that FIG. 6 is different from FIG. The customer telephone 304 and PSTN 302 are the same as those shown in FIG. The embodiment shown in FIG. 6 is different from the embodiment shown in FIG. 5 in that two kinds of conversation recording devices, that is, a reception conversation recording device 10A and a person-in-charge conversation recording device 10B are provided in the store 306. The devices 10A and 10B have the same configuration as the conversation recording device 10 shown in FIG. 1, and may be personal computers as shown. The device 10A is used by a receptionist who first receives a call from a customer, and the device 10B is used by a person in charge according to the contents of customer consultation. Both the devices 10A and 10B are representatively shown one by one in FIG. 6, but may be increased according to the number of receptionists and persons in charge. When the number is increased, the connection method to the LAN 18 for performing data communication and the extension telephone line 314 may be the same as that of the devices 10A and 10B shown in FIG.

  Although the components included in the devices 10A and 10B shown in FIG. 6 are almost omitted, the displays 62A and 62B corresponding to the salesperson screen 62 in FIG. 1 are shown. Headset 310A, 310B is connected to apparatus 10A, 10B shown in FIG. Headsets 310A and 310B are connected to an extension telephone line 314 in the store, and are microphones (not shown in FIG. 6) of devices 10A and 10B, respectively, and correspond to microphones 24 and 38 shown in FIG. Connected to things. That is, when an incoming call is made from the customer telephone 304 to the headsets 310A and 310B via the private branch exchange (PBX) 320, the conversation between the customer and the receptionist and the conversation between the customer and the person in charge are: Voice recognition is performed by the devices 10A and 10B, and recording is also possible.

  A PBX 320 shown in FIG. 6 is installed in the store 306 and exchanges between an extension and an outside line. The PBX 320 is connected to the PSTN 302 and is also connected to the LAN 18 and the extension telephone line 314 in the store 306. ing. The PBX 320 controls the headsets 310A and 310B, which are extension telephones in the store 306, and enables outgoing and incoming calls by connecting to a telephone exchange through a public line. The PBX 320 has a number display function provided by Nippon Telegraph and Telephone (NTT) and receives the telephone number of the other party when an incoming call is received from an outside line. Can be sent to 330. The PBX 320 may be a digital PBX and may provide various functions such as dial-in, hold, transfer, speed dial, and callback. Further, the PBX 320 may incorporate functions such as ACD (Automatic Call Distribution) and CMS (Call Management System) used for a call center.

  The PBX 320 shown in FIG. 6 is connected to the PSTN 302, but accepts a call from an IP telephone or the like via an IP (Internet Protocol) network (not shown) and provides a private branch exchange function on the IP network. -It is good also as IP-PBX to manage.

  In the embodiment of FIG. 6, a server 330 is connected to the LAN 18, and this server has the customer information database 16 shown in FIG. As described above, when receiving an incoming call from the outside line, the PBX 320 transmits the telephone number of the other party to the server 330 through the number display. The telephone number is sent to the server 330 and is queried in the customer information database 16 of the server 330, except when the other party makes a call from a public telephone or uses the non-notification function. The server 330 searches the database 16 using the telephone number, and if the telephone number is already registered, the server 330 can transmit the customer information together with the telephone number to the conversation recording devices 10A and 10B. Therefore, the receptionist and the person in charge using the devices 10A and 10B can not only receive the call from the outside line by the headsets 310A and 310B, but also display the information of the other party who has transmitted the call by the displays 62A and 62B. May be referenced. As described above, in this embodiment, the computer telephony system is configured by the PBX 320, the server 330, and the conversation recording devices 10A and 10B.

  In the first and second embodiments, of the two recorders included in one conversation recording apparatus 10, the partial recording recorder 34 records a necessary part of the conversation, and the entire recording recorder 36 records the entire conversation. However, in the present embodiment, the entire conversation between the receptionist and the customer may be recorded by a single recorder (not shown) of the reception conversation recording apparatus 10A that receives a call from the customer. Thereafter, the entire conversation between the person in charge and the customer may be recorded by one recorder (not shown) of the person-in-charge conversation recording device 10B used by the person in charge. The audio files saved by the recorders of the devices 10A and 10B can be time stamped by an operator input, similar to those used in the first and second embodiments. The time stamp pushed on the audio file serves as a bookmark, and the audio file can be reproduced by jumping to the time indicated by the time stamp. In addition, in the audio file, for example, the type of customer's business can be recorded in the header portion. The type of requirement may be “claim” for complaints or the like, and “request” for requests. In this case, the acceptor records the voice file using the input means of the apparatus 10A.

  A feature of the present embodiment is that a printer 340 connected to the LAN 18 recognizes the contents of the customer's summary out of the conversation between the receptionist and the customer recorded by the reception conversation recording device 10A, and is connected to the LAN 18. To print out. The person in charge can quickly understand the customer's requirements by looking at the printed out document before discussing with the customer.

  Functions of the components of the apparatus 10A shown in FIG. 6 will be described. Since the device 10A has the same configuration as the device 10 of FIG. 1, for the sake of convenience, components (not shown) included in the device 10A of FIG. 6 will be described using the reference numerals in FIG. During the conversation between the customer and the receptionist, a print start keyword is issued from the receptionist's mouth and is recognized by the speech recognition unit 22 of the conversation recording device 10A via the headset 310A. Start converting to data. The voice recognition unit 22 sequentially passes the converted text data to the control unit 100 via the signal line 30. The control unit 100 stores the received text data. A memory for storage may be prepared in the control unit 100, or the memory 48 or the memory 82 in the bias calculation unit 80 may be used. The control unit 100 stores the text data until a print end keyword is issued from the receptionist's mouth. Then, the control unit 100 transfers the stored text data to the printer 340 via the LAN 18 and prints it out as a text document.

  In this embodiment, the print start keyword registered in advance in the dictionary 20 may be a word such as “I will confirm the contents of customer consultation”, and the print end keyword is, for example, “This "Is it all right for you to consult us?" By doing so, the receptionist summarizes and confirms the contents of the consultation spoken by the customer between the time when the print start keyword is issued and the time when the print end keyword is issued. Then, since the summarized consultation content is stored as text data by the control unit 100 as described above, it is printed out by the printer 340 and delivered to the person in charge as a text document. The person in charge can quickly and accurately grasp the contents of the consultation in advance and enter into a discussion with the customer by looking at the contents of the consultation that is summarized at most several lines.

  The operation of the present embodiment configured as described above will be described below with reference to FIG. FIG. 7 is a sequence diagram for explaining the operation of this embodiment. When the customer calls the store 306 using the customer telephone 304, the PBX 320 (not shown in FIG. 7) that receives the incoming call S400 transfers the call to the reception conversation recording device 10A. If the telephone number of the customer telephone 304 is known from the number display, the PBX 320 transmits this to the server 330. The receptionist answers the call by off-hooking headset 310A. Then, the recorder of the apparatus 10A automatically starts recording the conversation between the customer and the receptionist. The receptionist first states the words of the reception greeting S402. The customer outlines the requirements (claims) (step S404). The acceptor operates the input device of the device 10A at an arbitrary time, and records the time stamp in the audio file 414 being recorded on the recorder 36A (step S406). This time stamp may be determined and pressed by the receptionist during the conversation of the customer, at a point in time when the customer feels angry and the atmosphere. The time stamp may be pushed again and again if a scene later appears that the customer's situation is more noticeably transmitted.

  The receptionist who has listened to the customer's story as a whole performs confirmation 408 of the requirement. That is, the customer's requirements are summarized between the printing start keyword “I will confirm the contents of the customer's consultation” and the printing end keyword “Are you sure you would like this consultation?”. The conversation made by the receptionist between these keywords is voice-recognized S410 by the reception conversation recording apparatus 10A and printed out as a text document 420 by the printer 340. On the other hand, the recorder 36A of the reception conversation recording apparatus 10A ends the conversation recording when the printing end keyword is issued. For the audio file 414 created in the recorder 36A, the acceptor operates the input means of the apparatus 10A to input information such as the type of customer requirement, for example, “claim” or “request” (step S412).

  In order to obtain the text document 420, it is not necessary to use voice recognition means. For example, when voice recognition fails, the receptionist may create the text document 420 manually.

  With the above operation, the recorder 36A stores the audio file 414 in which the conversation between the customer and the receptionist is recorded, and the audio file 414 shows the time when the atmosphere such as the customer's anger is clearly indicated. The stamp is stamped and includes the customer requirement type. On the other hand, the printer 340 obtains a text document 420 on which the contents summarizing customer requirements are printed. The receptionist plays music to the customer (step S421), and transfers the call from the customer to the conversation recording device 10B for the person in charge according to the customer's requirements (step S422). When the person in charge answers the call using the headset 310B, the receptionist informs the person in charge that the customer is calling and transfers the audio file 414 to the device 10B via the LAN 18 and also a text document. 420 is given to the person in charge (step S424). Note that the content printed on the text document 420 does not necessarily have to be given to the person in charge in a state printed on the text document. For example, the text data may be transferred via the LAN 18 to the person in charge 10B.

  The person in charge who has received the audio file 414 and the text document 420 first confirms the type of customer request in the audio file 414. If the type of customer request is just a “request”, for example, the receptionist may or may not hear the customer's voice directly by playing the audio file 414 before consulting with the customer. . This is because in the case of “request”, the receptionist determines that the customer is not angry. On the other hand, if the customer's business type is “claim”, the receptionist has determined that the customer is angry, so the person in charge will start talking to the customer from the time when the time stamp of the audio file 414 is pressed. And quickly grasp the customer's psychological state. In this embodiment, the recorder 36A records the entire conversation, but in the audio file 414, the customer's conversation can be reproduced from the time when the time stamp is pressed, and the customer's psychological state can be grasped in a good manner. . That is, the person in charge does not need to listen to the entire contents of the recorded audio file 414.

  In addition, the person in charge glances at the text document 420 to understand the customer's requirements accurately and quickly. As described above, when the person in charge understands the customer's psychological state and requirements quickly and accurately, the person in charge answers the telephone transferred from the receptionist and starts a conversation (session) with the customer (step S426). At the same time, the recorder 36B of the person-in-charge conversation recording device 10B starts recording the conversation between the customer and the person in charge. The person in charge greets and asks the customer to explain the requirement in detail while paying attention to the customer's psychological state (step S428). The customer explains the details of the business (step S430).

  Note that the conversation between the customer and the person in charge, that is, the detailed contents of the complaint, may be voice-recognized in whole or in part by the apparatus 10B and used as a text document 440 using the printer 340.

  Conventionally, as a means for dealing with a telephone request from a customer, there is a technique using an IVR (Interactive Voice Response) installed in a call center. According to IVR, the customer can narrow down the requirements by operating the dial buttons, and the call is automatically transferred to the person in charge who seems to be appropriate. In this way, IVR does not require a recipient, so the cost of response can be reduced. However, the customer refrains from the mechanical response and the time until the call is transferred to the person in charge is long, so the customer often disconnects the call during the operation of the IVR. In addition, if the customer's requirements are complex and the IVR does not belong to any of the categories prepared in advance, the “Other” category must be used, and the person in charge is not necessarily well organized. It was necessary to listen to the customer from the beginning and confirm the contents.

  Conventionally, instead of IVR, there has been a means of arranging a receptionist at a call center or a store to deal with customers. Although there is no problem that the customer tends to refrain from dealing with human beings, how to quickly and accurately convey the customer's requirements to the person in charge has been a problem. If the receptionist does not summarize the customer's requirements in a precise and accurate manner and tells the person in charge, the person in charge will end up listening to the customer from the beginning, and the customer must repeat the same explanation. Will occur.

  In this embodiment, the above-mentioned problems of the prior art are solved, and the customer does not refrain from dealing with human beings, and the person in charge looks at the text document summarizing the customer's requirements. It has the advantage that the customer's requirements can be quickly and quickly transmitted to the person in charge. In addition, since the person in charge can hear the recorded content of the conversation between the customer and the receptionist, the person in charge can infer the psychological state of the customer and the nuances of words in advance. With these advantages, it is possible to prevent the initial response error from adversely affecting the subsequent session of the customer / contact person. Note that the present embodiment is not limited to the use of accepting customer consultation from a remote place by telephone as shown in FIG. Even when a customer comes directly to the store 306 for consultation, it is possible to quickly transmit the contents of the consultation to the person in charge using this embodiment.

It is a block diagram which shows the Example of the conversation recording apparatus by this invention. It is a figure which shows notionally the recording content of the two recorders shown in FIG. 3 is a time chart conceptually showing automatic correction and manual correction shown in FIG. 2. It is a figure which shows notionally the process of the bias calculation which the bias calculation part shown in FIG. 1 performs. It is a figure which shows the Example of the remote conversation recording system using the conversation recording apparatus shown in FIG. It is a figure which shows the other Example of the remote conversation recording system using the conversation recording apparatus shown in FIG. It is a sequence diagram explaining operation | movement of the remote conversation recording system shown in FIG.

Explanation of symbols

10 Conversation recorder
22 Voice recognition unit
32 Recording control section
34, 36 recorder
60 Navigation support department
80 Bias calculator
200, 300 remote conversation recording system

Claims (12)

A topic storage means for recording one or more topics constituting a conversation and recording information indicating any incomplete topic for each of the one or more topics;
For each of the one or more topics, keyword storage means for recording a keyword indicating that the topic has started and a keyword indicating that the topic has ended;
Voice recognition means for voice recognition of a conversation and detecting a keyword indicating that any topic has started and ended from the conversation;
A navigation unit that records information indicating the end of the topic in the topic storage unit when both of the keywords indicating that any topic has started and the topic has ended are detected by the voice recognition unit; ,
A navigation support apparatus, comprising: display means for displaying the one or more topics and displaying information indicating whether the topic is incomplete for each of the one or more topics. 2. The apparatus according to claim 1 , wherein when a keyword indicating that any topic is started is detected by the voice recognition means, the navigation unit displays information indicating that the topic is in progress. A navigation support apparatus characterized by being displayed on a display means. 3. The apparatus according to claim 1 or 2, wherein when a keyword indicating that any topic is started is detected by the voice recognition unit, the apparatus starts recording the conversation and ends the topic. A navigation support apparatus, comprising: a recording control unit that terminates recording of the conversation and stores it as a voice file when the voice recognition unit detects a keyword indicating that the recording has been performed. 4. The apparatus according to claim 3, wherein the apparatus includes conversation importance calculation means for calculating importance of the conversation, and the conversation importance calculation means includes:
Keyword storage means for recording one or more keywords;
Voice recognition means for voice recognition of a conversation and detecting the one or more keywords from the conversation;
A first recording means for recording;
Importance recording means for recording a value indicating the importance of the conversation;
Wherein when it is detected one of the keywords by the speech recognition means, the navigation support apparatus according to claim and calculating means for increasing a predetermined amount a value indicating the importance of the conversation including it. 4. The apparatus according to claim 3, wherein the apparatus includes conversation importance calculation means for calculating importance of the conversation, and the conversation importance calculation means includes:
Time storage means for recording a time estimated to elapse from the start to the end of each topic for one or more topics constituting the conversation;
A first recording means for recording;
Importance recording means for recording a value indicating the importance of the conversation;
For each of the one or more topics recorded in the first recording means, the importance of the conversation is determined according to the difference between the time from the recording start time to the recording end time of the conversation and the predicted time. navigation support apparatus according to claim including that the calculating means for changing a value indicating. 6. The navigation support apparatus according to claim 4, wherein the apparatus includes a control unit that determines a storage period of the audio file in accordance with the importance of the conversation. Using the navigation support device
Displaying on the display means information representing any incomplete topic for each of one or more topics constituting the conversation;
Detecting a keyword indicating that any topic has been started from the conversation by voice recognition of the conversation;
Detecting a keyword indicating that one of the topics has ended from the conversation by voice recognition of the conversation;
And a step of displaying information indicating the end of any one of the topics on the display means. 8. The method of claim 7 , wherein the method is further in the middle of any of the topics upon detecting a keyword indicating that the topic has started from the conversation by voice recognition of the conversation. A navigation support method comprising a step of displaying information indicating that on the display means. 9. The method according to claim 7 or 8, further comprising detecting a keyword indicating that any topic has been started from the conversation by voice recognition of the conversation, and recording the conversation. A navigation support method comprising a step of ending recording of the conversation and storing it as an audio file when a keyword indicating that the topic has ended is detected. 10. The method of claim 9, further comprising:
Detecting any one of the one or more keywords by voice recognition of the conversation, and increasing the value indicating the importance of the conversation by a certain amount,
A navigation support method characterized by calculating the importance of the conversation. 10. The method of claim 9, further comprising:
For each one or more topics that make up the conversation, setting a time that is expected to elapse from the start to the end of the topic;
Recording the one or more topics from a conversation into a first recording means;
For each of the one or more topics recorded in the first recording means, the importance of the conversation is determined according to the difference between the time from the recording start time to the recording end time of the conversation and the predicted time. Changing the indicated value,
A navigation support method characterized by calculating the importance of the conversation. 12. The navigation support method according to claim 10, further comprising a step of determining a storage period of the audio file according to the importance of the conversation.

Images