Operator support system, operator support device, and operator support method

Abstract

To use a voice recognition technique and a generative AI to extract registration information of each item of a customer service record from a voice recognition result related to a call voice between a customer and an operator, and appropriately reflect an item extraction result on a registration screen while avoiding manual input of the operator from being wasted due to overwriting when reflecting the item extraction result on the registration screen.SOLUTION: An operator support system instructs a voice recognition section to execute voice recognition processing for converting a call voice into text by voice recognition, instructs a text analysis section to execute item extraction processing for extracting a draft of registration information of each item of a customer service record from a voice recognition result, displays a registration screen 51 on which an operator performs operation of registration regarding each item of the customer service record on an operator terminal, determines a state of manual input of the operator to an input field 52 of each item on the registration screen when acquiring an item extraction result from the text analysis section, and changes a mode of reflecting the item extraction result in the input field according to the state of the manual input.SELECTED DRAWING: Figure 5

Description

[Technical Field]

[0001] The present invention relates to an operator support system, an operator support device, and an operator support method for supporting the work of an operator who creates a customer interaction record based on the content of an inquiry made by a customer over the phone. [Background technology]

[0002] In a call center, after an operator has finished answering a customer inquiry by phone, they create a customer interaction record (ticket) in which the customer's inquiry is divided into items according to a predetermined format. Because creating this customer interaction record is time-consuming, it is desirable to support the operator's work and reduce the burden on the operator.

[0003] As a technology for supporting the work of such operators and reducing their burden, a technology has been known that uses voice recognition technology to convert the voice of a conversation between a customer and an operator into text and obtain customer information from the voice recognition results, thereby eliminating the need for the operator to manually enter information (see Patent Document 1). [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Patent No. 7237507 Summary of the Invention [Problem to be solved by the invention]

[0005] When creating a customer interaction record, a registration screen is displayed on the operator's terminal, allowing the operator to input information to be registered in each item of the customer interaction record. If, as with conventional technology, speech recognition results for the conversation between the customer and the operator can be used, the operator's manual input work can be eliminated. It is also possible to use a generation AI for the item extraction process, which extracts registration information for each item of the customer interaction record from the speech recognition results. In this case, once the speech recognition and item extraction processes are complete, the item extraction results can be reflected on the registration screen, allowing the operator to confirm the registration details.

[0006] On the other hand, during a call or while waiting for the completion of the voice recognition process and the item extraction process, an operator may manually input information into the registration screen using an input device. For example, proper nouns such as customer names are prone to errors in voice recognition, so the operator may manually input them. In this case, when the voice recognition process and the item extraction process are completed and the item extraction results are to be reflected on the registration screen, the operator's manual input results and the item extraction results will conflict. This can result in the operator's manual input being wasted due to overwriting, or the item extraction results cannot be properly reflected on the registration screen.

[0007] Therefore, the main object of the present invention is to provide an operator support system, an operator support device, and an operator support method that, when using voice recognition technology and generation AI to extract registration information for each item in a customer interaction record from the voice recognition results regarding the conversation voice between a customer and an operator, and when reflecting the item extraction results on a registration screen, can appropriately reflect the item extraction results on the registration screen while avoiding the operator's manual input being wasted by overwriting. [Means for solving the problem]

[0008] The operator support system of the present invention is an operator support system in which one or more processors execute processes to support the work of an operator who creates a customer interaction record based on the content of an inquiry made by a customer over the phone, and the processor instructs a speech recognition unit to execute a speech recognition process that converts the conversation voice between the customer and the operator into text using speech recognition to obtain the speech recognition result, instructs a text analysis unit to execute an item extraction process that extracts a draft of the registration information for each item of the customer interaction record from the speech recognition result, displays on the operator terminal a registration screen on which the operator performs registration operations for each item of the customer interaction record, and upon obtaining the item extraction result from the text analysis unit, determines the state of manual input by the operator for each item in the registration screen, and changes the manner in which the item extraction result is reflected in the input field depending on the state of manual input.

[0009] Furthermore, the operator support device of the present invention is an operator support device in which a processor executes a process to support the work of an operator who creates a customer interaction record based on the content of an inquiry made by a customer over the phone, and the processor instructs a speech recognition unit to execute a speech recognition process that converts the conversation voice between the customer and the operator into text using speech recognition to obtain the speech recognition result, instructs a text analysis unit to execute an item extraction process that extracts a draft of the registration information for each item of the customer interaction record from the speech recognition result, instructs a GUI control unit to display on the operator terminal a registration screen on which the operator performs registration operations for each item of the customer interaction record, and, upon obtaining the item extraction result from the text analysis unit, determines the state of manual input by the operator for each item in the registration screen, and changes the manner in which the item extraction result is reflected in the input field depending on the state of manual input.

[0010] Furthermore, the operator support method of the present invention is an operator support method that causes one or more processors to execute a process to support the work of an operator who creates a customer interaction record based on the content of an inquiry made by a customer over the phone, and is configured to instruct a speech recognition unit to execute a speech recognition process that converts the conversation voice between the customer and the operator into text using speech recognition, obtain the speech recognition result, instruct a text analysis unit to execute an item extraction process that extracts from the speech recognition result a draft of the registration information for each item of the customer interaction record, display on the operator terminal a registration screen on which the operator performs registration operations for each item of the customer interaction record, and when the item extraction result is obtained from the text analysis unit, determine the state of manual input by the operator for each item in the registration screen, and change the manner in which the item extraction result is reflected in the input field depending on the state of manual input. [Effects of the Invention]

[0011] According to the present invention, the manner in which the item extraction results are reflected in the input fields on the registration screen is changed depending on the state of the operator's manual input into the input fields for each item. Therefore, even if a conflict occurs between the operator's manual input results and the item extraction results, the item extraction results can be appropriately reflected on the registration screen while avoiding the operator's manual input from being wasted due to overwriting. [Brief explanation of the drawings]

[0012] [Figure 1] Overall configuration of an operator support system according to the present embodiment. [Figure 2] FIG. 1 is a block diagram showing the schematic configuration of a main control unit and a GUI control unit according to the present embodiment. [Figure 3] FIG. 10 is an explanatory diagram showing an overview of the process of reflecting the item extraction results performed by the main control unit according to the present embodiment; [Figure 4] FIG. 10 is an explanatory diagram showing a registration screen displayed on an operator terminal according to the embodiment; [Figure 5] FIG. 10 is an explanatory diagram showing a registration screen displayed on an operator terminal according to the embodiment; [Figure 6]FIG. 10 is an explanatory diagram showing an outline of a second example of an input completion determination process performed by the main control unit according to the embodiment; [Figure 7] FIG. 10 is an explanatory diagram showing an example of a prompt and its response in a third example of the input completion determination process performed by the main control unit according to the embodiment. [Figure 8] FIG. 10 is an explanatory diagram showing a registration screen displayed on an operator terminal according to a first modified example. [Figure 9] FIG. 10 is an explanatory diagram showing a registration screen displayed on an operator terminal according to a first modified example. [Figure 10] FIG. 10 is an explanatory diagram showing a registration screen displayed on an operator terminal according to a second modified example. [Figure 11] FIG. 10 is an explanatory diagram showing the transition status of the main part of the registration screen displayed on the operator terminal according to the second modified example. [Figure 12] FIG. 11 is an explanatory diagram showing an example of a table of correction history stored in a storage unit of a main control unit according to a third modified example. [Figure 13] FIG. 11 is an explanatory diagram showing a registration screen displayed on an operator terminal according to a third modified example. DETAILED DESCRIPTION OF THE INVENTION

[0013] The first invention made to solve the above problem is an operator support system in which one or more processors execute processes to support the work of an operator who creates customer interaction records based on the content of inquiries made by customers over the phone, wherein the processor instructs a speech recognition unit to execute a speech recognition process that converts the conversation voice between the customer and the operator into text using speech recognition, obtains the speech recognition result, instructs a text analysis unit to execute an item extraction process that extracts a draft of the registration information for each item of the customer interaction record from the speech recognition result, displays on the operator terminal a registration screen on which the operator performs registration operations for each item of the customer interaction record, and upon obtaining the item extraction result from the text analysis unit, determines the state of manual input by the operator for each item in the registration screen, and changes the manner in which the item extraction result is reflected in the input field depending on the state of manual input.

[0014] This allows the manner in which the item extraction results are reflected in the input fields to be changed depending on the state of the operator's manual input into the input fields for each item on the registration screen. Therefore, even if a conflict occurs between the operator's manual input results and the item extraction results, the item extraction results can be appropriately reflected on the registration screen while avoiding the operator's manual input from being wasted due to overwriting.

[0015] In addition, in a second invention, the processor is configured such that, in the case of an uninputted state where no manual input has been made by the operator yet, the processor reflects the item extraction result in the input field; in the case of an input already made by the operator, the processor determines whether the manual input by the operator has been completed, and in the case of an input-completed state where the manual input by the operator has been completed, the processor does not reflect the item extraction result in the input field; and in the case of an input-incompleted state where the manual input by the operator has not been completed, the processor reflects the item extraction result in the input field.

[0016] This allows the item extraction results to be reflected in the input fields appropriately according to the state of manual input by the operator (no input state, input completed state, input incomplete state).

[0017] In addition, a third invention is configured such that, when the processor obtains the item extraction result from the text analysis unit, the processor reflects the item extraction result in the input field in an unconfirmed state, and changes the item extraction result in the input field to a confirmed state in response to an operator's operation to select the input field.

[0018] This means that in order to confirm the item extraction results reflected in the input field, an operator's operation to approve the item extraction results is required, thereby ensuring that the operator can confirm whether the item extraction results are appropriate.

[0019] In addition, a fourth invention is configured such that, when the processor obtains the item extraction result from the text analysis unit, the processor reflects the item extraction result in the input field in an unconfirmed state, and displays an operable button on the registration screen to approve the item extraction result, and changes the item extraction result in the input field to a confirmed state in response to an operator's operation of the button.

[0020] This means that in order to confirm the item extraction results reflected in the input field, an operator's operation to approve the item extraction results is required, thereby ensuring that the operator can confirm whether the item extraction results are appropriate.

[0021] In a fifth aspect of the present invention, the processor displays the item extraction results in the input field in different modes depending on whether the item is in an unconfirmed state or a confirmed state.

[0022] This allows the operator to easily distinguish whether the item extraction result displayed in the input field is in an unconfirmed state or a confirmed state. In this case, for example, the unconfirmed state and the confirmed state may be distinguished by changing the character form, such as italics, character size, or character color.

[0023] In addition, in a sixth invention, the processor is configured to display, on the registration screen, display columns for displaying the item extraction results, arranged by item, separately from the input columns, and to display, on the registration screen, a button for approving the item extraction results, and, when the item extraction results are obtained from the text analysis unit, to display the item extraction results in the display columns, and to reflect the item extraction results displayed in the display columns in the input columns in response to an operator's operation of the button.

[0024] This means that in order to confirm the item extraction results reflected in the input field, an operator's operation to approve the item extraction results is required, thereby ensuring that the operator can confirm whether the item extraction results are appropriate.

[0025] In addition, a seventh invention is configured such that, when the input is complete and manual input by the operator has been completed, the processor does not reflect the item extraction results in the input field, and when the operator performs an operation to select the input field, the processor displays the item extraction results as input candidates in a pop-up on the registration screen, and reflects the item extraction results in the input field in response to the operator's operation to select the item extraction result.

[0026] According to this, even if the item extraction results are not reflected in the input field in the input completion state where the operator has completed manual input, the item extraction results can be reflected in the input field in response to the operator's operation.

[0027] In addition, an eighth invention is configured such that, in response to an operator's operation to select the input field, if a caret is positioned at the beginning of the input field, the processor displays the item extraction results as input candidates in a pop-up on the registration screen, and if a caret is not positioned at the beginning of the input field, in addition to the item extraction results, words related to the customer's inquiry as input candidates in a pop-up on the registration screen.

[0028] This allows the operator to be presented with appropriate input candidates according to the situation when editing the item extraction results displayed in the input field. Here, the words related to the customer's inquiry include, for example, words related to the word where the caret is located, words extracted from the speech recognition results, words related to the customer, words related to the type of equipment or system used by the customer, etc. The caret (cursor) indicates the text input position and is displayed when the operator selects the input field.

[0029] In addition, a ninth invention is configured such that the processor displays, on the registration screen, display columns in which the item extraction results are displayed, arranged by item, separately from the input columns, and after selecting a portion of the item extraction results displayed in the display columns, reflects only the selected portion of the item extraction results in the input columns in response to an operator's operation to instruct the reflection of the item extraction results.

[0030] This allows the operator to edit the character string in the input field by quoting a portion of the item extraction result displayed in the display field, thereby reducing the operator's input operation. In this case, when a caret is displayed in the input field, the operator may perform an operation to instruct the reflection of the item extraction result, thereby inserting a portion of the item extraction result at the position of the caret. Also, when a portion of the operator's manual input result is selected in the input field, the operator may perform an operation to instruct the reflection of the item extraction result, thereby replacing the manual input result selected in the input field with the portion of the item extraction result selected in the display field.

[0031] In addition, a tenth invention is configured such that the processor stores the results of manual corrections made by the operator to the item extraction result as a correction history in a memory unit, and in response to the operator's operation of selecting a word or phrase from the item extraction result in the input field, if the selected word or phrase matches a word or phrase that was previously corrected in the item extraction result based on the correction history, the processor displays the correction result at that time as an input candidate on the registration screen.

[0032] This allows the operator to correct any errors that may occur in the item extraction results, simply by selecting an input candidate, thereby reducing the burden on the operator.

[0033] In addition, an eleventh invention is configured such that the processor stores the results of manual corrections made by an operator to the item extraction results as a correction history in a memory unit, selects items that are likely to cause errors in the item extraction results based on the correction history, and highlights the input fields on the registration screen related to items that are likely to cause errors in the item extraction results.

[0034] This allows the operator to easily recognize items that are likely to cause errors in the item extraction results.

[0035] In addition, a twelfth invention is configured such that, when the call unit detects an incoming call from a customer making an inquiry, the processor newly registers the customer response record related to the inquiry, notifies the operator terminal of this, and, in response to a request from the operator terminal, displays the registration screen on the operator terminal in a state where the operator can manually enter data.

[0036] This allows the operator to appropriately input text on the registration screen to take notes on the call with the customer during the call. The operator can then input notes during the call and revise the item extraction results after the call ends, all on the same screen. This reduces the time required to switch screens, improving operator convenience.

[0037] Further, a thirteenth invention is an operator support device in which a processor executes a process to support the work of an operator who creates a customer interaction record based on the content of an inquiry made by a customer over the phone, wherein the processor instructs a speech recognition unit to execute a speech recognition process that converts the conversation voice between the customer and the operator into text using speech recognition, obtains the speech recognition result, instructs a text analysis unit to execute an item extraction process that extracts a draft of registration information for each item of the customer interaction record from the speech recognition result, instructs a GUI control unit to display on an operator terminal a registration screen on which the operator performs registration operations for each item of the customer interaction record, and upon obtaining the item extraction result from the text analysis unit, determines the state of manual input by the operator for each item in the registration screen, and changes the manner in which the item extraction result is reflected in the input field depending on the state of manual input.

[0038] According to this, as in the first invention, the manner in which the item extraction results are reflected in the input fields on the registration screen is changed depending on the state of the operator's manual input into the input fields for each item on the registration screen. Therefore, even if a conflict occurs between the operator's manual input results and the item extraction results, the item extraction results can be appropriately reflected on the registration screen while avoiding the operator's manual input from being wasted due to overwriting.

[0039] Further, a fourteenth invention is an operator support method that causes one or more processors to execute a process to support the work of an operator who creates a customer interaction record based on the content of an inquiry made by a customer over the phone, the method instructing a speech recognition unit to execute a speech recognition process that converts the conversation voice between the customer and the operator into text using speech recognition to obtain the speech recognition result, instructing a text analysis unit to execute an item extraction process that extracts a draft of the registration information for each item of the customer interaction record from the speech recognition result, displaying on an operator terminal a registration screen on which the operator performs registration operations for each item of the customer interaction record, and upon obtaining the item extraction result from the text analysis unit, determining the state of manual input by the operator for each item in the registration screen, and changing the manner in which the item extraction result is reflected in the input field depending on the state of manual input.

[0040] According to this, as in the first invention, the manner in which the item extraction results are reflected in the input fields on the registration screen is changed depending on the state of the operator's manual input into the input fields for each item on the registration screen. Therefore, even if a conflict occurs between the operator's manual input results and the item extraction results, the item extraction results can be appropriately reflected on the registration screen while avoiding the operator's manual input from being wasted due to overwriting.

[0041] Hereinafter, embodiments of the present invention will be described with reference to the drawings.

[0042] FIG. 1 is a diagram showing the overall configuration of an operator support system according to this embodiment.

[0043] The operator support system supports the work of call center operators who create customer interaction records (tickets) based on the details of customer inquiries made over the phone. Customer interaction records are entries in which the details of customer inquiries are divided into items according to a specified format.

[0044] The operator support system includes a call unit 1, a voice interception unit 2, a main control unit 3 (operator support device), a voice recognition unit 4, a text analysis unit 5, a management unit 6, a GUI control unit 7, and an operator terminal 8.

[0045] The call unit 1 relays the transmission and reception of voice between the telephone 9 used by the customer and the operator terminal 8 used by the operator, and realizes a call between the customer and the operator. The call unit 1 may be realized by a known cloud service, or may be realized by a telephone system in which the functions of a "PBX (main unit)" are built on the cloud and business phone functions can be used over an internet line.

[0046] The voice intercepting unit 2 intercepts (collects) the voice of the call between the customer and the operator made by the call unit 1. The voice intercepting unit 2 may be realized by a known cloud service.

[0047] The voice recognition unit 4 converts the voice of the call acquired by the voice interception unit 2 into text by voice recognition. The voice recognition unit 4 may be realized by a known cloud service.

[0048] The text analysis unit 5 executes an item extraction process on the speech recognition results output from the speech recognition unit 4. In the item extraction process, a generation AI is used to extract character strings that serve as a draft (original plan) of the registration information for each item in the customer interaction record from the speech recognition results. The text analysis unit 5 also executes a summary generation process on the speech recognition results. In the summary generation process, a summary of the inquiry content is generated using the generation AI. The text analysis unit 5 may be realized by a cloud service such as Google Bard (registered trademark), Microsoft Bing AI (registered trademark), or ChatGPT (registered trademark).

[0049] The management unit 6 manages various data such as customer correspondence records using a database. The management unit 6 may be realized, for example, by a tool called a CRM tool that manages contact history with customers, or a general relational database. It may also be realized by a known incident management tool or a cloud service including web server functions.

[0050] The main control unit 3 controls each part of the operator support system, generates registration information for each item of the customer interaction record from the voice of the call between the customer and the operator, and performs the necessary processing to register the registration information in the database of the management unit 6.

[0051] The GUI control unit 7 generates display information (e.g., HTML data) for various screens to be displayed on the operator terminal 8, and displays the various screens on the operator terminal 8. The GUI control unit 7 also acquires operation information corresponding to an operator's operation performed on the operator terminal 8. The GUI control unit 7 may be realized by, for example, a CRM tool or an incident management tool, which is a software package including a web server function, or may be realized by a cloud service including a web server function.

[0052] The operator terminal 8 is configured as a PC, a smartphone, a tablet terminal, etc. The operator terminal 8 includes a display device 11 such as a liquid crystal panel, an input device 12 such as a touch panel, a keyboard, and a mouse, a microphone 13, and a speaker 14. The microphone 13 and the speaker 14 may be headsets.

[0053] At the operator terminal 8, the operator can make a call with the customer using a microphone 13 and a speaker 14. At the operator terminal 8, various screens to be viewed by the operator are displayed on the display device 11 using an application such as a browser, based on display information transmitted from the GUI control unit 7. The operator terminal 8 also detects screen operations performed by the operator using the input device 12, and transmits the operation information to the GUI control unit 7.

[0054] Next, a description will be given of the general configuration of the main control unit 3 and the GUI control unit 7 according to this embodiment. Fig. 2 is a block diagram showing the general configuration of the main control unit 3 and the GUI control unit 7.

[0055] The main control unit 3 includes a communication unit 21, a storage unit 22, and a processor 23.

[0056] The communication unit 21 communicates with the voice interception unit 2, the voice recognition unit 4, the text analysis unit 5, the management unit 6, and the GUI control unit 7.

[0057] The storage unit 22 stores programs executed by the processor 23 and the like.

[0058] The processor 23 performs various processes by executing programs stored in the storage unit 22. In this embodiment, the processor 23 performs processes such as voice recognition control, item extraction control, input state determination, erroneous input determination, item extraction result reflection, correction history management, selection of items where errors frequently occur, and screen control.

[0059] In the voice recognition control process, the processor 23 instructs the voice recognition unit 4 to execute a voice recognition process for converting the voice of a call between a customer and an operator into text by voice recognition, and obtains the voice recognition result.

[0060] In the item extraction control process, the processor 23 instructs the text analysis unit 5 to execute an item extraction process that uses a generation AI to extract a draft (original plan) of registration information for each item of the customer interaction record based on the speech recognition results, and obtains the item extraction results.

[0061] In the input state determination process, the processor 23 determines the state of manual input performed by the operator using the input device 12 for the input field 52 of each item on the registration screen 51 (see FIGS. 4 and 5). Specifically, first, it is determined whether or not the operator has already performed manual input. If the operator has already performed manual input, it is then determined whether or not the operator has completed manual input (input completion determination process). This distinguishes between a state in which the operator has not performed manual input (uninput state), a state in which the operator has not completed manual input (input incomplete state), and a state in which the operator has completed manual input (input completed state).

[0062] In the process of determining whether an operator has manually input an incorrect character, the processor 23 determines whether the operator has manually input an incorrect character. Specifically, if the characters manually input by the operator are meaningless and are not a word or a sentence, the operator's manual input is determined to be an incorrect input.

[0063] In the process of reflecting the item extraction results, the processor 23 reflects the item extraction results output from the text analysis unit 5 on the registration screen 51 (see FIGS. 4 and 5) displayed on the operator terminal 8. In particular, the manner in which the item extraction results are reflected on the registration screen 51 is changed depending on the state of manual input by the operator.

[0064] In the process of managing the revision history, the processor 23 stores, as a revision history, cases in which the item extraction results reflected on the registration screen 51 (see FIGS. 4 and 5) have been manually revised by the operator in the storage unit 22. The revision history may also be stored in the management unit 6.

[0065] In the process of selecting frequently error-prone items, the processor 23 selects items that are likely to cause errors in the item extraction results based on the correction history. Specifically, items whose item extraction results include words that are likely to cause errors in speech recognition, such as proper nouns such as the name of a customer's representative, are selected as items that are likely to cause errors in the item extraction results. Note that items that are likely to cause errors in the item extraction results are highlighted on the registration screen 61 by the GUI control unit 7.

[0066] In the screen control process, the processor 23 instructs the GUI control unit 7 to display on the operator terminal 8 a registration screen 51 (see FIGS. 4 and 5) on which the operator performs a registration operation for each item of the customer interaction record. Note that a screen that presents the operator with the voice recognition results of the call voice during the call, or a screen that allows the operator to confirm the voice recognition results after the call ends, may also be displayed. Furthermore, when the operator performs an operation to instruct execution of the item extraction process after confirming the voice recognition results, the screen may transition to a registration screen 61.

[0067] The GUI control unit 7 includes a communication unit 31, a storage unit 32, and a processor 33.

[0068] The communication unit 31 communicates with the main control unit 3 .

[0069] The storage unit 32 stores programs executed by the processor 33 and the like.

[0070] The processor 33 performs various processes by executing programs stored in the storage unit 32. In this embodiment, the processor 33 performs processes such as generating screen information and acquiring operation information.

[0071] In the screen information generation process, the processor 33 generates display information for a screen to be displayed on the operator terminal 8. In this embodiment, display information for the registration screen 51 (see FIGS. 4 and 5) is generated.

[0072] In the operation information acquisition process, the processor 33 acquires operation information generated by detecting an operation by the operator at the operator terminal 8.

[0073] Next, a description will be given of the processing for reflecting the item extraction results, which is carried out by the main control unit 3 according to this embodiment. Fig. 3 is an explanatory diagram showing an overview of the processing for reflecting the item extraction results.

[0074] In this embodiment, for the input field 52 of each item on the registration screen 51 (see FIGS. 4 and 5), a no-input state, an incomplete input state, and an input-complete state are determined as states of manual input performed by the operator using the input device 12. Also, in this embodiment, an incorrect input state is determined for the input field 52 of each item.

[0075] In an uninputted state, i.e., when no manual input has been made by the operator, there is no string of characters manually input by the operator (manual input result) in the input field 52, and the item extraction results output from the text analysis unit 5 are automatically reflected in the input field 52.

[0076] In an incomplete input state, that is, in a state where the operator has already performed manual input but has not yet completed the manual input, the manual input results remain in the input field 52, and the item extraction results are automatically reflected in a manner that adds them to the manual input results. In this case, since confirmation by the operator is required, both the manual input results and the item extraction results are displayed in the input field 52, and the operator is prompted to make manual corrections.

[0077] In the input completion state, that is, when manual input by the operator is complete, the manual input results remain in the input field 52, and the item extraction results are not automatically reflected. In this case, there is no need to reflect the item extraction results, and no confirmation by the operator is required, so the item extraction results are not reflected.

[0078] In the case of an incorrect input, that is, in the case where the operator has made an incorrect manual input, the item extraction results are automatically reflected in the input field 52, and the manual input results in the input field 52 are overwritten and erased. In this case, there is no need to leave the manual input results, and no confirmation from the operator is required, so they are forcibly overwritten and erased.

[0079] Next, a description will be given of the registration screen 51 displayed on the operator terminal 8 according to this embodiment.

[0080] 4(A) shows the initial state of the registration screen 51. The registration screen 51 has input fields 52 for the following items: customer number, customer name, person in charge, reception category, causing device, inquiry content, failure content, and lamp status. In the input field 52 for each item, a character string is entered as registration information for each item in the customer interaction record.

[0081] The example shown in FIG. 4(B) is a case where an operator manually inputs data before the item extraction results are output from the text analysis unit 5.

[0082] In the input fields 52 for each of the items of person in charge, inquiry content, and problem content, characters are manually entered by the operator using the input device 12. In the input field 52 for the person in charge item, the operator enters the name of the person in charge based on what the customer said. In the input field 52 for the inquiry content item, the operator starts to enter characters, but is unable to enter the appropriate content immediately, so the characters entered are incomplete. In the input field 52 for the problem content item, meaningless characters have been entered due to an incorrect input by the operator.

[0083] Furthermore, on the registration screen 51, input fields 52 for items that are likely to produce errors in the item extraction results are highlighted. In the example shown in FIG. 4(B), a caution mark 54 is displayed in the input field 52 for the person in charge field. This allows the operator to recognize that errors are likely to occur in the item extraction results for the person in charge field, and enter the name of the person in charge based on the content of the customer's utterance without waiting for the item extraction results to be reflected. Furthermore, on the registration screen 51, a bold frame is displayed around input fields 52 that are selected by the operator's operation.

[0084] The example shown in FIG. 5(A) shows a state in which, after manual input by an operator, the text analysis unit 5 outputs an item extraction result, and the item extraction result is reflected on the registration screen 51.

[0085] The customer number and customer name determined from the customer's telephone number are entered in the input fields 52 for the customer number and customer name. When an inquiry call is received from a customer, the main control unit 3 determines the customer number and customer name based on the telephone number by referring to the customer database.

[0086] In the input field 52 for the person in charge item, manual input by the operator has been completed (input completed state). In this case, even if the item extraction result is output from the text analysis unit 5, the item extraction result is not reflected on the registration screen 51, and the state in which the manually input character string (manual input result) is displayed in the input field 52 is maintained.

[0087] The operator has not yet manually entered any information into the input fields 52 for the reception category, the causing device, and the lamp status (blank state). In this case, the item extraction results are reflected on the registration screen 51, and the character strings of the item extraction results are displayed in the input fields 52.

[0088] In the input field 52 for the inquiry content item, a character string has been manually entered by the operator, but because the character string does not match the attribute of the item, the operator's manual input is determined to be incomplete (input incomplete state). Specifically, in the inquiry content item, the character string must be entered as a sentence, but only the word "signage" has been entered. In this case, since it is expected that the operator will edit the character string, the character string manually entered by the operator (manual input result) remains, and the item extraction result is reflected on the registration screen 51. Specifically, in the input field 52, the character string of the item extraction result is added after the character string of the manual input result.

[0089] In the input field 52 for the fault content item, characters have already been manually entered by the operator, but because the characters are meaningless and not a word or a sentence, the operator's manual input is determined to be an incorrect input (incorrect input state). Specifically, the meaningless character "a" has been entered in the input field 52 (see FIG. 4(B)). In this case, the characters manually entered by the operator (manual input result) are discarded, and the item extraction result is reflected in the registration screen 51. Specifically, in the input field 52, the character string of the item extraction result is reflected in a state in which it overwrites the characters of the manual input result.

[0090] Furthermore, in the input field 52 for each item, the displayed character string is displayed in a different manner depending on whether the item is in an unconfirmed state or a confirmed state.

[0091] In the example shown in Figure 5(A), the character strings in the input fields 52 for the customer number and customer name are confirmed and do not need to be corrected, so they are displayed in a standard font. Also, the character strings in the input field 52 for the person in charge are confirmed and do not need to be corrected, so they are displayed in a standard font.

[0092] On the other hand, in the input fields 52 for the reception category, the causative device, the fault content, and the lamp status, the character strings of the item extraction results are in an unconfirmed state, and the character strings are displayed in a non-standard format to prompt the operator to check whether there are any errors in the item extraction results. In the example shown in Fig. 5(A), the character strings of the item extraction results are displayed in the input field 52 in a gray, italicized font. Alternatively, for example, the character strings may be displayed in a font smaller than the standard.

[0093] In the input field 52 for the inquiry item, the character string manually entered by the operator (manual input result) is displayed in a confirmed state, and the character string of the item extraction result is displayed in an unconfirmed state. Specifically, the character string of the manual input result is displayed in a standard font, and the character string of the item extraction result is displayed in a gray, italicized font.

[0094] The example shown in FIG. 5(B) is a case where, after the item extraction results output from the text analysis unit 5 are reflected on the registration screen 51, the character string is corrected by manual input by the operator.

[0095] In the input field 52 for the inquiry item, as shown in FIG. 5(A), when the item extraction result is reflected in the input field 52, the character string of the item extraction result is added after the character string of the manual input result, but in the example shown in FIG. 5(B), the character string displayed in the input field 52 has been corrected by the operator. Specifically, the character string "signage" has been deleted by the operator's operation. Furthermore, since the character string has been corrected by manual input by the operator in the input field 52 for the inquiry item, the character string displayed in the input field 52 changes to a confirmed state, and the corrected character string is displayed in a standard font.

[0096] Furthermore, in this embodiment, when the operator performs an operation to select the input field 52 for each item (for example, a tap operation or a click operation), the character string displayed in the input field 52 in an unconfirmed state changes to a confirmed state. In the example shown in FIG. 5(B), the character strings are displayed in an unconfirmed state in the input fields 52 for each of the items, namely, reception category, causative device, fault content, and lamp status, but when the operator performs an operation to select the input field 52 for each item, the character strings displayed in the input field 52 change to a confirmed state. This enables the operator to reliably check whether there are any errors in the item extraction results.

[0097] Furthermore, the registration screen 51 is provided with a registration button 53. When the operator operates the registration button 53, a registration process is performed in which the character strings displayed in the input fields 52 for each item are registered in the database of the management unit 6 as registration information for each item of the customer interaction record. At this time, an autocorrect function may be used to correct typos, etc. Furthermore, when the registration button 53 is operated while the character strings displayed in the input fields 52 are still unconfirmed, a message may be displayed on the screen urging the operator to confirm the character strings in the input fields 52.

[0098] In this embodiment, the registration screen 51 is displayed on the operator terminal 8, but a registration confirmation screen that allows the operator to confirm the registration contents may be displayed after the registration screen 51. That is, when the operator operates the registration button 53 on the registration screen 51, the screen may transition to the registration confirmation screen, and when the operator operates the registration confirmation screen to approve the registration contents, the registration process may be performed.

[0099] Incidentally, when the call unit 1 detects an incoming call from a customer making an inquiry, a contact ID is assigned to the incoming call, and the contact ID, the customer's telephone number, and the operator ID are registered in the database of the management unit 6. At this time, a new customer interaction record (ticket) may be registered.

[0100] In this case, items other than the contact ID, telephone number, and operator ID may be left blank. Also, items related to the contact ID, telephone number, and operator ID items may be registered. For example, if there is a separate database for managing customer information, the customer's telephone number is associated with the customer number and the store name. Also, if there is a separate database for managing operators, the operator ID is associated with the operator's name. In other words, taking the registration screen 51 (see FIG. 4) as an example, the items related to the contact ID, telephone number, and operator ID items are the customer number item, the customer name (store name), and the person in charge (operator name).

[0101] Furthermore, when a new customer interaction record (ticket) is registered in the management unit 6, the operator may be notified of this. Specifically, when the main control unit 3 receives a response from the management unit 6 indicating that the new customer interaction record has been successfully registered, the main control unit 3 may request external software to send a notification indicating that the new customer interaction record has been registered to the operator's contact information (operator terminal 8) associated with the operator ID. Examples of this external software include email and communication tools with business chat functions.

[0102] The notification that a new customer interaction record has been registered may also include access information for displaying a customer interaction record viewing screen on the operator terminal 8. An example of the access information is a URL. The operator can view the contents of the newly registered customer interaction record by performing an operation to open the customer interaction record viewing screen on the operator terminal 8.

[0103] Furthermore, the operator can manually input text on the customer interaction record viewing screen using the input device 12. This allows the operator to appropriately input text that they want to note down during a call with a customer.

[0104] Furthermore, the screen for viewing the customer interaction record may be, for example, a registration screen 51 (see FIG. 4(A)). In this case, the registration screen 51 is displayed when the call starts. On the registration screen 51, the operator can input text of the content that the operator wants to note down as a memo in the input field 52 for the relevant item. Furthermore, on the registration screen 51, the item extraction results are reflected after the call ends. At this time, the operator can correct the item extraction results while checking the text that was input as a memo. In this way, the operator can input memos during the call and correct the item extraction results after the call ends, all on the same screen. This reduces the effort required to switch screens, thereby improving convenience for the operator.

[0105] Next, we will explain a first example of the input completion determination process performed by the main control unit 3. In the input completion determination process, when the operator has already performed manual input, it is determined whether the operator's manual input has been completed, that is, whether the input is in an incomplete state or an input completed state.

[0106] In this example, predetermined attributes are set for each item in the customer interaction record, and the input completion determination process is performed based on whether the character string (manual input result) manually entered by the operator into the input field 52 on the registration screen 51 matches the attributes for each item (compatibility determination process).

[0107] First, for a specific item, if the character string to be registered is equal to or exceeds a predetermined minimum number of characters, the minimum number of characters is set as an attribute of that item. In the conformance determination process, the conformance of the manual input result is determined depending on whether the manual input result is equal to or exceeds the minimum number of characters. For example, for the "inquiry content" item, the minimum number of characters is set to 10, and it is determined whether the manual input result is equal to or exceeds 10 characters. Here, if the manual input result is equal to or exceeds 10 characters, it is determined that the manual input by the operator is complete, and if the manual input result is less than 10 characters, it is determined that the manual input by the operator is incomplete.

[0108] Furthermore, for a specific item, if the character string to be registered is written using a pattern of a predetermined combination of character types, the pattern of the character type combination (notation rule) is set as an attribute of that item. In the compatibility determination process, it is determined whether the manually input result matches the pattern of the character type combination. For example, in the "Model Number" item, if the model number is set based on the notation rule of "two uppercase alphabets + numbers," a pattern using regular expressions is set as follows, and it is determined whether the manually input result matches this predetermined pattern using regular expressions. "Pattern=[AZ][AZ]\d * " Here, if the manual input result matches a predetermined pattern, the manual input by the operator is determined to be complete, and if the manual input result does not match the predetermined pattern, the manual input by the operator is determined to be incomplete.

[0109] Furthermore, for a specific item, if the character string to be registered contains a predetermined type of character, the inclusion of the predetermined type of character is set as an attribute of the item. In the compatibility determination process, it is determined whether or not the manual input result contains the predetermined type of character. For example, for the item "lamp status," it is determined whether or not the manual input result contains a character representing a color. Here, if the manual input result contains a character representing a color, it is determined that the manual input by the operator is complete, and if the manual input result does not contain a character representing a color, it is determined that the manual input by the operator is incomplete.

[0110] Next, a second example of the input completion determination process performed by the main control unit 3 according to this embodiment will be described. Fig. 6 is an explanatory diagram showing an overview of the second example of the input completion determination process.

[0111] In this example, the input completion determination process is performed based on the similarity between the character string (manual input result) manually input by the operator into the input field 52 of the registration screen 51 and the character string (item extraction result) acquired by the generation AI in the item extraction process (similarity determination process). That is, if the manual input result is similar to the item extraction result, the manual input by the operator is determined to be complete, and if the manual input result is not similar to the item extraction result, the manual input by the operator is determined to be incomplete.

[0112] In the similarity determination process, for example, first, the manual input result and the item extraction result are decomposed into words by morphological analysis. Next, each of the manual input result and the item extraction result is vectorized based on the word appearance status. Here, for example, a word that appears in the sentence is assigned a value of 1, and a word that does not appear in the sentence is assigned a value of 0. Note that the tf-idf (term frequency-inverse document frequency) method may also be used. Next, cosine similarity is calculated based on the vectorization results for each of the manual input result and the item extraction result.

[0113] Next, by comparing the cosine similarity with a predetermined threshold (e.g., 0.5), it is determined whether the manual input result is similar to the item extraction result, and depending on this determination result, it is determined whether the operator's manual input has been completed. That is, if the cosine similarity is equal to or greater than the threshold, it is determined that the manual input result is similar to the item extraction result, and the operator's manual input is completed. On the other hand, if the cosine similarity is less than the threshold, it is determined that the manual input result is not similar to the item extraction result, and the operator's manual input is incomplete.

[0114] Here, if the threshold value for cosine similarity is set to, for example, 0.5, then in the example shown in Fig. 6, the cosine similarity is 0.408, which means that the similarity between the manual input result and the item extraction result is low, and it is determined that the manual input by the operator is incomplete. In this case, the item extraction result is reflected on the registration screen 51 (see Fig. 5). Specifically, the character string "The signage screen is not displayed" which is the manual input result, is followed by the character string "The signage screen is not displayed and the screen remains blue", and the character string "The signage screen is not displayed and the screen remains blue" is displayed in the input field 52 on the registration screen 51.

[0115] Next, a description will be given of a third example of the input completion determination process performed by the main control unit 3. Fig. 7 is an explanatory diagram showing an example of a prompt and its answer input to the text analysis unit 5.

[0116] In this example, the input completion determination process is performed using a generation AI in the text analysis unit 5. That is, in the text analysis unit 5, in addition to a process of extracting a draft (original plan) of registration information for each item of the customer interaction record (item extraction process), a process of determining whether or not the operator has completed manual input for each item of the customer interaction record (input completion determination process) is performed.

[0117] In this case, the main control unit 3 instructs the text analysis unit 5 to execute an input completion determination process. At this time, for example, a prompt (instruction sentence) as shown in Fig. 7(A) is generated by the main control unit 3, the prompt is input to the text analysis unit 5, and the answer (determination result) shown in Fig. 7(B) is output from the text analysis unit 5.

[0118] In the example shown in Figure 7(A), a string of characters (manually entered result) manually entered by the operator in input field 52 of a specific item (in this example, the result summary item) on registration screen 51 and a voice recognition result regarding the voice of the conversation between the customer and the operator are added to the prompt, and based on the voice recognition result, it is determined whether the manually entered result is valid as registration information for the specific item.

[0119] The answer (determination result) output from the text analysis unit 5 can be either "Yes" or "No." If the answer is "Yes," it is determined that the operator's manual input has been completed. On the other hand, if the answer is "No," it is determined that the operator's manual input has not been completed.

[0120] (First Modification) Next, a first modified example will be described. Note that the points not specifically mentioned here are the same as those in the above embodiment. Figures 8 and 9 are explanatory diagrams showing a registration screen 61 displayed on the operator terminal 8 according to the first modified example.

[0121] 8 and 9 is displayed on the operator terminal 8. As in the above embodiment, the registration screen 61 is provided with an input field 52 for each item. Also, in this modification, as in the above embodiment, when an item extraction result is output from the text analysis unit 5, the item extraction result is reflected in the input field 52 for each item in an unconfirmed state.

[0122] On the other hand, in this modified example, an assist button 62 is provided on the registration screen 61. The assist button 62 is provided for each of the following items: person in charge, reception category, causative device, inquiry content, fault content, and lamp status. When the operator operates the assist button 62, the item extraction results reflected in the input field 52 in an unconfirmed state change to a confirmed state.

[0123] In the customer number and customer name items, the customer number and customer name determined from the customer's telephone number are reflected, so the assist button 62 is not provided.

[0124] The example shown in Figure 8(A) is the stage before the item extraction results are output from the text analysis unit 5. In this case, since the item extraction results have not been output from the text analysis unit 5, the item extraction results cannot be reflected on the registration screen 61, and the assist button 62 is therefore in an inoperable state. In the example shown in Figure 8(A), the assist button 62 is displayed grayed out.

[0125] Furthermore, the input fields 52 for the customer number and customer name display the customer number and customer name calculated from the customer's telephone number. The input field 52 for the person in charge displays the character string manually entered by the operator during the call. The input fields 52 for the other items are blank.

[0126] The example shown in Fig. 8(B) is a case where an item extraction result is output from the text analysis unit 5. In this case, the item extraction result can be reflected on the registration screen 61, and the assist button 62 changes to an operable state.

[0127] Furthermore, in this embodiment, when the input is complete, that is, when a character string has already been manually entered into the input field 52 by the operator, if the operator performs an operation to select the input field 52, the character string resulting from the item extraction is presented as an input candidate.

[0128] In the example shown in FIG. 8(B), the operator performs an operation to select input field 52 for the person in charge item. In this case, a caret 63 is displayed in input field 52. In addition, a selection box 64 for selecting a character string from the item extraction result as an input candidate is popped up near caret 63. When the operator selects a character string from the item extraction result in selection box 64, the character string from the item extraction result is reflected in input field 52.

[0129] On the other hand, the item extraction results are displayed in an unconfirmed state in the input fields 52 for the reception category, the causing device, the inquiry content, the fault content, and the lamp status. Specifically, the character strings of the item extraction results are displayed in the input fields 52 in a light gray italic font.

[0130] Here, when the operator operates the assist button 62, the character string of the item extraction result displayed in the input field 52 in an unconfirmed state changes to a confirmed state, thereby enabling the operator to reliably confirm whether the item extraction result is appropriate.

[0131] The examples shown in Figures 9(A) and (B) are for a case where the operator selects input field 52 in order to edit the unconfirmed item extraction result displayed in input field 52. In this example, input field 52 for the inquiry item is selected. In this case, as shown in Figure 9(A), first, a caret 63 is displayed at the top of input field 52. In addition, a selection box 64 is popped up near caret 63. In selection box 64, character strings from the item extraction result are displayed as input candidates.

[0132] The selection box 64 may display the entire string of the item extraction result, or may display only a portion of the string of the item extraction result, for example, only the beginning of the string of the item extraction result.

[0133] 9(B) shows an example in which the operator moves caret 63 displayed in input field 52 from the top of input field 52. In this case, in addition to the item extraction results, words and phrases related to the customer's inquiry are displayed as input candidates in selection box 64. For example, words and phrases related to the word where caret 63 is located, words and phrases extracted from the voice recognition results, words and phrases related to the customer, words and phrases related to the types of equipment and systems used by the customer, etc. are displayed in selection box 64.

[0134] In this manner, in this modified example, the character strings of the item extraction results are displayed as input candidates in the selection box 64. Therefore, if the operator has already manually input a character string, and a conflict occurs between the character strings of the manual input and the character strings of the item extraction results, the operator can select either of them, so that the item extraction results can be appropriately reflected in the registration screen 61 while preventing the operator's manual input from being wasted due to overwriting.

[0135] The selection box 64 may be popped up as a predictive conversion function that displays words containing the characters as conversion candidates when the operator manually inputs the characters. In this case, the character strings of the item extraction results are displayed in the selection box 64 together with the predicted conversion candidate words.

[0136] In this modification, an assist button 62 is provided for each item, allowing the operator to instruct the reflection of the item extraction results for each item. Alternatively, an assist button 62 corresponding to all items may be provided, and the item extraction results may be reflected for all items collectively when the operator operates the assist button 62. Alternatively, multiple items may be grouped, an assist button 62 may be provided for each group, and the item extraction results may be reflected for each group.

[0137] Furthermore, in this modification, the display mode of the assist button 62 changes depending on whether the item extraction results can be reflected in the input field 52, i.e., whether the assist button 62 is in an inoperable state or an operable state, and therefore the assist button 62 functions as an assist status display unit that displays whether the item extraction results can be reflected in the input field 52. On the other hand, an assist status display unit may be provided that has the same form as the assist button 62, does not have the function as an operation unit for reflecting the item extraction results in the input field 52, and only has the function of displaying whether the item extraction results can be reflected in the input field 52.

[0138] In the above-described embodiment and first modification, the registration screen 61 is configured to fit a small, portrait-oriented display device 11 when the operator terminal 8 is a smartphone. Furthermore, in the present embodiment and first modification, the burden on the operator's input operations can be reduced by utilizing AI, making it possible to register customer interaction records even on a small display device 11, such as a smartphone. This reduces the cost of the operator terminal 8 provided to the operator when having the operator perform operations at a home-based call center, thereby reducing the system's operating costs. Furthermore, when the operator terminal 8 is a smartphone, it becomes easy to use voice input for character input, thereby reducing the burden on the operator's input operations.

[0139] (Second Modification) Next, a second modified example will be described. Note that points not particularly mentioned here are the same as those in the above embodiment. Fig. 10 is an explanatory diagram showing a registration screen 71 displayed on an operator terminal 8 according to the second modified example. Fig. 11 is an explanatory diagram showing the transition status of the main parts of the registration screen 71.

[0140] In this modified example, a registration screen 71 shown in Fig. 10 is displayed on the operator terminal 8. The registration screen 71 has an input field 52 for each item, as in the above embodiment, but in this modified example, a display field 72 for displaying the item extraction results is provided next to the input field 52, in addition to the input field 52.

[0141] 10 shows an example in which the item extraction results are output from the text analysis unit 5 and reflected in the display field 72. Before the item extraction results are output, the display fields 72 for the following items are blank: person in charge, reception category, device causing the problem, inquiry content, fault content, and lamp status. On the other hand, the customer number and customer name are obtained based on the customer's telephone number, so the customer number and customer name are displayed in the display field 72.

[0142] Furthermore, in the above embodiment, the state in which the item extraction results are reflected in the input field 52 differs depending on the state of manual input by the operator (input complete state, input incomplete state), but in this modified example, the item extraction results are reflected in the display field 72 for each item regardless of the state of manual input by the operator. Furthermore, while the operator can manually input character strings in the input field 52, the display field 72 only displays the item extraction results and does not allow the operator to manually edit the character strings displayed in the display field 72.

[0143] In this modified example, the input field 52 and the display field 72 are arranged side by side for each item, and the item extraction results are displayed in the display field 72, so that the operator can manually input strings into the input field 52 while referring to the item extraction results displayed in the display field 72.

[0144] Furthermore, the registration screen 71 is provided with an assist button 73. By operating the assist button 73, the operator can reflect the item extraction results displayed in the display field 72 in the input field 52. This allows the operator to select whether or not to reflect the item extraction results displayed in the display field 72 in the input field 52. Note that in the above embodiment and the first modified example, the item extraction results are reflected in the input field 52 in an unconfirmed state, but in this modified example, the item extraction results are reflected in the input field 52 in a confirmed state.

[0145] Here, when an operator manually inputs an appropriate character string into input field 52 during a call and then the item extraction result is displayed in display field 72, if the item extraction result is more appropriate than the manual input result, the operator can operate assist button 73 to overwrite the manual input result entirely in input field 52 with the item extraction result.

[0146] On the other hand, if there are appropriate words or expressions in part of the item extraction results, the effort of inputting them can be reduced by quoting part of the item extraction results. In this modification, the operator can select part of the character string of the item extraction results in the display field 72, and the item extraction results can be partially reflected in the input field 52.

[0147] Specifically, when the operator selects a portion of the character string resulting from the item extraction in the display field 72 with a caret displayed in the input field 52 and then operates the assist button 73, the character string selected in the display field 72 is inserted at the position of the caret in the input field 52 (insertion operation). Also, when the operator selects a portion of the character string resulting from the manual input in the input field 52 and also selects a portion of the character string resulting from the item extraction in the display field 72 and then operates the assist button 73, the character string selected in the input field 52 is replaced with the character string selected in the display field 72 (replacement operation).

[0148] The example shown in Fig. 11 is a case of a replacement operation. As shown in Fig. 11(A), a character string manually entered by the operator is displayed in input field 52. A character string resulting from item extraction is displayed in display field 72. As shown in Fig. 11(B), the operator selects a portion of the character string manually entered in input field 52 and a portion of the character string resulting from item extraction in display field 72, and then operates assist button 73. As a result, the character string selected in input field 52 is replaced with the character string selected in display field 72, as shown in Fig. 11(C).

[0149] In this modified example, the item extraction results are partially reflected in the input field 52 in response to the operator's operation of selecting a portion of the item extraction results in the display field 72, thereby eliminating the need for the operator to manually enter the information.

[0150] (Third Modification) Next, a third modified example will be described. Note that points not particularly mentioned here are the same as those in the above embodiment. Fig. 12 is an explanatory diagram showing an example of a table of correction history stored in the memory unit 22 of the main control unit 3 according to the third modified example. Fig. 13 is an explanatory diagram showing a registration screen 61 displayed on the operator terminal 8 according to the third modified example.

[0151] If there is an error in the item extraction results output from the text analysis unit 5, the item extraction results are corrected by the operator. In this modification, cases in which the item extraction results have been corrected by the operator in this way are accumulated as a correction history in the storage unit 22 of the main control unit 3. Specifically, as shown in FIG. 12 , the item extraction results that were the subject of correction by the operator and the correction results manually input by the operator are registered in a correction history table in association with each other. The correction history may also be accumulated in the management unit 6.

[0152] In this modified example, as shown in FIG. 13, in response to the operator's operation to select a phrase from the item extraction results in the input field 52 of the registration screen 61, if the selected phrase matches a phrase that has been corrected in the past regarding the item extraction results based on the correction history, the correction result at that time is displayed on the registration screen 61 as an input candidate.

[0153] The example shown in Figure 13(A) is the state immediately after the item extraction results output from the text analysis unit 5 are reflected on the registration screen 61. The example shown in Figure 13(B) is a case where the operator determines that there is an error in some of the item extraction results displayed in the input fields 52 for each item and that correction is required, and performs an operation to select the input field 52 that requires correction. In the example shown in Figure 13(B), the input field 52 for the inquiry content item is selected. When a selection operation is performed on the input field 52, the input field 52 transitions to an editable state, and a caret 63 is displayed in the input field 52.

[0154] Here, when the operator points the caret 63 at a word that needs to be corrected, if the word where the caret 63 is located is registered in the correction history, that is, if it matches a word in an item extraction result that was corrected by the operator in the past, a selection box 64 is displayed near the caret 63. In the selection box 64, based on the correction history, words that were manually input by the operator when the item extraction result was corrected in the past are displayed as input candidates.

[0155] Specifically, in the input field 52 for the inquiry item, the phrase "reference product" included in the reflected item extraction result was previously corrected to "3 products", and therefore the phrase "3 products" is displayed as an input candidate in the selection box 64. When the operator performs an operation to select the phrase "3 products" in the selection box 64, the phrase "reference product" displayed in the input field 52 is replaced with the phrase "3 products".

[0156] It should be noted that a plurality of words and phrases resulting from manual correction by the operator may be displayed as input candidates in the selection box 64 for the word and phrase resulting from the item extraction where the caret 63 is located.

[0157] In this way, in this modified example, if an error similar to that in the past occurs in the item extraction results, the operator can correct it simply by selecting an input candidate, thereby reducing the operational burden on the operator.

[0158] In the example shown in Figure 13(B), an input field 52 and an assist button 62 are provided for each item, as in the first modified example (see Figures 8 and 9), but this modified example can also be applied in the same way when an input field 52, a display field 72, and an assist button 73 are provided for each item, as in the second modified example (see Figure 10).

[0159] As described above, the embodiments have been described as examples of the technology disclosed in this application. However, the technology in this disclosure is not limited to these, and can be applied to embodiments in which modifications, substitutions, additions, omissions, etc. are made. Furthermore, it is also possible to combine the components described in the above embodiments to create new embodiments. [Industrial Applicability]

[0160] The operator support system, operator support device, and operator support method according to the present invention utilize speech recognition technology and generation AI to extract registration information for each item in a customer interaction record from the speech recognition results regarding the conversation between a customer and an operator, and when the item extraction results are reflected on a registration screen, they have the effect of being able to appropriately reflect the item extraction results on the registration screen while avoiding the operator's manual input being wasted by overwriting, and are useful as an operator support system, operator support device, and operator support method that supports the work of an operator who creates a customer interaction record based on the details of an inquiry made by a customer over the phone. [Explanation of symbols]

[0161] 1: Call section 2: Audio interception unit 3: Main control unit (operator support device) 4: Voice recognition section 5: Text analysis section 6: Management Department 7: GUI control section 8: Operator terminal 9: Telephone 11: Display device 12: Input device 13: Microphone 14: Speaker 23: Processor 33: Processor 51: Registration screen 52: Input field 53:Registration button 61: Registration screen 62: Assist button 63: Callet 64: Selection box 71: Registration screen 72:Display field 73: Assist button

Claims

1. In an operator support system, one or more processors execute a process for supporting an operator's work of creating a customer interaction record based on the content of a customer's telephone inquiry, The processor: Instruct the speech recognition unit to execute a speech recognition process to convert the speech between the customer and the operator into text by speech recognition, and obtain the speech recognition result. instructing a text analysis unit to execute an item extraction process for extracting a draft of registration information for each item of the customer interaction record from the speech recognition result; displaying on an operator terminal a registration screen for an operator to perform a registration operation for each item of the customer interaction record; When the item extraction results are obtained from the text analysis unit, the operator support system determines the state of manual input by the operator into the input fields for each item on the registration screen, and changes the manner in which the item extraction results are reflected in the input fields depending on the state of manual input.

2. The processor: If the operator has not yet manually entered any information, the item extraction results are reflected in the input field. If the operator has already manually input the data, it is determined whether the operator has completed the manual input; In the case of an input completion state where manual input by an operator has been completed, the item extraction results are not reflected in the input field, 2. The operator support system according to claim 1, wherein, when the operator has not yet completed manual input, the item extraction results are reflected in the input fields.

3. The processor: When the item extraction result is acquired from the text analysis unit, the item extraction result is reflected in the input field in an unconfirmed state; 2. The operator support system according to claim 1, wherein the item extraction result of the input field is changed to a confirmed state in response to an operator's operation to select the input field.

4. The processor: When the item extraction result is acquired from the text analysis unit, the item extraction result is reflected in the input field in an unconfirmed state, and a button for approving the item extraction result is displayed on the registration screen in an operable manner; 2. The operator support system according to claim 1, wherein the item extraction results in the input fields are changed to a confirmed state in response to an operator's operation on the button.

5. The processor:

5. The operator support system according to claim 3, wherein the item extraction results are displayed in the input field in different modes depending on whether the item is in an unconfirmed state or a confirmed state.

6. The processor: a display field for displaying the item extraction results arranged for each item on the registration screen, separate from the input field, and a button for approving the item extraction results is displayed on the registration screen; When the item extraction result is obtained from the text analysis unit, the item extraction result is displayed in the display field; 2. The operator support system according to claim 1, wherein the item extraction results displayed in the display field are reflected in the input field in response to an operator's operation of the button.

7. The processor: In the case of an input completion state where manual input by an operator has been completed, the item extraction result is not reflected in the input field, When an operator performs an operation to select the input field, the item extraction results are displayed as input candidates in a pop-up on the registration screen, 2. The operator support system according to claim 1, wherein the item extraction result is reflected in the input field in response to an operator's operation to select the item extraction result.

8. The processor: The operator support system according to claim 1, characterized in that, in response to an operator's operation to select the input field, if a caret is positioned at the beginning of the input field, the item extraction results are displayed as input candidates in a pop-up on the registration screen, and if a caret is not positioned at the beginning of the input field, in addition to the item extraction results, words related to the customer's inquiry are displayed as input candidates in a pop-up on the registration screen.

9. The processor: a display field for displaying the item extraction results, arranged by item, on the registration screen, separate from the input field; 2. The operator support system according to claim 1, wherein after a portion of the item extraction results displayed in the display field is selected, only the selected portion of the item extraction results is reflected in the input field in response to an operator's operation to instruct reflection of the item extraction results.

10. The processor: A correction result manually input by an operator to the item extraction result is stored in a storage unit as a correction history; The operator support system according to claim 1, characterized in that, in response to an operator's operation of selecting a phrase from the item extraction result in the input field, if the selected phrase matches a phrase that was previously corrected in the item extraction result based on the correction history, the correction result at that time is displayed on the registration screen as an input candidate.

11. The processor: A correction result manually input by an operator to the item extraction result is stored in a storage unit as a correction history; selecting items that are likely to cause errors in the item extraction results based on the correction history; 2. The operator support system according to claim 1, wherein the entry fields relating to items that are likely to cause errors in the item extraction results are highlighted on the registration screen.

12. The processor: When the call unit detects an incoming call from a customer making an inquiry, it registers a new customer response record related to the inquiry and notifies the operator terminal of the fact; 2. The operator support system according to claim 1, wherein the registration screen is displayed on the operator terminal in a state in which the operator can manually input data in response to a request from the operator terminal.

13. An operator support device in which a processor executes a process for supporting an operator's work of creating a customer interaction record based on the content of an inquiry made by a customer over the phone, The processor: Instruct the speech recognition unit to execute a speech recognition process to convert the speech between the customer and the operator into text by speech recognition, and obtain the speech recognition result. instructing a text analysis unit to execute an item extraction process for extracting a draft of registration information for each item of the customer interaction record from the speech recognition result; instructing the GUI control unit to display on the operator terminal a registration screen on which the operator performs a registration operation for each item of the customer interaction record; When the item extraction results are obtained from the text analysis unit, the operator support device determines the state of manual input by the operator into the input field for each item on the registration screen, and changes the manner in which the item extraction results are reflected in the input field depending on the state of manual input.

14. 1. An operator support method for causing one or more processors to execute a process for supporting an operator's work of creating a customer interaction record based on the content of an inquiry made by a customer over the phone, comprising: Instruct the speech recognition unit to execute a speech recognition process to convert the speech between the customer and the operator into text by speech recognition, and obtain the speech recognition result. instructing a text analysis unit to execute an item extraction process for extracting a draft of registration information for each item of the customer interaction record from the speech recognition result; displaying on an operator terminal a registration screen for an operator to perform a registration operation for each item of the customer interaction record; When the item extraction results are obtained from the text analysis unit, the operator's manual input status for each item input field on the registration screen is determined, and the manner in which the item extraction results are reflected in the input field is changed depending on the manual input status.

Images