Processing system, processing method and program

Abstract

To provide a processing system, a processing method and a program which can appropriately process sentences such as inquiries.SOLUTION: A processing system comprises: a sentence correction unit which corrects an inputted sentence according to a predetermined rule, and outputs it as a corrected sentence; an emotion determination unit which discriminates emotion from the inputted sentence or the corrected sentence to output it as an emotion discrimination result; and a selection unit which selects a processing method related to the inputted sentence, on the basis of the corrected sentence, the emotion discrimination result, and a predetermined state related to a creation source of the inputted sentence.SELECTED DRAWING: Figure 9

Description

[Technical Field]

[0001] The present disclosure relates to a processing system, a processing method, and a program. [Background technology]

[0002] Patent Document 1 describes a maintenance worker allocation support system as follows. Specifically, the maintenance worker allocation support system described in this document is a system that responds to calls from customers who have signed equipment maintenance contracts. It includes a recording unit and a dispatch candidate selection unit. The recording unit stores emotional information classified as anger, calm, or joy extracted from the content of the call with the customer. The dispatch candidate selection unit calculates a customer characteristic score that quantifies the customer's characteristics based on the emotional information, a dispatcher characteristic score that quantifies each maintenance worker's ability to respond to customers, and an arrival time score that quantifies the time the maintenance worker can arrive at the equipment. The dispatch candidate selection unit also adjusts the importance of at least one of the dispatcher characteristic score or the arrival time score according to the customer characteristic score, and then selects candidates for maintenance workers to be dispatched to perform equipment maintenance work based on the dispatcher characteristic score and the arrival time score. This system also automatically answers calls from users via voice and automatically converts the call content into text to extract emotional information. If the extracted emotional information is negative, the system switches to a human operator response; if it is positive, the system continues the automatic response. When the user's purpose and the content to be dealt with become clear through an automatic or manned response, the dispatch candidate selection unit calculates a dispatch candidate score by adding up the weighted dispatcher characteristic score and the arrival time score, and selects the maintenance worker with the highest dispatch candidate score as the dispatch candidate. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2023-89480 Summary of the Invention [Problem to be solved by the invention]

[0004] In the maintenance worker assignment support system described in Patent Document 1, emotions are extracted based on text data generated from speech. As a result, in some cases, the text may contain emotional words that have no direct connection to the subject matter or the content to be addressed, or may contain technical terms that are not widely understood. In such cases, the accuracy of understanding the subject matter and extracting emotions may deteriorate, resulting in a problem in which the inquiry cannot be handled appropriately.

[0005] An object of the present disclosure is to provide a processing system, a processing method, and a program that solve the above problems. [Means for solving the problem]

[0006] The processing system of the present disclosure includes a sentence correction unit that corrects an input sentence according to predetermined rules and outputs the corrected sentence, an emotion determination unit that determines an emotion from the input sentence or the corrected sentence and outputs the emotion determination result, and a selection unit that selects how to process the input sentence based on the corrected sentence, the emotion determination result, and a predetermined situation related to the source of the input sentence.

[0007] The processing method of the present disclosure includes the steps of correcting an input sentence in accordance with predetermined rules and outputting the corrected sentence, determining an emotion from the input sentence or the corrected sentence and outputting the emotion determination result, and selecting a method of processing the input sentence based on the corrected sentence, the emotion determination result, and a predetermined situation related to the source of the input sentence.

[0008] The program disclosed herein causes a computer to execute the steps of: correcting an input sentence according to predetermined rules and outputting the corrected sentence; determining an emotion from the input sentence or the corrected sentence and outputting the emotion determination result; and selecting a method of processing the input sentence based on the corrected sentence, the emotion determination result, and a predetermined situation related to the source of the input sentence. [Effects of the Invention]

[0009] According to the processing system, processing method, and program of the present disclosure, text such as inquiries can be processed appropriately. [Brief explanation of the drawings]

[0010] [Figure 1] 1 is a block diagram illustrating an example configuration of a processing system according to an embodiment of the present disclosure. [Figure 2] FIG. 2 is a schematic diagram illustrating an emotion determination unit according to an embodiment of the present disclosure. [Figure 3] FIG. 2 is a schematic diagram illustrating an emotion determination unit according to an embodiment of the present disclosure. [Figure 4] FIG. 2 is a schematic diagram for explaining an AI function unit (2) according to an embodiment of the present disclosure. [Figure 5] FIG. 2 is a schematic diagram for explaining an AI function unit (3) according to an embodiment of the present disclosure. [Figure 6] 10 is a flowchart illustrating an example of the operation of the processing system according to an embodiment of the present disclosure. [Figure 7] 1 is a block diagram illustrating an example configuration of a processing system according to an embodiment of the present disclosure. [Figure 8] 1 is a block diagram illustrating an example configuration of a processing system according to an embodiment of the present disclosure. [Figure 9] 1 is a block diagram illustrating an example configuration of a processing system according to an embodiment of the present disclosure. [Figure 10] 10 is a flowchart illustrating an example of the operation of the processing system according to an embodiment of the present disclosure. [Figure 11]FIG. 1 is a block diagram illustrating an example of a schematic configuration of a computer according to an embodiment of the present disclosure. DETAILED DESCRIPTION OF THE INVENTION

[0011] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. In each drawing, the same or corresponding components are designated by the same reference numerals and the description thereof will be omitted as appropriate.

[0012] First Embodiment FIG. 1 is a block diagram illustrating an example configuration of a processing system according to an embodiment of the present disclosure. The processing system 1 illustrated in FIG. 1 includes an inquiry response system 10, an AI (Artificial Intelligence) function unit (1) 20, an AI function unit (2) 30, an AI function unit (3) 40, a search system (1) 50, and a search system (2) 60. The inquiry response system 10, the AI ​​function unit (1) 20, the AI ​​function unit (2) 30, the AI ​​function unit (3) 40, the search system (1) 50, and the search system (2) 60 can be configured using, for example, one or more computers. Note that each computer may be a virtual computer. Also, for example, multiple components such as the AI ​​function unit (1) 20, the AI ​​function unit (2) 30, and the AI ​​function unit (3) 40 may be configured using a common computer. The processing system 1 shown in FIG. 1 is a system for supporting or executing processing of text, such as inquiries, using web forms, email (hereinafter also referred to as email), etc., in, for example, BtoB (Business to Business) hardware, software, etc. support. In this embodiment, a text is composed of one or more sentences and expresses thoughts or emotions. In this embodiment, text data (digital data representing a text) is also simply referred to as text. The processing system 1 according to this embodiment may be any system for supporting or executing processing of text, such as inquiries, and is not limited to the above examples. For example, it may be applied to fields such as maintenance technology, inquiry response, and customer support. In addition, text, such as inquiries, is not limited to those created by inquirers using web forms, email, etc., but may also be text that has been converted into text from the contents of a telephone conversation, etc. In the following, the content of the text may also be referred to as inquiry content.

[0013] The inquiry response system 10 includes, as functional blocks, an inquiry reception unit 11, an inquiry response unit 12, an information display unit 13, and an inquiry history database 14, which are configured, for example, by a combination of hardware and software. The inquiry reception unit 11 has a function for an inquirer to input inquiry details and a function for displaying the input inquiry details on the inquirer's information terminal, such as a personal computer, smartphone, or tablet terminal (not shown). The inquiry response system 10 may also have a function for inputting (receiving) and outputting (sending) information using email or the like. In this embodiment, the text of an inquiry input by an inquirer (or sent by email or the like) is also referred to as the original inquiry text. The inquiry reception unit 11 is connected to an AI function unit (1) 20 and inputs and outputs predetermined information to and from the AI ​​function unit (1) 20. The inquiry response system 10 provides a function as a contact point for inquiries and functions as a support center.

[0014] The inquiry response unit 12 has a function for the inquiry response staff member (hereinafter simply referred to as the staff member) to input the response content on their information terminal, such as a personal computer, smartphone, or tablet terminal (not shown). The inquiry response unit 12 also has a function for displaying the response content input by the staff member on, for example, the inquirer's information terminal. The inquiry response unit 12 is also connected to the response generation unit 42 of the AI ​​function unit (3) 40, and has a function for displaying the response content automatically generated by the response generation unit 42 on, for example, the inquirer's information terminal.

[0015] The information display unit 13 is connected to the staff assignment unit 41 of the AI ​​function unit (3) 40. The information display unit 13 displays the outputs of the AI ​​function units (1) 20, (2) 30, and (3) 40, as well as the original text of the inquiry sentence, on, for example, the information terminal of the staff member. The outputs of the AI ​​function units (1) 20, (2) 30, and (3) 40 are, for example, the revised inquiry sentence (hereinafter also referred to as the revised sentence (or revised sentence)), the emotion determination result, the contract status and usage status of the inquirer, the tag and rank of the inquiry, the name of the staff member handling the inquiry, etc., which will be described in detail later.

[0016] The response history database 14 has a function to store past inquiry response history (including the original text, revised text, and replies for all inquiries received in the past, and the exchanges up to the time the inquiry is closed). The response history database 14 also has a function to store related information for the response history (information related to the inquiry as the output of the AI ​​function units (1) 20, (2) 30, and (3) 40 (emotion determination results, the inquirer's contract status and usage status, inquiry tags and ranks, and the name and attribute information of the person handling the inquiry)). The response history database 14 is also connected to the history acquisition unit 221 in the emotion determination unit 22 of the AI ​​function unit (1) 20, and provides the past original text or revised text of the inquiry of the inquirer whose emotion is being determined.

[0017] The AI ​​function unit (1) 20 is composed of a text correction unit 21, an emotion discrimination unit 22, and a situation search unit 23, and generates input data for the AI ​​function unit (2) 30. The text correction unit 21 corrects the original inquiry text. When the original inquiry text is input to the text correction unit 21, a text corrected based on rules is output. That is, the text correction unit 21 corrects the input text according to predetermined rules and outputs the corrected text. In addition, the rules can be created using a machine learning model that has undergone deep learning using information representing the content of processing related to past input text as input data. In this embodiment, domain knowledge regarding software and hardware failures is built in through deep learning using past inquiry response history stored in the response history database 14 as input data, enabling text correction based on specialized knowledge. The text correction unit 21 aims to extract the inquirer's requirements and correct inconsistencies in the expression of technical content. To achieve this, it performs the following processes (C1) and (C2).

[0018] (C1) Standardization of words and phrases that express technical content Synonyms and alternative expressions for words and phrases (called "representative words") that express technical content are registered in advance as groups in a table. The registered words and phrases are read from the table and compared to see if there are matching words or phrases in the original query. If a matching word or phrase is found but is not a representative word, the word or expression is replaced with a representative word registered in the table as the same group. Note that the trained machine learning model described above can, for example, input a word or phrase and output which group it belongs to (or does not belong to), and the table can be updated based on the input / output results. Standardizing the words and phrases that express technical content makes it possible to accurately determine, for example, whether an automatic answer is possible, as described below.

[0019] (C2) Elimination of emotional expressions Words and phrases expressing emotions are registered in a table in advance. The registered words and phrases are read from the table and compared to see if there are any matching words or phrases in the original query. If there is a matching word or phrase, that word or phrase is deleted. Here, the words and phrases expressing emotions may be, for example, emotional words and phrases, that is, words and phrases that express a state of losing reason and being overwhelmed by emotion, excitement, etc. By eliminating emotional expressions, it is possible to accurately determine, for example, whether an automatic response, as described below, is possible.

[0020] The emotion determination unit 22 is composed of a history acquisition unit 221 and an emotion analysis unit 222, and determines the emotion of the inquirer in the inquiry from the revised text or original text corrected by the text correction unit 21 and outputs the emotion determination result. In this case, the emotion determination unit 22 can determine the emotion by reflecting the results of comparing the revised text or original text with past revised text or original text created by the same source. The history acquisition unit 221 acquires the revised text or original text of a past inquiry made by the inquirer from the response history database 14. The emotion determination unit 22 performs emotion analysis using the revised text or original text of the inquiry and the revised text or original text of the past inquiry acquired by the history acquisition unit 221 as input data, and outputs the emotion determination result. Note that when the emotion determination unit 22 performs emotion analysis using the original text of the inquiry and the original text of the past inquiry acquired by the history acquisition unit as input data and outputs the emotion determination result, emotion determination can be performed based on the text before emotional expressions are removed (the original text or the revised text after process C1). The emotion determination unit 22 does not necessarily have to perform either emotion determination based on the corrected sentence or emotion determination based on the original sentence.

[0021] The emotion determination unit 22 aims to measure the level of the inquirer's anger. In hardware and software inquiries, the unit defines the characteristics that appear in the writing of angry people, as well as the points assigned to each characteristic, such as words, phrases, and symbols (e.g., "?" marks and "!" marks) that fit those characteristics, and registers them in a table in advance. An example of the defined characteristics and points is shown in Figure 2. The emotion determination unit 22 reads words, phrases, symbols, etc. from a pre-created table and compares them with the words, phrases, symbols, etc. used in the revised or original inquiry. If there are any matching words, phrases, symbols, etc., the unit assigns the points defined in the table to each matching element. At this time, the unit refers to past inquiries by the inquirer, and if there are any differences in the characteristics used, the unit assigns points that are, for example, doubled. For example, as shown in Figure 3, for the same feature "Could you please give me ~?" (normally scored as 1 point because it is a question), a person who has used "Could you please give me ~?" in past inquiries will receive different scores for "Could you please give me ~?" in the current inquiry (a new inquiry) than a person who has used "It would be helpful if you could give me ~." (0 point because it does not contain any emotional expression). The use of an unusual phrase is interpreted as indicating a strong feeling of anger, and this is reflected in the score. In this way, points are assigned each time a feature appears, and finally, the total score is calculated by adding up the points assigned to the revised inquiry or the entire original sentence. A total score of 0 points is classified as "calm," 1 or 2 points as "dissatisfied," and 3 or more points as "furious," and the classification result is output as the emotion identification result for the inquiry.

[0022] FIG. 2 is a schematic diagram illustrating an emotion determination unit according to an embodiment of the present disclosure. FIG. 2 shows an example of the relationship between defined features and points. In the example shown in FIG. 2, for example, if a question is used, the point is 1. The higher the point, the greater the degree of emotionality. If the feature is an imperative tone, the point is 3. If any of the following words ("deal," "contract extension," "contract termination," "will consider," "quality," or "doubt") are used as a set, the point is 3. Furthermore, if a deadline is presented, the point is 2. Furthermore, FIG. 3 is a schematic diagram illustrating an emotion determination unit according to an embodiment of the present disclosure. FIG. 3 shows an example of the relationship between past inquiries, the inquiry (the inquiry for which emotion determination is to be performed), and the points.

[0023] The status search unit 23 is composed of a key extraction unit 231 and an automatic search unit 232, and collects the contract information and usage information of the inquirer. The key extraction unit 231 extracts items to be used as search conditions (the name of the company to which the inquirer belongs, the product being inquired about) from the data received by the inquiry reception unit 11. The automatic search unit 232 is connected to the search system (1) 50, and executes a search in the search system (1) 50 based on the information extracted by the key extraction unit 231 and saves the results obtained.

[0024] The search system (1) 50 is composed of a contract information database 51 and a usage information database 52. The contract information database 51 stores product contract information (product name, contract date, contract renewal date, cumulative years of use, contract amount, etc.) and maintenance service contract information (maintenance service name, contract date, contract renewal date, cumulative years of use, contract amount, maintenance service usage record, etc.) for all customers. The usage information database 52 stores data showing the product usage status of all customers (number of licenses, number of accesses, log, and other performance data).

[0025] The AI ​​function unit (2) 30 is composed of a tagging unit 31 and a list creation unit 32. Using the data generated by the AI ​​function unit (1) 20, tags are assigned to queries and queries are listed for each tag of the same type.

[0026] The tagging unit 31 assigns a score to the inquiry based on the data generated by the AI ​​function unit (1) 20, and assigns one of four types of tags depending on the score. The list creation unit 32 sorts the inquiries in order of the scores assigned by the tagging unit 31, and creates a list showing the response priority for inquiries with the same tag.

[0027] The tagging unit 31 assigns a score to the inquiry text based on the revised inquiry text, the result of the inquirer's emotion determination, and the inquirer's contract status and usage status of the product / maintenance service. The urgency and importance of the inquiry are determined from this score, and one of four types of tags is assigned to the inquiry as shown in FIG. 4 (this is the absolute evaluation of the inquiry). FIG. 4 is a schematic diagram for explaining the AI ​​function unit (2) according to an embodiment of the present disclosure. Furthermore, the list creation unit 32 sorts inquiries assigned with the same type of tag in descending order of score and creates a list indicating the response priority within the same tag (this is the relative evaluation of the inquiry). The tagging unit 31 may select tags to assign based on at least one of the contract status and usage status. The meanings of each tag are as follows:

[0028] The "black" tag is given to unreasonable requests (emotional requests that do not constitute a proper support inquiry, etc.) and inquiries that are unlikely to result in new contracts or contract renewals, and is given the lowest priority. The level of the responder and the schedule are also adjusted to lower the priority.

[0029] The "red" tag is given to problems that are deemed to require an immediate response, inquiries from people who are likely to sign new contracts or contract renewals, and legitimate requests that cause great anger on the part of the inquirer. The highest priority is given to responding to these issues. The level and schedule of the responders are also adjusted with the highest priority.

[0030] The "yellow" tag is given to problems or inquiries that require manual response but are of low urgency, and to inquiries from calm individuals. The response priority is medium. The level and schedule of the responders are adjusted with a priority second only to the "red" tag.

[0031] The "green" tag is assigned to inquiries that are anticipated as frequently asked questions. An automatic response is immediately provided by AI. That is, if an automatic response can be provided for the corrected sentence, the "green" tag is assigned, and an automatic response is selected as the processing method. That is, if the assigned tag is the "green" tag, the response generation unit 42 of the AI ​​function unit (3) 40 automatically creates a response to the inquiry.

[0032] The AI ​​function unit (2) 30 assigns tags to inquiries based on the judgment axes shown in Figure 4 and the total points allocated to each judgment axis. In the figure, (1), (2), (3), and (4) are judgment axes, and the order of the numbers indicates the priority of the judgment axes. Judgment axis (1) is know-how for the inquiry content, and there are two patterns: "yes" and "no." Know-how means that there is past data for the same or nearly identical (considered identical) combination of inquiry and answer, and an automatic answer is possible. Judgment axis (2) is the emotion determination result, and there are three patterns: "furious," "dissatisfied," and "calm." Judgment axis (3) is contract status, and there are four patterns: "short time since contract date and no prospect of contract renewal," "long time since contract date but no prospect of contract renewal," "short time since contract date but prospect of contract renewal," and "long time since contract date and prospect of contract renewal." Judgment axis (4) refers to usage status, and there are two patterns: "not much use" and "heavy user." Regarding the judgment axis (2) emotion judgment result and judgment axis (3) contract status and judgment axis (4) usage status, points are assigned for each judgment axis and registered in a table in advance. Tagging of inquiries is performed in the following manner. First, inquiries for which the content of the inquiry is judged as "present" based on judgment axis (1) are classified as green tags. Next, of the remaining inquiries, those for which the emotion judgment result based on judgment axis (2) is "calm" are classified as yellow tags. After that, the judgment results of judgment axis (2) emotion judgment result and judgment axis (3) contract status and judgment axis (4) usage status are compared with the table, and points are assigned according to each judgment result. For each inquiry, the points assigned for the three judgment axes are added up, and those with a high total score are classified as red tags, and those with a low total score are classified as black tags.

[0033] Also, in the example shown in Figure 4, the "black" tag has the lowest priority of the four tags, low urgency, and low importance. The "red" tag has the highest priority of the four tags, high urgency, and high importance. The "yellow" tag has the second highest priority of the four tags, low urgency, and high importance. The "green" tag has the third highest priority of the four tags, low urgency, and no definition of importance is required (no definition is required).

[0034] The AI ​​function unit (3) 40 is connected to the AI ​​function unit (2) 30 and is composed of a staff member assignment unit 41 and a response generation unit 42. The staff member assignment unit 41 classifies inquiries according to the tags assigned by the tag assignment unit 31 of the AI ​​function unit (2) 30, and transmits the inquiry data to the response generation unit 42 or assigns a staff member using an optimization function. The staff member assignment unit 41 is connected to the search system (2) 60, and when assigning a staff member, assigns the staff member most suitable for the inquiry based on the tags assigned by the AI ​​function unit (2) 30 and the list created. The staff member assignment unit 41 transmits the assignment results, the deliverables of the AI ​​function units (1) 20 and (2) 30 (the revised inquiry text and emotion determination results, the inquirer's contract status and usage status, the inquiry tag and rank), and the original inquiry text to the information display unit 13 in the inquiry response system 10.

[0035] The staff member assignment by the staff member assignment unit 41 is performed based on the judgment axis shown in FIG. 5. FIG. 5 is a schematic diagram for explaining the AI ​​function unit (3) according to an embodiment of the present disclosure. The staff member assignment unit 41 assigns a black tag to a complainer staff member, a red tag to a high-level responder, and a yellow tag to a normal responder. The tag to which each staff member is assigned is determined based on the staff member's seniority, skill level, position, and personality assessment results. Basically, staff members with seniority, skill level, and position are assigned the red tag. In addition, staff members who are found to have characteristics such as stress tolerance and an emotion-oriented communication style based on the personality assessment results are assigned the black or red tag. In addition, staff members with attentive listening skills are assigned the black tag.

[0036] The response generation unit 42 automatically creates a response to an inquiry received from the person in charge allocation unit 41 based on previously registered example responses to the inquiry, and sends the response to the inquiry response unit 12 in the inquiry response system 10.

[0037] The search system (2) 60 has a database of staff members 61, which stores information on staff members who handle inquiries (years, skill level, position, personality test results, free time, and other data required for allocating inquiries) and updates it as needed.

[0038] Next, an example of operation of the processing system 1 according to an embodiment of the present disclosure will be described with reference to Fig. 6. Fig. 6 is a flowchart showing an example of operation of the processing system according to an embodiment of the present disclosure. Fig. 6 shows the flow of processing by the processing system 1 in response to one inquiry.

[0039] In the process shown in FIG. 6, first, the inquiry receiving unit 11 receives an inquiry (step S11). Next, the text correcting unit 21 corrects the inquiry sentence (step S12). Next, the emotion determining unit 22 determines the emotion of the inquirer (step S13). Next, the situation searching unit 23 acquires the contract status and usage status data of the inquirer (step S14). Next, the tagging unit 31 assigns a score to the inquiry based on the corrected inquiry sentence, the emotion determining result of the inquirer, and the contract status and usage status of the inquirer's products and maintenance services. The tagging unit 31 determines the urgency and importance of the inquiry from the score and assigns one of four types of tags to the inquiry (step S15). Next, the list creating unit 32 sorts the inquiries assigned with the same type of tag in descending order of score and creates a list indicating the response priority within the same tag (step S16).

[0040] Next, the person in charge allocation unit 41 determines whether the assigned tag is a green tag (step S17), and if it is a green tag (step S17: YES), the answer generation unit 42 automatically generates an answer, and the inquiry answering unit 12 automatically answers (step S18). After that, if there is an additional inquiry to the inquiry receiving unit 11, the process returns to inquiry reception (step S11), and the flow continues until there are no more additional inquiries (until step S19: NO (if automatic answers are possible for all)).

[0041] If the assigned tag is other than the "green" tag (step S17: NO), the agent assignment unit 41 of the AI ​​function unit (3) 40 assigns a response agent according to the type of tag and the order of the list (step S20). The agent refers to the corrected inquiry text, the emotion determination result, the contract status and usage status of the inquirer, the inquiry tag, etc. in the inquiry response system 10, and answers the inquiry (step S21). If there are any additional inquiries after that (step S22: YES), that agent continues to answer (step S21), and the flow ends when there are no more additional inquiries (step S22: NO).

[0042] The specific operation of the present invention will be described below using a hypothetical inquiry as an example. A bug of unknown cause occurred in the software (SW) that a certain company had introduced, causing the work efficiency of 20% of the company's employees to drop by 30% compared to normal times. A person in the information systems department of the company hastily made an inquiry to the inquiry response system 10 requesting that the cause of the bug be investigated and a solution be provided.

[0043] This inquiry was immediately sent to the AI ​​Function Unit (1) 20, which corrected the inquiry, determined the inquirer's emotions, and collected contract and usage data for the inquiring company. The inquirer wrote the inquiry in a state of urgency and impatience, and used many expressions that had not been used in previous inquiries, resulting in an emotion determination of rage. Furthermore, the contract and usage data confirmed that the company had been using the software for quite some time, was an important customer with a large number of contracted licenses and a large contract amount, and that significant abnormalities had been observed in the software.

[0044] The data generated by AI function unit (1) 20 was immediately sent to AI function unit (2) 30, and a red tag was assigned based on the data. In addition, due to the high urgency of the problem and the importance of the company's contract, the inquiry was ranked high among other red-tagged inquiries.

[0045] The information generated by the AI ​​function unit (2) 30 is sent to the AI ​​function unit (3) 40, and the person in charge allocation unit 41 assigns an available expert to the AI ​​function unit (3).

[0046] The person in charge assigned to this inquiry receives a notification from the inquiry response system 10 and refers to the information display unit 13. After reading the revised inquiry text, the person in charge confirms that the inquiry is red-tagged, that the company in question is an important customer, and that the inquirer's emotion is "furious." The person in charge then decides that it would be best to provide a first response as soon as possible, while still empathizing with the inquirer's emotions, even if it is not yet possible to answer all of the inquirer's requests. The person in charge also decides that it would be necessary to take the time to provide a thorough and logical explanation once the cause of the bug and how to deal with it are all clear, and so responds accordingly.

[0047] The person who made the request received a first response 15 minutes after sending it, which conveyed that the situation had been accurately understood, and their sense of urgency and impatience was reduced. They were also able to inform the relevant departments and executives within the company that the issue was currently being handled. After receiving a detailed explanation from the person in charge, they were able to begin considering ways to improve their operations in the future. This concludes the response to the inquiry.

[0048] As described above, according to this embodiment, the efficiency of support responses can be improved. Furthermore, by replacing the understanding of inquiry content and priority determination performed by each individual person with a unified and comprehensive judgment based on data by AI, the time from understanding the inquiry content to responding to the inquiry can be shortened. Furthermore, inquiries that are determined by AI to be able to be automatically answered can be responded to without human intervention, which also leads to improved efficiency in support responses.

[0049] Furthermore, this embodiment contributes to improving the quality and effectiveness of support responses. Because AI determines the response method based on the content of the inquiry and the situation of the inquirer, it is possible to provide the optimal support response for each inquiry from the perspective of business expansion and customer satisfaction. For example, by identifying inquiries that may lead to new contracts or contract renewals and assigning them to highly experienced staff, support responses can be used as an opportunity to increase sales.

[0050] The processing system 1 of this embodiment includes a text correction unit 21, an emotion determination unit 22, and a person in charge allocation unit 41 (one example of the configuration of the "selection unit" according to the present disclosure). The text correction unit 21 corrects the input text according to predetermined rules and outputs the corrected text. The emotion determination unit 22 determines the emotion from the input text or the corrected text and outputs the emotion determination result. The person in charge allocation unit 41 (selection unit) selects a method of processing the input text based on the corrected text, the emotion determination result, and a predetermined situation related to the creator of the input text. Here, the method of processing the input text means, for example, a method of replying to the input text. With this configuration, text such as an inquiry can be appropriately processed.

[0051] The rule for the above-mentioned correction includes deleting predetermined emotional words. This configuration allows the person in charge assignment unit 41 (selection unit) to make a more appropriate selection.

[0052] Furthermore, the rule for the above modification includes replacing a predetermined character string with another predetermined character string. This configuration allows the person in charge assignment unit 41 (selection unit) to make a more appropriate selection.

[0053] Furthermore, the rules for the above corrections can be created using a machine learning model that has undergone deep learning using information representing the content of processing related to past input sentences as input data. With this configuration, the rules to be applied can be updated to more appropriate ones based on the content of past processing, allowing the person in charge assignment unit 41 (selection unit) to make more appropriate selections.

[0054] Furthermore, the emotion determination unit 22 determines emotions by reflecting the results of a comparison between the input or revised text and previous input or revised texts created by the same source. This configuration allows for more accurate emotion determination. The same source includes, for example, the same creator; different creators but the same affiliation, department, workplace, etc.; and different creators but the same organization, such as a corporation or union, for which each creator works. Reflecting the comparison results means, in the above example, changing the determination result to either rage or calm based on the comparison results.

[0055] Furthermore, if an automatic reply can be made to the corrected text, the person in charge assignment unit 41 (selection unit) can select the automatic reply as the processing method.

[0056] Furthermore, when an automatic response cannot be made to the corrected sentence, the person in charge assignment unit 41 (selection unit) can select a response from a person in charge selected based on the emotion determination result and a predetermined situation as a processing method.

[0057] Furthermore, if the input text is a text of an inquiry to a support center, the predetermined situation may include at least one of the contract situation or the usage situation related to the subject of support.

[0058] Second Embodiment Next, a processing system according to a second embodiment of the present disclosure will be described with reference to FIG. 7. FIG. 7 is a block diagram illustrating an example configuration of a processing system according to an embodiment of the present disclosure. The processing system 1a illustrated in FIG. 7 differs from the processing system 1 according to the first embodiment illustrated in FIG. 1 in that it newly includes an AI function unit (4) 70. Note that, apart from the configuration of the AI ​​function unit (4) 70, the input and output data are partially different. However, the configuration and operation of the processing system 1a illustrated in FIG. 7 are essentially the same as those of the processing system 1 illustrated in FIG. 1. By adding the AI ​​function unit (4) 70 to the processing system 1a illustrated in FIG. 7, the content of newly received inquiries and their response results can also be subjected to machine learning, thereby enabling the functions of the AI ​​function units (1) 20, (2) 30, and (3) 40 to be enhanced and updated. The AI ​​function unit (4) 70 includes a learning and analysis unit 71 and is connected to the response history database 14 in the inquiry response system 10 and the AI ​​function units (1) 20, (2) 30, and (3) 40. The AI ​​function unit (4) 70 performs machine learning based on the data in the response history database 14 and transmits the results to the AI ​​function units (1) 20, (2) 30, and (3) 40.

[0059] Third Embodiment Next, a processing system according to a third embodiment of the present disclosure will be described with reference to FIG. 8. FIG. 8 is a block diagram illustrating an example configuration of a processing system according to an embodiment of the present disclosure. The processing system 1b illustrated in FIG. 8 differs from the processing system 1 according to the first embodiment illustrated in FIG. 1 in that a response system 10a illustrated in FIG. 8, which corresponds to the inquiry response system 10 illustrated in FIG. 1, is newly equipped with a chatbot 15. As illustrated in FIG. 8, the chatbot 15 may be installed in the inquiry response system 10a. The chatbot 15 is connected to an inquiry reception unit 11 and an inquiry answering unit 12. The chatbot 15 determines whether an automatic response is possible for an inquiry received by the inquiry reception unit 11, automatically generates a response for an inquiry that can be automatically answered, and transmits the response to the inquiry answering unit. An inquiry that cannot be automatically answered is transmitted to the AI ​​function unit (1) 20.

[0060] <Fourth embodiment> Next, a fourth embodiment of the present disclosure will be described with reference to FIGS. 9 and 10. FIG. 9 is a block diagram showing an example configuration of a processing system according to an embodiment of the present disclosure. FIG. 10 is a flowchart showing an example operation of the processing system according to an embodiment of the present disclosure. The processing system 100 according to the fourth embodiment of the present disclosure shown in FIG. 9 includes a sentence correction unit 101, an emotion determination unit 102, and a selection unit 103. The sentence correction unit 101 corrects an input sentence according to a predetermined rule and outputs the corrected sentence. The emotion determination unit 102 determines an emotion from the input sentence or the corrected sentence and outputs the emotion determination result. The selection unit 103 selects a method of processing the input sentence based on the corrected sentence, the emotion determination result, and a predetermined situation related to the creator of the input sentence.

[0061] Moreover, the processing method according to this embodiment shown in FIG. 10 includes a step (S101) of correcting an input sentence in accordance with predetermined rules and outputting the corrected sentence, a step (S102) of determining an emotion from the input sentence or the corrected sentence and outputting the determined emotion result, and a step (S103) of selecting how to process the input sentence based on the corrected sentence, the emotion determination result, and a predetermined situation related to the creator of the input sentence.

[0062] <Computer configuration> 11 is a block diagram showing a schematic configuration example of a computer according to an embodiment of the present disclosure. The computer 700 includes a CPU (Central Processing Unit) 710, a main memory device 720, an auxiliary memory device 730, and an interface 740. A non-volatile memory medium 750, for example, is connected to the interface 740. The processing systems 1, 1a, and 1b described above are implemented in the computer 700. The operations of the above-described processing units are stored in the auxiliary memory device 730 in the form of a program. The CPU 710 reads the program from the auxiliary memory device 730, loads it into the main memory device 720, and executes the above-described processing in accordance with the program. The CPU 710 also allocates storage areas in the main memory device 720 corresponding to the above-described storage units in accordance with the program.

[0063] The program may be for realizing some of the functions to be performed by the computer 700. For example, the program may be combined with other programs already stored in the auxiliary storage device 730 or other programs implemented in other devices to perform the functions. In other embodiments, the computer may include a custom LSI (Large Scale Integrated Circuit) such as a PLD (Programmable Logic Device) in addition to or instead of the above configuration. Examples of PLDs include PAL (Programmable Array Logic), GAL (Generic Array Logic), CPLD (Complex Programmable Logic Device), and FPGA (Field Programmable Gate Array). In this case, some or all of the functions realized by the CPU 710 may be realized by the integrated circuit.

[0064] Examples of auxiliary storage device 730 include a hard disk drive (HDD), a solid state drive (SSD), a magnetic disk, a magneto-optical disk, a compact disc read-only memory (CD-ROM), a digital versatile disc read-only memory (DVD-ROM), and semiconductor memory. Auxiliary storage device 730 may be internal media directly connected to the bus of computer 700, or may be external media connected to computer 700 via interface 740 or a communication line. Furthermore, if this program is distributed to computer 700 via a communication line, computer 700 that receives the program may load the program into main storage device 720 and execute the above-described processing. In at least one embodiment, auxiliary storage device 730 is a non-transitory tangible storage medium.

[0065] Although the present disclosure has been described above with reference to the embodiments, the present disclosure is not limited to the above-described embodiments. Various modifications that can be understood by those skilled in the art can be made to the configuration and details of the present disclosure within the scope of the present disclosure. Each embodiment can be combined with other embodiments as appropriate. Furthermore, part or all of the programs executed by the computer in the above-described embodiments can be distributed via a computer-readable recording medium or a communication line.

[0066] Some or all of the above embodiments can be described as, but are not limited to, the following supplementary notes.

[0067] (Appendix 1) a sentence correction unit that corrects an input sentence according to a predetermined rule and outputs the corrected sentence; an emotion determination unit that determines an emotion from the input sentence or the corrected sentence and outputs the determined emotion result; a selection unit that selects a method of processing the input sentence based on the corrected sentence, the emotion determination result, and a predetermined situation related to the creator of the input sentence; A processing system comprising:

[0068] (Appendix 2) The rules include removing predetermined emotional words. 10. The processing system of claim 1.

[0069] (Appendix 3) The rule includes replacing a predetermined string with another predetermined string. 3. The processing system of claim 1 or 2.

[0070] (Appendix 4) The rules are created using a machine learning model that is deep-learned using information representing the content of processing related to the past input sentences as input data. 4. A processing system according to any one of appendices 1 to 3.

[0071] (Appendix 5) The emotion determination unit determines the emotion by reflecting a result of comparing the input or revised sentence with a previous input or revised sentence created by the same creator. 5. A processing system according to any one of appendices 1 to 4.

[0072] (Appendix 6) The selection unit selects automatic reply as the processing method when an automatic reply can be made to the corrected sentence. 6. A processing system according to any one of appendices 1 to 5.

[0073] (Appendix 7) When an automatic reply cannot be made to the corrected sentence, the selection unit selects a reply by a person selected based on the emotion determination result and the predetermined situation as the processing method. 7. A processing system according to any one of appendices 1 to 6.

[0074] (Appendix 8) The input sentence is a sentence of an inquiry to a support center, The predetermined situation includes at least one of a contract situation and a usage situation related to the object of support. 8. A processing system according to any one of appendices 1 to 7.

[0075] (Appendix 9) A first step of correcting an input sentence according to a predetermined rule and outputting the corrected sentence; a second step of determining an emotion from the input sentence or the corrected sentence and outputting the determined emotion result; a third step of selecting a method of processing the input sentence based on the corrected sentence, the emotion determination result, and a predetermined situation related to the creator of the input sentence; A processing method comprising:

[0076] (Appendix 10) The rules include removing predetermined emotional words. The processing method described in Appendix 9.

[0077] (Appendix 11) The rule includes replacing a predetermined string with another predetermined string. 11. The processing method according to claim 9 or 10.

[0078] (Appendix 12) The rules are created using a machine learning model that is deep-learned using information representing the content of processing related to the past input sentences as input data. A processing method according to any one of appendices 9 to 11.

[0079] (Appendix 13) In the second step, the emotion is determined by reflecting a result of comparing the input or revised sentence with a previous input or revised sentence created by the same creator. A processing method according to any one of appendices 9 to 12.

[0080] (Appendix 14) In the third step, if an automatic reply can be made to the corrected sentence, the automatic reply is selected as the processing method. A processing method according to any one of appendices 9 to 13.

[0081] (Appendix 15) In the third step, if an automatic reply cannot be made to the corrected sentence, a reply by a person selected based on the emotion determination result and the predetermined situation is selected as the processing method. A processing method according to any one of appendices 9 to 14.

[0082] (Appendix 16) The input sentence is a sentence of an inquiry to a support center, The predetermined situation includes at least one of a contract situation and a usage situation related to the object of support. A processing method according to any one of appendices 9 to 15.

[0083] (Appendix 17) a step of correcting the input sentence in accordance with a predetermined rule and outputting the corrected sentence; a step of determining an emotion from the input sentence or the corrected sentence and outputting the determined emotion result; selecting a method of processing the input sentence based on the corrected sentence, the emotion determination result, and a predetermined situation related to the creator of the input sentence; A program that causes a computer to execute the following.

[0084] (Appendix 18) The rules include removing predetermined emotional words. 17. The program described in Appendix 17.

[0085] (Appendix 19) The rule includes replacing a predetermined string with another predetermined string. 19. The program of claim 17 or 18.

[0086] (Appendix 20) The rules are created using a machine learning model that is deep-learned using information representing the content of processing related to the past input sentences as input data. A program according to any one of appendices 17 to 19.

[0087] (Appendix 21) In the second step, the emotion is determined by reflecting a result of comparing the input or revised sentence with a previous input or revised sentence created by the same creator. A program according to any one of appendices 17 to 20.

[0088] (Appendix 22) In the third step, if an automatic reply can be made to the corrected sentence, the automatic reply is selected as the processing method. A program according to any one of appendices 17 to 21.

[0089] (Appendix 23) In the third step, if an automatic reply cannot be made to the corrected sentence, a reply by a person selected based on the emotion determination result and the predetermined situation is selected as the processing method. A program according to any one of appendices 17 to 22.

[0090] (Appendix 24) The input sentence is a sentence of an inquiry to a support center, The predetermined situation includes at least one of a contract situation and a usage situation related to the object of support. A program according to any one of appendices 17 to 23. [Explanation of symbols]

[0091] 1, 1a, 1b Processing Systems 21 Text correction section 22 Emotion Discriminator 41 Personnel Allocation Department 100 Processing Systems 101 Text correction section 102 Emotion Discriminator 103 Selection section

Claims

1. a sentence correction unit that corrects an input sentence according to a predetermined rule and outputs the corrected sentence; an emotion determination unit that determines an emotion from the input sentence or the corrected sentence and outputs the determined emotion result; a selection unit that selects a method of processing the input sentence based on the corrected sentence, the emotion determination result, and a predetermined situation related to the creator of the input sentence; A processing system comprising:

2. The rules include removing predetermined emotional words. The processing system of claim 1 .

3. The rule includes replacing a predetermined string with another predetermined string. The processing system of claim 2 .

4. The rules are created using a machine learning model that is deep-learned using information representing the content of processing related to the past input sentences as input data. The processing system of claim 3 .

5. The emotion determination unit determines the emotion by reflecting a result of comparing the input or revised sentence with a previous input or revised sentence created by the same creator. The processing system of claim 4 .

6. The selection unit selects automatic reply as the processing method when an automatic reply can be made to the corrected sentence. The processing system of claim 5 .

7. When an automatic reply cannot be made to the corrected sentence, the selection unit selects a reply by a person selected based on the emotion determination result and the predetermined situation as the processing method. The processing system of claim 6 .

8. The input sentence is a sentence of an inquiry to a support center, The predetermined situation includes at least one of a contract situation and a usage situation related to the object of support. The processing system according to any one of claims 1 to 7.

9. a step of correcting the input sentence in accordance with a predetermined rule and outputting the corrected sentence; a step of determining an emotion from the input sentence or the corrected sentence and outputting the determined emotion result; selecting a method of processing the input sentence based on the corrected sentence, the emotion determination result, and a predetermined situation related to the creator of the input sentence; A processing method comprising:

10. a step of correcting the input sentence in accordance with a predetermined rule and outputting the corrected sentence; a step of determining an emotion from the input sentence or the corrected sentence and outputting the determined emotion result; selecting a method of processing the input sentence based on the corrected sentence, the emotion determination result, and a predetermined situation related to the creator of the input sentence; A program that causes a computer to execute the following.

Images