Method and device for determining reply information in departure scene and storage medium

Abstract

The invention discloses a method and device for determining reply information in a departure scene and a storage medium. The method comprises the following steps: acquiring inquiry information in a departure scene; the inquiry information is classified, a classification result is obtained, and the classification result is used for indicating the information emergency degree and / or inquiry type corresponding to the inquiry information; the classification result and the inquiry information are input into an intelligent customer service model for analysis, initial reply information matched with the inquiry information is obtained, and the intelligent customer service model is obtained by training inquiry information samples in the departure scene and reply information samples matched with the inquiry information samples; outputting initial reply information, and obtaining a feedback result of the initial reply information; and adjusting the initial reply information based on the feedback result to obtain target reply information. The technical problem that the efficiency of determining the reply information in the departure scene is low is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of airport departure services, and in particular to a method, device and storage medium for determining reply information in a departure scenario. Background Art

[0002] At present, in the departure business scenario, due to the complexity of the business process, the uncertainty of the system environment and the diversity of passengers, staff often encounter various situations that they cannot handle when handling business. For example, when the check-in counter staff handles passenger check-in, the passenger check-in may fail due to factors such as lack of familiarity with the business system, lack of proficiency in the business process or incorrect passenger information. As a key link in the civil aviation industry, departure business needs to deal with a large number of passengers. Therefore, quickly and efficiently solving on-site problems is an important means to improve the travel experience of passengers and improve the service quality of airports and airlines.

[0003] In the relevant technologies, there is a lack of effective filtering mechanisms to distinguish the severity of on-site problems, which wastes valuable human resources and time to a large extent; limited customer service resources cannot solve unlimited on-site problems in a timely manner. Actual surveys show that most staff members need to wait for a long time when calling customer service, which greatly affects passengers' satisfaction with airport and airline services. Therefore, there is a technical problem of low efficiency in determining reply information in departure scenarios.

[0004] With respect to the above-mentioned technical problem of low efficiency in determining reply information in the departure scenario, no effective solution has been proposed so far. Summary of the invention

[0005] The embodiments of the present invention provide a method, device and storage medium for determining reply information in a departure scenario, so as to at least solve the technical problem of low efficiency in determining reply information in a departure scenario.

[0006] According to one aspect of an embodiment of the present invention, a method for determining reply information in a departure scenario is provided. The method may include: obtaining inquiry information in a departure scenario; classifying the inquiry information to obtain a classification result, wherein the classification result is used to indicate the urgency of information and / or the inquiry type corresponding to the inquiry information; inputting the classification result and the inquiry information into an intelligent customer service model for analysis to obtain initial reply information matching the inquiry information, wherein the intelligent customer service model is trained using inquiry information samples in a departure scenario and reply information samples matching the inquiry information samples; outputting the initial reply information and obtaining feedback results of the initial reply information, wherein the feedback results are used to indicate the satisfaction of the question object corresponding to the inquiry information with the initial reply information; adjusting the initial reply information based on the feedback results to obtain the target reply information.

[0007] Optionally, the inquiry information is classified to obtain a classification result, including: extracting keywords in the inquiry information; matching the keywords with information in a database to obtain a matching result, wherein the information in the database is historical inquiry information and information urgency and / or inquiry type corresponding to the historical inquiry information; and determining a classification result based on the matching result.

[0008] Optionally, the method for determining the reply information in the departure scenario further includes: inputting the inquiry information sample into the initial customer service model for training to obtain an intelligent customer service model.

[0009] Optionally, based on the feedback result, the initial reply information is adjusted to obtain target reply information, including: in response to the feedback result being a satisfactory feedback result, determining the initial reply information as the target reply information; in response to the feedback result being an unsatisfactory feedback result, outputting prompt information, wherein the prompt information is used to prompt the questioner that the inquiry information needs to be adjusted; based on the prompt information, determining the target reply information.

[0010] Optionally, based on the prompt information, target reply information is determined, including: adjusting the inquiry information in response to the questioner, and obtaining the adjusted inquiry information; inputting the adjusted inquiry information into the intelligent customer service model for analysis to obtain the target reply information.

[0011] Optionally, before adjusting the initial reply information based on the feedback results to obtain the target reply information, the method for determining the reply information in the departure scenario also includes: obtaining the number of times the feedback results are unsatisfactory; in response to the number being greater than a preset threshold, outputting the inquiry information to the manual customer service end; and storing the inquiry information and the manual reply information output by the manual customer service end in a database.

[0012] According to another aspect of an embodiment of the present invention, a device for determining reply information in a departure scenario is also provided. The device may include: an acquisition unit for acquiring inquiry information in a departure scenario; a classification unit for classifying the inquiry information to obtain a classification result, wherein the classification result is used to indicate the urgency of the information and / or the inquiry type corresponding to the inquiry information; an input unit for inputting the classification result and the inquiry information into an intelligent customer service model for analysis to obtain initial reply information matching the inquiry information, wherein the intelligent customer service model is trained using inquiry information samples in a departure scenario and reply information samples matching the inquiry information samples; an output unit for outputting the initial reply information and obtaining a feedback result of the initial reply information, wherein the feedback result is used to indicate the satisfaction of the question object corresponding to the inquiry information with the initial reply information; and an adjustment unit for adjusting the initial reply information based on the feedback result to obtain the target reply information.

[0013] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored program, wherein when the program is executed by a processor, the device where the storage medium is located is controlled to execute the method for determining reply information in a departure scenario in an embodiment of the present invention.

[0014] According to another aspect of an embodiment of the present invention, a processor is further provided. The processor is used to run a program, wherein when the program is run, the method for determining reply information in a departure scenario in an embodiment of the present invention is executed.

[0015] According to another aspect of an embodiment of the present invention, a computer program product is provided, which includes computer instructions, and when the computer instructions are executed by a processor, the method for determining reply information in a departure scenario in an embodiment of the present invention is implemented.

[0016] In an embodiment of the present invention, inquiry information in a departure scenario is obtained; the inquiry information is classified and processed to obtain a classification result; the classification result and the inquiry information are input into an intelligent customer service model for analysis to obtain initial reply information matching the inquiry information, wherein the intelligent customer service model is trained using inquiry information samples in a departure scenario and reply information samples matching the inquiry information samples; the initial reply information is output, and feedback results of the initial reply information are obtained; based on the feedback results, the initial reply information is adjusted to obtain target reply information. That is, the embodiment of the present invention obtains inquiry information, classifies the inquiry information, processes the inquiry information using an intelligent customer service model, and obtains reply information. At the same time, the present invention can adjust the reply information according to the satisfaction of the questioning object with the reply information, thereby solving the technical problem of low efficiency in determining reply information in a departure scenario, and achieving the technical effect of improving the efficiency of determining reply information in a departure scenario. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0018] Figure 1 is a flow chart of a method for determining reply information in a departure scenario according to an embodiment of the present invention;

[0019] Figure 2 is a flow chart of an intelligent matching method of a departure intelligent customer service system according to an embodiment of the present invention;

[0020] Figure 3 is a schematic diagram of the architecture of an intelligent matching device of a departure intelligent customer service system according to an embodiment of the present invention;

[0021] Figure 4 It is a schematic diagram of a device for determining reply information in a departure scenario according to an embodiment of the present invention. DETAILED DESCRIPTION

[0022] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0023] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, functional component or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, functional components or devices.

[0024] According to an embodiment of the present invention, an embodiment of a method for determining reply information in a departure scenario is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in an order different from that shown here.

[0025] Figure 1 is a flow chart of a method for determining reply information in a departure scenario according to an embodiment of the present invention. Figure 1 As shown, the method may include the following steps:

[0026] Step S101, obtaining inquiry information related to the departure scenario.

[0027] In the technical solution provided in the above step S101 of the present invention, the inquiry information may also be referred to as a departure business question.

[0028] In this embodiment, inquiry information in the departure scenario is obtained. For example, the departure business question input by the departure user in the front-end interface of the departure intelligent customer service system is obtained. This is only an illustrative example and does not limit the specific method of inquiry information in the departure scenario.

[0029] Step S102: classify the query information to obtain a classification result.

[0030] In the technical solution provided in the above step S102 of the present invention, the classification result is used to indicate the information urgency and / or inquiry type corresponding to the inquiry information.

[0031] In this embodiment, after obtaining the inquiry information in the departure scenario in step S101, the inquiry information is classified and processed to obtain a classification result. For example, the inquiry information is classified and processed using a keyword extraction rule and other methods to obtain a classification result. This is only an illustrative example and does not limit the specific method for obtaining the classification result.

[0032] For example, according to the keyword extraction rules, the outbound intelligent customer service system extracts keywords from the input questions. The extracted keywords are matched and classified with the data categories in the outbound customer service database, and the outbound customer service data is classified and processed into data sets of different dimensions.

[0033] Optionally, the inquiry information is classified and processed according to three dimensions: urgency E, impact scale Q, and problem frequency F of the departure system business scenario problem.

[0034] For example, the departure problem level table is obtained based on the three dimensions of urgency E, impact scale Q, and problem frequency F of the departure system business scenario problem. The level is used as the weight value M for extracting the problem keyword. Urgency E defines: the urgency of the departure problem, which mainly has three values: general, medium, and urgent; impact scale Q defines: the impact scope of the departure problem, which mainly has five values: airport, airline, flight, passenger, and counter; problem frequency F defines: the frequency of occurrence of departure problems, that is, the frequency of occurrence of a certain type of departure problem among all departure problems, which mainly has three values: low, medium, and high.

[0035] Step S103: input the classification result and the inquiry information into the intelligent customer service model for analysis to obtain initial response information matching the inquiry information.

[0036] In the technical solution provided in step S103 of the present invention, the intelligent customer service model is trained using inquiry information samples in the departure scenario and response information samples matching the inquiry information samples. The intelligent customer service model can also be called a departure customer service model.

[0037] In this embodiment, after the classification result is obtained in step S102, the classification result and the inquiry information are input into the intelligent customer service model for analysis to obtain initial response information matching the inquiry information. The intelligent customer service model can be trained using the departure customer service data set.

[0038] For example, the outbound customer service data set is divided into a basic data set and an incremental data set. The pre-processed outbound customer service data set is used as training data. The outbound customer service model is trained using a public supervised learning method, and the model is trained and fine-tuned. The trained outbound customer service model is used to answer questions after classification and matching.

[0039] Step S104, outputting the initial reply information and obtaining the feedback result of the initial reply information.

[0040] In the technical solution provided in the above step S104 of the present invention, the feedback result is used to indicate the satisfaction degree of the question object corresponding to the inquiry information with the initial response information.

[0041] In this embodiment, after obtaining the initial reply information matching the inquiry information in step S103, the initial reply information is output and the feedback result of the initial reply information is obtained. For example, the initial reply information is output to the front-end interface of the departure intelligent customer service system. This is only an example and does not limit the specific method of outputting the initial reply information.

[0042] For example, the departure intelligent customer service system outputs the reply generated by the departure intelligent customer service, that is, the initial reply information, to the front-end interface to form a complete conversation. The departure intelligent customer service system replies to the user and sets the reply result option to feedback the reply result.

[0043] Step S105: Based on the feedback result, the initial response information is adjusted to obtain the target response information.

[0044] In the technical solution provided in the above step S105 of the present invention, after the feedback result of the initial reply information is obtained in step S104, the initial reply information is adjusted to obtain the target reply information.

[0045] In this embodiment, when the user's feedback result is satisfactory, it means that the initial reply information can solve the user's problem. Based on this, the initial reply information is determined as the target reply information.

[0046] Optionally, when the feedback result is unsatisfactory, it means that the initial reply information did not solve the user's problem, that is, the initial reply information is incorrect. Based on this, the departure intelligent customer service system prompts the user to readjust the question, and the departure intelligent customer service re-replies.

[0047] It should be noted that the above embodiments can be executed through the departure intelligent customer service system.

[0048] In the above steps S101 to S105 of the present invention, inquiry information in the departure scenario is obtained; the inquiry information is classified and processed to obtain the classification result; the classification result and the inquiry information are input into the intelligent customer service model for analysis to obtain the initial reply information matching the inquiry information, wherein the intelligent customer service model is trained using the inquiry information sample in the departure scenario and the reply information sample matching the inquiry information sample; the initial reply information is output, and the feedback result of the initial reply information is obtained; based on the feedback result, the initial reply information is adjusted to obtain the target reply information. That is to say, the embodiment of the present invention obtains the inquiry information, classifies the inquiry information, processes the inquiry information using the intelligent customer service model, and obtains the reply information. At the same time, the present invention can adjust the reply information according to the satisfaction of the questioning object with the reply information, thereby solving the technical problem of low efficiency in determining the reply information in the departure scenario, and achieving the technical effect of improving the efficiency of determining the reply information in the departure scenario.

[0049] The above method of this embodiment is further introduced below.

[0050] As an optional implementation method, the inquiry information is classified and processed to obtain a classification result, including: extracting keywords in the inquiry information; matching the keywords with information in a database to obtain a matching result, wherein the information in the database is historical inquiry information and the information urgency and / or inquiry type corresponding to the historical inquiry information; and determining a classification result based on the matching result.

[0051] In this embodiment, keywords in the query information are extracted, for example, keywords in the query information are extracted using a keyword extraction rule.

[0052] For example, according to the keyword extraction rules set by the system, the keyword tuple K of the question text Q is extracted = (K1, K2, ..., K n ). The specific implementation method is: word segmentation processing: use the open source word segmentation system ICTCLAS to perform word segmentation processing on the text Q, and obtain the initial keyword tuple K0 = ((K1 ', f1), (K2 ', f2), ..., (K m ',f m )).

[0053] For another example, the initial keyword tuple after word segmentation of the question text Q is: K0 = (('use', 1), ('work number', 1), ('login', 1), ('time', 1), ('prompt', 1), ('account', 1), ('equipment', 1), ('information', 1), ('abnormal', 1)).

[0054] Optionally, the extracted keywords are matched with data categories in the departure customer service database to determine a classification result.

[0055] For example, the Jaccard similarity algorithm is used to calculate the initial keyword tuple K0 = ((K1 ', f1), (K2 ', f2), ..., (K m ',f m The keywords in the departure problem level table are respectively related to the problem classification T j The similarity S ij As shown in the following formula (1):

[0056]

[0057] Among them, when S ij >=0.5, then the keyword K i 'Get the weight value L i .

[0058] As an optional implementation manner, the method for determining the reply information in the departure scenario also includes: inputting the inquiry information sample into the initial customer service model for training to obtain an intelligent customer service model.

[0059] In this embodiment, the inquiry information sample is input into the initial customer service model for training to obtain an intelligent customer service model, wherein the inquiry information sample may be a data sample obtained by processing the departure customer service data set.

[0060] Optionally, by training an intelligent customer service model, questions raised by users can be answered through the intelligent customer service model, avoiding the technical problem of low efficiency in determining reply information caused by manual resolution of all problems, thereby achieving the technical effect of improving the efficiency of determining reply information in departure scenarios.

[0061] As an optional implementation method, the initial reply information is adjusted to obtain the target reply information, including: in response to the feedback result being a satisfactory feedback result, the initial reply information is determined as the target reply information; in response to the feedback result being an unsatisfactory feedback result, prompt information is output, wherein the prompt information is used to prompt the questioner that the inquiry information needs to be adjusted; based on the prompt information, the target reply information is determined.

[0062] In this embodiment, when the user's feedback result is satisfactory, it means that the initial reply information can solve the user's problem. Based on this, the initial reply information is determined as the target reply information.

[0063] Optionally, when the feedback result is unsatisfactory, it means that the initial reply information does not solve the question raised by the user, based on which, a prompt message is output.

[0064] Optionally, the target reply information is determined according to the output prompt information. The specific determination method is as follows.

[0065] As an optional implementation method, target reply information is determined based on prompt information, including: adjusting the inquiry information in response to the questioner to obtain the adjusted inquiry information; inputting the adjusted inquiry information into the intelligent customer service model for analysis to obtain the target reply information.

[0066] In this embodiment, after the questioner adjusts the inquiry information, the adjusted inquiry information is obtained, and the adjusted inquiry information is re-input into the intelligent customer service model for analysis to obtain target answer information.

[0067] For example, when a user reports that a problem has not been successfully solved, the Departure Intelligent Customer Service System will prompt the user to readjust the problem and the Departure Intelligent Customer Service will respond again.

[0068] As an optional implementation method, before adjusting the initial reply information based on the feedback results to obtain the target reply information, the method for determining the reply information in the departure scenario also includes: obtaining the number of times the feedback results are unsatisfactory; in response to the number being greater than a preset threshold, outputting the inquiry information to the manual customer service end; and storing the inquiry information and the manual reply information output by the manual customer service end in a database.

[0069] In this embodiment, the number of times the feedback result is unsatisfactory is obtained. When the number is greater than a preset threshold, it means that the inquiry information exceeds the answerable range of the intelligent customer service. Based on this, the inquiry information is output to the manual customer service.

[0070] For example, during a conversation, when the frequency of users reporting unresolved issues exceeds the threshold set by the departure intelligent customer service system, the departure intelligent customer service system will mark the issue as a labeled issue and feedback it to Air China's manual customer service through various possible channels (telephone, email, text message, web message, etc.).

[0071] Optionally, after the manual customer service gives the reply information, the inquiry information and the manual reply information output by the manual customer service are stored in a database.

[0072] For example, Air China's manual customer service connects to the departure intelligent customer service system and responds to identified questions; at the same time, the current question and response data are updated to the incremental data set of the departure customer service data subset of the corresponding category in the departure customer service database to serve as a data source for training and fine-tuning the departure customer service model.

[0073] It should be noted that the above embodiments can be executed through the departure intelligent customer service system.

[0074] In this embodiment, inquiry information in the departure scenario is obtained; the inquiry information is classified and processed to obtain a classification result; the classification result and the inquiry information are input into the intelligent customer service model for analysis to obtain initial reply information matching the inquiry information, wherein the intelligent customer service model is trained using inquiry information samples in the departure scenario and reply information samples matching the inquiry information samples; the initial reply information is output, and the feedback result of the initial reply information is obtained; based on the feedback result, the initial reply information is adjusted to obtain the target reply information. That is to say, the embodiment of the present invention obtains inquiry information, classifies the inquiry information, processes the inquiry information using the intelligent customer service model, and obtains reply information. At the same time, the present invention can adjust the reply information according to the satisfaction of the questioning object with the reply information, thereby solving the technical problem of low efficiency in determining reply information in the departure scenario, and achieving the technical effect of improving the efficiency of determining reply information in the departure scenario.

[0075] The technical solution of the embodiment of the present invention is illustrated below in conjunction with preferred implementation modes.

[0076] At present, in the departure business scenario, due to the complexity of the business process, the uncertainty of the system environment and the diversity of passengers, staff often encounter various situations that they cannot handle when handling business. For example, when the check-in counter staff handles passenger check-in, the passenger check-in may fail due to factors such as lack of familiarity with the business system, lack of proficiency in the business process or incorrect passenger information. As a key link in the civil aviation industry, departure business needs to deal with a large number of passengers. Therefore, quickly and efficiently solving on-site problems is an important means to improve the travel experience of passengers and improve the service quality of airports and airlines.

[0077] In the relevant technology, there is a lack of effective filtering mechanism to distinguish the severity of on-site problems, which wastes precious human resources and time to a large extent; limited customer service resources cannot solve unlimited on-site problems in a timely manner. Actual surveys show that most staff members need to wait for a long time when calling customer service, which greatly affects passengers' satisfaction with airport and airline services. Therefore, there is a technical problem of low efficiency in determining reply information in departure scenarios. In view of the above-mentioned technical problem of low efficiency in determining reply information in departure scenarios, no effective solution has been proposed.

[0078] However, an embodiment of the present invention proposes an intelligent matching method for a departure intelligent customer service system, in which a departure user inputs a departure business question on the front-end interface; the departure intelligent customer service system extracts keywords for the input question according to the keyword extraction rules. The extracted keywords are matched with the data categories in the departure customer service database, and the departure customer service data is classified and processed into data sets of different dimensions; each data set is further divided into a basic data set and an incremental data set, and a trained departure customer service model is used to respond to the questions after classification and matching processing. The departure intelligent customer service system outputs the response generated by the departure intelligent customer service to the front-end interface to form a complete conversation. The technical problem of low efficiency in determining response information in departure scenarios is solved, and the technical effect of improving the efficiency of determining response information in departure scenarios is achieved.

[0079] The embodiments of the present invention are further described below.

[0080] Figure 2 is a flow chart of an intelligent matching method of a departure intelligent customer service system according to an embodiment of the present invention. Figure 2 As shown, the intelligent matching method of the departure intelligent customer service system includes the following steps:

[0081] Step S201: The departing user inputs a business question.

[0082] In this embodiment, the departure user inputs departure business questions in the front-end interface.

[0083] Optionally, the departing user enters a question in the front-end interface; the system obtains the question text Q.

[0084] For example, the user input problem is: when logging in using the work ID, it prompts that the account device information is abnormal.

[0085] Step S202, classify and match the questions.

[0086] In this embodiment, according to the keyword extraction rules, the system extracts keywords from the input question, uses the extracted keywords to match the data categories in the departure customer service database, and transmits the matching results to the departure customer service module for subsequent system processing.

[0087] Optionally, according to the keyword extraction rules set by the system, a keyword tuple K=(K1, K2, ..., K n ).

[0088] Optionally, word segmentation processing: use the open source word segmentation system ICTCLAS to perform word segmentation processing on the text Q to obtain an initial keyword tuple K0 = ((K1 ', f1), (K2 ', f2), ..., (K m ',f m )).

[0089] For example, the initial keyword tuple after word segmentation of the question text Q is: K0 = (('use', 1), ('work number', 1), ('login', 1), ('time', 1), ('prompt', 1), ('account', 1), ('equipment', 1), ('information', 1), ('abnormality', 1)).

[0090] Optionally, keyword extraction processing: This link is divided into the following steps. A departure problem level table is obtained based on the three dimensions of urgency E, impact scale Q, and problem frequency F of the departure system business scenario problem. The level is used as the weight value M for extracting the problem keyword.

[0091] Optionally, the urgency level E defines: the urgency level of the departure problem, which mainly includes three values: general, medium, and urgent.

[0092] Optionally, the impact scale Q is defined as the impact scope of the departure problem, which mainly includes five values: airport, airline, flight, passenger, and counter.

[0093] Optionally, the problem frequency F is defined as the frequency of departure problems, that is, the frequency of a certain type of departure problem in all departure problems, which mainly has three values: low, medium and high. Table 1 is a departure problem level table, and the type, urgency and other information corresponding to each problem are shown in Table 1.

[0094] Table 1 Departure problem rating table

[0095] Question Type Urgency E Image scale Q Problem frequency F Level L Flight Class urgent flight middle 1 Passenger medium Travelers high 2 System Class urgent Airport middle 3 Account generally counter Low 4 Using Classes generally counter Low 5

[0096] Optionally, the Jaccard similarity algorithm is used to calculate the initial keyword tuple K0=((K1′, f1), (K2′, f2), ..., (K m ',f m The keywords in the departure problem level table are respectively related to the problem classification T j The similarity S ij , as shown in the above formula (1), which will not be repeated here.

[0097] Optionally, an improved TF-IDF algorithm is used to identify the initial keyword tuple K0=((K1′, f1), (K2′, f2), ..., (K m ',f m )) calculates the TF-IDF value of each keyword in the tuple, and finally obtains the keyword K of question Q = (K1, K2, ..., K n ). The definition of TF-IDF algorithm is as follows:

[0098] Optionally, TF term frequency represents the frequency of the keyword appearing in the text Q, as shown in the following formula (2):

[0099] TF i =f i (2)

[0100] Optionally, the IDF inverse document frequency represents the inverse document frequency of the keyword in the departure customer service database, as shown in the following formula (3):

[0101] IDF i =log(M / N i ) (3)

[0102] Where M represents the total number of questions in the departure customer service database, N i Indicates the total number of questions containing this keyword in the outbound customer service database.

[0103] Optionally, this patent improves the TF-IDF algorithm: a specific weight value is assigned to the original TF-IDF algorithm to obtain keywords suitable for departure business issues, as shown in the following formula (4):

[0104] TF i ×IDF i ×L i =f i ×log(M / N i )×L i (4)

[0105] Optionally, the improved TF-IDF value of each keyword is calculated, and the important keyword tuple K=(K1, K2, ..., K n ).

[0106] Optionally, the question is classified into T according to the category of the outbound customer service data set and the extracted keywords, and the result pair (K, T) is passed to the outbound customer service module.

[0107] For example, according to the dataset category and keyword K, the question text Q is classified as account class A. The (('work number', 'login', 'equipment', 'information', 'abnormal'), A) label is passed to the departure customer service module.

[0108] Step S203: pre-processing of outbound customer service data.

[0109] In this embodiment, the departure customer service data is classified into data sets of different dimensions; each data set is further divided into a basic data set and an incremental data set. The advantage of dividing the data set is that it is convenient to strengthen the model according to the response to specific types of questions in the future.

[0110] Optionally, the outbound customer service data is classified into four outbound customer service data subsets: account problem set A, system usage problem set B, business problem set C, and system failure problem set D;

[0111] Optionally, the size S of each subset, i.e., the amount of data in the subset, is defined as S A , S B , S C , S D ; S is composed of the basic data volume G and the incremental data volume E, so that: G A =G B =G C =G D , E A =S A -G A , E B =S B -G B , E C =S C -G C , E D =S D -G D .

[0112] Optionally, the pre-processed departure customer service data set is used as training data, and a departure customer service model is trained using a public supervised learning method, and the model is trained and fine-tuned. The trained departure customer service model is used to answer questions after classification and matching processing.

[0113] Step S204: the system outputs a reply.

[0114] In this embodiment, the system outputs the reply generated by the departure intelligent customer service to the front-end interface to form a complete conversation.

[0115] Optionally, a trained departure customer service model is used to reply to the result pair (K, T) after classification matching.

[0116] Step S205: The user provides feedback on whether the problem has been solved.

[0117] In this embodiment, the system replies to the user and sets the reply result option. If the user replies that the problem is solved, the process ends; if the user replies that the problem is not successfully solved, step S206 is executed. The system prompts the user to readjust the problem, and the departure intelligent customer service replies again.

[0118] Optionally, the system sets a reply result option (solved / unsolved), and the user receives the system reply and provides feedback of "solved" or "unsolved"; if the feedback is "solved", the process ends and the Departure Intelligent Customer Service System completes the reply to the question; if the feedback is "unsolved", the Departure Intelligent Customer Service continues to reply to the current question.

[0119] Step S206: whether the number of feedbacks is within a threshold range.

[0120] In this embodiment, during a conversation, when the frequency of users reporting unresolved issues exceeds a threshold set by the system, step S207 is executed; when the frequency of users reporting unresolved issues does not exceed the threshold set by the system, step S201 is executed.

[0121] Step S207, identifying the problem.

[0122] In this embodiment, the system identifies the question as a tagged question.

[0123] Optionally, the system sets a threshold N. When the frequency of receiving “unresolved” feedback is greater than N, the system identifies question Q as a labeled question Q. (tag，T) .

[0124] Step S208: manual customer service access.

[0125] In this embodiment, the system feeds back to the Air China customer service through various possible channels (telephone, email, text message, web message, etc.), and the Air China customer service accesses the departure intelligent customer service system and responds to the identified issues.

[0126] Optionally, a labeled question Q (tag，T) The Air China customer service will answer the R (tag，T) ; At the same time, answer the question correctly (Q (tag，T) , R (tag，T) ) is updated to the incremental data set of the corresponding category of the outbound customer service data subset in the outbound customer service database to serve as the data source for outbound customer service model training and fine-tuning.

[0127] Step S209, pre-processing the departure database.

[0128] In this embodiment, the current question and answer data are updated to the incremental data set of the corresponding category of the outbound customer service data subset in the outbound customer service database to serve as a data source for training and fine-tuning the outbound customer service model.

[0129] Step S210, training a departure customer service model.

[0130] In this embodiment, the incremental data set is used to train the departure customer service model.

[0131] Optionally, the KNN algorithm is a model based on labeled training data, which is suitable for a departure question answering model with clear classification. Therefore, this patent uses the KNN algorithm to train the departure customer service model.

[0132] Optionally, a training data set is selected, that is, the previously pre-processed departure customer service data set J, where J can be A, B, C, or D.

[0133] Optionally, the Euclidean distance L(Q, J) between the input question data Q and the data set J is calculated.

[0134] Optionally, a k value is determined, and k data with the smallest L(Q, J) values ​​are selected. The data labels with the largest number are used as labels for the problem data.

[0135] Optionally, after the customer service model is trained, the subsets are sorted according to their sizes, and the trained model is fine-tuned using the incremental data volume E of the largest subset.

[0136] Figure 3 FIG. 1 is a schematic diagram of the architecture of an intelligent matching device of a departure intelligent customer service system according to an embodiment of the present invention. Figure 3 As shown, the intelligent matching device 300 includes: an I / O module 301, a customer service model training module 302, a departure customer service module 303, a problem matching module 304, a problem identification module 305, a manual processing module 306, a departure customer service data integration module 307 and a log management module 308.

[0137] The I / O module 301 is the input and output interface module of the system. When a user inputs a departure service question, this module is responsible for collecting the user's question and sending it to the question matching module 304 .

[0138] Optionally, after the departure customer service module 303 generates a reply, the module outputs the reply; ③ The reply effect (successfully solved / unsolved) fed back by the user is transmitted to the problem identification module 305. This module provides a data source for subsequent modules of the system and outputs system processing results.

[0139] The customer service model training module 302 trains the model based on the data provided by the outbound customer service data integration module 307 to provide an intelligent customer service model for the outbound customer service module 303 .

[0140] The departure customer service module 303 receives the question output by the question matching module and inputs it into the intelligent customer service model, which then responds and outputs the answer to the I / O module to generate a session.

[0141] The problem matching module 304 processes the collected problems through a classification algorithm according to the existing classification of departure business problems, and transmits the processing results to the departure customer service module 303 for subsequent system processing.

[0142] The problem identification module 305 obtains the response result (solved / unsolved) of the user feedback from the I / O module 301, identifies the current problem, and stores the identified problem in the departure customer service data integration module 307.

[0143] Manual processing module 306, based on the user feedback obtained by I / O module 301, Air China manual customer service obtains the problem and responds; at the same time, the problem response is maintained in departure customer service data integration module 307.

[0144] The departure customer service data integration module 307 is based on the departure customer service database of Air China, and integrates the departure customer service database data and the user manual data of the departure front-end business system. The system inputs the problems processed by the problem identification module 305 into this module, updates the data in the module, and provides a training data set for the training module.

[0145] Log management module 308, this module records the log of the entire process of the departure intelligent customer service system processing and answering departure user questions. This module supports query operations to facilitate subsequent system maintenance and optimization.

[0146] In this embodiment, the departing user inputs a departure business question in the front-end interface; according to the keyword extraction rules, the departure intelligent customer service system extracts keywords from the input question. The extracted keywords are matched with the data categories in the departure customer service database, and the departure customer service data is classified and processed into data sets of different dimensions; each data set is further divided into a basic data set and an incremental data set, and the trained departure customer service model is used to answer the questions after classification and matching. The departure intelligent customer service system outputs the answer generated by the departure intelligent customer service to the front-end interface to form a complete conversation. The technical problem of low efficiency in determining the answer information in the departure scenario is solved, and the technical effect of improving the efficiency of determining the answer information in the departure scenario is achieved.

[0147] According to an embodiment of the present invention, a device for determining reply information in a departure scenario is also provided. It should be noted that the device for determining reply information in a departure scenario can be used to execute the method for determining reply information in a departure scenario in the method embodiment.

[0148] Figure 4 2 is a schematic diagram of a device for determining reply information in a departure scenario according to an embodiment of the present invention. Figure 4 As shown, the device 400 for determining the reply information in the departure scenario may include: an acquisition unit 401 , a classification unit 402 , an input unit 403 , an output unit 404 and an adjustment unit 405 .

[0149] The acquisition unit 401 is used to acquire inquiry information in the departure scenario.

[0150] The classification unit 402 is used to classify the inquiry information to obtain a classification result, wherein the classification result is used to indicate the information urgency and / or inquiry type corresponding to the inquiry information.

[0151] The input unit 403 is used to input the classification result and the inquiry information into the intelligent customer service model for analysis to obtain initial reply information matching the inquiry information, wherein the intelligent customer service model is trained using inquiry information samples in the departure scenario and reply information samples matching the inquiry information samples.

[0152] The output unit 404 is used to output the initial reply information and obtain the feedback result of the initial reply information, wherein the feedback result is used to indicate the satisfaction degree of the question object corresponding to the inquiry information with the initial reply information.

[0153] The adjustment unit 405 is used to adjust the initial response information based on the feedback result to obtain the target response information.

[0154] Optionally, the classification unit 402 may include: an extraction module for extracting keywords from the inquiry information; a matching module for matching the keywords with information in a database to obtain a matching result, wherein the information in the database is historical inquiry information and the information urgency and / or inquiry type corresponding to the historical inquiry information; and a first determination module for determining a classification result based on the matching result.

[0155] Optionally, the device 400 for determining reply information in the departure scenario may further include: a training unit, configured to input the inquiry information sample into the initial customer service model for training to obtain an intelligent customer service model.

[0156] Optionally, the adjustment unit 405 may include: a second determination module, used to determine the initial reply information as the target reply information in response to the feedback result being a satisfactory feedback result; an output module, used to output prompt information in response to the feedback result being an unsatisfactory feedback result, wherein the prompt information is used to prompt the questioner that the inquiry information needs to be adjusted; and a third determination module, used to determine the target reply information based on the prompt information.

[0157] Optionally, the third determination module may include: an acquisition submodule, used to adjust the inquiry information in response to the questioner and obtain the adjusted inquiry information; an input submodule, used to input the adjusted inquiry information into the intelligent customer service model for analysis to obtain target reply information.

[0158] Optionally, the device 400 for determining the reply information in the departure scenario may further include: a first acquisition unit, used to obtain the number of times the feedback result is unsatisfactory; a first output unit, used to output the inquiry information to the manual customer service end in response to the number of times being greater than a preset threshold; and a storage unit, used to store the inquiry information and the manual reply information output by the manual customer service end in a database.

[0159] In this embodiment, inquiry information in the departure scenario is obtained; the inquiry information is classified and processed to obtain a classification result; the classification result and the inquiry information are input into the intelligent customer service model for analysis to obtain initial reply information matching the inquiry information, wherein the intelligent customer service model is trained using inquiry information samples in the departure scenario and reply information samples matching the inquiry information samples; the initial reply information is output, and the feedback result of the initial reply information is obtained; based on the feedback result, the initial reply information is adjusted to obtain the target reply information. That is to say, the embodiment of the present invention obtains inquiry information, classifies the inquiry information, processes the inquiry information using the intelligent customer service model, and obtains reply information. At the same time, the present invention can adjust the reply information according to the satisfaction of the questioning object with the reply information, thereby solving the technical problem of low efficiency in determining reply information in the departure scenario, and achieving the technical effect of improving the efficiency of determining reply information in the departure scenario.

[0160] According to an embodiment of the present invention, a computer-readable storage medium is further provided, the storage medium comprising a stored program, wherein the program executes the method for determining reply information in a departure scenario in an embodiment of the method.

[0161] According to an embodiment of the present invention, a processor is further provided, the processor being used to run a program, wherein when the program is run, the method for determining reply information in a departure scenario in the method embodiment is executed.

[0162] According to an embodiment of the present invention, a computer program product is also provided. The computer program product includes computer instructions. When the computer instructions are executed by a processor, the method for determining reply information in a departure scenario in the method embodiment is implemented.

[0163] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0164] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0165] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0166] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed over multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0167] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0168] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent functional component, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software functional component, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk, etc., which can store program codes.

[0169] The above are only preferred embodiments of the present invention. It should be pointed out that, for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for determining reply information in a departure scenario, characterized in that: include: Obtain inquiry information in departure scenarios; Classify the inquiry information to obtain a classification result, wherein the classification result is used to indicate the information urgency and / or inquiry type corresponding to the inquiry information; Inputting the classification result and the inquiry information into an intelligent customer service model for analysis to obtain initial reply information matching the inquiry information, wherein the intelligent customer service model is trained using inquiry information samples in the departure scenario and reply information samples matching the inquiry information samples; Outputting the initial reply information and obtaining feedback results of the initial reply information, wherein the feedback results are used to indicate the satisfaction degree of the question object corresponding to the inquiry information with the initial reply information; Based on the feedback result, the initial response information is adjusted to obtain target response information.

2. The method according to claim 1, characterized in that: The query information is classified to obtain classification results, including: Extracting keywords from the query information; Matching the keyword with information in a database to obtain a matching result, wherein the information in the database is historical inquiry information and information urgency and / or inquiry type corresponding to the historical inquiry information; Based on the matching result, the classification result is determined.

3. The method according to claim 1, characterized in that The method further comprises: The inquiry information sample is input into the initial customer service model for training to obtain the intelligent customer service model.

4. The method according to claim 1, characterized in that Based on the feedback result, the initial response information is adjusted to obtain target response information, including: In response to the feedback result being a satisfactory feedback result, determining the initial reply information as the target reply information; In response to the feedback result being an unsatisfactory feedback result, outputting prompt information, wherein the prompt information is used to prompt the questioning object that the inquiry information needs to be adjusted; Based on the prompt information, the target reply information is determined.

5. The method according to claim 4, characterized in that Determining the target response information based on the prompt information includes: In response to the question subject adjusting the query information, acquiring the adjusted query information; The adjusted inquiry information is input into the intelligent customer service model for analysis to obtain the target reply information.

6. The method according to claim 1, characterized in that Before adjusting the initial response information based on the feedback result to obtain the target response information, the method further includes: Obtain the number of times the feedback result is unsatisfactory; In response to the number of times being greater than a preset threshold, outputting the inquiry information to a manual customer service terminal; The inquiry information and the manual reply information output by the manual customer service end are stored in a database.

7. A device for determining reply information in a departure scenario, characterized in that: include: An acquisition unit, used to acquire inquiry information in a departure scenario; A classification unit, used to classify the inquiry information to obtain a classification result, wherein the classification result is used to indicate the information urgency and / or inquiry type corresponding to the inquiry information; an input unit, configured to input the classification result and the inquiry information into an intelligent customer service model for analysis to obtain initial reply information matching the inquiry information, wherein the intelligent customer service model is trained using inquiry information samples in the departure scenario and reply information samples matching the inquiry information samples; an output unit, configured to output the initial reply information and obtain a feedback result of the initial reply information, wherein the feedback result is used to indicate the satisfaction degree of the question object corresponding to the inquiry information with the initial reply information; An adjustment unit is used to adjust the initial response information based on the feedback result to obtain target response information.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein when the program is executed by a processor, the device where the storage medium is located is controlled to execute the method according to any one of claims 1 to 6.

9. A processor, characterized in that: The processor is used to run a program, wherein the program executes the method according to any one of claims 1 to 6 when running.

10. A computer program product, characterized in that The computer program product comprises computer instructions, which implement the method according to any one of claims 1 to 6 when executed by a processor.

Images