CN112650829A

CN112650829A - Customer service processing method and device

Info

Publication number: CN112650829A
Application number: CN201910965162.1A
Authority: CN
Inventors: 李嘉特
Original assignee: Alibaba Group Holding Ltd
Current assignee: Alibaba Group Holding Ltd
Priority date: 2019-10-11
Filing date: 2019-10-11
Publication date: 2021-04-13

Abstract

The application relates to a customer service processing method and device. The method comprises the following steps: acquiring a plurality of user query sentences in a customer service session in real time; aggregating the question keywords corresponding to the plurality of user query sentences to determine hot question keywords; and determining response measures matched with the hot spot question keywords. Through the method of the embodiments of the application, the customer service processing method of the embodiments of the application can help a user to quickly find some abnormal problems, especially some sudden problems in a short time, so as to further make defense.

Description

Customer service processing method and device

Technical Field

The application relates to the technical field of big data processing, in particular to a customer service processing method and device.

Background

At present, an intelligent customer service technology is widely applied to enterprises, is an industry-oriented technology developed on the basis of large-scale knowledge processing, and is suitable for various industries such as large-scale knowledge processing, natural language understanding, knowledge management, automatic question and answer systems, reasoning and the like. The intelligent customer service not only provides a fine-grained knowledge management technology for enterprises, but also establishes a quick and effective technical means based on natural language for communication between the enterprises and massive users, and can provide statistical analysis information required by fine management for the enterprises.

In the related art, the intelligent customer service product usually answers the questions provided by the user in a knowledge base mode, and the method can only be generally used for solving the related questions with the answers configured in advance and cannot obtain the answers of the real-time hot-spot questions. In some practical application scenarios, due to the situation that the user traffic is increased suddenly or the system is abnormal suddenly, some abnormal problems may occur, and such problems may occur temporarily, so that the knowledge base does not have answers to such problems.

Therefore, there is a need in the art for a way to enable.

Disclosure of Invention

In order to overcome the problems in the related art, the application provides a customer service processing method and a customer service processing device.

The customer service processing method and the customer service processing device provided by the embodiment of the application are specifically realized as follows:

a method of customer service processing, the method comprising:

acquiring a plurality of user query sentences in a customer service session in real time;

aggregating the question keywords corresponding to the plurality of user query sentences to determine hot question keywords;

and determining response measures matched with the hot spot question keywords.

A customer service processing apparatus comprising a processor and a memory for storing processor-executable instructions, the instructions when executed by the processor effecting:

and determining response measures matched with the hot spot question keywords.

A non-transitory computer readable storage medium, instructions in the storage medium, when executed by a processor, enable the processor to perform the customer service handling method.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the application.

The customer service processing method and the customer service processing device can acquire a plurality of user query sentences from a customer service environment in real time, aggregate the plurality of user query sentences to obtain hot problem keywords, and finally determine response measures matched with the problem keywords. The customer service processing method provided by each embodiment of the application can find hot spots and high-frequency problems in the customer service environment in real time, and particularly can help a user to quickly find some abnormal problems, particularly some sudden problems in short time, in the customer service environment with a rapidly increased consultation amount in short time so as to further make defense.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and together with the description, serve to explain the principles of the application.

FIG. 1 is a schematic diagram illustrating an application scenario in accordance with an exemplary embodiment.

FIG. 2 is a schematic diagram illustrating an application scenario in accordance with an exemplary embodiment.

FIG. 3 is a schematic diagram illustrating an application scenario in accordance with an exemplary embodiment.

FIG. 4 is a schematic diagram illustrating an application scenario in accordance with an exemplary embodiment.

FIG. 5 is a flow diagram illustrating a customer service processing method according to an exemplary embodiment.

FIG. 6 is a block diagram illustrating a customer service processing device according to an example embodiment.

Detailed Description

Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, like numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present application, as detailed in the appended claims.

For the convenience of those skilled in the art to understand the technical solutions provided in the embodiments of the present application, an application environment for implementing the technical solutions is described below.

At present, many enterprises apply intelligent customer service robots, and particularly in the e-commerce industry, the intelligent customer service robots can quickly and accurately answer questions contained in a question bank based on the existing question bank. However, for a question that does not exist in the question bank, the intelligent customer service robot can determine a similar question of the question according to the degree of similarity between the question and the question in the question bank, and display the answer of the similar question. However, in some sudden application scenarios, problems never appeared in some problem libraries often appear, and due to the real-time nature of the problems, enterprises may have difficulty finding the problems in a short time.

In one exemplary scenario, an e-commerce initiates a promotional program starting at 10 months, 30 days, 0. After the preheating period of one week, the user traffic of the e-commerce is increased rapidly at the time starting from 0 of 10 months and 30 days, and the type of the related service is also complicated. However, problems that cannot be predicted in advance, such as the coupon being unusable or the address being unable to be changed, occur very quickly. Moreover, the user may choose to consult the intelligent customer service robot the first time when encountering a problem. When the intelligent customer service robot receives the unpredictable questions, the intelligent customer service robot cannot find answers to the questions, possibly answers questions with low similarity, or does not answer the questions with delay. Then, from the perspective of an enterprise, some hot problems occurring in real time cannot be found in time, and the efficiency of the enterprise for processing emergencies is influenced; from the perspective of the user, a satisfactory answer is not obtained, the problem is not solved, and the shopping experience is influenced.

Based on the above technical requirements, the customer service processing method provided by each embodiment of the application can determine the hot problem keywords from the query sentences of the real-time users, and the hot problem keywords can help enterprises to find real-time hot problems in time, so that further defense measures can be taken. The following describes a customer service processing method provided in various embodiments of the present application through an exemplary scenario, where fig. 1 is a flowchart of a method in an overall technical solution, and fig. 2 is a flowchart of a method for extracting a question keyword and counting a hot question keyword.

An e-commerce initiates a promotional program starting at 10 months, 30 days, 0. After the preheating of one week, starting from 0 of 10 months and 30 days, the query sentences sent by the user to the intelligent customer service are collected, and the customer service processing method steps shown in the figure 1 are set to be executed every 5 minutes. As the flow of the promotion activities is increased, correspondingly, the query sentences sent by the users to the intelligent customer service are also in an increased state. Up to 10 months, 30 days, 0 point 5 points, a total of 200 ten thousand user query statements are collected, as shown in fig. 1, query 2, query 3, … …, and query N. The user query statement may include "shipping address not modifiable, online, etc.! "," pay 5 minutes, still can not pay successfully, sell no you on the end "," say that the first 15 minutes was good, how 3 minutes of the activity ended? "," how cannot red packages? "and the like.

After 200 ten thousand user query sentences are collected, the question keywords in the user query sentences can be extracted respectively. In particular, in practice, these user query statements may first be matched to a question database, which may include a plurality of question keywords, which may include such phrases as "seller Do not deal," "postage undertaker dispute," "complaint entry consultation," "expedited delivery," "seller attitudinal," and so forth. And matching the user query sentence with the question keywords in the question database, wherein if the matching is successful, the user query sentence and the question keywords express the same meaning, and at this time, the question keywords of the user query sentence, namely the first batch of question keywords in fig. 2, can be determined. If some problems that never occur or have a low probability of occurring historically occur, the user query statement may not match any of the problem keywords in the problem database, in which case the user query statement may be input to a problem recognition model component, and the problem keywords in the user query statement may be recognized by the problem recognition model component. The question recognition model component may be obtained by training using data in the question database, and specifically, may be obtained by training using a question keyword in the question database and a corresponding relationship between the question keyword and a historical user query statement corresponding to the question keyword.

Since the problem identification model component outputs the likelihood of corresponding to each problem keyword, the likelihood can be measured by a probability value. For the problem keyword with a higher probability value (e.g., greater than 0.65), the problem keyword with the higher probability value may be determined as the problem keyword of the user query sentence, i.e., the second group of problem keywords in fig. 1. On the contrary, for example, the probability of belonging to the question keyword "seller is not administrative", the probability of belonging to the question keyword "postage undertaker dispute" is 0.1, the probability of belonging to the question keyword "complaint entry consultation" is 0.03, the probability of belonging to the question keyword "complaint entry consultation" is 0.07, and the probability of belonging to the question keyword "delivery hasten" is 0.01, so that it can be seen that the above probability values are all low and are not significant, and in this case, it can be determined that the question recognition model component cannot recognize the question keyword in the user query sentence. Based on this, keyword extraction may be performed on the user query sentence, and the extracted algorithm may include Textrank algorithm, etc., so as to generate the third batch of question keywords.

After passing through a relatively comprehensive multi-level algorithm, the coverage rate of extracting the question keywords of the above 200 ten thousand user query sentences can be greatly improved, for example, a total of 190 ten thousand question keywords are extracted. Then, the 190 ten thousand question keywords can be counted, for example, 3000 question keywords are counted. Finally, the first 50 question keywords with relatively high frequencies can be selected as the final hot spot question keywords.

Of course, after the hot spot question keywords are determined, some response measures may be set according to the hot spot question keywords. In the above scenario, the first-ranked hotspot problem keyword is determined to be "red packet unusable". Based on this, some response measures for "red package not available" may be set, which may include issuing some notification message to the buyer or sending a reminder message to the merchant. The user interface shown in fig. 3 is an app home page of XX treasure on the buyer client, and as shown in fig. 3, a floating window may be set up in the app home page to show a notification message about "red package is not available", and the content of the notification message may include "respect user, good! Due to the rapid increase of the transaction amount during the activity, the system is out of order, and part of the red packets cannot be used. For the unusable red envelope, after you receive the goods, please contact the customer service to apply for refund, thank you for understanding! "by this notification message, the user XX treasure platform can be informed about the" red pack is unavailable "failure resolution.

On the other hand, response measures for the hot spot question keywords of the merchant can be set, and fig. 4 is a user interface of the XX treasure of the merchant client. After the hotspot problem keyword of 'red package unavailable' is acquired, a notification message about the hotspot problem keyword can be pushed to the merchant. The notification message may include "respect Merchant, good! Due to the rapid increase of the transaction amount during the activity, the system is out of order, and part of the red packets cannot be used. At that point, the buyer is notified and, after receiving the package, the buyer can apply for a return red envelope. "through the notification message, the merchant XX treasure platform can be informed about the resolution of the fault of" red packet is unavailable ". After receiving the notification message, the merchant may notify the buyer of the processing measure of the failure of "no use of red envelope" after receiving the inquiry of the buyer, so as to further refine the response to the inquiry of the user.

Of course, in other implementation scenarios, after determining the hotspot problem keyword, the method is not limited to sending a notification message to the buyer or the merchant client, and measures such as repairing a system fault, closing a fault entrance, and the like may also be taken, and the application is not limited herein.

The customer service processing method described in the present application is described in detail below with reference to the accompanying drawings. Fig. 5 is a flowchart illustrating a method of an embodiment of a method for determining a hotspot problem provided by the present application. Although the present application provides method steps as shown in the following examples or figures, more or fewer steps may be included in the method based on conventional or non-inventive efforts. In the case of steps where no necessary causal relationship exists logically, the order of execution of the steps is not limited to that provided by the embodiments of the present application. The method can be executed in sequence or in parallel according to the method shown in the embodiment or the figures (for example, in the environment of a parallel processor or a multi-thread processing) in the actual customer service process or when the device executes.

Specifically, an embodiment of the method for determining a hotspot problem provided by the present application is shown in fig. 5, where the method may include:

s501: and acquiring a plurality of user query sentences in the customer service session in real time.

S503: and aggregating the question keywords corresponding to the plurality of user query sentences to determine the hot question keywords.

S505: and determining response measures matched with the hot spot question keywords.

In the embodiment of the present application, the multiple user query sentences may be obtained in real time, and in an example, if the current time is 16: 39/7/9/2019, the obtaining of the multiple user query sentences in real time may include obtaining multiple user query sentences in a time period before 16: 39/7/9/7/2019, for example, user query sentences within 16:38-16:39 minutes. The time period may be a fixed preset time period, such as 1 minute, 3 minutes, 5 minutes, and so on, i.e., S201-S207 are performed every other preset time period. Of course, the time period may also be not fixed, for example, the time interval for executing S201 to S207 is determined by the number of the obtained user query statements, for example, when the number of the user query statements is 8000, S201 to S207 may be executed once, and the time interval for executing S201 to S207 is not limited in the embodiment of the present application.

The user query statement in the embodiment of the present application may include a statement transmitted by a user through a question consultation interface, which may include, for example, an interface of an intelligent customer service robot, and the like. Through the question consultation interface, the user can express the meaning of some question consultation. In some examples, the user query statement may include "chat she does not care with the seller", "i am this order seller says that the warranty was a year, is now out of question for a short time", "make a logistic call unavailable", and so on.

In an embodiment of the present application, at least one clause corresponding to the same session may be used as a user query statement. In an actual application scenario, a user expresses a plurality of sentences, but often only reflects one or two core problems. For example, in the following example, a user expresses N words in one dialog: sentence 1, "last week of this order", sentence 2 "monday me received the goods, found that one of the bags walnut has grown wool", sentence 3, "monday afternoon did not reply with seller feedback all the time", sentence 4 "seller don't know how to handle" sentence 5 "i can contact him how", sentence 6 "i want to complain him", … …, sentence N "shogao chen", according to the above user expression sentences, two core problems can be reflected: the quality of the goods is problematic and the seller is not connected. However, if each of the above user expression sentences is used as a user query sentence, the same problem may be counted many times, for example, sentences 3 to 5 are all problems expressing that the seller is not connected. Based on this, at least one clause corresponding to the same session may be taken as one user query statement. A session may comprise a complete session, for example, the query statements that the user has expressed in an uninterrupted conversation may all be the user query statements in the same session. For example, in the above example, the N sentences of the user may be expressed as the contents of the same user query sentence.

In this embodiment, after the plurality of user query sentences are acquired, the plurality of user query sentences may be respectively matched with question keywords in a question database. The question database may include a plurality of question keywords, which may include phrases or phrases for briefly expressing the question content, and in some examples, the question keywords may include such phrases as "seller disregarder", "postal fee undertaker dispute", "complaint entrance consultation", "urging to ship", "seller with bad attitude", and the like.

The problem database in the embodiment of the present application may be constructed and generated as follows: a plurality of historical user query sentences are acquired, the historical user query sentences may include user query sentences which have been generated historically, for example, the user query sentences in the intelligent customer service robot between 1 and 1 day of 2018 and 30 and 6 months of 2019 may be acquired. After the plurality of user query sentences are acquired, word segmentation processing may be performed on the plurality of historical user query sentences, respectively, to determine at least one word segmentation in the historical user query sentences. In one example, after the query statement "Monday afternoon seller feedbacks all the time" participles are not replied, "a plurality of participles such as" Monday | afternoon | seller | feedbacks | all the time | not replied "are obtained. Before or after the word segmentation process, some stop words of the historical query statement may be removed, leaving some useful words. After determining the participles in each historical query sentence, clustering at least one participle corresponding to each of the plurality of historical user query sentences, and generating at least one question keyword according to a result of the clustering. In one embodiment, after 1 ten thousand historical user query sentences are acquired and word segmentation processing is performed on the historical user query sentences, a total of 8 ten thousand segmented words can be acquired, and then clustering processing can be performed on the 8 ten thousand segmented words to determine a plurality of problem keywords. In an embodiment, the clustering processing mode may include a clustering algorithm such as an N-gram, after the clustering processing, a plurality of classifications may be obtained, and then, the participles corresponding to the larger number of classifications may be used as the problem keywords, for example, 500 problem keywords may be generated finally.

In an actual application environment, the expressions with the same meaning by the user are often different, for example, in the embodiment of the present application, after the problem keyword is determined, synonym expansion may be performed on the problem keyword respectively. The "no use of red envelope", "no use of red envelope" and "no use of red envelope" all mean the same meaning, and therefore, after synonym expansion is performed on the question keyword in the question database, if the user query sentence can be matched with any one of the question keyword and its synonym, it can be determined that the user query sentence is matched with the question keyword.

In the embodiment of the present application, in the process of determining whether the user query sentence matches the question keyword, it may be determined whether a phrase or a phrase combination identical or similar to the question keyword and a synonym thereof appears in the user query sentence, for example, "please help me to find, how can red envelope not be used? The phrase "the user inquires" includes two phrases "red envelope" and "cannot use" and after combination, the phrase "cannot use red envelope" is matched with the question keywords "cannot use red envelope", "cannot use red envelope" and "cannot use red envelope" in the question center library.

With the development of business, more and more problems may occur on the enterprise platform, including some problems that have never occurred before. Based on this, in the embodiment of the present application, in order to make the problem database more suitable for the continuously updated service development requirement, the problem database may be updated. In the embodiment of the present application, the updating manner may include offline data updating and real-time data updating. The problem database may be updated with some offline data, such as the previous week or month. The real-time data update may be, for example, to update the question database with data within a time period before the current time, where the time period may be set to 3 minutes, 5 minutes, 10 minutes, and so on, i.e., to update the question database in real time every other time period. Of course, the problem database may also be updated according to a time period, and the application is not limited herein.

In the embodiment of the present application, before performing offline data update or real-time data update on the question database, a plurality of offline user query data or instant user query statements may be obtained, where an instant user query statement is a user query statement obtained in real time. Then, the offline user query data or the instant user query sentence may be subjected to clustering processing after word segmentation processing, and at least one candidate question keyword may be generated. The specific word segmentation processing and clustering processing are the same as the processing mode of the historical user query data when the problem database is constructed, and the application is not limited herein. After the candidate problem keywords are obtained, the similarity between the candidate problem keywords and existing problem keywords in the problem database can be calculated, and if the similarity is greater than a preset threshold, the candidate problem keywords can be removed, so that the problem database does not include repeated problem keywords. If the similarity is less than or equal to a preset threshold, the candidate question keywords may be added to the question database.

In a practical application environment, it may be difficult to add the questions involved in the historical user query statement, the offline user query statement, and the instant user query statement to the question database, so as to avoid increasing the calculation time for matching the user query statement with the question keywords in the question database. For example, in the above example, after 8 ten thousand of participles are clustered, 4000 question phrases or phrases are formed together, but only 500 of the participles with higher occurrence frequency are finally selected as the question keywords and added to the question data. However, some questions, which have a low probability of being temporarily generated and a high frequency of being generated later, may remain in the remaining 3500 question phrases or phrases. Thus, to promote comprehensiveness of finding a question, a question recognition model component can be utilized to determine question keywords in the user query statement. Specifically, at least part of the user query sentences may be input into a question recognition model component, and question keywords corresponding to the user query sentences may be output through the question recognition model component.

In the embodiment of the present application, the question identification model component may be obtained by training using correspondence between a plurality of user query sample sentences and question keywords of the user query sample sentences. Wherein the user query sample statement may include some user query statements that have occurred historically. In one embodiment, the user query sample statement may include a historical user query statement used for generating the question database, or an offline user query statement or a real-time user query statement used for updating the question database, so that some data can be reused, the training cost of the question identification model component is saved, and on the other hand, the historical user query statement, the offline user query statement or the real-time user query statement added to the question database after being screened tends to have higher use value, so that the question identification model component has more accurate identification capability.

In one embodiment of the present application, the user query sample sentence may be further labeled with a word segmentation and a word segmentation part of speech to train the problem recognition model component to have the capability of word segmentation and word segmentation part of speech determination. In this embodiment, the problem recognition model component may be further constructed, and training parameters may be set in the problem recognition model component. The user query sample statements may then be input into the problem recognition model component, respectively, to generate prediction results. Finally, iterative adjustment may be performed on the training parameters based on a difference between the prediction result and the problem keyword until the difference meets a preset requirement. It should be noted that, in the case where the problem recognition model component is a multi-layer neural network. The word segmentation and word part-of-speech prediction part may be disposed in an intermediate layer of the multilayer neural network, or may be disposed in an output layer of the multilayer neural network, and the present application is not limited herein.

It should be noted that the problem recognition model component may include a model component trained by using a machine learning method. The machine learning mode can also comprise a K nearest neighbor algorithm, a perception machine algorithm, a decision tree, a support vector machine, a logistic background regression, a maximum entropy and the like, and correspondingly, the generated model components such as naive Bayes, hidden Markov and the like. Of course, in other embodiments, the machine learning manner may further include a deep learning manner, a reinforcement learning manner, and the like, and the generated model component may include a Convolutional Neural Network model Component (CNN), a Recurrent Neural Network model component (RNN), LeNet, ResNet, a Long-Short Term Memory Network model component (LSTM), a bidirectional Long-Short Term Memory Network model component (Bi-LSTM), and the like, which is not limited herein.

Further, in the embodiment of the present application, the problem identification model component may also be updated. In some embodiments, the problem identification model component may be updated at a fixed frequency, for example, every other week or 15 days. Of course, in other embodiments, the update may also be performed according to the number of accumulated user query sample sentences after the last update, and in one example, when the number of accumulated user query sample sentences reaches 3000, for example, the problem identification model component may be updated once. Of course, in other embodiments, the problem identification model component may be updated at any point in time, and the application is not limited thereto.

It should be noted that, for some user query sentences, it is possible that the problem keyword is not matched, and the result probability of the output of the model component identified by the problem is also low (because the output of the model component is the probability value of each problem keyword). In this case, the algorithm such as Textrank can be used to extract the question keywords in the user query sentence, so as to improve the coverage of the mining of the question keywords in the user query sentence.

In the embodiment of the present application, after at least one question keyword in the plurality of user query sentences is determined, statistics may be performed on the question keywords, that is, the number of times each question keyword appears is determined. Then, the hotspot problem keyword with a higher occurrence frequency may be determined according to the occurrence frequency of the problem keyword, for example, the first 50 problem keywords may be used as the hotspot problem keyword, and of course, any other number of the problem keywords may be used, which is not limited herein. In addition, a problem keyword whose occurrence frequency is greater than a preset frequency may be used as the hot problem keyword, for example, a problem keyword whose occurrence frequency is greater than 1000 times may be used as the hot problem keyword. Of course, the hotspot problem keywords may be determined from the statistical problem keywords in any other screening manner, which is not limited herein.

In the embodiment of the application, after the hot spot question keyword is determined, the answer of the hot spot question keyword can be configured, and the answer is displayed under the condition that a user query sentence contains the hot spot question keyword. In one example, after the above-described S21-S25 is performed for 1 ten thousand user query sentences acquired in real time, it is determined that the hotspot problem of the first name is "no red envelope used" ("no red envelope used", "no red envelope expired"). The answer to the hot question may then be configured, and in one example, the answer may be "sorry, because the system crashed so that the red purse cannot be used for this day's activities, and after you have received the good, the system will automatically return the red purse to your account". Then, subsequent users asking such questions can present the answer to the asking user.

In the embodiment of the application, whether the hot spot problem keyword belongs to the problem of the fault type can be further judged, and the fault early warning message is sent under the condition that the hot spot problem keyword belongs to the problem of the fault type. The fault type problem may include some types of problems related to server faults, client faults, business logic faults, and the like. By sending the fault early warning message to related personnel, the method can help the related personnel to quickly find some fault problems so as to quickly solve the fault problems.

In this embodiment of the present application, after determining the hot problem keyword, at least one of the following indicators of the user query statement corresponding to the hot problem keyword in the preset time period may be counted: total number, increment and increment rate relative to hot spot problem keywords in the last preset time period. The preset time period is the time period of the obtained query sentences of the users. In one example, the first question keyword in the first 5 minutes of the current time is determined to be "merchant unawary", 83 times of the first question keyword are appeared in the 5 minutes, and 105 times of the first question keyword are appeared in the 5 minutes, so that the rate of increase of the appearance of the question is-20.95%. Through the statistical mode, the occurrence frequency and the growth rate of various problem keywords in each statistical period can be determined. Then, some of the question keywords with high occurrence frequency or growth rate may be defended, such as setting the answer to the question, issuing a unified bulletin, and so on, as described in the above embodiments.

The customer service processing method can acquire a plurality of user query sentences from a customer service environment in real time, aggregate the plurality of user query sentences to obtain hot problem keywords, and finally determine response measures matched with the problem keywords. The customer service processing method provided by each embodiment of the application can find hot spots and high-frequency problems in the customer service environment in real time, and particularly can help a user to quickly find some abnormal problems, particularly some sudden problems in short time, in the customer service environment with a rapidly increased consultation amount in short time so as to further make defense.

Corresponding to the above customer service processing method, as shown in fig. 6, the present application further provides a customer service processing apparatus, including a processor and a memory for storing processor executable instructions, where the processor executes the instructions to implement:

and determining response measures matched with the hot spot question keywords.

Optionally, in an embodiment of the present application, the aggregating, by the processor, the question keywords corresponding to the plurality of query sentences in the implementing step, and determining the hot question keyword includes:

matching the plurality of user query sentences with question keywords in a question database respectively, and if the matching is successful, determining the question keywords corresponding to the user query sentences;

respectively inputting at least part of the user query sentences into a question identification model component, and outputting question keywords corresponding to the user query sentences through the question identification model component;

and counting the problem keywords corresponding to the plurality of user query sentences, and determining the hot problem keywords.

Optionally, in an embodiment of the present application, the question database is configured to be generated as follows:

acquiring a plurality of historical user query sentences;

performing word segmentation processing on the plurality of historical user query sentences respectively, and determining at least one word segmentation in the historical user query sentences;

and clustering at least one participle corresponding to each of the plurality of historical user query sentences, and generating at least one problem keyword according to the clustering result.

Optionally, in an embodiment of the present application, the question database is configured to be updated in the following manner:

acquiring a plurality of off-line user query sentences or real-time user query sentences;

performing word segmentation processing on the plurality of offline user query sentences or real-time user query sentences respectively, and determining at least one word segmentation in the plurality of offline user query sentences or real-time user query sentences;

clustering at least one participle corresponding to each of the off-line user query sentences or the real-time user query sentences, and generating at least one candidate problem keyword according to the result of clustering;

and removing keywords with similarity larger than a preset threshold value with the question keywords in the question database from the at least one candidate question keyword, and adding the remaining candidate question keywords into the question database.

Optionally, in an embodiment of the present application, the user query statement includes at least one clause in the same session.

Optionally, in an embodiment of the present application, the problem recognition model component is configured to be trained in the following manner:

acquiring a plurality of user query sample sentences and question keywords corresponding to the user query sample sentences, wherein the user query sample sentences are marked with word segments and word parts of words;

constructing a problem recognition model component, wherein training parameters are set in the problem recognition model component;

inputting the user query sample sentences into the problem identification model component respectively to generate a prediction result;

and iteratively adjusting the training parameters based on the difference between the prediction result and the question keyword until the difference meets a preset requirement.

Optionally, in an embodiment of the present application, the processor further implements the following steps:

and configuring answers of the hot problem keywords, and displaying the answers under the condition that a user query statement contains the hot problem keywords.

Optionally, in an embodiment of the present application, when the processor determines the response measure matching the hotspot question keyword in the implementing step, the processor includes:

judging whether the hot spot problem keywords belong to the problem of the fault type;

and sending a fault early warning message under the condition that the hot spot problem keywords belong to the fault type problem.

Optionally, in an embodiment of the present application, when the implementing step obtains the query statements of the plurality of users in the customer service session in real time, the processor includes:

and acquiring a plurality of user inquiry sentences in the customer service session within a preset time period before the current moment, wherein the preset time period is a set time interval for determining the hot problem keywords.

counting at least one of the following indexes of the user query statement corresponding to the hot problem keyword in the preset time period: total number, increment and increment rate relative to hot spot problem keywords in the last preset time period.

Optionally, in an embodiment of the application, the processor, in the step of implementing, counts the question keywords corresponding to the plurality of user query sentences, and when determining the hotspot question keyword, includes:

determining a probability value corresponding to a problem keyword output by the problem identification model component;

under the condition that the probability value is smaller than a preset threshold value, extracting keywords of the corresponding user query sentence, and re-determining the problem keywords of the user query sentence;

In another aspect, the present application further provides a computer-readable storage medium, on which computer instructions are stored, and the instructions, when executed, implement the steps of the method according to any of the above embodiments.

The computer readable storage medium may include physical means for storing information, typically by digitizing the information for storage on a medium using electrical, magnetic or optical means. The computer-readable storage medium according to this embodiment may include: devices that store information using electrical energy, such as various types of memory, e.g., RAM, ROM, etc.; devices that store information using magnetic energy, such as hard disks, floppy disks, tapes, core memories, bubble memories, and usb disks; devices that store information optically, such as CDs or DVDs. Of course, there are other ways of storing media that can be read, such as quantum memory, graphene memory, and so forth.

In the 90 s of the 20 th century, improvements in a technology could clearly distinguish between improvements in hardware (e.g., improvements in circuit structures such as diodes, transistors, switches, etc.) and improvements in software (improvements in process flow). However, as technology advances, many of today's process flow improvements have been seen as direct improvements in hardware circuit architecture. Designers almost always obtain the corresponding hardware circuit structure by programming an improved method flow into the hardware circuit. Thus, it cannot be said that an improvement in the process flow cannot be realized by hardware physical modules. For example, a Programmable Logic Device (PLD), such as a Field Programmable Gate Array (FPGA), is an integrated circuit whose Logic functions are determined by programming the Device by a user. A digital system is "integrated" on a PLD by the designer's own programming without requiring the chip manufacturer to design and fabricate application-specific integrated circuit chips. Furthermore, nowadays, instead of manually making an Integrated Circuit chip, such Programming is often implemented by "logic compiler" software, which is similar to a software compiler used in program development and writing, but the original code before compiling is also written by a specific Programming Language, which is called Hardware Description Language (HDL), and HDL is not only one but many, such as abel (advanced Boolean Expression Language), ahdl (alternate Hardware Description Language), traffic, pl (core universal Programming Language), HDCal (jhdware Description Language), lang, Lola, HDL, laspam, hardward Description Language (vhr Description Language), vhal (Hardware Description Language), and vhigh-Language, which are currently used in most common. It will also be apparent to those skilled in the art that hardware circuitry that implements the logical method flows can be readily obtained by merely slightly programming the method flows into an integrated circuit using the hardware description languages described above.

The controller may be implemented in any suitable manner, for example, the controller may take the form of, for example, a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro) processor, logic gates, switches, an Application Specific Integrated Circuit (ASIC), a programmable logic controller, and an embedded microcontroller, examples of which include, but are not limited to, the following microcontrollers: the ARC625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320, the memory controller may also be implemented as part of the control logic for the memory. Those skilled in the art will also appreciate that, in addition to implementing the controller as pure computer readable program code, the same functionality can be implemented by logically programming method steps such that the controller is in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers and the like. Such a controller may thus be considered a hardware component, and the means included therein for performing the various functions may also be considered as a structure within the hardware component. Or even means for performing the functions may be regarded as being both a software module for performing the method and a structure within a hardware component.

The systems, devices, modules or units illustrated in the above embodiments may be implemented by a computer chip or an entity, or by a product with certain functions. One typical implementation device is a computer. In particular, the computer may be, for example, a personal computer, a laptop computer, a cellular telephone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

For convenience of description, the above devices are described as being divided into various units by function, and are described separately. Of course, the functionality of the units may be implemented in one or more software and/or hardware when implementing the present application.

As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

In a typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory. The memory may include forms of volatile memory in a computer readable medium, Random Access Memory (RAM) and/or non-volatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). Memory is an example of a computer-readable medium.

Computer-readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), Digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device. As defined herein, a computer readable medium does not include a transitory computer readable medium such as a modulated data signal and a carrier wave.

It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.

As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

The application may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The application may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.

The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the system embodiment, since it is substantially similar to the method embodiment, the description is simple, and for the relevant points, reference may be made to the partial description of the method embodiment.

The above description is only an example of the present application and is not intended to limit the present application. Various modifications and changes may occur to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A method of customer service processing, the method comprising:

and determining response measures matched with the hot spot question keywords.

2. The method of claim 1, wherein aggregating the question keywords corresponding to the plurality of query sentences to determine hot question keywords comprises:

3. The method of claim 2, wherein the question database is configured to be generated by:

acquiring a plurality of historical user query sentences;

4. A method according to claim 2, wherein the question database is arranged to be updated in the following manner:

5. The method of claim 1, wherein the user query statement comprises at least one clause in the same session.

6. The method of claim 2, wherein the problem recognition model component is configured to be trained in the following manner:

7. The method of claim 1, further comprising:

8. The method of claim 1, wherein determining the response measure matching the hotspot question keyword comprises:

9. The method of claim 1, wherein the obtaining a plurality of user query statements in a customer service session in real-time comprises:

10. The method of claim 9, further comprising:

11. The method of claim 2, wherein the counting the question keywords corresponding to the plurality of user query sentences and determining the hotspot question keywords comprises:

12. A customer service processing apparatus comprising a processor and a memory for storing processor-executable instructions, the instructions when executed by the processor performing:

and determining response measures matched with the hot spot question keywords.

13. The apparatus of claim 12, wherein the processor, when implementing the aggregating the question keywords corresponding to the plurality of query sentences, determines the hot question keyword, comprises:

14. The apparatus of claim 13, wherein the question database is configured to be generated as follows:

acquiring a plurality of historical user query sentences;

15. The apparatus of claim 13, wherein the question database is configured to be updated as follows:

16. The apparatus of claim 12, wherein the user query statement comprises at least one clause in the same session.

17. The apparatus of claim 13, wherein the problem recognition model component is configured to be trained in the following manner:

18. The apparatus of claim 12, wherein the processor further implements the steps of:

19. The apparatus of claim 12, wherein the processor, when implementing the step of determining the response measure matching the hotspot question keyword, comprises:

20. The apparatus of claim 12, wherein the processor, when implementing the steps to obtain the plurality of user query statements in the customer service session in real time, comprises:

21. The apparatus of claim 20, wherein the processor further implements the steps of:

22. The apparatus of claim 13, wherein the processor, when performing the step of counting the question keywords corresponding to the plurality of user query sentences and determining the hot question keywords, comprises:

23. A non-transitory computer readable storage medium, wherein instructions in the storage medium, when executed by a processor, enable the processor to perform the customer service handling method of any of claims 1-11.