Online customer service evaluation system for cross-border e-commerce

Abstract

The invention provides an online customer service evaluation system for cross-border e-commerce, and relates to the field of customer service evaluation. The online customer service evaluation system for cross-border e-commerce comprises a customer evaluation module, a customer sensitive word evaluation module, a customer service sensitive word evaluation module, a repeated content module, a response time module, an emotion analysis module and a collection evaluation module, and performs weight distribution according to scores obtained by the emotion analysis module. The customer service is comprehensively evaluated through a weighted average method by integrating scores of the customer evaluation module, the customer sensitive word evaluation module, the customer service sensitive word evaluation module, the repeated content module and the response time module, and a total score is obtained. The weights of response time modules of the customer evaluation module, the customer sensitive word evaluation module, the customer service sensitive word evaluation module and the repeated content module for customers with different emotions are fluctuated, and the customer service evaluation fairness is improved.

Description

Technical Field

[0001] The present invention relates to the field of customer service evaluation, and in particular to an online customer service evaluation system for cross-border e-commerce. Background Art

[0002] Cross-border e-commerce refers to an international business activity in which trading entities in different customs territories reach transactions through e-commerce platforms, conduct electronic payment settlements, and deliver goods through cross-border e-commerce logistics and remote warehousing to complete transactions.

[0003] Cross-border e-commerce refers to a new e-commerce application model in which trading entities in different countries or regions use e-commerce platforms to carry out various activities of commodity trading, and use cross-border logistics to realize the flow of commodities from sellers to buyers and other related activities. Cross-border e-commerce originated from e-commerce and is a new application model of e-commerce, including overseas shopping, purchasing on behalf of others, cross-border retail, and cross-border B2B models. All commercial activities that cross borders with the help of e-commerce models fall into the category of cross-border e-commerce.

[0004] In the field of e-commerce, customer service is mainly used to connect with customers and solve their problems. In order to improve customer satisfaction, after the conversation between the customer and the customer service, it is generally necessary to evaluate the customer service service accordingly. The existing customer service evaluation is generally evaluated through parameters such as customer feedback, response time, and the number of sensitive words. However, this evaluation method is affected to a certain extent by the customer's emotions. For example, when a customer is in a bad mood, the customer feedback at this time is mostly subjective rather than objective, and the feedback at this time is mostly negative. Then this type of customer service evaluation is one-sided and cannot accurately reflect the customer service service.

[0005] Therefore, it is necessary to provide a new online customer service evaluation system for cross-border e-commerce to solve the above technical problems. Summary of the invention

[0006] In order to solve the above technical problems, the present invention provides an online customer service evaluation system for cross-border e-commerce.

[0007] The online customer service evaluation system for cross-border e-commerce provided by the present invention includes:

[0008] Customer evaluation module: used to analyze the evaluations given by customers and give ratings;

[0009] Customer sensitive word evaluation module: import the data of customer sent information into the pre-established sensitive word database, analyze the frequency of sensitive words, and make scores;

[0010] Customer service sensitive word evaluation module: imports the data sent by customer service into the pre-established sensitive word database, analyzes the frequency of sensitive words, and performs scoring;

[0011] Duplicate content module: Analyze the same content based on the information sent by the customer, obtain the number of times the same content is repeated, and score it;

[0012] Response time module: Analyzes the time from when the customer service replies to the last message sent by the customer before the reply, calculates the time difference between the two, obtains the response time, and performs scoring;

[0013] Sentiment analysis module: import the information sent by customers before customer service goes online into the sentiment quality analysis model, analyze customer sentiment, and score it;

[0014] Collection and evaluation module: Collect the scores of the customer evaluation module, customer sensitive word evaluation module, customer service sensitive word evaluation module, repeated content module, response time module and sentiment analysis module, assign weights based on the scores obtained from the sentiment analysis module, and use the weighted average method to comprehensively evaluate customer service based on the scores of the customer evaluation module, customer sensitive word evaluation module, customer service sensitive word evaluation module, repeated content module and response time module to obtain a total score.

[0015] Preferably, obtaining the emotion quality analysis model comprises the following steps:

[0016] S201: Collect text data with sentiment labels from social media comments, product reviews, customer service conversation records, and news comment channels to build a dataset;

[0017] S202: Cleaning the data set;

[0018] S203: Divide the data set into a training set and a test set;

[0019] S204: Use TF-IDF to perform feature extraction and convert text data into numerical features;

[0020] S205: Build a random forest model;

[0021] S206: input the training set into the model and perform multiple rounds of iterative training until the model converges;

[0022] S207: Verify the performance of the converged model based on the test set to obtain a sentiment quality analysis model.

[0023] Preferably, the scores of the customer evaluation module, customer sensitive word evaluation module, customer service sensitive word evaluation module, duplicate content module, response time module, and sentiment analysis module are all 1-10. The higher the score, the higher the corresponding customer evaluation, the fewer the number of customer sensitive words, the fewer the number of customer service sensitive words, the fewer the number of repeated identical contents, the shorter the response time, and the better the customer sentiment.

[0024] Preferably, the sensitive word analysis of the customer service or customer includes the following steps:

[0025] S401: Collect the conversation text of the customer service or customer, including the information sent by the customer and the information sent by the customer service;

[0026] S402: Delete the punctuation marks and special characters in the text;

[0027] S403: Use the string matching method to check whether the text contains sensitive words;

[0028] S404: Count the number of sensitive words appearing in the conversation text sent by the customer service or customer.

[0029] Preferably, the data set is divided according to the ratio of 70% training set and 30% test set.

[0030] Preferably, the cleaning of the data set includes the following steps:

[0031] S601: Delete the punctuation marks and special characters in the text;

[0032] S602: Convert all text data to lowercase to avoid inconsistencies caused by case;

[0033] S603: Remove common stop words, such as "de", "shi", "zai", etc.;

[0034] S604: Stem extraction and lemmatization.

[0035] Preferably, the response time module is scored as follows:

[0036] If the response time is greater than 120s, the score is 1;

[0037] If the response time is 91-120s, the score is 2;

[0038] If the response time is 61-90s, the score is 3;

[0039] If the response time is 46-60s, the score is 4;

[0040] If the response time is 31-45s, the score is 5;

[0041] If the response time is 21-30s, the score is 6;

[0042] The response time is 11-20s, and the score is 7;

[0043] The response time is 6-10s, and the score is 8;

[0044] The response time is 3-5s, and the score is 9;

[0045] The response time is less than 3s, with a score of 10.

[0046] Compared with related technologies, the online customer service evaluation system for cross-border e-commerce provided by the present invention has the following beneficial effects:

[0047] 1. When evaluating online customer service, the present invention introduces a sentiment analysis module, imports the information sent by the customer before the customer service goes online into the sentiment quality analysis model, analyzes the customer's sentiment, obtains the corresponding sentiment level, and performs scoring. Through the sentiment analysis module, the evaluation dimension is increased, the evaluation result is more comprehensive and objective, the interference of human factors is reduced, and the evaluation result is more accurate.

[0048] 2. In the present invention, by scoring the corresponding emotions and flexibly allocating weights according to the emotion levels, the weights of the response time modules of the customer evaluation module, customer sensitive word evaluation module, customer service sensitive word evaluation module, and repeated content module for customers with different emotions will fluctuate to a certain extent. By using different weight allocations, the evaluation is made more personalized and targeted, and the fairness of the customer service evaluation is also improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 A process wireframe diagram of the online customer service evaluation system for cross-border e-commerce provided by the present invention;

[0050] Figure 2 A wireframe diagram of the process of obtaining the emotion quality analysis model provided by the present invention;

[0051] Figure 3 A wireframe diagram of the process of performing the sensitive word analysis on customer service or customers provided by the present invention. DETAILED DESCRIPTION

[0052] The present invention will be further described below in conjunction with the accompanying drawings and implementation modes.

[0053] Please refer to Figure 1-Figure 3 ,in, Figure 1 A process wireframe diagram of the online customer service evaluation system for cross-border e-commerce provided by the present invention; Figure 2 A wireframe diagram of the process of obtaining the emotion quality analysis model provided by the present invention; Figure 3A wireframe diagram of the process of performing the sensitive word analysis on customer service or customers provided by the present invention.

[0054] In the specific implementation process, Figure 1-Figure 3 As shown, including:

[0055] Customer evaluation module: used to analyze the evaluations given by customers and give ratings;

[0056] Customer sensitive word evaluation module: import the data of customer sent information into the pre-established sensitive word database, analyze the frequency of sensitive words, and make scores;

[0057] Customer service sensitive word evaluation module: imports the data sent by customer service into the pre-established sensitive word database, analyzes the frequency of sensitive words, and performs scoring;

[0058] Duplicate content module: Analyze the same content based on the information sent by the customer, obtain the number of times the same content is repeated, and score it;

[0059] Response time module: Analyzes the time from when the customer service replies to the last message sent by the customer before the reply, calculates the time difference between the two, obtains the response time, and performs scoring;

[0060] Sentiment analysis module: import the information sent by customers before customer service goes online into the sentiment quality analysis model, analyze customer sentiment, and score it;

[0061] When scoring each module, the customer evaluation module, customer sensitive word evaluation module, customer service sensitive word evaluation module, repeated content module, response time module, and sentiment analysis module are all scored on a scale of 1-10, with higher scores corresponding to higher customer evaluation, fewer customer sensitive words, fewer customer service sensitive words, fewer repetitions of the same content, shorter response time, and better customer sentiment;

[0062] Collection and evaluation module: collects the scores of the customer evaluation module, customer sensitive word evaluation module, customer service sensitive word evaluation module, repeated content module, response time module and sentiment analysis module, assigns weights according to the scores obtained by the sentiment analysis module, and comprehensively evaluates the customer service service through the weighted average method based on the scores of the customer evaluation module, customer sensitive word evaluation module, customer service sensitive word evaluation module, repeated content module and response time module to obtain a total score;

[0063] Acquiring the emotion quality analysis model comprises the following steps:

[0064] S201: Collect text data with sentiment labels from social media comments, product reviews, customer service conversation records, and news comment channels to build a dataset;

[0065] S202: Remove punctuation marks and special characters in the text, convert all text data to lowercase to avoid inconsistencies caused by case, remove common stop words such as "的", "是", "在", etc., perform stemming and lemmatization;

[0066] S203: Divide the dataset into a training set and a test set, and divide the dataset according to the ratio of 70% training set and 30% test set;

[0067] S204: Use TF-IDF for feature extraction and convert the text data into numerical features;

[0068] S205: Build a random forest model;

[0069] When building the random forest model, the initial number of trees can be 100. Starting from 100, gradually increase to 200, 300, 400, 500, and observe the changes in the model performance;

[0070] The initial maximum depth is 10. Starting from 10, gradually increase to 15, 20, and observe the changes in the model performance;

[0071] The initial minimum number of samples required to further divide an internal node can be 2. Starting from 2, gradually increase;

[0072] The minimum number of samples required for a leaf node is 1. Starting from 1, gradually increase;

[0073] S206: Input the training set into the model and perform multiple rounds of iterative training until the model converges;

[0074] S207: Verify the performance of the converged model according to the test set to obtain an emotion quality analysis model;

[0075] The analysis of the sensitive words for the customer service or the customer includes the following steps:

[0076] S401: Collect the conversation text of the customer service or the customer, including the information sent by the customer and the information sent by the customer service;

[0077] S402: Remove punctuation marks and special characters in the text;

[0078] S403: Use the string matching method to check whether the text contains sensitive words;

[0079] S404: Count the number of times the sensitive words appear in the conversation text sent by the customer service or the customer;

[0080] The response time module is scored in the following way:

[0081] If the response time is greater than 120s, the score is 1;

[0082] The response time is 91-120s, with a score of 2;

[0083] The response time is 61-90s, with a score of 3;

[0084] The response time is 46-60s, with a score of 4;

[0085] The response time is 31-45s, with a score of 5;

[0086] The response time is 21-30s, and the score is 6;

[0087] The response time is 11-20s, and the score is 7;

[0088] The response time is 6-10s, and the score is 8;

[0089] The response time is 3-5s, and the score is 9;

[0090] The response time is less than 3 seconds, with a score of 10;

[0091] When allocating weights, the following criteria may be used:

[0092] Response time module: When the customer sentiment is poor, such as 1-3 points, the customer pays more attention to the response time, so the weight of the response time module should be relatively high. As the sentiment score increases, the customer's tolerance for response time increases, and the weight gradually decreases;

[0093] Customer evaluation module: As the sentiment score increases, customers are more willing to give detailed and positive reviews, so the weight of the customer evaluation module gradually increases;

[0094] Sensitive word evaluation module: The weight of the sensitive word evaluation module for customers and customer service staff also increases with the improvement of the sentiment score. When the customer is in a good mood, the customer pays more attention to the quality and politeness of the communication, so the use of sensitive words becomes more important;

[0095] Duplicate content module: When customers are in a good mood, they tend to express their questions clearly and concisely. If duplicate content appears, it means, to a certain extent, that the customer service staff did not accurately understand or respond to the customer's questions. Therefore, the weight of the duplicate content module increases accordingly when the customer is in a good mood.

[0096] Total score = customer evaluation module score × customer evaluation module weight + customer sensitive word evaluation module score × customer sensitive word evaluation module weight + customer service sensitive word evaluation module score × customer service sensitive word evaluation module weight + duplicate content module score × duplicate content module weight + response time module score × response time module weight.

[0097] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. An online customer service evaluation system for cross-border e-commerce, characterized in that: include: Customer evaluation module: used to analyze the evaluations given by customers and give ratings; Customer sensitive word evaluation module: import the data of customer sent information into the pre-established sensitive word database, analyze the frequency of sensitive words, and make scores; Customer service sensitive word evaluation module: imports the data sent by customer service into the pre-established sensitive word database, analyzes the frequency of sensitive words, and performs scoring; Duplicate content module: Analyze the same content based on the information sent by the customer, obtain the number of times the same content is repeated, and score it; Response time module: Analyzes the time from when the customer service replies to the last message sent by the customer before the reply, calculates the time difference between the two, obtains the response time, and performs scoring; Sentiment analysis module: import the information sent by customers before customer service goes online into the sentiment quality analysis model, analyze customer sentiment, and score it; Collection and evaluation module: Collect the scores of the customer evaluation module, customer sensitive word evaluation module, customer service sensitive word evaluation module, repeated content module, response time module and sentiment analysis module, assign weights based on the scores obtained from the sentiment analysis module, and use the weighted average method to comprehensively evaluate customer service based on the scores of the customer evaluation module, customer sensitive word evaluation module, customer service sensitive word evaluation module, repeated content module and response time module to obtain a total score.

2. The online customer service evaluation system for cross-border e-commerce according to claim 1 is characterized in that: Acquiring the emotion quality analysis model comprises the following steps: S201: Collect text data with sentiment labels from social media comments, product reviews, customer service conversation records, and news comment channels to build a dataset; S202: Cleaning the data set; S203: Divide the data set into a training set and a test set; S204: Use TF-IDF to perform feature extraction and convert text data into numerical features; S205: Build a random forest model; S206: Input the training set into the model and perform multiple rounds of iterative training until the model converges; S207: Verify the performance of the converged model based on the test set to obtain a sentiment quality analysis model.

3. The online customer service evaluation system for cross-border e-commerce according to claim 2 is characterized in that: The scores of the customer evaluation module, customer sensitive word evaluation module, customer service sensitive word evaluation module, repeated content module, response time module, and sentiment analysis module are all 1-10. The higher the score, the higher the customer evaluation, the fewer the number of customer sensitive words, the fewer the number of customer service sensitive words, the fewer the number of repetitions of the same content, the shorter the response time, and the better the customer sentiment.

4. The online customer service evaluation system for cross-border e-commerce according to claim 3 is characterized in that: The sensitive word analysis for customer service or customers includes the following steps: S401: Collecting conversation texts between customer service or customers, including information sent by customers and information sent by customer service; S402: Delete punctuation marks and special characters in the text; S403: Use a string matching method to check whether the text contains sensitive words; S404: Count the number of times sensitive words appear in the conversation text sent by the customer service or the customer.

5. The online customer service evaluation system for cross-border e-commerce according to claim 4 is characterized in that: The dataset is divided into 70% training set and 30% test set.

6. The online customer service evaluation system for cross-border e-commerce according to claim 5 is characterized in that: The cleaning data set comprises the following steps: S601: Delete punctuation marks and special characters in the text; S602: convert all text data into lowercase to avoid inconsistency caused by uppercase and lowercase letters; S603: Remove common stop words; S604: Stemming and word form restoration.

7. The online customer service evaluation system for cross-border e-commerce according to claim 6 is characterized in that: The response time module is scored as follows: If the response time is greater than 120 seconds, the score is 1; The response time is 91-120s, with a score of 2; The response time is 61-90s, with a score of 3; The response time is 46-60s, with a score of 4; The response time is 31-45s, with a score of 5; The response time is 21-30s, and the score is 6; The response time is 11-20s, and the score is 7; The response time is 6-10s, and the score is 8; The response time is 3-5s, and the score is 9; The response time is less than 3s, with a score of 10.

Images