CN102073687A - Method and device for identifying quality of customer service through text tendency analysis - Google Patents
Method and device for identifying quality of customer service through text tendency analysis Download PDFInfo
- Publication number
- CN102073687A CN102073687A CN 201010603981 CN201010603981A CN102073687A CN 102073687 A CN102073687 A CN 102073687A CN 201010603981 CN201010603981 CN 201010603981 CN 201010603981 A CN201010603981 A CN 201010603981A CN 102073687 A CN102073687 A CN 102073687A
- Authority
- CN
- China
- Prior art keywords
- customer service
- answer
- answers
- utmost point
- quality
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a method for identifying the quality of customer service through text tendency analysis, which comprises the following steps: A) aquiring multiple questions of the customer service and multiple answers of the multiple questions and storing into a customer service database; B) carrying out the real-time indexing of the multiple questions and the multiple answers and storing an index list into an index database; C) carrying out the text tendency analysis on the multiple questions and the multiple answers and obtaining the tendency values of the multiple questions and the multiple answers; and D) inputting a search item, wherein the search results are sequenced according to the tendency values. The invention also provides a device for identifying the quality of the customer service through the text tendency analysis. The method and device provided by the invention can evaluate the service quality of customer service personnel, the satisfaction degree of customers, the quality of a product and the relative service quality of the product in real time, thereby enhancing the satisfaction degree of the customers.
Description
Technical field
The present invention relates to information retrieval field and customer service realm, relate in particular to a kind of by text based on sentiment classification discriminating customer service method for quality and device.
Background technology
The text based on sentiment classification has been widely used in the network technology, by the text based on sentiment classification, can differentiate that the Internet user is the viewpoint of holding commendation or derogatory sense to certain product, incident or policy etc.
And in existing mainstream customers service system, the text based on sentiment classification is not carried out in the answer that problem that the client proposed and customer service personnel are answered, certainly, more can not list the tendentiousness value is problem and the answer thereof that the utmost point is demoted or the utmost point is praised, also nature can not list that the tendentiousness value is that the utmost point is demoted or the utmost point is praised through sorted hot issue and answer thereof.Therefore, the supervisor of customer service quality can't obtain according to the ordering of tendentiousness value or is that the utmost point is demoted or the utmost point is praised the Search Results of ordering according to the tendentiousness value when current problem of search and answer thereof.
Owing to can't obtain Search Results according to the ordering of tendentiousness value, the supervisor of customer service quality also just can not obtain the real-time feedback of customer service personnel's service quality, thereby can't search out product in real time or serve unsatisfied client, perhaps the flaw of product or service is learnt in search.
Summary of the invention
Because the above-mentioned defective of prior art, technical matters to be solved by this invention provides a kind of by text based on sentiment classification discriminating customer service method for quality and device, obtaining each problem of customer service and the tendentiousness value of each answer, can the acquired tendency value be problem and the answer that the utmost point is demoted and the utmost point is praised also.By ordering, on the one hand, find out contact staff in poor service in real time, find out unsatisfied client, find out poor product or service, so that next step improves these rough sledding to problem and/or answer tendentiousness value; On the other hand, find out the good contact staff of service quality in real time, find out satisfied client, the product of finding out or service, so that next step unifies the best practice in employee's service regulation, customer satisfaction criterion, product or the service regulation, thereby finally improve the satisfaction of customer service.
For achieving the above object, the invention provides a kind of by text based on sentiment classification discriminating customer service method for quality, may further comprise the steps: A), obtain some answers of some problems and described some problems of customer service, and be stored to the Customer Service Information storehouse; B), described some problems and described some answers are carried out real time indexing, and with index table stores to index data base; C), described some problems and described some answers are carried out the text based on sentiment classification, obtain the tendentiousness value of described some problems and described some answers; D), the inputted search item, Search Results is by the ordering of described tendentiousness value.
Further, the preceding step D ' that also comprises described step D)), described some problems and described some answers are classified described step D) described Search Results be through sorted problem and/or answer by the ordering of described tendentiousness value.
Further, the preceding step D that also comprises described step D) "), described some problems and described some answers are carried out analysis of central issue, described step D) described Search Results be problem and/or the answer that belongs to focus by the ordering of described tendentiousness value.
Further, described step C) further may further comprise the steps:
C1), each described some problem and described some answers are done the subordinate sentence processing;
C2), according to objective deictic words dictionary and objective sentence evaluation algorithm, distinguish the objective expression sentence in described some problems and the described some answers, and filter and remove described objective expression sentence;
C3), according to dictionary for word segmentation and branch word algorithm, for the word cutting done in the remaining subjective sentence of expressing in described some problems and the described some answers;
C4), basis is passed judgement on polarity speech dictionary, degree speech dictionary, negative word dictionary, vocabulary dictionary of collocations, the judgement of sentence formula, paragraph structure, is that the tendentiousness values are set in each described some problem and each described some answer.
Further, described step D) further may further comprise the steps:
D1), selecting query type, described query type is key word, problem, customer name, customer ID number, customer service personnel name or customer service personnel identification number;
D2) if query type is a key word, search for the index of described some problems and described some answers, obtain to include the problem and/or the answer of described key word; If query type is a problem, earlier described problem is done the word cutting, the index of the described some problems of removal search and described some answers returns the set of relevant problem again, the problem and the problem in the described set of input are done the similarity contrast, return the problem that similarity is higher than predetermined threshold value; If query type is customer name or customer ID number, return the problem that the client inquired and/or the answer of described problem; If query type is customer service personnel name or customer service personnel identification number, return the problem that the customer service personnel answered and/or the answer of described problem.
For achieving the above object, the present invention also provides another kind of and has differentiated the customer service method for quality by the text based on sentiment classification, may further comprise the steps:
A), obtain some answers of some problems and described some problems of customer service, and be stored to the Customer Service Information storehouse;
B), described some problems and described some answers are carried out real time indexing, and with index table stores to index data base;
C), described some problems and described some answers are carried out the text based on sentiment classification, obtain the tendentiousness value of described some problems and described some answers;
D), demote threshold value and the utmost point according to the default utmost point and praise threshold value, the acquired tendency value is problem and/or the answer that the utmost point is demoted and the utmost point is praised;
E), the inputted search item, the tendentiousness value ordering that the tendentiousness value that Search Results is demoted by the utmost point and/or the utmost point are praised.
For achieving the above object, the present invention also provides a kind of the discriminating by the text based on sentiment classification to comprise the device of customer service quality: the Customer Service Information storehouse is used to store some answers of some problems and described some problems of customer service; Index is used for described some problems and described some answers are carried out real time indexing; Index data base is used to store concordance list; The based on sentiment classification device is used for described some problems and described some answers are carried out the text based on sentiment classification, obtains the tendentiousness value of described some problems and described some answers; Searcher is used for the search terms according to input, and Search Results is sorted by described tendentiousness value.
Further, described device also comprises sorter, is used for described some problems and described some answers are classified, and described Search Results is through sorted problem and/or answer by described tendentiousness value ordering.
Further, described device also comprises the analysis of central issue device, is used for described some problems and described some answers are carried out analysis of central issue, and described Search Results is problem and/or the answer by described tendentiousness value ordering that belongs to focus.
For achieving the above object, the present invention also provides a kind of the discriminating by the text based on sentiment classification to comprise the device of customer service quality: the Customer Service Information storehouse is used to store some answers of some problems and described some problems of customer service; Index is used for described some problems and described some answers are carried out real time indexing; Index data base is used to store concordance list; The based on sentiment classification device is used for described some problems and described some answers are carried out the text based on sentiment classification, obtains the tendentiousness value of described some problems and described some answers; The utmost point is praised the utmost point and is demoted filtrator, is used for demoting threshold value and the utmost point according to the default utmost point and praises threshold value, and the acquired tendency value is problem and the answer that the utmost point is demoted and the utmost point is praised; Searcher is used for the search terms according to input, the tendentiousness value ordering that the tendentiousness value that Search Results is demoted by the utmost point and/or the utmost point are praised.
Further, described device also comprises sorter, is used for described some problems and described some answers are classified, and described Search Results is the problem and/or the answer of the tendentiousness value ordering of praising through sorted tendentiousness value of demoting by the described utmost point and/or the described utmost point.
Further, described device also comprises the analysis of central issue device, be used for described some problems and described some answers are carried out analysis of central issue, described Search Results is problem and/or the answer that belongs to the tendentiousness value of demoting by the described utmost point of focus and/or the tendentiousness value ordering that the described utmost point is praised.
Further, the described utmost point is praised the utmost point and is demoted filtrator and be used to obtain demote threshold value or praise the problem and/or the answer of threshold value greater than the described utmost point less than the described utmost point.
Beneficial effect of the present invention is:
By method and apparatus of the present invention, client's problem and/or customer service personnel's answer can be sorted by the tendentiousness value, find out the flaw of contact staff in poor service, unsatisfied client, product or service in real time, so that in time improve the quality of product or service.Otherwise, also can find out client, quality product or the service of the good customer service employee of service quality, satisfaction in real time, so that the best practice in the employee's service regulation that upgrades in time, customer satisfaction criterion, product or the service regulation.Thereby finally promote client's satisfaction.
Method and apparatus of the present invention can also be classified client's problem and/or customer service personnel's answer, more expediently contact staff's service quality, client's satisfaction, product quality and product related service quality are assessed.
Further, method and apparatus of the present invention can also carry out analysis of central issue to client's problem and/or customer service personnel's answer, finds out in real time belonging to the problem and/or the answer of focus in the recent period, more directly to find out the improved problem of demanding urgently.
Method and apparatus of the present invention is when searching for, and the selection of query type is more, includes but not limited to key word, problem, customer name, customer ID number, customer service personnel name or customer service personnel identification number.Various selection makes the searchers, and the supervisor of for example customer service quality searches for more expediently.
The tendentiousness value ordering that the tendentiousness value that method and apparatus of the present invention can be demoted client's problem and/or customer service personnel's answer by the utmost point and/or the utmost point are praised, thus more effectively contact staff's service quality, client's satisfaction, product quality and product related service quality are assessed.
Description of drawings
Fig. 1 is a process flow diagram of differentiating first embodiment of customer service method for quality by the text based on sentiment classification of the present invention;
Fig. 2 is the forward direction concordance list of problem;
Fig. 3 is the forward direction concordance list of answer;
Fig. 4 is the inverted index table of problem;
Fig. 5 is the inverted index table of answer;
Fig. 6 is a process flow diagram of differentiating second embodiment of customer service method for quality by the text based on sentiment classification of the present invention;
Fig. 7 is a process flow diagram of differentiating the 3rd embodiment of customer service method for quality by the text based on sentiment classification of the present invention;
Fig. 8 is a process flow diagram of differentiating the 4th embodiment of customer service method for quality by the text based on sentiment classification of the present invention;
Fig. 9 is a process flow diagram of differentiating the 5th embodiment of customer service method for quality by the text based on sentiment classification of the present invention;
Figure 10 is a process flow diagram of differentiating the 6th embodiment of customer service method for quality by the text based on sentiment classification of the present invention;
Figure 11 is the structural representation of first embodiment of the device of differentiating the customer service quality by the text based on sentiment classification of the present invention;
Figure 12 is the structural representation of second embodiment of the device of differentiating the customer service quality by the text based on sentiment classification of the present invention;
Figure 13 is the structural representation of the 3rd embodiment of the device of differentiating the customer service quality by the text based on sentiment classification of the present invention.
Embodiment
Be described further below with reference to the technique effect of accompanying drawing, to understand purpose of the present invention, feature and effect fully design of the present invention, concrete structure and generation.
As shown in Figure 1, the invention provides a kind of method that the text based on sentiment classification is differentiated first embodiment of customer service quality of passing through, may further comprise the steps:
Step 101), obtain some answers of some problems and these some problems of customer service, and be stored to the Customer Service Information storehouse;
Customer Service Department for example answers the customer service phone when services client, during to the doorstep for customer service, obtain some problems from client's part; From the product of enterprises, service, knowledge, for example in contact staff's the answer, obtain some answers of these problems again.These problems and answer thereof are stored in the computing machine of Customer Service Department after arrangement, also are in the Customer Service Information storehouse.
Step 102), these problems in the Customer Service Information storehouse and answer thereof are carried out real time indexing, and with index table stores to index data base;
At first, index scans each problem and each answer, for the word cutting is done in each problem and each answer.
Secondly, make the forward direction concordance list of these problems and answer thereof.See also Fig. 2, the forward direction concordance list of problem comprises the sequence number questionid of problem, each key word word1, word2 in the problem, word3 ... sequence number word id1, the word id2 of each key word, word id3 ... the position of each key word in problem, i.e. side-play amount word offset1, word offset2, word offset3 ...The sequence number of the sequence number of problem, each key word, each key word is unique.But the side-play amount of each key word can because a key word can repeat in the sentence in a problem, also can repeat in a plurality of sentences in the problem for one or more.
Similarly, see also Fig. 3, the forward direction concordance list of answer comprises the sequence number answerid of answer, and other ingredients and Fig. 2 are similar, also comprises each key word in the answer, the sequence number and the position of each key word in answer of each key word.
Once more, make the inverted index table of these problems and answer thereof.See also Fig. 4, the inverted index table of problem comprises each key word word1 of problem, word2, word3 ... the sequence number word id1 of each key word, word id2, word id3 ... the quantity nquestions1 that comprises the problem of each key word, nquestions2, nquestions3 ... the sequence number questionid1 that comprises the problem of each key word, questionid2, questionid3, questionid4, questionid5, questionid6 ..., the side-play amount word offset1 of each key word, word offset2, word offset3, word offset4, wordoffset5, word offset6 ...The sequence number of each key word, each key word, the quantity that comprises the problem of each key word, the sequence number that comprises the problem of each key word are unique.But the side-play amount of each key word can because a key word can repeat in the sentence in a problem, also can repeat in a plurality of sentences in the problem for one or more.
Similarly, see also Fig. 5, the inverted index table of answer comprises the sequence number answerid of answer of quantity nanswer, each key word of the answer of each key word, other ingredients and Fig. 4 are similar, also comprising each key word in the answer, the sequence number and the position of each key word in answer of each key word, also is the side-play amount of each key word.
Forward direction concordance list and inverted index table belong to ripe prior art, repeat no more herein.
Step 103), these problems in the Customer Service Information storehouse and answer thereof are carried out the text based on sentiment classification, obtain the tendentiousness value of these problems and answer thereof;
Step 103) further may further comprise the steps:
Step 103A), each problem and each answer are done the subordinate sentence processing;
Step 103B), according to objective deictic words dictionary and objective sentence evaluation algorithm, distinguish the objective expression sentence in these problems and the answer thereof, and filter and remove described objective expression sentence;
Step 103C), according to dictionary for word segmentation and branch word algorithm, for the word cutting done in the remaining subjective sentence of expressing in these problems and the answer thereof;
Step 103D), basis is passed judgement on polarity speech dictionary, degree speech dictionary, negative word dictionary, vocabulary dictionary of collocations, the judgement of sentence formula, paragraph structure, is that the tendentiousness value is set in each problem and each answer.
These problems and answer thereof are carried out the text based on sentiment classification, can obtain the value of passing judgement on of each problem and answer thereof.The threshold range of tendentiousness value can be set arbitrarily, such as being set at-3 to+3, and the negative number representation derogatory sense, positive number is represented commendation, and the numerical value of tendentiousness value is big more, and the degree of commendation is high more.The tendentiousness value is by commendatory term, derogatory term, and negative word, the degree speech, the vocabulary collocation, the sentence formula is judged, comprehensive decision such as paragraph structure.
Step 104), the inputted search item, Search Results is by tendentiousness value ordering.
Step 104) further may further comprise the steps:
Step 104A), select query type, query type can be key word, certain problem, customer name, customer ID number, customer service personnel name or customer service personnel identification number;
Certainly, query type is not limited to the above-mentioned type, and the type of any problem that can find out desire inquiry and answer can.Various selection makes the searchers, and the supervisor of for example customer service quality searches for more expediently.
Step 104B) if query type is a key word, searches for the index of these problems and answer thereof, obtain to include the problem and/or the answer of described key word; If query type is a problem, earlier this problem is done the word cutting, the index of these problems of removal search and answer thereof returns the set of relevant problem again, the problem and the problem in this set of input are done the similarity contrast, return the problem that similarity is higher than predetermined threshold value; If query type is customer name or customer ID number, return the problem that the client inquired and/or the answer of this problem; If query type is customer service personnel name or customer service personnel identification number, return the problem that the customer service personnel answered and/or the answer of this problem.
By method of the present invention, client's problem and/or customer service personnel's answer can be sorted by the tendentiousness value, find out the flaw of contact staff in poor service, unsatisfied client, product or service in real time, so that in time improve the quality of product or service.Otherwise, also can find out client, quality product or the service of the good customer service employee of service quality, satisfaction in real time, so that the best practice in the employee's service regulation that upgrades in time, customer satisfaction criterion, product or the service regulation.Thereby finally promote client's satisfaction.
Fig. 6 is a process flow diagram of differentiating second embodiment of customer service method for quality by the text based on sentiment classification of the present invention.As shown in Figure 6, the difference of first embodiment of present embodiment and Fig. 1 is: in step 104) preceding also in steps 104 '), these problems and answer thereof are classified step 104 then) Search Results be through sorted problem and/or answer by the ordering of tendentiousness value.
These problems and answer thereof are to produce for services client or in the process of services client.Because client's demand is many-sided, and product or the service that provides may be eurypalynous, therefore, can classify to these problems and answer.The classification basis that is to say that by the unit that service is provided the unit at customer service system place determines.The process of classification is carried out automatically by sorter.Classification can make the supervisor more expediently contact staff's service quality, client's satisfaction, product quality and product related service quality be assessed.
In the present embodiment, step 104 ') be positioned at step 102) and step 103) between, but the present invention is not limited to this, step 104 ') can be positioned at step 101) and step 104) between the optional position.
Fig. 7 is a process flow diagram of differentiating the 3rd embodiment of customer service method for quality by the text based on sentiment classification of the present invention.As shown in Figure 7, the difference of first embodiment of present embodiment and Fig. 1 is: in step 104) preceding also in steps 104 "), these problems and answer thereof are carried out analysis of central issue, then step 104) Search Results be problem and/or the answer that belongs to focus by the ordering of tendentiousness value.
These problems and answer thereof are to produce in the process of services client.Owing to product type, COS, product quality, service quality etc. all change along with the time, therefore, need carry out analysis of central issue, to understand the hot issue and/or the answer of user, product, service, customer service etc. to these problems and answer thereof.Analysis of central issue can be found out in real time belonging to the problem and/or the answer of focus in the recent period, more directly to find out the improved problem of demanding urgently.
In the present embodiment, step 104 ") be positioned at step 102) and step 103) between, but the present invention is not limited to this, step 104 ") can be positioned at step 101) and step 104) between the optional position.
Fig. 8 is a process flow diagram of differentiating the 4th embodiment of customer service method for quality by the text based on sentiment classification of the present invention.As shown in Figure 8, the difference of first embodiment of present embodiment and Fig. 1 is: in step 104) the preceding step 104 that comprises simultaneously ') and step 104 ").Step 104) Search Results is problem and/or the answer by the ordering of tendentiousness value that belong to focus through classification.
In the present embodiment, step 104 '), step 104 ") be positioned at step 102) and step 103) between; but the present invention is not limited to this; step 104 '), step 104 ") can be positioned at step 101) and step 104) between the optional position, and step 104 '), step 104 ") between ordinal relation before and after also not having, also be step 104 ') also can be in step 104 ") before.
The present invention also provides another kind of and has differentiated the customer service method for quality by the text based on sentiment classification, as shown in Figure 9, may further comprise the steps:
501), obtain some answers of some problems and these some problems of customer service, and be stored to the Customer Service Information storehouse;
502), these some problems and answer thereof are carried out real time indexing, and with index table stores to index data base;
503), these some problems and answer thereof are carried out the text based on sentiment classification, obtain the tendentiousness value of these some problems and answer thereof;
Step 501), step 502), step 503) correspond respectively to the step 101 of first embodiment of Fig. 1), step 102), step 103).
504), demote threshold value and the utmost point according to the default utmost point and praise threshold value, the acquired tendency value is problem and/or the answer that the utmost point is demoted and the utmost point is praised;
After each problem and each answer acquired tendency value, demote threshold value and the utmost point according to the utmost point of setting and praise threshold value (demoting threshold value such as the utmost point be-2, and it is+2 that the utmost point is praised threshold value),, be the utmost point and demote problem or the utmost point and demote answer if demote threshold value-2 less than the utmost point; If praise threshold value+2, be that the utmost point is praised problem or the utmost point is praised answer greater than the utmost point.
505), the inputted search item, the tendentiousness value ordering that the tendentiousness value that Search Results is demoted by the utmost point and/or the utmost point are praised.
The tendentiousness value ordering that the tendentiousness value that Search Results is demoted by the utmost point and/or the utmost point are praised certainly, when ordering, also need take into account the correlativity of these problems and these answers and ageing.
In addition, Search Results can be listed problems all in the Customer Service Information storehouse and answer, problem that the utmost point is demoted and/or answer, and problem and/or answer that the utmost point is praised come the front.Certainly, Search Results also can only be listed problem and/or the answer that the utmost point is demoted, and the utmost point problem and/or the answer of praising.
The tendentiousness value ordering that the tendentiousness value that method of the present invention can be demoted client's problem and/or customer service personnel's answer by the utmost point and/or the utmost point are praised, thus more effectively contact staff's service quality, client's satisfaction, product quality and product related service quality are assessed.
Figure 10 is a process flow diagram of differentiating the 6th embodiment of customer service method for quality by the text based on sentiment classification of the present invention.As shown in figure 10, the difference of the 5th embodiment of present embodiment and Fig. 9 is:
In step 505) preceding also in steps 505 '), these problems and answer thereof are classified;
With step 505 "), these problems and answer thereof are carried out analysis of central issue.
Therefore, Search Results step 505) is for through the tendentiousness value of demoting by the utmost point that belongs to focus of classification and/or the problem and/or the answer of the tendentiousness value ordering that the utmost point is praised.
In the present embodiment, step 505 '), step 505 ") be positioned at step 503) and step 504) between; but the present invention is not limited to this; step 505 '), step 505 ") can be positioned at step 501) and step 505) between the optional position, and step 505 '), step 505 ") between ordinal relation before and after also not having, also be step 505 ") also can be in step 505 ') before.
In addition, as another embodiment of the present invention, of the present inventionly differentiate that by the text based on sentiment classification customer service method for quality can include only step 505 ') or include only step 505 "); promptly do not comprise step 505 simultaneously yet ') and step 505 "), be similar to second embodiment of Fig. 6 and the 3rd embodiment of Fig. 7.The present invention does not repeat them here.
As shown in figure 11, the present invention also provides a kind of device of differentiating customer service quality first embodiment by the text based on sentiment classification, also be customer service system 70, comprise: Customer Service Information storehouse 701 is used to store some answers of some problems He these problems of customer service; Index 702 is used for these problems and answer thereof are carried out real time indexing; Index data base 703 is used to store forward direction concordance list and inverted index table; Based on sentiment classification device 704 is used for these problems and answer thereof are carried out the text based on sentiment classification, obtains the tendentiousness value of these problems and answer thereof; Searcher 705 is used for the search terms according to searchers's 706 inputs, and Search Results is sorted by the tendentiousness value.
Figure 12 is the structural representation of second embodiment of the device of differentiating the customer service quality by the text based on sentiment classification of the present invention.As shown in figure 12, the difference of first embodiment of present embodiment and Figure 11 is: customer service system 70 ' also comprises sorter 707, be used for these problems and answer thereof are classified, Search Results is through sorted problem and/or answer by the ordering of tendentiousness value.
Certainly, customer service system 70 ' also can include only analysis of central issue device 708, perhaps comprises sorter 707 and analysis of central issue device 708 simultaneously.Analysis of central issue device 708 is used for these problems and answer thereof are carried out analysis of central issue.
When customer service system 70 ' included only analysis of central issue device 708, Search Results was problem and/or the answer by the ordering of tendentiousness value that belongs to focus.When customer service system 70 ' comprised sorter 707 and analysis of central issue device 708 simultaneously, Search Results was problem and/or the answer by the ordering of tendentiousness value that belongs to focus through sorted.
Figure 13 is the structural representation of the 3rd embodiment of the device of differentiating the customer service quality by the text based on sentiment classification of the present invention.As shown in figure 13, the difference of first embodiment of present embodiment and Figure 11 is: customer service system 70 " comprise that also sorter 707, analysis of central issue device 708 and the utmost point praise the utmost point and demote filtrator 709.The utmost point is praised the utmost point and is demoted filtrator 709 and be used to obtain demote threshold value or praise the problem and the answer of threshold value greater than the utmost point less than the utmost point.Search Results is for through the tendentiousness value of demoting by the utmost point that belongs to focus of classification and/or the problem and/or the answer of the tendentiousness value ordering that the utmost point is praised.
Certainly, customer service system 70 " also can include only the utmost point and praise the utmost point and demote filtrator 709, perhaps the utmost point is praised the utmost point and is demoted filtrator 709 and sorter 707, and perhaps the utmost point is praised the utmost point and is demoted filtrator 709 and analysis of central issue device 708.
Below, differentiate that by the text based on sentiment classification customer service method for quality and device are elaborated to of the present invention with a concrete example.
Theing contents are as follows of the first pair of problem and answer thereof:
Store two pairs of problems and answer thereof in the Customer Service Information storehouse 701, be respectively:
Problem and answer sequence number thereof: 1
Customer name: Wang Xiaohua
Client id:20091023021001
Customer issue: the quality of this brand WCKFR-62LJ air-conditioning is too poor, and the new air-conditioning of just having bought has not just freezed.People's attitude general, the circuit board of saying so has been broken with touching on lightly, and then allow I etc. accessory.The whole story is more than 10 day like this.When Qingdao is the hottest, original just irritated, more indignant, or else repair, just require to return goods!
Customer service personnel name: Liu Xin
Customer service personnel id:028
Customer service personnel answer: you are good, and I am company's customer service, job number 028, and being very glad is your service.The ac control circuit plate of this model that you mention approximately needs week age to deliver to the area at your place also in transit, woulds you please wait.Whether I will attempt adjusting the ac control circuit plate from the surrounding area to you as possible, see and can change to you as soon as possible, it is refrigerant to strive for allowing you enjoy as early as possible.Company apologizes to the inconvenience that you bring to the flaw of this product.
Theing contents are as follows of the second pair of problem and answer thereof:
Problem and answer sequence number thereof: 2
Customer name: Li Ming
Client id:20100322028007
Customer issue: the YK06 refrigerator is bought 2 years, has keeped in repair 3 times, and current fault 7 days has in the past been made a call several times, and is still out of care, and there has been problem in your company very much after sale really.
Customer service personnel name: Zhang Ning
Customer service personnel id:039
Customer service personnel answer: you are good, and I am company's customer service, job number 039.The after sale service problem that you mention mainly is because heat period just now, and after maintenance personal's work of your location all had been discharged to 7 days, we also had no idea for this, and your fortune is poor, can only wait.
First pair and second pair problem of index 702 scannings and answer thereof are for it sets up real time indexing.At first, every pair of problem of index scanning and answer thereof are for the word cutting is done in each problem and each answer; Secondly, make the forward direction concordance list of two pairs of problems and answer thereof, write down the position of each key word in every pair of problem and answer thereof, and write down the sequence number of every pair of problem and answer thereof; Once more, make the inverted index table of two pairs of problems and answer thereof.Inverted index table comprises the sequence number of each key word, the problem at each key word place and the logarithm of answer thereof, the problem at each key word place and the sequence number of answer thereof; At last, inverted index table is write index data base 703.
In addition, inverted index table also can comprise the position of each key word in problem and answer thereof, also is information such as side-play amount.Because forward direction concordance list, inverted index table are widely-used in existing search engine, so do not repeat them here.
Classify by 707 pairs first pair of sorter and the second pair of problem and answer thereof.The first pair of problem and answer thereof belong to the quality problems of the air-conditioning class in the product classification.The second pair of problem and answer thereof belong to the quality problems of the refrigerator class in the product class.
Carry out analysis of central issue by 708 pairs first pair of analysis of central issue device and the second pair of problem and answer thereof.According to before hot spot data, existing many to problem and answer thereof about WCKFR-62LJ air-conditioning quality, so the first pair of problem and answer thereof are focus.And about also fewer before the problem of the quality problems of refrigerator product and the answer thereof, so the second pair of problem and answer thereof also are not focus at present.
Carry out based on sentiment classification by 704 pairs first pair of based on sentiment classification device and the second pair of problem and answer thereof.About first pair of problem and answer thereof, the tendentiousness of first problem analysis part.Problem partly contains following derogatory sense vocabulary or phrase: poor, bad, touch on lightly, indignant; Do not contain commendation vocabulary or phrase; Contain the degree speech: " too " in " too poor "; Contain negative word: " no " in " not freezing ".Analyze the tendentiousness of answer part in first pair of problem and the answer thereof again, contain following derogatory sense vocabulary or phrase: flaw, inconvenience; Contain following commendation vocabulary or phrase: glad, do the best, strive for, enjoy.
Factors such as commendatory term, derogatory term, negative word, degree speech, the vocabulary of taking all factors into consideration first pair of problem and answer thereof is arranged in pairs or groups, sentence formula, the tendentiousness value of the problem part in the first pair of problem and the answer thereof is-2, (derogatory sense is defined as [3 to the threshold range of commendation to the tendentiousness value of answer part in this example for+2, + 3],-3 tendentiousness value is for demoting most, and+3 tendentiousness value is for praising most).In like manner, the tendentiousness value of the problem part in the second pair of problem and the answer thereof be-1, and answer tendentiousness value partly is-1.
Praising the utmost point by the utmost point demotes 709 pairs first pair and second pair problem of filtrator and answer thereof and carries out the utmost point and praise the filtration of demoting with the utmost point.The tendentiousness value of the problem part of the first pair of problem and answer thereof is that the utmost point is demoted (2).The tendentiousness value of the answer part of the first pair of problem and answer thereof is that the utmost point is praised (+2).
By searcher 705, search in Customer Service Information storehouse 701.Such as the searchers, the supervisor of for example customer service quality, search customer name " Wang Xiaohua " or customer ID number find that the tendentiousness value of the problem that this client asks be-2, just know that client " Wang Xiaohua " is unsatisfied with product, service or the customer service etc. of company.Answer part such as search problem and answer thereof, search word is not set, return all answer parts, the tendentiousness value that the answer part of supposing to return is demoted according to the utmost point is arranged, No. 039 contact staff can rank the first (the tendentiousness value is-1), and the supervisor can know that this contact staff's service quality may be not good; The tendentiousness value that the answer part of supposing to return is praised according to the utmost point is arranged, No. 028 contact staff can rank the first (the tendentiousness value is+1), the supervisor can know that this customer service personnel's service quality is better, this contact staff's service expression can be used for reference for other customer services personnel, thereby can improve the integrity service level.Search " air-conditioning " for another example, suppose the tendentiousness value arrangement that returning part (can be problem and/or answer) is demoted according to the utmost point, " WCKFR-62LJ air-conditioning " ranks the first (the tendentiousness value is-2), and the supervisor knows that then may there be some problems in the air-conditioning of this model.
Method and apparatus of the present invention also can be applicable to the information retrieval of the customer service system of various natural languages such as English, German, Russian, Japanese, Spanish.The present invention can be applicable to problem and the answer thereof in the customer service system, or in the data such as the relevant structured text of customer service, non-structured text.
More than describe preferred embodiment of the present invention in detail.Should be appreciated that those of ordinary skill in the art need not creative work and just can design according to the present invention make many modifications and variations.Therefore, all those skilled in the art all should be in claim protection domain of the present invention under this invention's idea on the basis of existing technology by the available technical scheme of logical analysis, reasoning, or a limited experiment.
Claims (20)
1. one kind is passed through text based on sentiment classification discriminating customer service method for quality, it is characterized in that, may further comprise the steps:
A), obtain some answers of some problems and described some problems of customer service, and be stored to the Customer Service Information storehouse;
B), described some problems and described some answers are carried out real time indexing, and with index table stores to index data base;
C), described some problems and described some answers are carried out the text based on sentiment classification, obtain the tendentiousness value of described some problems and described some answers;
D), the inputted search item, Search Results is by the ordering of described tendentiousness value.
2. as claimed in claim 1 by text based on sentiment classification discriminating customer service method for quality, it is characterized in that, described step D) the preceding step D ' that also comprises), described some problems and described some answers are classified described step D) described Search Results be through sorted problem and/or answer by the ordering of described tendentiousness value.
3. as claimed in claim 1 or 2 by text based on sentiment classification discriminating customer service method for quality, it is characterized in that, described step D) the preceding step D that also comprises "), described some problems and described some answers are carried out analysis of central issue, described step D) described Search Results be problem and/or the answer that belongs to focus by the ordering of described tendentiousness value.
4. as claimed in claim 1 by text based on sentiment classification discriminating customer service method for quality, it is characterized in that described step C) further may further comprise the steps:
C1), each described some problem and described some answers are done the subordinate sentence processing;
C2), according to objective deictic words dictionary and objective sentence evaluation algorithm, distinguish the objective expression sentence in described some problems and the described some answers, and filter and remove described objective expression sentence;
C3), according to dictionary for word segmentation and branch word algorithm, for the word cutting done in the remaining subjective sentence of expressing in described some problems and the described some answers;
C4), basis is passed judgement on polarity speech dictionary, degree speech dictionary, negative word dictionary, vocabulary dictionary of collocations, the judgement of sentence formula, paragraph structure, is that the tendentiousness values are set in each described some problem and each described some answer.
5. as claimed in claim 1 or 2 by text based on sentiment classification discriminating customer service method for quality, it is characterized in that described step D) further may further comprise the steps:
D1), selecting query type, described query type is key word, problem, customer name, customer ID number, customer service personnel name or customer service personnel identification number;
D2) if query type is a key word, search for the index of described some problems and described some answers, obtain to include the problem and/or the answer of described key word; If query type is a problem, earlier described problem is done the word cutting, the index of the described some problems of removal search and described some answers returns the set of relevant problem again, the problem and the problem in the described set of input are done the similarity contrast, return the problem that similarity is higher than predetermined threshold value; If query type is customer name or customer ID number, return the problem that the client inquired and/or the answer of described problem; If query type is customer service personnel name or customer service personnel identification number, return the problem that the customer service personnel answered and/or the answer of described problem.
6. as claimed in claim 3 by text based on sentiment classification discriminating customer service method for quality, it is characterized in that described step D) further may further comprise the steps:
D1), selecting query type, described query type is key word, problem, customer name, customer ID number, customer service personnel name or customer service personnel identification number;
D2) if query type is a key word, search for the index of described some problems and described some answers, obtain to include the problem and/or the answer of described key word; If query type is a problem, earlier described problem is done the word cutting, the index of the described some problems of removal search and described some answers returns the set of relevant problem again, the problem and the problem in the described set of input are done the similarity contrast, return the problem that similarity is higher than predetermined threshold value; If query type is customer name or customer ID number, return the problem that the client inquired and/or the answer of described problem; If query type is customer service personnel name or customer service personnel identification number, return the problem that the customer service personnel answered and/or the answer of described problem.
7. as claimed in claim 1 by text based on sentiment classification discriminating customer service method for quality, it is characterized in that, described step C) and step D) between also comprise step C '), demote threshold value and the utmost point according to the default utmost point and praise threshold value, the acquired tendency value is problem and/or the answer that the utmost point is demoted and the utmost point is praised; Then described step D) described tendentiousness value is the tendentiousness value that the utmost point tendentiousness value of demoting and/or the utmost point are praised.
8. as claimed in claim 7 by text based on sentiment classification discriminating customer service method for quality, it is characterized in that, described step D) the preceding step C that also comprises "), described some problems and described some answers are classified described step D) described Search Results be the problem and/or the answer of the tendentiousness value ordering of praising through sorted tendentiousness value of demoting by the described utmost point and/or the described utmost point.
9. differentiate the customer service method for quality as claim 7 or the 8 described text based on sentiment classification that pass through, it is characterized in that, described step D) the preceding step C ' that also comprises "), described some problems and described some answers are carried out analysis of central issue, described step D) described Search Results be problem and/or the answer that belongs to the tendentiousness value of demoting by the described utmost point of focus and/or the tendentiousness value ordering that the described utmost point is praised.
10. as claimed in claim 7 by text based on sentiment classification discriminating customer service method for quality, it is characterized in that described step C) further may further comprise the steps:
C1), each described some problem and described some answers are done the subordinate sentence processing;
C2), according to objective deictic words dictionary and objective sentence evaluation algorithm, distinguish the objective expression sentence in described some problems and the described some answers, and filter and remove described objective expression sentence;
C3), according to dictionary for word segmentation and branch word algorithm, for the word cutting done in the remaining subjective sentence of expressing in described some problems and the described some answers;
C4), basis is passed judgement on polarity speech dictionary, degree speech dictionary, negative word dictionary, vocabulary dictionary of collocations, the judgement of sentence formula, paragraph structure, is that the tendentiousness values are set in each described some problem and each described some answer.
11. differentiate the customer service method for quality as claim 7 or the 8 described text based on sentiment classification that pass through, it is characterized in that described step D) further may further comprise the steps:
D1), selecting query type, described query type is key word, problem, customer name, customer ID number, customer service personnel name or customer service personnel identification number;
D2) if query type is a key word, search for the index of described some problems and described some answers, obtain to include the problem and/or the answer of described key word; If query type is a problem, earlier described problem is done the word cutting, the index of the described some problems of removal search and described some answers returns the set of relevant problem again, the problem and the problem in the described set of input are done the similarity contrast, return the problem that similarity is higher than predetermined threshold value; If query type is customer name or customer ID number, return the problem that the client inquired and/or the answer of described problem; If query type is customer service personnel name or customer service personnel identification number, return the problem that the customer service personnel answered and/or the answer of described problem.
12. as claimed in claim 9 by text based on sentiment classification discriminating customer service method for quality, it is characterized in that described step D) further may further comprise the steps:
D1), selecting query type, described query type is key word, problem, customer name, customer ID number, customer service personnel name or customer service personnel identification number;
D2) if query type is a key word, search for the index of described some problems and described some answers, obtain to include the problem and/or the answer of described key word; If query type is a problem, earlier described problem is done the word cutting, the index of the described some problems of removal search and described some answers returns the set of relevant problem again, the problem and the problem in the described set of input are done the similarity contrast, return the problem that similarity is higher than predetermined threshold value; If query type is customer name or customer ID number, return the problem that the client inquired and/or the answer of described problem; If query type is customer service personnel name or customer service personnel identification number, return the problem that the customer service personnel answered and/or the answer of described problem.
13. as claimed in claim 7ly differentiate the customer service method for quality, it is characterized in that described step C ' by the text based on sentiment classification) further be: obtain to demote threshold value or praise the problem and/or the answer of threshold value greater than the described utmost point less than the described utmost point.
14. the device by text based on sentiment classification discriminating customer service quality is characterized in that, comprising:
The Customer Service Information storehouse is used to store some answers of some problems and described some problems of customer service;
Index is used for described some problems and described some answers are carried out real time indexing;
Index data base is used to store concordance list;
The based on sentiment classification device is used for described some problems and described some answers are carried out the text based on sentiment classification, obtains the tendentiousness value of described some problems and described some answers;
Searcher is used for the search terms according to input, and Search Results is sorted by described tendentiousness value.
15. the device of differentiating the customer service quality by the text based on sentiment classification as claimed in claim 14, it is characterized in that, described device also comprises sorter, be used for described some problems and described some answers are classified, described Search Results is through sorted problem and/or answer by described tendentiousness value ordering.
16. as claim 14 or the 15 described devices of differentiating the customer service quality by the text based on sentiment classification, it is characterized in that, described device also comprises the analysis of central issue device, be used for described some problems and described some answers are carried out analysis of central issue, described Search Results is problem and/or the answer by described tendentiousness value ordering that belongs to focus.
17. the device of differentiating the customer service quality by the text based on sentiment classification as claimed in claim 14, it is characterized in that, described device comprises that also the utmost point praises the utmost point and demote filtrator, is used for demoting threshold value and the utmost point according to the default utmost point and praises threshold value, and the acquired tendency value is problem and/or the answer that the utmost point is demoted and the utmost point is praised; Then the described tendentiousness value that is used to sort of described searcher is the tendentiousness value that the utmost point tendentiousness value of demoting and/or the utmost point are praised.
18. the device of differentiating the customer service quality by the text based on sentiment classification as claimed in claim 17, it is characterized in that, described device also comprises sorter, be used for described some problems and described some answers are classified, described Search Results is the problem and/or the answer of the tendentiousness value ordering of praising through sorted tendentiousness value of demoting by the described utmost point and/or the described utmost point.
19. as claim 17 or the 18 described devices of differentiating the customer service quality by the text based on sentiment classification, it is characterized in that, described device also comprises the analysis of central issue device, be used for described some problems and described some answers are carried out analysis of central issue, described Search Results is problem and/or the answer that belongs to the tendentiousness value of demoting by the described utmost point of focus and/or the tendentiousness value ordering that the described utmost point is praised.
20. as claimed in claim 17 the discriminating by the text based on sentiment classification is characterized in that the device of customer service quality, the described utmost point is praised the utmost point and is demoted filtrator and be used to obtain demote threshold value or praise the problem and/or the answer of threshold value greater than the described utmost point less than the described utmost point.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201010603981 CN102073687A (en) | 2010-12-21 | 2010-12-21 | Method and device for identifying quality of customer service through text tendency analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201010603981 CN102073687A (en) | 2010-12-21 | 2010-12-21 | Method and device for identifying quality of customer service through text tendency analysis |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102073687A true CN102073687A (en) | 2011-05-25 |
Family
ID=44032226
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 201010603981 Pending CN102073687A (en) | 2010-12-21 | 2010-12-21 | Method and device for identifying quality of customer service through text tendency analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102073687A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103678720A (en) * | 2014-01-02 | 2014-03-26 | 中国标准化研究院 | Method and device for processing user feedback data |
| CN106340308A (en) * | 2016-08-25 | 2017-01-18 | 北京云知声信息技术有限公司 | Speech reply method and device |
| CN110543587A (en) * | 2019-07-24 | 2019-12-06 | 湖北爱运动体育信息科技有限公司 | sports fitness communication platform based on Internet of things |
| CN114238598A (en) * | 2021-12-07 | 2022-03-25 | 北京妙医佳健康科技集团有限公司 | Question-answering system and labeling, auditing and model training method thereof |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101639824A (en) * | 2009-08-27 | 2010-02-03 | 北京理工大学 | Text filtering method based on emotional orientation analysis against malicious information |
-
2010
- 2010-12-21 CN CN 201010603981 patent/CN102073687A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101639824A (en) * | 2009-08-27 | 2010-02-03 | 北京理工大学 | Text filtering method based on emotional orientation analysis against malicious information |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103678720A (en) * | 2014-01-02 | 2014-03-26 | 中国标准化研究院 | Method and device for processing user feedback data |
| CN103678720B (en) * | 2014-01-02 | 2017-02-22 | 中国标准化研究院 | Method and device for processing user feedback data |
| CN106340308A (en) * | 2016-08-25 | 2017-01-18 | 北京云知声信息技术有限公司 | Speech reply method and device |
| CN110543587A (en) * | 2019-07-24 | 2019-12-06 | 湖北爱运动体育信息科技有限公司 | sports fitness communication platform based on Internet of things |
| CN114238598A (en) * | 2021-12-07 | 2022-03-25 | 北京妙医佳健康科技集团有限公司 | Question-answering system and labeling, auditing and model training method thereof |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106776574B (en) | User comment text mining method and device | |
| CN106874292B (en) | Topic processing method and device | |
| CN111767716B (en) | Method and device for determining enterprise multi-level industry information and computer equipment | |
| US20040049499A1 (en) | Document retrieval system and question answering system | |
| CN104516902A (en) | Semantic information acquisition method and corresponding keyword extension method and search method | |
| CN110298028A (en) | A kind of the critical sentence extracting method and device of text fragment | |
| Rosa et al. | Twitter topic fuzzy fingerprints | |
| CN110597978B (en) | Article abstract generation method, system, electronic equipment and readable storage medium | |
| CN111611356A (en) | Information searching method and device, electronic equipment and readable storage medium | |
| CN104834651A (en) | Method and apparatus for providing answers to frequently asked questions | |
| CN110717041A (en) | Case retrieval method and system | |
| CN102073687A (en) | Method and device for identifying quality of customer service through text tendency analysis | |
| CN109978624A (en) | Information processing method, electronic equipment and computer readable storage medium | |
| CN109508456B (en) | Text processing method and device | |
| CN115269959A (en) | Search association recommendation method and device, electronic equipment and storage medium | |
| CN114722086A (en) | Method and device for determining search rearrangement model | |
| CN117592489B (en) | Method and system for realizing electronic commerce commodity information interaction by using large language model | |
| CN114372122A (en) | Information acquisition method, computing device and storage medium | |
| Owoeye et al. | Classification of extremist text on the web using sentiment analysis approach | |
| Khatoon | Real-time twitter data analysis of Saudi telecom companies for enhanced customer relationship management | |
| Blümel et al. | Comparative analysis of classical and deep learning-based natural language processing for prioritizing customer complaints | |
| Palliyali et al. | Comparative study of extractive text summarization techniques | |
| Moodley | Language identification with decision trees: Identification of individual words in the south african languages | |
| Cai et al. | Music autotagging as captioning | |
| CN108509449A (en) | A kind of method and server of information processing |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C12 | Rejection of a patent application after its publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20110525 |