CN109063000A - Question sentence recommended method, customer service system and computer readable storage medium - Google Patents
Question sentence recommended method, customer service system and computer readable storage medium Download PDFInfo
- Publication number
- CN109063000A CN109063000A CN201810744322.5A CN201810744322A CN109063000A CN 109063000 A CN109063000 A CN 109063000A CN 201810744322 A CN201810744322 A CN 201810744322A CN 109063000 A CN109063000 A CN 109063000A
- Authority
- CN
- China
- Prior art keywords
- question sentence
- answered
- similar
- sentence
- question
- 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.)
- Granted
Links
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a kind of question sentence recommended method, customer service system and computer readable storage mediums, and the method comprising the steps of: after exporting answer according to question sentence to be answered, each similar question sentence similar to the question sentence to be answered is searched in corpus；According to similarity calculation algorithm, the question sentence to be answered is subjected to similarity calculation to each similar question sentence in the corpus, obtains each similarity value；According to the similarity value of each similar question sentence and feedback data is prestored, calculates recommendation scores of each similar question sentence relative to the question sentence to be answered, score is higher than the similar question sentence of given value as recommendation question sentence；Export the recommendation question sentence.The present invention improves the accuracy rate that customer service system recommends question sentence.
Description
Technical field
The present invention relates to field of artificial intelligence more particularly to a kind of question sentence recommended methods, customer service system and calculating
Machine readable storage medium storing program for executing.
Background technique
With the development of economy, artificial intelligence technology is also more and more mature, and the scene of human-computer interaction is also more and more, common
Human-computer interaction scene be read statement after, obtain recommending sentence by Keywords matching.
Since the prior art is when identifying user's question sentence, obtains recommending sentence only with the mode of Keywords matching, deposit
In biggish error, when the question sentence of user's input is too short and there are recommendation sentence when ambiguity, frequently resulted in is also not accurate enough.
Summary of the invention
The main purpose of the present invention is to provide a kind of question sentence recommended method, customer service system and computer-readable storage mediums
Matter, it is intended to solve the technical problem that existing customer service system recommends the accuracy rate of question sentence low.
To achieve the above object, the present invention provides a kind of question sentence recommended method, the question sentence recommended method comprising steps of
After exporting answer according to question sentence to be answered, searched in corpus similar each with the question sentence to be answered
Similar question sentence；
According to similarity calculation algorithm, the question sentence to be answered is subjected to phase to each similar question sentence in the corpus
It is calculated like degree, obtains each similarity value；
According to the similarity value of each similar question sentence and prestore feedback data, calculate each similar question sentence relative to it is described to
Score is higher than the similar question sentence of given value as recommendation question sentence by the recommendation scores for answering question sentence；
Export the recommendation question sentence.
Preferably, it the similarity value according to each similar question sentence and prestores feedback data, calculates each similar question sentence
Relative to the recommendation scores of the question sentence to be answered, the similar question sentence that score is higher than given value is wrapped as the step of recommending question sentence
It includes:
Determine next question sentence similar with the question sentence to be answered, in the feedback data to count each next question sentence
The number of appearance, obtain all next question sentences similar with the question sentence to be answered occur in the feedback data it is total secondary
Number；
Calculate accounting of the number of each next question sentence appearance in the total degree；
According to the corresponding similarity value of each next question sentence, preset similarity weighted value, accounting in the total degree
Than and preset accounting weighted value, the recommendation scores of each similar question sentence relative to the question sentence to be answered are calculated, by score
Similar question sentence higher than given value is used as recommendation question sentence.
Preferably, in the feedback data, each for corpus prestores the relationship of question sentence foundation between any two, and is arranged
The weighted value for prestoring question sentence relationship between any two, shows as (A ← B:a, b), indicates that question sentence B is in same user conversation
The number of next question sentence of question sentence A is a, and the weight of this relationship is b between question sentence A and B.
Preferably, described after exporting answer according to question sentence to be answered, it is searched and the phase to be answered in corpus
As each similar question sentence the step of before, the method also includes:
After the user's question sentence for receiving user's input, judge whether user's question sentence has ambiguity；
If user's question sentence has ambiguity, user's question sentence is compared with the question sentence that prestores in corpus；
When being matched at least two in the corpus and prestoring question sentence, according to the feedback data, judge to be matched to
It is described to prestore whether question sentence meets preset condition；
Meet the question sentence that prestores of the preset condition if it exists, then using meet the preset condition prestore question sentence as to
It answers question sentence and replaces user's question sentence.
Preferably, described when being matched at least two in the corpus and prestoring question sentence, according to the feedback data, sentence
It is disconnected be matched to described prestore that the step of whether question sentence meets preset condition includes:
When being matched at least two and prestoring question sentence, obtain every be matched to and prestore question sentence to go out in the feedback data
Existing number, and obtain all total degrees for prestoring question sentence and occurring in the feedback data being matched to；
It calculates described every and prestores accounting of the number of question sentence appearance in the total degree；
The number for prestoring question sentence appearance according to described every and the accounting in the total degree, judge that every prestores sentence
Whether preset condition is met；
Wherein, when the number for prestoring question sentence and occurring in the feedback data is greater than preset times, and the question sentence that prestores goes out
When accounting of the existing number in total degree is greater than target value, it is determined that the question sentence that prestores meets preset condition.
Preferably, described when being matched at least two in the corpus and prestoring question sentence, according to the feedback data, sentence
It is disconnected be matched to it is described prestore the step of whether question sentence meets preset condition after, the method also includes:
The question sentence that prestores for meeting the preset condition if it does not exist exports all question sentences that prestore being matched to then for user
Selection.
Preferably, described after exporting answer according to question sentence to be answered, it searches in corpus and is asked with described wait answer
After the step of sentence similar each similar question sentence, the method also includes:
According to the question sentence to be answered and the feedback data, obtain and the associated next question sentence of question sentence to be answered；
The next question sentence that will acquire is exported as recommendation question sentence.
Preferably, the question sentence to be answered according to and the feedback data, acquisition are associated with the question sentence to be answered
Next question sentence the step of include:
In the feedback data prestored, next question sentence of the question sentence to be answered is inquired, if next question sentence is asked there are multiple
Sentence then obtains the most question sentence of frequency of occurrence as the associated next question sentence of question sentence to be answered.
In addition, to achieve the above object, the present invention also provides a kind of customer service system, the customer service system include memory,
Processor recommends journey with the question sentence recommended program that can be run on the memory and on the processor, the question sentence is stored in
Sequence realizes the step of question sentence recommended method as described above when being executed by the processor.
In addition, to achieve the above object, it is described computer-readable the present invention also provides a kind of computer readable storage medium
Question sentence recommended program is stored on storage medium, the question sentence recommended program realizes question sentence as described above when being executed by processor
The step of recommended method.
The present invention searches similar to the question sentence to be answered after exporting answer according to question sentence to be answered in corpus
Each similar question sentence, then according to similarity calculation algorithm, by the question sentence to be answered and each phase in the corpus
Like question sentence carry out similarity calculation, obtain each similarity value, further according to each similar question sentence similarity value and prestore feedback
Data calculate recommendation scores of each similar question sentence relative to the question sentence to be answered, and score is higher than the similar of given value and is asked
Sentence recommends question sentence described in final output as question sentence is recommended.The present invention when recommending sentence, be first filtered out in corpus with
The similar each similar question sentence of question sentence to be answered, the recommendation scores of each similar question sentence are calculated in conjunction with the feedback data prestored,
Similar question sentence score to be higher than to given value only recommends question sentence by keyword compared with prior art as question sentence is recommended, this
Invention improves the accuracy rate that customer service system recommends question sentence.
Detailed description of the invention
Fig. 1 is the system structure diagram for the hardware running environment that the embodiment of the present invention is related to；
Fig. 2 is the flow diagram of question sentence recommended method first embodiment of the present invention；
Fig. 3 is a kind of feedback matrix schematic diagram in the embodiment of the present invention；
Fig. 4 is in the embodiment of the present invention according to similarity calculation algorithm, will be in the question sentence be answered and the corpus
Each similar question sentence the step of carrying out similarity calculation, obtaining each similarity value refined flow chart；
Fig. 5 is the flow diagram of question sentence recommended method second embodiment of the present invention.
The embodiments will be further described with reference to the accompanying drawings for the realization, the function and the advantages of the object of the present invention.
Specific embodiment
It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, it is not intended to limit the present invention.
The solution of the embodiment of the present invention is mainly: after exporting answer according to question sentence to be answered, in corpus
Each similar question sentence similar to the question sentence to be answered is searched to ask described wait answer then according to similarity calculation algorithm
Sentence carries out similarity calculation to each similar question sentence in the corpus, each similarity value is obtained, further according to each similar
It the similarity value of question sentence and prestores feedback data, calculates recommendation scores of each similar question sentence relative to the question sentence to be answered,
The similar question sentence that score is higher than given value recommends question sentence described in final output as question sentence is recommended.It is pushed away with solving customer service system
Recommend the low problem of the accuracy rate of question sentence.
As shown in Figure 1, Fig. 1 is the customer service system structural representation for the hardware running environment that the embodiment of the present invention is related to
Figure.
As shown in Figure 1, the customer service system may include: processor 1001, such as CPU, network interface 1004, user interface
1003, memory 1005, communication bus 1002.Wherein, communication bus 1002 is for realizing the connection communication between these components.
User interface 1003 may include display screen (Display), input unit, such as keyboard (Keyboard), optional user interface
1003 can also include standard wireline interface and wireless interface.Network interface 1004 optionally may include that the wired of standard connects
Mouth, wireless interface (such as WI-FI interface).Memory 1005 can be high speed RAM memory, be also possible to stable memory
(non-volatile memory), such as magnetic disk storage.Memory 1005 optionally can also be independently of aforementioned processor
1001 storage device.
Optionally, customer service system can also include camera, RF (Radio Frequency, radio frequency) circuit, sensor,
Voicefrequency circuit, WiFi module etc..
It will be understood by those skilled in the art that the restriction of the not structure paired terminal of customer service system structure shown in Fig. 1, it can
To include perhaps combining certain components or different component layouts than illustrating more or fewer components.
As shown in Figure 1, as may include operating system in a kind of memory 1005 of computer storage medium and ask
Sentence recommended program.Wherein, operating system is to manage and control the program of customer service system hardware and software resource, and question sentence is supported to recommend
The operation of program and other softwares and/or program.
In customer service system shown in Fig. 1, network interface 1004 is mainly used for accessing network；User interface 1003 is mainly used
In obtaining question sentence to be answered, and processor 1001 can be used for calling the question sentence recommended program stored in memory 1005, and hold
The step of row described below question sentence recommended method.
Based on above-mentioned hardware configuration, each embodiment of question sentence recommended method is proposed.
Customer service system in the embodiment of the present invention can be PC, be also possible to smart phone, tablet computer, portable computer
Equal terminal devices.
It is the flow diagram of question sentence recommended method first embodiment of the present invention referring to Fig. 2, Fig. 2.
In the present embodiment, the embodiment of question sentence recommended method is provided, it should be noted that although showing in flow charts
Go out logical order, but in some cases, it can be with the steps shown or described are performed in an order that is different from the one herein.
The question sentence recommended method includes:
Step S10 is searched and the question sentence phase to be answered in corpus after exporting answer according to question sentence to be answered
As each similar question sentence；
Step S20 similar asks the question sentence to be answered to each in the corpus according to similarity calculation algorithm
Sentence carries out similarity calculation, obtains each similarity value；
Step S30 according to the similarity value of each similar question sentence and prestores feedback data, it is opposite to calculate each similar question sentence
In the recommendation scores of the question sentence to be answered, score is higher than the similar question sentence of given value as recommendation question sentence；
Step S40 exports the recommendation question sentence.
Each step of question sentence recommended method of the present invention described below:
Step S10 is searched and the question sentence phase to be answered in corpus after exporting answer according to question sentence to be answered
As each similar question sentence；
In the present embodiment, the user's question sentence for first receiving user's input, then checks language according to the keyword of user's question sentence
Whether material is stored with corresponding question sentence to be answered in library, if so, then basis is somebody's turn to do question sentence output answer to be answered, it should be appreciated that corpus
Each question sentence in library is all corresponding with answer, it is thus determined that is, exportable question sentence to be answered is corresponding to be answered after wait answer question sentence
Case.
Later, it for the question sentence to be answered, is further searched in corpus with the presence or absence of similar to the question sentence to be answered
Each similar question sentence, in the present embodiment, the optional keyword by question sentence to be answered is searched in corpus, will with to
The similarity for answering question sentence reaches the similar question sentence for prestoring question sentence as question sentence to be answered of preset threshold.
After customer service system searches similar question sentence similar to question sentence to be answered in corpus, customer service system according to
While question sentence to be answered exports corresponding answer, shown using the similar question sentence of the higher N item of similarity as recommendation question sentence in answer
Show that the page is shown, wherein N can be set as five or ten according to actual needs.In the present embodiment, customer service system will match
To similar question sentence be shown in answer show the page lower section, specifically, be located at answer lower section.
Step S20 similar asks the question sentence to be answered to each in the corpus according to similarity calculation algorithm
Sentence carries out similarity calculation, obtains each similarity value；
The each similar question sentence of question sentence to be answered and other in corpus is carried out similarity calculation by customer service system, obtains every
The similarity value of similar question sentence and question sentence to be answered, wherein calculate the similarity value of every similar question sentence and question sentence to be answered
Method can be Jacobi's similarity algorithm, cosine similarity algorithm, one of cosine similarity algorithm with TF-IDF or
It is several, it is not limited here.
Step S30 according to the similarity value of each similar question sentence and prestores feedback data, it is opposite to calculate each similar question sentence
In the recommendation scores of the question sentence to be answered, score is higher than the similar question sentence of given value as recommendation question sentence；
Customer service system is according to the similarity value of every similar question sentence and question sentence to be answered, in conjunction with feedback data and default score
Recommended formula, calculates the recommendation scores of every similar question sentence, and customer service system is chosen according to score height and is higher than the similar of given value
Question sentence is as recommendation question sentence.
In the present embodiment, it is a feedback matrix that feedback data is optional, for collecting between user and customer service system
Session information, referring to Fig. 3, specific manifestation are as follows: in feedback matrix, customer service system is that each question sentence that prestores in corpus is established
Relationship between any two, and the weighted value for prestoring question sentence relationship between any two is set, it shows as (A ← B:a, b), indicates question sentence B
Occurring the number for the next question sentence for being question sentence A in same user conversation is a, and the weight of this relationship is between question sentence A and B
B, such as (A ← B:10,1), indicating that the number for the next question sentence for being question sentence A occurs in same user conversation in question sentence B is 10, and
The weight of this relationship is 1 between question sentence A and B, wherein the assignment of weighted value is optional to determine according to frequency of occurrence, such as this implementation
1-99 weighted value is 1 in example.After user proposes question sentence A in the same session, if next question sentence is question sentence B, customer service system
The number of (A ← B) relationship then can be increased by 1 in feedback matrix by system.In user conversation, question sentence B is either user is autonomous
Input, is also possible to what user selected in the recommendation question sentence of question sentence A.
Specifically, include: referring to Fig. 4, the step S20
Step S21, determining next question sentence similar with the question sentence to be answered in the feedback data, each with statistics
The number that next question sentence occurs obtains all next question sentences similar with the question sentence to be answered and occurs in the feedback data
Total degree；
Step S22 calculates accounting of the number of each next question sentence appearance in the total degree；
Step S23, according to the corresponding similarity value of each next question sentence, preset similarity weighted value, at described total time
Accounting and preset accounting weighted value in number calculate each similar question sentence and recommend relative to the question sentence to be answered
Point, score is higher than the similar question sentence of given value as recommendation question sentence.
During calculating the recommendation scores of every similar question sentence, customer service system obtains in feedback data customer service system
Next question sentence similar with question sentence to be answered, the number that every next question sentence of statistics occurs in feedback data, and obtain all
The total degree that next question sentence occurs in feedback data calculates the ratio of number and total degree that every next question sentence occurs, root
According to every be calculated next Question sentence parsing value, preset similarity weighted value, every next question sentence frequency of occurrence total
The accounting of number and preset accounting weighted value calculate the recommendation scores of every similar question sentence, will be above the similar of given value
Question sentence is as output object.In the present embodiment, score formula specific manifestation are as follows:
Wherein, sim_scorexFor the similarity value of similar question sentence x and question sentence y to be answered, sim_weightxFor similarity
The weighted value of score, county←xThere is the number that next question sentence is similar question sentence x for question sentence y to be answered,There is the accounting weight of the total degree of next question sentence in y question sentence to there is the number of similar question sentence x
Value, the value range of similarity weighted value and accounting weighted value are 0-1, and the two
For the weighted value of next question sentence of question sentence y, sta_weighty ← x and sta_weighty ← i can be according to practical feelings
The artificial assignment of condition, in the present embodiment, the fixed value of the two is 1.A given value β is preset, is pushed away when customer service system is calculated
When recommending score α > given value β, the similar question sentence is exported, β is optional in the present embodiment is set as 0.5.
It should be noted that in the present embodiment, given value β can not be set, customer service system be calculated every it is similar
After the recommendation scores of question sentence, according to score height, high first five of score or preceding ten are chosen as recommending sentences.
To avoid Matthew effect, in addition to the similar question sentence being calculated, customer service system can also be from similar to question sentence to be answered
Question sentence in select at random a similar question sentence as output question sentence so that all similar question sentences have the possibility being demonstrated.
Step S40 exports the recommendation question sentence.
In the present embodiment, after exporting answer according to question sentence to be answered, searches in corpus and asked with described wait answer
The similar each similar question sentence of sentence will be in the question sentence be answered and the corpus then according to similarity calculation algorithm
Each similar question sentence carries out similarity calculation, obtains each similarity value, similarity value further according to each similar question sentence and pre-
Feedback data is deposited, recommendation scores of each similar question sentence relative to the question sentence to be answered are calculated, score is higher than given value
Similar question sentence recommends question sentence described in final output as question sentence is recommended.The present invention is first to sieve in corpus when recommending sentence
Each similar question sentence similar to question sentence to be answered is selected, the recommendation of each similar question sentence is calculated in conjunction with the feedback data prestored
Score, the similar question sentence that score is higher than given value are only asked by keyword recommendation compared with prior art as question sentence is recommended
Sentence, the present invention improve the accuracy rate that customer service system recommends question sentence.
Further, question sentence recommended method second embodiment of the present invention is proposed.
The difference of the question sentence recommended method second embodiment and the question sentence recommended method first embodiment is, reference
Before Fig. 5, step S10, question sentence recommended method further include:
Step S50 judges whether user's question sentence has ambiguity after the user's question sentence for receiving user's input；
User can be manually entered user's question sentence in customer service system display interface, and user can also be by way of voice
Input user's question sentence can convert user's question sentence of speech form when customer service system gets user's question sentence of speech form
Judge that user's question sentence is when customer service system receives user's question sentence of user's input for user's question sentence of written form
It is no to have ambiguity, judge whether user's question sentence has ambiguity, it is more exactly to see whether the keyword of user's question sentence is matched in corpus
A sentence, such as the keyword of user's question sentence are matched to two sentences by the keyword for " amount " in corpus, " borrow
Amount of money degree be how many ", " amount of refund is how many ", at this time, it is believed that user's sentence has ambiguity.
Step S60 carries out the question sentence that prestores in user's question sentence and corpus if user's question sentence has ambiguity
Compare；
If user's question sentence has ambiguity, user's question sentence is compared by customer service system with the question sentence that prestores in corpus,
Wherein, customer service system includes corpus, prestores question sentence in corpus.
Specifically, customer service system is by user's question sentence and the process for prestoring question sentence and being compared in corpus are as follows: work as customer service
When system gets user's question sentence, word segmentation processing is carried out to user's question sentence, wherein customer service system can pass through participle tool pair of increasing income
User's question sentence carries out word segmentation processing, the question sentence to be answered after being segmented.After user's question sentence after customer service system obtains participle,
Customer service system extracts the keyword in user's question sentence, and the keyword prestored in question sentence in the keyword and corpus of extraction is carried out
Compare, to calculate the similarity for prestoring question sentence in user's question sentence and corpus, wherein the method for calculating similarity can be refined
Than similarity algorithm, cosine similarity algorithm, band TF-IDF (term frequency-inverse document
Frequency, a kind of common weighting technique for information retrieval and data mining) one of cosine similarity algorithm or
It is several, it is not limited here.
When user's question sentence is calculated and prestores between question sentence similarity be greater than preset value when, then illustrate user's question sentence and language
Question matching is prestored in material library, preset value is arranged according to specific needs, such as may be configured as 60%, or be set as 76%
Deng.
It should be noted that customer service system can filter nonsense words, such as when carrying out word segmentation processing to user's question sentence
" thanks ", " manual service ", " not having ", " hello ", " turn artificial " and " the problem of " etc. phrases.Such as when user's question sentence is " loan
What is, thanks " when, after carrying out word segmentation processing to user's question sentence, obtained vocabulary is " loan ", "Yes", " what ",
" thing " and " thanks " filters out meaningless phrase " thanks " first.
Further, for customer service system before carrying out word segmentation processing to user's question sentence, customer service system first detects user's question sentence
In with the presence or absence of number or punctuation mark, if customer service system detect in user's question sentence exist number or punctuation mark, delete
Number or punctuation mark in user's question sentence；If number or punctuation mark is not detected in customer service system in user's question sentence, right
User's question sentence carries out word segmentation processing.
Step S70, when being matched at least two in the corpus and prestoring question sentence, according to the feedback data, judgement
Prestore whether question sentence meets preset condition described in being matched to；
Customer service system is by the user's question sentence received and prestores question sentence in corpus and matches, if obtaining being matched at least
Two prestore question sentence as a result, illustrating user's question sentence there are ambiguity, customer service system is matched to according to the feedback data prestored, judgement
Prestore whether question sentence meets preset condition, wherein feedback data has been described in detail in foregoing embodiments, is not repeated herein.
Further, the embodiment of step S70 includes:
1) mode one, the step S70 include:
Step a1 obtains every be matched to and prestores question sentence in the feedback when being matched at least two and prestoring question sentence
The number occurred in data, and obtain all total degrees for prestoring question sentence and occurring in the feedback data being matched to；
Step b1 calculates described every and prestores accounting of the number of question sentence appearance in the total degree；
Step c1 prestores accounting of the question sentence in the total degree according to described every, judges whether every prestore sentence
Meet preset condition；
Wherein, when accounting of the number for prestoring question sentence appearance in total degree is greater than target value, it is determined that described
It prestores question sentence and meets preset condition.
2) mode two, the step S70 include:
Step a2 obtains every be matched to and prestores question sentence in the feedback when being matched at least two and prestoring question sentence
The number occurred in data, and obtain all total degrees for prestoring question sentence and occurring in the feedback data being matched to；
Step b2 calculates described every and prestores accounting of the number of question sentence appearance in the total degree；
Step c2, the number for prestoring question sentence appearance according to described every and the accounting in the total degree, judge every
Prestore whether sentence meets preset condition；
Wherein, when the number for prestoring question sentence and occurring in the feedback data is greater than preset times, and the question sentence that prestores goes out
When accounting of the existing number in total degree is greater than target value, it is determined that the question sentence that prestores meets preset condition.
Judge to be matched to described prestores the detailed process whether question sentence meets preset condition are as follows: when user's question sentence is in corpus
When being matched at least two in library and prestoring question sentence, customer service system obtains every be matched to and prestores question sentence to be occurred in feedback data
Number, and calculate it is all be matched to prestore the sum of the number that question sentence occurs in feedback data, obtain all be matched to
The total degree that question sentence occurs in feedback data is prestored, then, every be matched to is calculated and prestores the number of question sentence appearance and total
The ratio of number obtains every and prestores accounting of the number of question sentence appearance in total degree.Question sentence is prestored according to every feeding back
The number that data occur, and the accounting in total degree, judge that every prestores whether sentence meets preset condition.
Specific judgment basis are as follows: be greater than given threshold value when prestoring the number that question sentence occurs in feedback data, and send one's regards in advance
When accounting of the number that sentence occurs in feedback data in total degree is greater than target value, it is determined that it is default that this prestores question sentence satisfaction
Condition, in the present embodiment, customer service system preset condition is embodied in formula:
Wherein, count1To prestore the number that question sentence occurs in feedback data, θ is given threshold value, and θ can be according to reality
Situation is set as 100 or 500, and target value is set as γ, value be [0.5,1), it therefore meets the pre- of preset condition is sent one's regards to
Sentence at most only one.
Step S80 meets the question sentence that prestores of the preset condition if it exists, then will meet the pre- of the preset condition and send one's regards to
Sentence replaces user's question sentence as question sentence to be answered.
If customer service system judgement, which prestores question sentence, meets preset condition, it is default that customer service system obtains satisfaction from corpus
Condition prestores question sentence, will meet the question sentence to be answered for prestoring question sentence as customer service system of preset condition, i.e., subsequent customer service system
System need to only reply the question sentence that prestores for meeting preset condition.
Further, the question sentence that prestores for meeting preset condition is got in corpus when customer service system, and will be met pre-
If the question sentence that prestores of condition is used as after answering question sentence, customer service system obtains that question sentence to be answered is corresponding answers with this in corpus
Case, and the answer is exported into display interface, so that user checks.It should be noted that each being sent one's regards in advance in corpus
The all corresponding answer associated storage of sentence.Therefore, after some for determining in corpus prestores question sentence, that is, it can determine its correspondence
Answer and output.
The present embodiment, which realizes, to be got when answering question sentence, is first filtered out in corpus and question matching to be answered
Prestore question sentence, further screened in conjunction with the feedback data prestored to prestoring question sentence, with search filter out wait answer
Question sentence improves the accuracy rate that customer service system recommends question sentence.
Further, question sentence recommended method 3rd embodiment of the present invention is proposed.
The difference of the question sentence recommended method 3rd embodiment and the question sentence recommended method second embodiment is, step
After S70, question sentence recommended method further include:
Step A meets the question sentence that prestores of the preset condition if it does not exist, then export be matched to all prestore question sentence with
It is selected for user.
If what customer service system output was matched to all prestores there is no meeting the question sentence that prestores of preset condition in corpus
Question sentence prestores question sentence so that user manually selects as input question sentence.
Specifically, meeting the question sentence that prestores of preset condition if it does not exist, then customer service system pops up pop-up, the institute that will match to
Prestore question sentence in pop-up in the form of a list arrangement output, wherein arrangement foundation optionally according to similarity from height to
Low, arranged in sequence, user selects the question sentence of certain serial number then to indicate that the corresponding question sentence of the serial number receives question sentence as customer service system.
When user, which selectes certain, prestores question sentence as input question sentence, the question sentence to be answered that client originally inputs cancels, customer service
System prestores question sentence according to what user selected, this is searched in corpus and prestores the corresponding answer of question sentence as the recommendation of this question sentence
Final result, and export the answer.
It should be noted that pop-up in addition to show that customer service system is matched to it is all prestore question sentence other than, be additionally provided with and return
Hui Jian, when user to customer service system export it is all prestore question sentence it is all dissatisfied when, click return key, customer service system exports at this time
Default answer, the recommendation of this question sentence terminates, wherein default answer can be arranged according to the actual situation, such as may be configured as " sorry not
Satisfactory answer is found, we understand continuous learning ".
The present embodiment when there is no meeting when prestoring question sentence of preset condition, exported in corpus be matched to it is all pre-
It sends one's regards to sentence for selection by the user, the question sentence that prestores that user selects is received into question sentence as customer service system, and export reception question sentence
Corresponding answer improves the success rate that customer service system answers user's question sentence.
Further, question sentence recommended method fourth embodiment of the present invention is proposed.
The question sentence recommended method fourth embodiment and the question sentence recommended method first embodiment to 3rd embodiment
Difference is, after step S10, the question sentence recommended method further include:
Step B is obtained associated next with the question sentence to be answered according to the question sentence to be answered and the feedback data
Question sentence；
Step C, the next question sentence that will acquire are exported as recommendation question sentence.
Further, step B includes:
In the feedback data prestored, next question sentence of the question sentence to be answered is inquired, if next question sentence is asked there are multiple
Sentence then obtains the most question sentence of frequency of occurrence as the associated next question sentence of question sentence to be answered.
When customer service system is according to question sentence to be answered, after corresponding answer is found in corpus, customer service system is according to anti-
Present data query and the associated next question sentence of question sentence to be answered, wherein with the associated next question sentence of question sentence to be answered be customer service system
System will be collected into the question sentence to be answered in user conversation, the frequent item set found out by the mining algorithm of correlation rule.If
With the associated next question sentence of question sentence to be answered there are multiple question sentences, then obtain the frequency of occurrence in feedback data it is most with wait return
The associated next question sentence of sentence is answered a question, the next question sentence that will acquire exports this on answer display interface and push away as recommendation question sentence
Recommend question sentence.
It should be noted that the place that the present embodiment is different from the third embodiment is, it is anti-that the present embodiment only passes through inquiry
The number that occurs in feedback data in feedback data with the associated next question sentence of question sentence to be answered screens recommendation question sentence, without
It will be by calculating similarity, i.e., recommendation question sentence can not have similarity relation with question sentence to be answered.
When feedback data collection is smaller, in order to guarantee the accuracy rate recommended, the frequent item set needs excavated by algorithm
Recommendation question sentence can be used as after carrying out manual examination and verification.
The present embodiment proposes a kind of only pass through in query feedback data with the reception associated next question sentence of question sentence in feedback coefficient
Recommendation question sentence is screened according to the number of middle appearance, without user being realized and obtaining oneself input by calculating similarity
It, can be associated next in the question sentence that the display interface of customer service system is directly selected with it is inputted during the corresponding answer of question sentence
Question sentence improves the intelligence of customer service system.
Further, question sentence recommended method further include:
Step D updates time that question sentence to be answered in feedback data occurs in feedback data according to the question sentence to be answered
Number.
When customer service system obtains after answering question sentence, the number that customer service system occurs question sentence to be answered in feedback data
Add 1, to update feedback data.
In addition, the embodiment of the present invention also proposes a kind of computer readable storage medium, the computer readable storage medium
On be stored with question sentence recommended program, the question sentence recommended program realizes question sentence recommendation side as described above when being executed by processor
The step of method.
It should be noted that, in this document, the terms "include", "comprise" or its any other variant are intended to non-row
His property includes, so that the process, method, article or the device that include a series of elements not only include those elements, and
And further include other elements that are not explicitly listed, or further include for this process, method, article or device institute it is intrinsic
Element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that including being somebody's turn to do
There is also other identical elements in the process, method of element, article or device.
The serial number of the above embodiments of the invention is only for description, does not represent the advantages or disadvantages of the embodiments.
Through the above description of the embodiments, those skilled in the art can be understood that above-described embodiment side
Method can be realized by means of software and necessary general hardware platform, naturally it is also possible to by hardware, but in many cases
The former is more preferably embodiment.Based on this understanding, technical solution of the present invention substantially in other words does the prior art
The part contributed out can be embodied in the form of software products, which is stored in a storage medium
In (such as ROM/RAM, magnetic disk, CD), including some instructions are used so that a terminal device (can be mobile phone, computer, clothes
Business device, air conditioner or the network equipment etc.) execute method described in each embodiment of the present invention.
The above is only a preferred embodiment of the present invention, is not intended to limit the scope of the invention, all to utilize this hair
Equivalent structure or equivalent flow shift made by bright specification and accompanying drawing content is applied directly or indirectly in other relevant skills
Art field, is included within the scope of the present invention.
Claims (10)
1. a kind of question sentence recommended method, which is characterized in that the question sentence recommended method the following steps are included:
After exporting answer according to question sentence to be answered, searched in corpus similar each similar to the question sentence to be answered
Question sentence；
According to similarity calculation algorithm, the question sentence to be answered is subjected to similarity to each similar question sentence in the corpus
It calculates, obtains each similarity value；
According to the similarity value of each similar question sentence and feedback data is prestored, calculates each similar question sentence relative to described wait answer
Score is higher than the similar question sentence of given value as recommendation question sentence by the recommendation scores of question sentence；
Export the recommendation question sentence.
2. question sentence recommended method as described in claim 1, which is characterized in that the similarity value according to each similar question sentence
With prestore feedback data, calculate recommendation scores of each similar question sentence relative to the question sentence to be answered, score be higher than given
The similar question sentence of value includes: as the step of recommending question sentence
Determining next question sentence similar with the question sentence to be answered in the feedback data, to count each next question sentence appearance
Number, obtain the total degree that all next question sentences similar with the question sentence to be answered occur in the feedback data；
Calculate accounting of the number of each next question sentence appearance in the total degree；
According to the corresponding similarity value of each next question sentence, preset similarity weighted value, the accounting in the total degree with
And preset accounting weighted value, recommendation scores of each similar question sentence relative to the question sentence to be answered are calculated, score is higher than
The similar question sentence of given value is as recommendation question sentence.
3. question sentence recommended method as described in claim 1, which is characterized in that be each of corpus in the feedback data
It prestores question sentence and establishes relationship between any two, and the weighted value for prestoring question sentence relationship between any two is set, show as (A ← B:a,
B), indicate that occur the number for the next question sentence for being question sentence A in same user conversation be a to question sentence B, and between question sentence A and B this
The weight of relationship is b.
4. question sentence recommended method as claimed in claim 3, which is characterized in that it is described according to question sentence to be answered export answer it
Afterwards, in corpus search to it is described wait answer similar each similar question sentence the step of before, the method also includes:
After the user's question sentence for receiving user's input, judge whether user's question sentence has ambiguity；
If user's question sentence has ambiguity, user's question sentence is compared with the question sentence that prestores in corpus；
When being matched at least two in the corpus and prestoring question sentence, according to the feedback data, judge to be matched to described
Prestore whether question sentence meets preset condition；
Meet the question sentence that prestores of the preset condition if it exists, then prestores question sentence as wait answer for meet the preset condition
Question sentence replaces user's question sentence.
5. question sentence recommended method as claimed in claim 4, which is characterized in that described to be matched at least two in the corpus
When item prestores question sentence, according to the feedback data, judge to be matched to described the step of whether question sentence meets preset condition is prestored
Include:
When being matched at least two and prestoring question sentence, obtains every be matched to and prestore what question sentence occurred in the feedback data
Number, and obtain all total degrees for prestoring question sentence and occurring in the feedback data being matched to；
It calculates described every and prestores accounting of the number of question sentence appearance in the total degree；
The number for prestoring question sentence appearance according to described every and the accounting in the total degree, judge whether every prestore sentence
Meet preset condition；
Wherein, when the number for prestoring question sentence and occurring in the feedback data is greater than preset times, and described question sentence appearance is prestored
When accounting of the number in total degree is greater than target value, it is determined that the question sentence that prestores meets preset condition.
6. question sentence recommended method as claimed in claim 4, which is characterized in that described to be matched at least two in the corpus
When item prestores question sentence, according to the feedback data, judge to be matched to described the step of whether question sentence meets preset condition is prestored
Later, the method also includes:
The question sentence that prestores for meeting the preset condition if it does not exist exports all question sentences that prestore being matched to then for user's choosing
It selects.
7. the question sentence recommended method as described in claim any one of 1-6, which is characterized in that described to be exported according to question sentence to be answered
After answer, after the step of each similar question sentence similar to the question sentence to be answered is searched in corpus, the method
Further include:
According to the question sentence to be answered and the feedback data, obtain and the associated next question sentence of question sentence to be answered；
The next question sentence that will acquire is exported as recommendation question sentence.
8. question sentence recommended method as claimed in claim 7, which is characterized in that the question sentence to be answered according to and described anti-
The step of presenting data, obtaining next question sentence associated with the question sentence to be answered include:
In the feedback data prestored, inquire next question sentence of the question sentence to be answered, if next question sentence there are multiple question sentences,
The most question sentence of frequency of occurrence is obtained as the associated next question sentence of question sentence to be answered.
9. a kind of customer service system, which is characterized in that the customer service system includes memory, processor and is stored in the memory
Question sentence recommended program that is upper and can running on the processor, the question sentence recommended program are realized when being executed by the processor
Such as the step of question sentence recommended method described in any item of the claim 1 to 8.
10. a kind of computer readable storage medium, which is characterized in that be stored with question sentence on the computer readable storage medium and push away
Program is recommended, realizes that question sentence described in any item of the claim 1 to 8 such as is recommended when the question sentence recommended program is executed by processor
The step of method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810744322.5A CN109063000B (en) | 2018-07-06 | 2018-07-06 | Question recommendation method, customer service system and computer-readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810744322.5A CN109063000B (en) | 2018-07-06 | 2018-07-06 | Question recommendation method, customer service system and computer-readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109063000A true CN109063000A (en) | 2018-12-21 |
| CN109063000B CN109063000B (en) | 2022-02-01 |
Family
ID=64819591
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810744322.5A Active CN109063000B (en) | 2018-07-06 | 2018-07-06 | Question recommendation method, customer service system and computer-readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109063000B (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109766013A (en) * | 2018-12-28 | 2019-05-17 | 北京金山安全软件有限公司 | Poetry sentence input recommendation method and device and electronic equipment |
| CN110008322A (en) * | 2019-03-25 | 2019-07-12 | 阿里巴巴集团控股有限公司 | Art recommended method and device under more wheel session operational scenarios |
| CN110765247A (en) * | 2019-09-30 | 2020-02-07 | 支付宝(杭州)信息技术有限公司 | Input prompting method and device for question-answering robot |
| CN111353015A (en) * | 2018-12-24 | 2020-06-30 | 阿里巴巴集团控股有限公司 | Crowdsourcing question recommendation method, device, equipment and storage medium |
| CN111475632A (en) * | 2020-04-09 | 2020-07-31 | 深圳追一科技有限公司 | Question processing method and device, electronic equipment and storage medium |
| CN111737449A (en) * | 2020-08-03 | 2020-10-02 | 腾讯科技（深圳）有限公司 | Method and device for determining similar problems, storage medium and electronic device |
| CN112541069A (en) * | 2020-12-24 | 2021-03-23 | 山东山大鸥玛软件股份有限公司 | Text matching method, system, terminal and storage medium combined with keywords |
| CN113139034A (en) * | 2020-01-17 | 2021-07-20 | 深圳市优必选科技股份有限公司 | Statement matching method, statement matching device and intelligent equipment |
| CN117132392A (en) * | 2023-10-23 | 2023-11-28 | 蓝色火焰科技成都有限公司 | Vehicle loan fraud risk early warning method and system |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6519578B1 (en) * | 1999-08-09 | 2003-02-11 | Mindflow Technologies, Inc. | System and method for processing knowledge items of a knowledge warehouse |
| CN102722558A (en) * | 2012-05-29 | 2012-10-10 | 百度在线网络技术（北京）有限公司 | User question recommending method and device |
| US8700558B1 (en) * | 2007-03-06 | 2014-04-15 | Patrick Laughlin Kelly | User interface for entering and viewing quantitatively weighted factors for decision choices |
| CN105653671A (en) * | 2015-12-29 | 2016-06-08 | 畅捷通信息技术股份有限公司 | Similar information recommendation method and system |
| CN107220380A (en) * | 2017-06-27 | 2017-09-29 | 北京百度网讯科技有限公司 | Question and answer based on artificial intelligence recommend method, device and computer equipment |
| CN107704563A (en) * | 2017-09-29 | 2018-02-16 | 广州多益网络股份有限公司 | A kind of question sentence recommends method and system |
-
2018
- 2018-07-06 CN CN201810744322.5A patent/CN109063000B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6519578B1 (en) * | 1999-08-09 | 2003-02-11 | Mindflow Technologies, Inc. | System and method for processing knowledge items of a knowledge warehouse |
| US8700558B1 (en) * | 2007-03-06 | 2014-04-15 | Patrick Laughlin Kelly | User interface for entering and viewing quantitatively weighted factors for decision choices |
| CN102722558A (en) * | 2012-05-29 | 2012-10-10 | 百度在线网络技术（北京）有限公司 | User question recommending method and device |
| CN105653671A (en) * | 2015-12-29 | 2016-06-08 | 畅捷通信息技术股份有限公司 | Similar information recommendation method and system |
| CN107220380A (en) * | 2017-06-27 | 2017-09-29 | 北京百度网讯科技有限公司 | Question and answer based on artificial intelligence recommend method, device and computer equipment |
| CN107704563A (en) * | 2017-09-29 | 2018-02-16 | 广州多益网络股份有限公司 | A kind of question sentence recommends method and system |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111353015A (en) * | 2018-12-24 | 2020-06-30 | 阿里巴巴集团控股有限公司 | Crowdsourcing question recommendation method, device, equipment and storage medium |
| CN111353015B (en) * | 2018-12-24 | 2024-03-15 | 阿里巴巴集团控股有限公司 | Crowd-sourced question recommendation method, device, equipment and storage medium |
| CN109766013A (en) * | 2018-12-28 | 2019-05-17 | 北京金山安全软件有限公司 | Poetry sentence input recommendation method and device and electronic equipment |
| CN110008322B (en) * | 2019-03-25 | 2023-04-07 | 创新先进技术有限公司 | Method and device for recommending dialogues in multi-turn conversation scene |
| CN110008322A (en) * | 2019-03-25 | 2019-07-12 | 阿里巴巴集团控股有限公司 | Art recommended method and device under more wheel session operational scenarios |
| CN110765247B (en) * | 2019-09-30 | 2022-10-25 | 支付宝(杭州)信息技术有限公司 | Input prompting method and device for question-answering robot |
| CN110765247A (en) * | 2019-09-30 | 2020-02-07 | 支付宝(杭州)信息技术有限公司 | Input prompting method and device for question-answering robot |
| CN113139034A (en) * | 2020-01-17 | 2021-07-20 | 深圳市优必选科技股份有限公司 | Statement matching method, statement matching device and intelligent equipment |
| CN111475632A (en) * | 2020-04-09 | 2020-07-31 | 深圳追一科技有限公司 | Question processing method and device, electronic equipment and storage medium |
| CN111737449A (en) * | 2020-08-03 | 2020-10-02 | 腾讯科技（深圳）有限公司 | Method and device for determining similar problems, storage medium and electronic device |
| CN112541069A (en) * | 2020-12-24 | 2021-03-23 | 山东山大鸥玛软件股份有限公司 | Text matching method, system, terminal and storage medium combined with keywords |
| CN117132392A (en) * | 2023-10-23 | 2023-11-28 | 蓝色火焰科技成都有限公司 | Vehicle loan fraud risk early warning method and system |
| CN117132392B (en) * | 2023-10-23 | 2024-01-30 | 蓝色火焰科技成都有限公司 | Vehicle loan fraud risk early warning method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109063000B (en) | 2022-02-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109063000A (en) | Question sentence recommended method, customer service system and computer readable storage medium | |
| CN107832286B (en) | Intelligent interaction method, equipment and storage medium | |
| CN107609101B (en) | Intelligent interaction method, equipment and storage medium | |
| JP6440732B2 (en) | Automatic task classification based on machine learning | |
| US11556572B2 (en) | Systems and methods for coverage analysis of textual queries | |
| CN104836720A (en) | Method for performing information recommendation in interactive communication, and device | |
| CN108681970A (en) | Finance product method for pushing, system and computer storage media based on big data | |
| US20110055241A1 (en) | High precision search system and method | |
| CN104899315A (en) | Method and device for pushing user information | |
| CN107562939A (en) | Vertical field news recommends method, apparatus and readable storage medium | |
| CN110472008B (en) | Intelligent interaction method and device | |
| CN109542956A (en) | Report form generation method, device, computer equipment and storage medium | |
| US20230244862A1 (en) | Form processing method and apparatus, device, and storage medium | |
| US20160253313A1 (en) | Updating language databases using crowd-sourced input | |
| CN104077327B (en) | The recognition methods of core word importance and equipment and search result ordering method and equipment | |
| CN107315833A (en) | Method and apparatus of the retrieval with downloading based on application program | |
| CN111666513A (en) | Page processing method and device, electronic equipment and readable storage medium | |
| CN116681561A (en) | Policy matching method and device, electronic equipment and storage medium | |
| CN110110143B (en) | Video classification method and device | |
| JP2016153998A (en) | Service evaluation device and method | |
| Wei et al. | Online education recommendation model based on user behavior data analysis | |
| CN106919593B (en) | Searching method and device | |
| KR101955920B1 (en) | Search method and apparatus using property language | |
| CN117420998A (en) | Client UI interaction component generation method, device, terminal and medium | |
| US20220027419A1 (en) | Smart search and recommendation method for content, storage medium, and terminal |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |