CN108170739A - Problem matching process, terminal and computer readable storage medium - Google Patents
Problem matching process, terminal and computer readable storage medium Download PDFInfo
- Publication number
- CN108170739A CN108170739A CN201711362221.3A CN201711362221A CN108170739A CN 108170739 A CN108170739 A CN 108170739A CN 201711362221 A CN201711362221 A CN 201711362221A CN 108170739 A CN108170739 A CN 108170739A
- Authority
- CN
- China
- Prior art keywords
- customer
- similarity
- prestores
- algorithm
- customer problem
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Human Computer Interaction (AREA)
- Computational Linguistics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a kind of problem matching process, terminal and computer readable storage mediums, the described method comprises the following steps:Customer problem is received, and the customer problem of reception is input in knowledge base;Each problem of prestoring is retrieved using searching algorithm in the knowledge base, to retrieve and the matched multiple problems of the customer problem;The problem of being determined in matched multiple problems with the customer problem matching degree highest using sorting algorithm;The associated answer of the problem of matching degree highest is searched in the knowledge base, and by the answer feedback of lookup to display interface.The present invention improves the matched accuracy of problem.
Description
Technical field
The present invention relates to a kind of artificial intelligence field more particularly to problem matching process, terminal and computer-readable storages
Medium.
Background technology
Many customer service robot systems are all based on the question answering system progress problem of knowledge base and the matching of answer at present, and
The problems in knowledge base and answer need to be ready in advance, when being interacted between user and robot, the problem of enquirement and
The problems in knowledge base is matched, and needs to involve how to the similitude between two texts of measurement.
The problem of traditional, matching process was identified by the algorithm of short text similarity, since short text similitude is calculated
Method has certain limitation, and there are certain errors, lead to not accurately measure the similitude between two texts, accordingly
, the matched accuracy of problem is caused also to decrease.
Invention content
It is a primary object of the present invention to provide a kind of problem matching process, terminal and computer readable storage medium, purport
Solving the problems, such as existing matching way, the technical issues of accuracy is relatively low.
To achieve the above object, the present invention provides a kind of problem matching process, and described problem matching process includes:
Customer problem is received, and the customer problem of reception is input in knowledge base;
Each problem of prestoring is retrieved using searching algorithm in the knowledge base, is asked with retrieving with the user
Inscribe matched multiple problems;
The problem of being determined in matched multiple problems with the customer problem matching degree highest using sorting algorithm;
The associated answer of the problem of matching degree highest is searched in the knowledge base, and by the answer feedback of lookup extremely
Display interface.
Optionally, it is described that each problem of prestoring is retrieved using searching algorithm in the knowledge base, to retrieve
The step of multiple problems matched with the customer problem, includes:
In the knowledge base, the customer problem is calculated using the first predetermined number searching algorithm respectively and is prestored with each
The similarity of problem;
In the similarity calculated in each searching algorithm, determine that the similarity with the customer problem reaches predetermined threshold value
Similarity;
Using with the similarity of the customer problem reach the corresponding multiple problems of predetermined threshold value as with the customer problem
Matched multiple problems.
Optionally, it is described in the knowledge base, the user is calculated using the first predetermined number searching algorithm respectively and is asked
The step of inscribing the similarity with each problem that prestores includes:
When searching algorithm is convolutional neural networks algorithm, the associated training pattern of convolutional neural networks algorithm is obtained;
Customer problem and each problem of prestoring are input to the training pattern, to calculate institute by the training pattern
State the similarity of customer problem and each problem that prestores, and export similarity reach the predetermined threshold value it is multiple prestore problem and
Corresponding similarity.
Optionally, it is described in the knowledge base, the user is calculated using the first predetermined number searching algorithm respectively and is asked
The step of inscribing the similarity with each problem that prestores includes:
When searching algorithm is editing distance short text Similarity algorithm, to the word of the customer problem and each problem that prestores
Symbol is divided;
When calculating each problem of prestoring and being converted to customer problem, the number of edit operation performed is needed;
The number that performs is needed when being converted to customer problem according to each problem of prestoring, determine the customer problem with it is each
The similarity for the problem that prestores.
Optionally, it is described in the knowledge base, the user is calculated using the first predetermined number searching algorithm respectively and is asked
The step of inscribing the similarity with each problem that prestores includes:
In searching algorithm document-frequency algorithm reverse for word frequency, word is carried out to the customer problem and each problem that prestores
Language is split, and the customer problem and each problem of prestoring are splitted into multiple words;
The word frequency of customer problem and each problem that prestores and reverse file frequency are calculated respectively according to multiple words after fractionation
Rate;
Word frequency-reverse document-frequency vector of customer problem is obtained according to the word frequency of customer problem and reverse document-frequency,
And according to the word frequency of each problem that prestores and reverse document-frequency obtain the word frequency of each problem that prestores-reverse document-frequency to
Amount;
Calculate word frequency-reverse document-frequency vector of customer problem and word frequency-reverse document-frequency of each problem that prestores
The corresponding cosine similarity of vector, to obtain the similarity of customer problem and each problem that prestores.
Optionally, it is described in the knowledge base, the user is calculated using the first predetermined number searching algorithm respectively and is asked
The step of inscribing the similarity with each problem that prestores includes:
When searching algorithm is term vector algorithm, the customer problem and each problem that prestores all are converted into term vector;
The term vector of the customer problem is multiplied with the term vector of each problem that prestores respectively, to calculate the use
The similarity of family problem and each problem that prestores.
Optionally, it is described in the knowledge base, the user is calculated using the first predetermined number searching algorithm respectively and is asked
The step of inscribing the similarity with each problem that prestores includes:
When searching algorithm is Recognition with Recurrent Neural Network algorithm, the associated training pattern of Recognition with Recurrent Neural Network algorithm is obtained;
Customer problem and each problem of prestoring are input to the training pattern, to calculate institute by the training pattern
State the similarity of customer problem and each problem that prestores, and export similarity reach the predetermined threshold value it is multiple prestore problem and
Corresponding similarity.
Optionally, it is described to be determined and the customer problem matching degree highest in matched multiple problems using sorting algorithm
The problem of the step of include:
It obtains and the matched multiple problems of the customer problem;
The matching rate of each problem and customer problem in multiple problems is calculated using the sorting algorithm of the second predetermined number;
The preset weights of each sorting algorithm are obtained, to the preset weights of each sorting algorithm and each problem and user
The matching rate of problem is weighted averagely, obtains the similarity value of each problem and customer problem;
The problem of using the problem of similarity value highest as with the customer problem matching degree highest.
In addition, to achieve the above object, the present invention also provides a kind of terminal, the terminal include processor, memory and
The problem of being stored on the memory and can running on the processor matcher, described problem matcher are described
The step of matching process the problem of as described above is realized when processor performs.
In addition, to achieve the above object, it is described computer-readable the present invention also provides a kind of computer readable storage medium
Problematic matcher is stored on storage medium, described problem matcher is applied to terminal, and described problem matcher is located
Manage the step of realizing matching process the problem of as described above when device performs.
Problem matching process proposed by the present invention first receives customer problem, and the customer problem of reception is input to
In knowledge base, then each problem of prestoring is retrieved using searching algorithm in the knowledge base, with retrieve with it is described
The matched multiple problems of customer problem, then determine to match with the customer problem in matched multiple problems using sorting algorithm
The problem of spending highest, the problem of matching degree highest is finally searched in the knowledge base associated answer, and by lookup
Answer feedback is to display interface.The matching for realizing problem is determined by searching algorithm and two class algorithm of sorting algorithm, relatively
The matching of problem is carried out in the single short text Similarity Algorithm of tradition, the present invention is more comprehensive to the matching of problem, improves
The matched accuracy of problem, so as to improve the accuracy of answer feedback.
Description of the drawings
Fig. 1 is the hardware architecture diagram of terminal of the present invention;
Fig. 2 is the flow diagram of problem matching process first embodiment of the present invention;
Fig. 3 is the refinement flow diagram of step S20 in Fig. 2;
The first refinement flow diagram that Fig. 4 is step S21 in Fig. 3;
The second refinement flow diagram that Fig. 5 is step S21 in Fig. 3;
The third that Fig. 6 is step S21 in Fig. 3 refines flow diagram;
The 4th refinement flow diagram that Fig. 7 is step S21 in Fig. 3;
The 5th refinement flow diagram that Fig. 8 is step S21 in Fig. 3;
Fig. 9 is the refinement flow diagram of step S30 in Fig. 2;
Figure 10 is the schematic diagram that problem of the present invention matches preferable implement scene.
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:Customer problem is first received, and the customer problem of reception is defeated
Enter into knowledge base, then each problem of prestoring retrieved using searching algorithm in the knowledge base, with retrieve with
The matched multiple problems of customer problem, then determined and the customer problem in matched multiple problems using sorting algorithm
The problem of matching degree highest, the problem of matching degree highest is finally searched in the knowledge base associated answer, and will look into
The answer feedback looked for is to display interface, to solve the problem of that existing matching way accuracy is relatively low.
As shown in Figure 1, the structure diagram of the terminal of hardware running environment that Fig. 1, which is the embodiment of the present invention, to be related to.
The terminal of the embodiment of the present invention can be PC, smart mobile phone, tablet computer, pocket computer or service
The equipment that device, virtual machine facility etc. have display function.
As shown in Figure 1, the terminal can include:Processor 1001, such as CPU, communication bus 1002, user interface
1003, network interface 1004, memory 1005.Wherein, communication bus 1002 is used to implement the connection communication between these components.
User interface 1003 can include display screen (Display), input unit such as keyboard (Keyboard), optional user interface
1003 can also include the wireline interface (such as connecting wired keyboard, wire mouse etc.) of standard, wireless interface (such as with
In connection Wireless Keyboard, wireless mouse).Network interface 1004 can optionally include the wireline interface of standard (for being connected with
Gauze network), wireless interface (such as WI-FI interfaces, blue tooth interface, infrared interface, for connecting wireless network).Memory
1005 can be high-speed RAM memory or the memory (non-volatile memory) of stabilization, such as disk are deposited
Reservoir.Memory 1005 optionally can also be the storage device independently of aforementioned processor 1001.
Optionally, terminal can also include camera, RF (Radio Frequency, radio frequency) circuit, sensor, audio
Circuit, WiFi module etc..
It will be understood by those skilled in the art that the restriction of the terminal structure shown in Fig. 1 not structure paired terminal, can wrap
It includes and either combines certain components or different components arrangement than illustrating more or fewer components.
As shown in Figure 1, as operating system, net can be included in a kind of memory 1005 of computer readable storage medium
Network communication module, Subscriber Interface Module SIM and problem matcher.Wherein, operating system is that management and control terminal are provided with software
The program in source supports the operation of network communication module, Subscriber Interface Module SIM, problem matcher and other programs or software;
Network communication module is used to managing and controlling network interface 1002;Subscriber Interface Module SIM is used to managing and controlling user interface
1003。
In terminal shown in Fig. 1, the terminal calls the problem of being stored in memory 1005 by processor 1001
With program, to realize following steps:
Customer problem is received, and the customer problem of reception is input in knowledge base;
Each problem of prestoring is retrieved using searching algorithm in the knowledge base, is asked with retrieving with the user
Inscribe matched multiple problems;
The problem of being determined in matched multiple problems with the customer problem matching degree highest using sorting algorithm;
The associated answer of the problem of matching degree highest is searched in the knowledge base, and by the answer feedback of lookup extremely
Display interface.
Further, the terminal calls the problem of being stored in memory 1005 matcher by processor 1001, with
Realization retrieves each problem of prestoring using searching algorithm in the knowledge base, to retrieve and the customer problem
The step of each problem matched:
In the knowledge base, the customer problem is calculated using the first predetermined number searching algorithm respectively and is prestored with each
The similarity of problem;
In the similarity calculated in each searching algorithm, determine that the similarity with the customer problem reaches predetermined threshold value
Similarity;
Using with the similarity of the customer problem reach the corresponding multiple problems of predetermined threshold value as with the customer problem
Matched multiple problems.
Further, the terminal calls the problem of being stored in memory 1005 matcher by processor 1001, with
It realizes in the knowledge base, the customer problem and each problem that prestores is calculated using the first predetermined number searching algorithm respectively
Similarity the step of:
When searching algorithm is convolutional neural networks algorithm, the associated training pattern of convolutional neural networks algorithm is obtained;
Customer problem and each problem of prestoring are input to the training pattern, to calculate institute by the training pattern
State the similarity of customer problem and each problem that prestores, and export similarity reach the predetermined threshold value it is multiple prestore problem and
Corresponding similarity.
Further, the terminal calls the problem of being stored in memory 1005 matcher by processor 1001, with
It realizes in the knowledge base, the customer problem and each problem that prestores is calculated using the first predetermined number searching algorithm respectively
Similarity the step of:
When searching algorithm is editing distance short text Similarity algorithm, to the word of the customer problem and each problem that prestores
Symbol is divided;
When calculating each problem of prestoring and being converted to customer problem, the number of edit operation performed is needed;
The number that performs is needed when being converted to customer problem according to each problem of prestoring, determine the customer problem with it is each
The similarity for the problem that prestores.
Further, the terminal calls the problem of being stored in memory 1005 matcher by processor 1001, with
It realizes in the knowledge base, the customer problem and each problem that prestores is calculated using the first predetermined number searching algorithm respectively
Similarity the step of:
In searching algorithm document-frequency algorithm reverse for word frequency, word is carried out to the customer problem and each problem that prestores
Language is split, and the customer problem and each problem of prestoring are splitted into multiple words;
The word frequency of customer problem and each problem that prestores and reverse file frequency are calculated respectively according to multiple words after fractionation
Rate;
Word frequency-reverse document-frequency vector of customer problem is obtained according to the word frequency of customer problem and reverse document-frequency,
And according to the word frequency of each problem that prestores and reverse document-frequency obtain the word frequency of each problem that prestores-reverse document-frequency to
Amount;
Calculate word frequency-reverse document-frequency vector of customer problem and word frequency-reverse document-frequency of each problem that prestores
The corresponding cosine similarity of vector, to obtain the similarity of customer problem and each problem that prestores.
Further, the terminal calls the problem of being stored in memory 1005 matcher by processor 1001, with
It realizes in the knowledge base, the customer problem and each problem that prestores is calculated using the first predetermined number searching algorithm respectively
Similarity the step of:
When searching algorithm is term vector algorithm, the customer problem and each problem that prestores all are converted into term vector;
The term vector of the customer problem is multiplied with the term vector of each problem that prestores respectively, to calculate the use
The similarity of family problem and each problem that prestores.
Further, the terminal calls the problem of being stored in memory 1005 matcher by processor 1001, with
It realizes in the knowledge base, the customer problem and each problem that prestores is calculated using the first predetermined number searching algorithm respectively
Similarity the step of:
When searching algorithm is Recognition with Recurrent Neural Network algorithm, the associated training pattern of Recognition with Recurrent Neural Network algorithm is obtained;
Customer problem and each problem of prestoring are input to the training pattern, to calculate institute by the training pattern
State the similarity of customer problem and each problem that prestores, and export similarity reach the predetermined threshold value it is multiple prestore problem and
Corresponding similarity.
Further, the terminal calls the problem of being stored in memory 1005 matcher by processor 1001, with
Realize the step of being determined in matched multiple problems with the problem of the customer problem matching degree highest using sorting algorithm:
It obtains and the matched multiple problems of the customer problem;
The matching rate of each problem and customer problem in multiple problems is calculated using the sorting algorithm of the second predetermined number;
The preset weights of each sorting algorithm are obtained, to the preset weights of each sorting algorithm and each problem and user
The matching rate of problem is weighted averagely, obtains the similarity value of each problem and customer problem;
The problem of using the problem of similarity value highest as with the customer problem matching degree highest.
Technical solution proposed by the present invention, the terminal call the problem of being stored in memory 1005 by processor 1001
Matcher, to realize step:Customer problem is first received, and the customer problem of reception is input in knowledge base, then
Each problem of prestoring is retrieved using searching algorithm in the knowledge base, it is matched with the customer problem to retrieve
Multiple problems, then using sorting algorithm in matched multiple problems determine with the customer problem matching degree highest the problem of,
The associated answer of the problem of matching degree highest is finally searched in the knowledge base, and the answer feedback of lookup is extremely shown
Interface.The matching for realizing problem is determined by searching algorithm and two class algorithm of sorting algorithm, single short relative to tradition
Text similarity algorithm carries out the matching of problem, and the present invention is more comprehensive to the matching of problem, and it is matched accurate to improve problem
Property, so as to improve the accuracy of answer feedback.
Based on the hardware configuration of above-mentioned terminal, each embodiment of problem matching process of the present invention is proposed.
With reference to Fig. 2, Fig. 2 is the flow diagram of problem matching process first embodiment of the present invention.
In the present embodiment, described problem matching process is applied to terminal, and described problem matching process includes:
Step S10 receives customer problem, and the customer problem of reception is input in knowledge base;
Step S20 retrieves each problem of prestoring using searching algorithm in the knowledge base, to retrieve and institute
State the matched multiple problems of customer problem;
Step S30, it is determining highest with the customer problem matching degree in matched multiple problems using sorting algorithm
Problem;
Step S40, the problem of matching degree highest is searched in the knowledge base associated answer, and answering lookup
Case feeds back to display interface.
The problem of described in the embodiment of the present invention, matching process was applied to terminal, and the terminal is chosen as the terminal described in Fig. 1,
The terminal integrated knowledge database stores largely prestore problem and the corresponding answer of each problem that prestores in the knowledge base.
In addition, the first predetermined number searching algorithm and the second predetermined number sorting algorithm are also stored in the knowledge base,
The specific number of the searching algorithm and sorting algorithm does not limit, and can set according to actual needs, in the present embodiment, optional institute
The number for stating searching algorithm is 5, and the number of sorting algorithm is 2, hereinafter as example.Wherein, the searching algorithm
Including:CNN (Convolutional Neural Network, convolutional neural networks) algorithm, the similar calculation of editing distance short text
(wherein, tf full name are Term Frequency, represent word frequency by method, tf-idf;Idf full name are Inverse Document
Frequency represents reverse document-frequency) algorithm, word2vec (term vector) algorithm, RNN (Recurrent Neural
Network, Recognition with Recurrent Neural Network) algorithm;The sorting algorithm includes:gbdt(Gradient Boosting Decision
Tree, iteration decision tree) algorithm, softmax (grader) algorithm.In the present embodiment, each searching algorithm is used in knowledge base
In the problem of retrieving corresponding and being matched with customer problem, each sorting algorithm for the matched multiple problems of customer problem
In, calculate with the most matched pre- problem of customer problem, finally, terminal can feed back the corresponding answer of the problem of this is most matched to use
Family improves the accuracy of answer feedback.
The matched each step of problem is done step-by-step in the present embodiment described in detail below:
Wherein, step S10 receives customer problem, and the customer problem of reception is input in knowledge base;
In the present embodiment, described problem matching process is applied to terminal, and the terminal is optional to be provided with display interface, should
Display interface is equipped with information input window, so that user inputs problem in described information input window.When terminal is on display circle
When the customer problem of input is received in the information input window in face, the customer problem received is input in knowledge base.This
The knowledge base of the knowledge base at place, that is, described above is stored with prestoring for predetermined number in the knowledge base and problem and each pre- sends one's regards to
Inscribe corresponding answer and the first predetermined number searching algorithm and the second predetermined number sorting algorithm.
Step S20 retrieves each problem of prestoring using searching algorithm in the knowledge base, to retrieve and institute
State the matched multiple problems of customer problem;
After the customer problem received is input to knowledge base by terminal, in the knowledge base, using searching algorithm
Each problem of prestoring is retrieved, with retrieve with the matched multiple problems of the customer problem, specifically, with reference to Fig. 3, institute
Step S20 is stated to include:
Step S21, in the knowledge base, using the first predetermined number searching algorithm calculate respectively the customer problem with
The similarity of each problem that prestores;
Step S22 in the similarity calculated in each searching algorithm, determines that the similarity with the customer problem reaches
The similarity of predetermined threshold value;
Step S23, using with the similarity of the customer problem reach the corresponding multiple problems of predetermined threshold value as with it is described
The matched multiple problems of customer problem.
That is, after the customer problem received is input to knowledge base by terminal, first first is extracted in the knowledge base
Predetermined number searching algorithm, the searching algorithm are:CNN algorithms, the searching algorithm of editing distance, tf-idf algorithms,
After each searching algorithm is extracted, the use is calculated by each searching algorithm respectively for word2vec algorithms, RNN algorithms
The similarity of family problem and each problem that prestores, wherein, in the case of searching algorithm difference, customer problem pre- is sent one's regards to each
The corresponding result of calculation of similarity of topic may be different, and specifically, the embodiment of the step S21 includes:
1) mode one, with reference to Fig. 4, the step S21 includes:
Step S21A when searching algorithm is convolutional neural networks algorithm, obtains the associated instruction of convolutional neural networks algorithm
Practice model;
Customer problem and each problem of prestoring are input to the training pattern by step S21B, to pass through the training
Model calculates the customer problem and the similarity of each problem that prestores, and exports similarity and reach the multiple of the predetermined threshold value
The problem that prestores and corresponding similarity.
In the present embodiment, when searching algorithm is convolutional neural networks algorithm (CNN algorithm), terminal first obtains convolution
The associated training pattern of neural network algorithm, the training pattern train to obtain according to mass data in advance, are getting trained mould
After type, customer problem and each problem of prestoring are input to the training pattern, to calculate institute by the training pattern
The similarity of customer problem and each problem that prestores is stated, final output similarity reaches multiple problems that prestore of the predetermined threshold value
With corresponding similarity.
2) mode two, with reference to Fig. 5, the step S21 includes:
Step S21C, searching algorithm be editing distance short text Similarity algorithm when, to the customer problem with it is each pre-
The character for sending one's regards to topic is divided;
Step S21D when each problem of prestoring of calculating is converted to customer problem, needs the number of edit operation performed;
Step S21E needs the number performed, determines the user when being converted to customer problem according to each problem of prestoring
The similarity of problem and each problem that prestores.
In the present embodiment, when searching algorithm is editing distance short text Similarity algorithm, first to customer problem and respectively
Then a problem that prestores, when each problem of prestoring of calculating is converted to customer problem, needs the editor performed into the division of line character
The number of operation needs the number performed during being converted to customer problem according to each problem of prestoring, determines the customer problem
With the similarity of each problem that prestores.It is to be understood that between editing distance refers to two characters, another word is changed by a character
Minimum edit operation number needed for symbol, the edit operation of execution include a character being substituted for another character, are inserted into one
A character, deletes a character, and the number of wherein edit operation is fewer, it is believed that two problems are more similar.Therefore, the present embodiment leads to
Editing distance short text Similarity algorithm is crossed, determines the similarity of customer problem and each problem that prestores.
3) mode three, with reference to Fig. 6, the step S21 includes:
Step S21F when searching algorithm is the reverse document-frequency algorithm of word frequency, to the customer problem and each prestores
Problem carries out word fractionation, and the customer problem and each problem of prestoring are splitted into multiple words;
Step S21G calculates the word frequency of customer problem and each problem that prestores and inverse respectively according to multiple words after fractionation
To document-frequency;
Step S21H obtains word frequency-reverse file of customer problem according to the word frequency of customer problem and reverse document-frequency
Frequency vector and the word frequency of each problem that prestores-reverse text is obtained according to the word frequency of each problem that prestores and reverse document-frequency
Part frequency vector;
Step S21I calculates the word frequency of the word frequency of customer problem-reverse document-frequency vector and each problem that prestores-reverse
The corresponding cosine similarity of document-frequency vector, to obtain the similarity of customer problem and each problem that prestores.
In the present embodiment, when searching algorithm is word frequency-reverse document-frequency algorithm (tf-idf algorithm), first to
Family problem and each problem that prestores carry out the fractionation of word, customer problem and each problem of prestoring are splitted into multiple words, so
The word frequency of customer problem and each problem that prestores and reverse document-frequency, then root are calculated respectively according to multiple words after fractionation afterwards
Word frequency-reverse document-frequency vector of customer problem is obtained according to the word frequency and reverse document-frequency of customer problem and according to each
The word frequency for the problem that prestores and reverse document-frequency obtain word frequency-reverse document-frequency vector of each problem that prestores, final to calculate
The word frequency of customer problem-reverse document-frequency vector and word frequency-reverse document-frequency vector of each problem that prestores are corresponding remaining
String similarity, to obtain the similarity of customer problem and each problem that prestores.
Specifically:Calculate frequency-word frequency of each word in sentence (including customer problem and the problem that prestores):Word
The number of all words in the number/sentence occurred in sentence;Occur the number of files of some word and taking the logarithm in language material-inverse
To document-frequency:(there is the different document number of certain word) in log in total number of documents/corpus of corpus, wherein, corpus
It is cut-and-dried.To be best understood from, it is exemplified below:
Trunk word is extracted, stop words is removed, similar word is converted into canonical form, such as:Hello, you are good, hello systems
One into " hello ".
Statistics calculates word frequency and reverse document-frequency obtains word frequency-reverse document-frequency vector, is calculated by cosine formula
Vector similarity.Such as:
Sentence 1:You get well me and want to borrow money, and how particle loan borrows money.
Participle:Hello, I thinks, borrow money, particle is borrowed, how
Sentence 2:How particle loan borrows money.
Participle:Particle borrow, how, borrow money
Each word in sentence 1 can obtain a word frequency-reverse document-frequency value, so as to obtain the word of sentence
Frequently-reverse document-frequency is vectorial:V1=(x1, x2, x3, x4 ..., xn)
Each word in sentence 2 can obtain a word frequency-reverse document-frequency value, so as to obtain the word of sentence
Frequently-reverse document-frequency is vectorial:V2=(y1, y2, y3, y4 ..., yn)
The similarity of two sentences can be obtained by calculating the similarity of v1, v2.
It is appreciated that there are multiple words in a word, word frequency-reverse document-frequency value composition sentence of multiple words is calculated
Word frequency-reverse document-frequency vector, customer problem and word frequency-reverse file frequency of each problem that prestores are obtained according to the method
Rate vector, finally according between word frequency-reverse document-frequency vector similarity (such as Euler's distance, cosine similarity,
The methods of Jaccard coefficients) similarity of customer problem and each problem that prestores is obtained.
4) mode four, with reference to Fig. 7, the step S21 includes:
When searching algorithm is term vector algorithm, the customer problem and each problem that prestores all are converted by step S21J
Term vector;
The term vector of the customer problem is multiplied by step S21K with the term vector of each problem that prestores respectively, with
Calculate the similarity of the customer problem and each problem that prestores.
In the present embodiment, when searching algorithm is term vector algorithm (word2vec algorithm), by the customer problem
Term vector is all converted to each problem that prestores, then respectively by the term vector of the customer problem and the word of each problem that prestores
Vector is multiplied, to calculate the similarity of the customer problem and each problem that prestores.To be best understood from, it is exemplified below:
Such as prototype statement is " computer software ", sentence for future reference has:" computer software ", " novel programmed tool ", " software
Maintenance tool ", " assembling computer ".Customer problem and the term vector of each problem that prestores are obtained using cbow models, is recycled
Word2vec models calculate similarity, and it is as follows to obtain result:
Source:Computer software
Result:
Vector:82183 similarities:1.000000 participle:Computer software
Vector:102563 similarities:0.987453 participle:Novel programmed tool
Vector:222835 similarities:0.813425 participle:Software maintenance tool
Vector:4435246 similarities:0.725321 participle:Assemble computer.
5) mode five, with reference to Fig. 8, the step S21 includes:
Step S21L when searching algorithm is Recognition with Recurrent Neural Network algorithm, obtains the associated instruction of Recognition with Recurrent Neural Network algorithm
Practice model;
Customer problem and each problem of prestoring are input to the training pattern by step S21M, to pass through the training
Model calculates the customer problem and the similarity of each problem that prestores, and exports similarity and reach the multiple of the predetermined threshold value
The problem that prestores and corresponding similarity.
In the present embodiment, when searching algorithm is Recognition with Recurrent Neural Network algorithm (RNN algorithm), terminal first obtains cycle
The associated training pattern of neural network algorithm, the training pattern train to obtain according to mass data in advance, are getting trained mould
After type, customer problem and each problem of prestoring are input to the training pattern, to calculate institute by the training pattern
State the similarity of customer problem and each problem that prestores, and export similarity reach the predetermined threshold value it is multiple prestore problem and
Corresponding similarity.
Pass through above-mentioned each searching algorithm, you can calculate the similarity of customer problem and each problem that prestores, Ran Hou
In the similarity that each searching algorithm calculates, determine that the similarity with the customer problem is asked more than the multiple of predetermined threshold value
Topic is most more than multiple problems of predetermined threshold value as matched with the customer problem with the similarity of the customer problem at last
Multiple problems.That is, after a customer problem is inputted, by each searching algorithm, it can obtain and multiple asked with user
The problem of topic matching.
Step S30, it is determining highest with the customer problem matching degree in matched multiple problems using sorting algorithm
Problem;
After multiple the problem of being matched with customer problem are obtained according to each searching algorithm, using the second predetermined number
The problem of sorting algorithm is determined in matched each problem with the customer problem matching degree highest, specifically, reference Fig. 9,
The step S30 includes:
Step S31 is obtained and the matched multiple problems of the customer problem;
Step S32 calculates each problem and customer problem in multiple problems using the sorting algorithm of the second predetermined number
Matching rate;
Step S33 obtains the preset weights of each sorting algorithm, to the preset weights of each sorting algorithm and each asks
The matching rate of topic and customer problem is weighted averagely, obtains the similarity value of each problem and customer problem;
Step S34, the problem of using the problem of similarity value highest as with the customer problem matching degree highest.
That is, first acquisition and the matched multiple problems of the customer problem, then using the sorting algorithm of the second predetermined number
Calculate the matching rate of each problem and customer problem in multiple problems, in the present embodiment, sorting algorithm for gbdt algorithms and
Softmax algorithms, wherein, gbdt is a kind of decision Tree algorithms of iteration, which is made of more decision trees, the knot of all trees
Final result is done by adding up;Softmax algorithms are to return classification function.By two classification functions calculate each problems with
After the matching rate of customer problem, the preset weights of each sorting algorithm are obtained, then according to the preset power of each sorting algorithm
Value and the matching rate of each problem and customer problem are weighted averagely, obtain each problem and the similarity of customer problem
Value.
To be best understood from the present embodiment, it is exemplified below:After problem is calculated by two sorting algorithms there are one current, obtain
Matching rate be respectively 0.9 and 0.8, obtain two preset weights of sorting algorithm, in the present embodiment, each sorting algorithm is preset
The corresponding concrete numerical value of weights be to be set in advance according to actual conditions, in the present embodiment the weights of optional gbdt algorithms be 0.7,
The weights of softmax algorithms are 0.8.So, by the preset weights of each sorting algorithm and each problem and customer problem
Matching rate is weighted averagely, and weighted results are (0.7*0.9+0.8*0.8)/2=0.635, finally obtain the problem and user
The similarity value of problem is 0.635, calculates the similarity value of each problem and customer problem successively by this algorithm.Most
Eventually, using the problem of similarity value highest as with the customer problem matching degree highest the problem of.
Step S40, the problem of matching degree highest is searched in the knowledge base associated answer, and answering lookup
Case feeds back to display interface.
It is found in knowledge base with being searched the problem of the customer problem matching degree highest and then from the knowledge base
The associated answer of the problem of matching degree highest, then by the answer feedback of lookup to display interface.
The problem of the present embodiment proposes matching process first receives customer problem, and the customer problem of reception is inputted
Into knowledge base, then each problem of prestoring is retrieved using searching algorithm in the knowledge base, to retrieve and institute
The matched multiple problems of customer problem are stated, then are determined and the customer problem in matched multiple problems using sorting algorithm
The problem of with degree highest, the problem of matching degree highest is finally searched in the knowledge base associated answer, and will search
Answer feedback to display interface.The matching for realizing problem is determined by searching algorithm and two class algorithm of sorting algorithm, phase
The matching of problem is carried out for the single short text Similarity Algorithm of tradition, the present invention is more comprehensive to the matching of problem, improves
The matched accuracy of problem, so as to improving the accuracy of answer feedback.
To be best understood from this programme, with reference to Figure 10, this programme is described with concrete scene.
After customer problem is input to knowledge base, determined in knowledge base respectively using five searching algorithms in Figure 10
Multiple the problem of reaching predetermined threshold value with customer problem similarity, the problem of multiple similarities then are reached predetermined threshold value, are defeated again
Enter into two sorting algorithms, to calculate the matching of each problem and customer problem in multiple problems by two sorting algorithms
Rate then according to the weights set before two sorting algorithms, calculates the similarity value of each problem and customer problem, most at last
The problem of obtaining after the problem of similarity value highest matches as the customer problem in knowledge base, and looked into the knowledge base
The corresponding answer of the problem is looked for, and feeds back the answer to user.
The present invention further provides a kind of computer readable storage mediums.
Problematic matcher is stored on the computer readable storage medium, described problem matcher is held by processor
Following steps are realized during row:
Customer problem is received, and the customer problem of reception is input in knowledge base;
Each problem of prestoring is retrieved using searching algorithm in the knowledge base, is asked with retrieving with the user
Inscribe matched multiple problems;
The problem of being determined in matched multiple problems with the customer problem matching degree highest using sorting algorithm;
The associated answer of the problem of matching degree highest is searched in the knowledge base, and by the answer feedback of lookup extremely
Display interface.
Further, it when described problem matcher is executed by processor, also realizes in the knowledge base using retrieval
Algorithm retrieves each problem of prestoring, the step of to retrieve multiple problems matched with the customer problem:
In the knowledge base, the customer problem is calculated using the first predetermined number searching algorithm respectively and is prestored with each
The similarity of problem;
In the similarity calculated in each searching algorithm, determine that the similarity with the customer problem reaches predetermined threshold value
Similarity;
Using with the similarity of the customer problem reach the corresponding multiple problems of predetermined threshold value as with the customer problem
Matched multiple problems.
Further, it when described problem matcher is executed by processor, also realizes in the knowledge base, using first
Predetermined number searching algorithm calculates the step of customer problem and similarity of each problem that prestores respectively:
When searching algorithm is convolutional neural networks algorithm, the associated training pattern of convolutional neural networks algorithm is obtained;
Customer problem and each problem of prestoring are input to the training pattern, to calculate institute by the training pattern
State the similarity of customer problem and each problem that prestores, and export similarity reach the predetermined threshold value it is multiple prestore problem and
Corresponding similarity.
Further, it when described problem matcher is executed by processor, also realizes in the knowledge base, using first
Predetermined number searching algorithm calculates the step of customer problem and similarity of each problem that prestores respectively:
When searching algorithm is editing distance short text Similarity algorithm, to the word of the customer problem and each problem that prestores
Symbol is divided;
When calculating each problem of prestoring and being converted to customer problem, the number of edit operation performed is needed;
The number that performs is needed when being converted to customer problem according to each problem of prestoring, determine the customer problem with it is each
The similarity for the problem that prestores.
Further, it when described problem matcher is executed by processor, also realizes in the knowledge base, using first
Predetermined number searching algorithm calculates the step of customer problem and similarity of each problem that prestores respectively:
In searching algorithm document-frequency algorithm reverse for word frequency, word is carried out to the customer problem and each problem that prestores
Language is split, and the customer problem and each problem of prestoring are splitted into multiple words;
The word frequency of customer problem and each problem that prestores and reverse file frequency are calculated respectively according to multiple words after fractionation
Rate;
Word frequency-reverse document-frequency vector of customer problem is obtained according to the word frequency of customer problem and reverse document-frequency,
And according to the word frequency of each problem that prestores and reverse document-frequency obtain the word frequency of each problem that prestores-reverse document-frequency to
Amount;
Calculate word frequency-reverse document-frequency vector of customer problem and word frequency-reverse document-frequency of each problem that prestores
The corresponding cosine similarity of vector, to obtain the similarity of customer problem and each problem that prestores.
Further, it when described problem matcher is executed by processor, also realizes in the knowledge base, using first
Predetermined number searching algorithm calculates the step of customer problem and similarity of each problem that prestores respectively:
When searching algorithm is term vector algorithm, the customer problem and each problem that prestores all are converted into term vector;
The term vector of the customer problem is multiplied with the term vector of each problem that prestores respectively, to calculate the use
The similarity of family problem and each problem that prestores.
Further, it when described problem matcher is executed by processor, also realizes in the knowledge base, using first
Predetermined number searching algorithm calculates the step of customer problem and similarity of each problem that prestores respectively:
When searching algorithm is Recognition with Recurrent Neural Network algorithm, the associated training pattern of Recognition with Recurrent Neural Network algorithm is obtained;
Customer problem and each problem of prestoring are input to the training pattern, to calculate institute by the training pattern
State the similarity of customer problem and each problem that prestores, and export similarity reach the predetermined threshold value it is multiple prestore problem and
Corresponding similarity.
Further, it when described problem matcher is executed by processor, also realizes using sorting algorithm matched more
The step of being determined in a problem with the problem of the customer problem matching degree highest:
It obtains and the matched multiple problems of the customer problem;
The matching rate of each problem and customer problem in multiple problems is calculated using the sorting algorithm of the second predetermined number;
The preset weights of each sorting algorithm are obtained, to the preset weights of each sorting algorithm and each problem and user
The matching rate of problem is weighted averagely, obtains the similarity value of each problem and customer problem;
The problem of using the problem of similarity value highest as with the customer problem matching degree highest.
Technical solution proposed by the present invention when described problem matcher is executed by processor, realizes following steps:First connect
Customer problem is received, and the customer problem of reception is input in knowledge base, is then calculated in the knowledge base using retrieval
Method retrieves each problem of prestoring, with retrieve with the matched multiple problems of the customer problem, then using sorting algorithm
The problem of determining in matched multiple problems with the customer problem matching degree highest, finally searches institute in the knowledge base
The associated answer of the problem of stating matching degree highest, and by the answer feedback of lookup to display interface.Realizing the matching of problem is
It is determined by searching algorithm and two class algorithm of sorting algorithm, problem is carried out relative to the single short text Similarity Algorithm of tradition
Matching, the present invention is more comprehensive to the matching of problem, the matched accuracy of problem is improved, so as to improve the standard of answer feedback
True property.
It should be noted that herein, term " comprising ", "comprising" or its any other variant are intended to non-row
His property includes, so that process, method, article or device including a series of elements not only include those elements, and
And it further includes the other elements being not explicitly listed or further includes intrinsic for this process, method, article or device institute
Element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that including this
Also there are other identical elements in the process of element, method, article or device.
The embodiments of the present invention are for illustration only, do not represent the quality of embodiment.
Through the above description of the embodiments, those skilled in the art can be understood that above-described embodiment side
Method can add the mode of required general hardware platform to realize by software, naturally it is also possible to by hardware, but in many cases
The former is more preferably embodiment.Based on such understanding, technical scheme of the present invention substantially in other words does the prior art
Going out the part of contribution can be embodied in the form of software product, which is stored in a storage medium
In (such as ROM/RAM, magnetic disc, CD), used including some instructions so that a station terminal equipment (can be mobile phone, computer takes
Be engaged in device, air conditioner or the network equipment etc.) perform method described in each embodiment of the present invention.
It these are only the preferred embodiment of the present invention, be not intended to limit the scope of the invention, it is every to utilize this hair
The equivalent structure or equivalent flow shift that bright specification and accompanying drawing content are made directly or indirectly is used in other relevant skills
Art field, is included within the scope of the present invention.
Claims (10)
1. a kind of problem matching process, which is characterized in that described problem matching process includes:
Customer problem is received, and the customer problem of reception is input in knowledge base;
Each problem of prestoring is retrieved using searching algorithm in the knowledge base, to retrieve and the customer problem
The multiple problems matched;
The problem of being determined in matched multiple problems with the customer problem matching degree highest using sorting algorithm;
The associated answer of the problem of matching degree highest is searched in the knowledge base, and the answer feedback of lookup is extremely shown
Interface.
2. problem matching process as described in claim 1, which is characterized in that described that searching algorithm is used in the knowledge base
Each problem of prestoring is retrieved, is included the step of multiple problems matched with the customer problem with retrieving:
In the knowledge base, the customer problem and each problem that prestores are calculated using the first predetermined number searching algorithm respectively
Similarity;
In the similarity calculated in each searching algorithm, determine that the similarity with the customer problem reaches the phase of predetermined threshold value
Like degree;
It is matched the corresponding multiple problems of predetermined threshold value are reached with the similarity of the customer problem as with the customer problem
Multiple problems.
3. problem matching process as claimed in claim 2, which is characterized in that it is described in the knowledge base, it is pre- using first
If number searching algorithm calculates the step of customer problem and similarity of each problem that prestores and includes respectively:
When searching algorithm is convolutional neural networks algorithm, the associated training pattern of convolutional neural networks algorithm is obtained;
Customer problem and each problem of prestoring are input to the training pattern, to calculate the use by the training pattern
The similarity of family problem and each problem that prestores, and export multiple problem and correspondences of prestoring that similarity reaches the predetermined threshold value
Similarity.
4. problem matching process as claimed in claim 2, which is characterized in that it is described in the knowledge base, it is pre- using first
If number searching algorithm calculates the step of customer problem and similarity of each problem that prestores and includes respectively:
When searching algorithm is editing distance short text Similarity algorithm, to the character of the customer problem and each problem that prestores into
Row divides;
When calculating each problem of prestoring and being converted to customer problem, the number of edit operation performed is needed;
The number performed is needed when being converted to customer problem according to each problem of prestoring, determines that the customer problem prestores with each
The similarity of problem.
5. problem matching process as claimed in claim 2, which is characterized in that it is described in the knowledge base, it is pre- using first
If number searching algorithm calculates the step of customer problem and similarity of each problem that prestores and includes respectively:
In searching algorithm document-frequency algorithm reverse for word frequency, are carried out by word and is torn open for the customer problem and each problem that prestores
Point, the customer problem and each problem of prestoring are splitted into multiple words;
The word frequency of customer problem and each problem that prestores and reverse document-frequency are calculated respectively according to multiple words after fractionation;
Word frequency-reverse document-frequency the vector and root of customer problem are obtained according to the word frequency of customer problem and reverse document-frequency
Word frequency-reverse document-frequency vector of each problem that prestores is obtained according to the word frequency and reverse document-frequency of each problem that prestores;
Calculate word frequency-reverse document-frequency vector of customer problem and word frequency-reverse document-frequency vector of each problem that prestores
Corresponding cosine similarity, to obtain the similarity of customer problem and each problem that prestores.
6. problem matching process as claimed in claim 2, which is characterized in that it is described in the knowledge base, it is pre- using first
If number searching algorithm calculates the step of customer problem and similarity of each problem that prestores and includes respectively:
When searching algorithm is term vector algorithm, the customer problem and each problem that prestores all are converted into term vector;
The term vector of the customer problem with the term vector of each problem that prestores is multiplied respectively, is asked with calculating the user
The similarity of topic and each problem that prestores.
7. problem matching process as claimed in claim 2, which is characterized in that it is described in the knowledge base, it is pre- using first
If number searching algorithm calculates the step of customer problem and similarity of each problem that prestores and includes respectively:
When searching algorithm is Recognition with Recurrent Neural Network algorithm, the associated training pattern of Recognition with Recurrent Neural Network algorithm is obtained;
Customer problem and each problem of prestoring are input to the training pattern, to calculate the use by the training pattern
The similarity of family problem and each problem that prestores, and export multiple problem and correspondences of prestoring that similarity reaches the predetermined threshold value
Similarity.
8. such as claim 1-7 any one of them problem matching process, which is characterized in that described to be matched using sorting algorithm
Multiple problems in determine with the problem of the customer problem matching degree highest the step of include:
It obtains and the matched multiple problems of the customer problem;
The matching rate of each problem and customer problem in multiple problems is calculated using the sorting algorithm of the second predetermined number;
The preset weights of each sorting algorithm are obtained, to the preset weights of each sorting algorithm and each problem and customer problem
Matching rate be weighted average, obtain the similarity value of each problem and customer problem;
The problem of using the problem of similarity value highest as with the customer problem matching degree highest.
9. a kind of terminal, which is characterized in that the terminal includes processor, memory and is stored on the memory and can be
The problem of being run on processor matcher realizes that right such as will when described problem matcher is performed by the processor
The step of the problem of asking any one of 1 to 8 described matching process.
10. a kind of computer readable storage medium, which is characterized in that problematic is stored on the computer readable storage medium
With program, such as claim 1 to 8 any one of them problem match party is realized when described problem matcher is executed by processor
The step of method.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711362221.3A CN108170739A (en) | 2017-12-18 | 2017-12-18 | Problem matching process, terminal and computer readable storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711362221.3A CN108170739A (en) | 2017-12-18 | 2017-12-18 | Problem matching process, terminal and computer readable storage medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108170739A true CN108170739A (en) | 2018-06-15 |
Family
ID=62522157
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711362221.3A Pending CN108170739A (en) | 2017-12-18 | 2017-12-18 | Problem matching process, terminal and computer readable storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108170739A (en) |
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108897888A (en) * | 2018-07-10 | 2018-11-27 | 四川淘金你我信息技术有限公司 | It is man-machine to white silk method under voice customer service training scene |
CN109033244A (en) * | 2018-07-05 | 2018-12-18 | 百度在线网络技术(北京)有限公司 | Search result ordering method and device |
CN109189881A (en) * | 2018-06-20 | 2019-01-11 | 网宿科技股份有限公司 | Man-machine interaction method and intelligent robot |
CN109376847A (en) * | 2018-08-31 | 2019-02-22 | 深圳壹账通智能科技有限公司 | User's intension recognizing method, device, terminal and computer readable storage medium |
CN109657232A (en) * | 2018-11-16 | 2019-04-19 | 北京九狐时代智能科技有限公司 | A kind of intension recognizing method |
CN109947909A (en) * | 2018-06-19 | 2019-06-28 | 平安科技(深圳)有限公司 | Intelligent customer service answer method, equipment, storage medium and device |
CN110047473A (en) * | 2019-04-19 | 2019-07-23 | 交通银行股份有限公司太平洋信用卡中心 | A kind of man-machine collaboration exchange method and system |
CN110188205A (en) * | 2019-05-08 | 2019-08-30 | 三角兽(北京)科技有限公司 | A kind of update method and device of intelligent customer service system knowledge base |
CN110209787A (en) * | 2019-05-29 | 2019-09-06 | 袁琦 | A kind of intelligent answer method and system based on pet knowledge mapping |
CN110674087A (en) * | 2019-09-03 | 2020-01-10 | 平安科技(深圳)有限公司 | File query method and device and computer readable storage medium |
CN110866095A (en) * | 2019-10-10 | 2020-03-06 | 重庆金融资产交易所有限责任公司 | Text similarity determination method and related equipment |
CN110941706A (en) * | 2018-09-21 | 2020-03-31 | 华为技术有限公司 | Answer determining method and system |
CN111008268A (en) * | 2019-10-31 | 2020-04-14 | 支付宝(杭州)信息技术有限公司 | Method and device for acquiring question reversing sentence corresponding to user question based on dialog system |
CN111881672A (en) * | 2020-06-18 | 2020-11-03 | 升智信息科技(南京)有限公司 | Intention identification method |
CN112765960A (en) * | 2021-02-07 | 2021-05-07 | 成都新潮传媒集团有限公司 | Text matching method and device and computer equipment |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2007136402A1 (en) * | 2006-05-16 | 2007-11-29 | Iac Search & Media, Inc. | Systems and methods for answering user questions |
CN101169780A (en) * | 2006-10-25 | 2008-04-30 | 华为技术有限公司 | Semantic ontology retrieval system and method |
CN102411580A (en) * | 2010-09-20 | 2012-04-11 | 腾讯科技(深圳)有限公司 | Retrieval method and device for extensible markup language (XML) files |
CN102542000A (en) * | 2011-12-07 | 2012-07-04 | 北京风灵创景科技有限公司 | Method and equipment for retrieving contacts |
CN105653559A (en) * | 2014-11-28 | 2016-06-08 | 国际商业机器公司 | Method and device for searching in database |
CN106776532A (en) * | 2015-11-25 | 2017-05-31 | 中国移动通信集团公司 | A kind of knowledge question answering method and device |
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 |
-
2017
- 2017-12-18 CN CN201711362221.3A patent/CN108170739A/en active Pending
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2007136402A1 (en) * | 2006-05-16 | 2007-11-29 | Iac Search & Media, Inc. | Systems and methods for answering user questions |
CN101169780A (en) * | 2006-10-25 | 2008-04-30 | 华为技术有限公司 | Semantic ontology retrieval system and method |
CN102411580A (en) * | 2010-09-20 | 2012-04-11 | 腾讯科技(深圳)有限公司 | Retrieval method and device for extensible markup language (XML) files |
CN102542000A (en) * | 2011-12-07 | 2012-07-04 | 北京风灵创景科技有限公司 | Method and equipment for retrieving contacts |
CN105653559A (en) * | 2014-11-28 | 2016-06-08 | 国际商业机器公司 | Method and device for searching in database |
CN106776532A (en) * | 2015-11-25 | 2017-05-31 | 中国移动通信集团公司 | A kind of knowledge question answering method and device |
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 |
Cited By (21)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109947909A (en) * | 2018-06-19 | 2019-06-28 | 平安科技(深圳)有限公司 | Intelligent customer service answer method, equipment, storage medium and device |
CN109947909B (en) * | 2018-06-19 | 2024-03-12 | 平安科技(深圳)有限公司 | Intelligent customer service response method, equipment, storage medium and device |
CN109189881A (en) * | 2018-06-20 | 2019-01-11 | 网宿科技股份有限公司 | Man-machine interaction method and intelligent robot |
CN109033244A (en) * | 2018-07-05 | 2018-12-18 | 百度在线网络技术(北京)有限公司 | Search result ordering method and device |
CN109033244B (en) * | 2018-07-05 | 2020-10-16 | 百度在线网络技术(北京)有限公司 | Search result ordering method and device |
CN108897888A (en) * | 2018-07-10 | 2018-11-27 | 四川淘金你我信息技术有限公司 | It is man-machine to white silk method under voice customer service training scene |
CN108897888B (en) * | 2018-07-10 | 2021-08-24 | 四川淘金你我信息技术有限公司 | Man-machine sparring method under voice customer service training scene |
CN109376847A (en) * | 2018-08-31 | 2019-02-22 | 深圳壹账通智能科技有限公司 | User's intension recognizing method, device, terminal and computer readable storage medium |
CN110941706A (en) * | 2018-09-21 | 2020-03-31 | 华为技术有限公司 | Answer determining method and system |
CN109657232A (en) * | 2018-11-16 | 2019-04-19 | 北京九狐时代智能科技有限公司 | A kind of intension recognizing method |
CN110047473A (en) * | 2019-04-19 | 2019-07-23 | 交通银行股份有限公司太平洋信用卡中心 | A kind of man-machine collaboration exchange method and system |
CN110188205A (en) * | 2019-05-08 | 2019-08-30 | 三角兽(北京)科技有限公司 | A kind of update method and device of intelligent customer service system knowledge base |
CN110209787A (en) * | 2019-05-29 | 2019-09-06 | 袁琦 | A kind of intelligent answer method and system based on pet knowledge mapping |
CN110209787B (en) * | 2019-05-29 | 2023-09-22 | 袁琦 | Intelligent question-answering method and system based on pet knowledge graph |
CN110674087A (en) * | 2019-09-03 | 2020-01-10 | 平安科技(深圳)有限公司 | File query method and device and computer readable storage medium |
CN110866095A (en) * | 2019-10-10 | 2020-03-06 | 重庆金融资产交易所有限责任公司 | Text similarity determination method and related equipment |
CN111008268B (en) * | 2019-10-31 | 2021-05-18 | 支付宝(杭州)信息技术有限公司 | Method and device for acquiring question reversing sentence corresponding to user question based on dialog system |
CN111008268A (en) * | 2019-10-31 | 2020-04-14 | 支付宝(杭州)信息技术有限公司 | Method and device for acquiring question reversing sentence corresponding to user question based on dialog system |
CN111881672A (en) * | 2020-06-18 | 2020-11-03 | 升智信息科技(南京)有限公司 | Intention identification method |
CN112765960A (en) * | 2021-02-07 | 2021-05-07 | 成都新潮传媒集团有限公司 | Text matching method and device and computer equipment |
CN112765960B (en) * | 2021-02-07 | 2022-11-25 | 成都新潮传媒集团有限公司 | Text matching method and device and computer equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108170739A (en) | Problem matching process, terminal and computer readable storage medium | |
CN109902672B (en) | Image labeling method and device, storage medium and computer equipment | |
CN108319599A (en) | A kind of interactive method and apparatus | |
CN110362667A (en) | Intelligent customer service method, apparatus, equipment and readable storage medium storing program for executing | |
CN109885810A (en) | Nan-machine interrogation's method, apparatus, equipment and storage medium based on semanteme parsing | |
CN109240637B (en) | Volume adjustment processing method, device, equipment and storage medium | |
CN109240901A (en) | Method for analyzing performance, performance evaluation device, storage medium and electronic equipment | |
CN109447958B (en) | Image processing method, image processing device, storage medium and computer equipment | |
CN107679082A (en) | Question and answer searching method, device and electronic equipment | |
CN109885828A (en) | Word error correction method, device, computer equipment and medium based on language model | |
CN107589828A (en) | The man-machine interaction method and system of knowledge based collection of illustrative plates | |
CN107818491A (en) | Electronic installation, Products Show method and storage medium based on user's Internet data | |
CN111062520B (en) | Hostname feature prediction method based on random forest algorithm | |
US10762902B2 (en) | Method and apparatus for synthesizing adaptive data visualizations | |
CN110019616A (en) | A kind of POI trend of the times state acquiring method and its equipment, storage medium, server | |
CN107280693A (en) | Psychoanalysis System and method based on VR interactive electronic sand tables | |
CN112231497B (en) | Information classification method and device, storage medium and electronic equipment | |
CN107688950A (en) | Household electrical appliances consultation method, electronic equipment and computer-readable recording medium after sale | |
CN108776677A (en) | Creation method, equipment and the computer readable storage medium of parallel statement library | |
CN111079001A (en) | Decoration recommendation information generation method and device, storage medium and electronic equipment | |
CN108319888A (en) | The recognition methods of video type and device, terminal | |
CN109062963A (en) | From Media Recommendation Method, device and electronic equipment | |
CN109165286A (en) | Automatic question-answering method, device and computer readable storage medium | |
CN108229640A (en) | The method, apparatus and robot of emotion expression service | |
CN103064967B (en) | A kind of method and apparatus for establishing user's binary crelation library |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180615 |
|
RJ01 | Rejection of invention patent application after publication |