CN105786794B - A kind of question and answer are to search method and community's question and answer searching system - Google Patents
A kind of question and answer are to search method and community's question and answer searching system Download PDFInfo
- Publication number
- CN105786794B CN105786794B CN201610082304.6A CN201610082304A CN105786794B CN 105786794 B CN105786794 B CN 105786794B CN 201610082304 A CN201610082304 A CN 201610082304A CN 105786794 B CN105786794 B CN 105786794B
- Authority
- CN
- China
- Prior art keywords
- lexical item
- question sentence
- weight
- keyword
- 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.)
- Active
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
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/951—Indexing; Web crawling techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/237—Lexical tools
- G06F40/247—Thesauruses; Synonyms
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a kind of question and answer to search method, including:At least one keyword is extracted from question sentence, and obtains the expansion word of each keyword and the final similarity of each expansion word and corresponding keyword;Analyze the dependence between each two lexical item with grammar association in question sentence;According to being in advance the importance weight of dependence setting, the distance weighting of tightness degree between each two lexical item in reflection question sentence is determined;The degree of association in question sentence between each two lexical item is determined according to distance weighting;The lexical item weight of each lexical item in question sentence is determined according to the degree of association, and according to the lexical item weight retrieval of each lexical item in question sentence and the relevant question and answer pair of question sentence;According to the lexical item weight of expansion word and the final similarity calculation expansion word of corresponding keyword, and according to the retrieval of the lexical item weight of expansion word and the relevant question and answer pair of question sentence;By all question and answer retrieved to being ranked up display according to preset rules.The invention also discloses a kind of community's question and answer searching systems.
Description
Technical field
The present invention relates to technical field of information retrieval more particularly to a kind of question and answer to search method and community's question and answer retrieval system
System.
Background technology
In recent years, community's question answering system has been increasingly becoming a kind of very popular and practical the Internet, applications, is asked with tradition
It answers unlike system, in community's question answering system, user not only can put question to and answer any field, any kind of ask
Topic, but also can make the answer of other users evaluation and ballot, or even can also directly search that system accumulated go through
Similar Problems in history problem answers library, greatly enrich and meet the information requirement of user.
When user wants to utilize the retrieval of community's question answering system and oneself proposition problem same or similar problem and its answer
When, due to it is input by user be the question sentence using natural language description, complicated structure and interminable clause so that from question sentence
The middle important key word item of extraction can be relatively difficult.Because can not accurately obtain the kernel keyword in question sentence, lead to retrieval result not
It is enough accurate.
Invention content
Search method and community's question and answer are examined in view of this, the main purpose of the embodiment of the present invention is to provide a kind of question and answer
Cable system, to realize the purpose for improving question and answer to the accuracy of retrieval result.
To achieve the above object, an embodiment of the present invention provides a kind of question and answer to search method, including:
At least one keyword is extracted from question sentence, and obtain each keyword expansion word and each expansion word with it is right
Answer the final similarity of keyword;
Analyze the dependence between each two lexical item with grammar association in the question sentence;
According to the importance weight for being in advance dependence setting, determine reflect in the question sentence each two lexical item it
Between tightness degree distance weighting;
The degree of association in the question sentence between each two lexical item is determined according to the distance weighting;
The lexical item weight of each lexical item in the question sentence is determined according to the degree of association, and according to each word in the question sentence
The retrieval of lexical item weight and the relevant question and answer pair of the question sentence of item;
According to the lexical item weight of the expansion word and expansion word described in the final similarity calculation of corresponding keyword, and according to
The retrieval of lexical item weight and the relevant question and answer pair of the question sentence of the expansion word;
By all question and answer retrieved to being ranked up display according to preset rules.
Optionally, the expansion word for obtaining each keyword and each expansion word are final similar with corresponding keyword
Degree, including:
Obtain at least one expansion word of each keyword respectively using Hownet HowNet, and define each expansion word with it is right
It is 1 to answer the initial similarity of keyword;
Obtain at least one expansion word of each keyword respectively using Chinese thesaurus, and define each expansion word with it is right
It is 1 to answer the initial similarity of keyword;
Using the text depth representing model word2vec after trained, at least one expansion of each keyword is obtained respectively
Open up the initial similarity of word and each expansion word and corresponding keyword;
Merge the identical expansion word got, calculates separately the most last phase of each expansion word and corresponding keyword after merging
Like degree SR, wherein SR=Ssum/ 3, SsumFor the sum of corresponding all initial similarities of the expansion word.
Optionally, the basis is the importance weight of dependence setting in advance, determines and reflects in the question sentence
The distance weighting of tightness degree between each two lexical item, including:
The distance between the first lexical item and each second lexical item weight D are calculated separately, first lexical item is the question sentence
In any one lexical item, second lexical item be there are the lexical items of the dependence with first lexical item;
Wherein,Y be that dependence between first lexical item and second lexical item is arranged in advance
Importance weight is worth on the basis of α;
The distance between first lexical item and each third lexical item weight Dis are calculated separately, the third lexical item is institute
Any one lexical item in addition to first word in question sentence is stated, Dis is deposited between first lexical item and the third lexical item
The sum of the corresponding at least one distance weighting D of at least one dependence.
Optionally, the degree of association determined according to the distance weighting in the question sentence between each two lexical item, including:
Lexical item t in the question sentence is calculated according to the following equationiWith lexical item tjBetween degree of association wrel(i,j):
wrel(i,j)=λ Dep (ti,tj)+(1-λ)Closepmi(ti,tj);
tiIndicate i-th of lexical item in the question sentence, tjIndicate j-th of lexical item in the question sentence, i=1,2 ... n, j
=1,2 ... n, n are the lexical item sum in the question sentence;
λ is regulatory factor;
B is a constant more than 1;
Dis(ti,tj) it is lexical item tiWith lexical item tjThe distance between weight;
Lexical item t is concentrated for question sentenceiWith lexical item tjThe probability occurred jointly, Nd(ti,tj) it is question sentence
Concentrate lexical item tiWith lexical item tjThe number of the question sentence occurred simultaneously, NDThe question sentence sum concentrated for question sentence;
WithLexical item t is indicated respectivelyiWith lexical item tjEach leisure question sentence concentrates appearance
Probability, Nd(ti) it is that question sentence is concentrated comprising lexical item tiQuestion sentence sum, Nd(tj) it is that question sentence is concentrated comprising lexical item tjQuestion sentence it is total
Number, NDThe question sentence sum concentrated for question sentence.
Optionally, the lexical item weight that each lexical item in the question sentence is determined according to the degree of association, including:
The weight matrix of each lexical item final weight composition in the question sentence is calculated according to the following equation
Wherein, α is given constant;
E is that incidence matrix M is carried out the random matrix after orthogonal transformation, and the incidence matrix M is every two in the question sentence
The symmetrical matrix that the degree of association between a lexical item is formed;
For the weight matrix of the original weight composition of each lexical item in the question sentence.
Optionally, the lexical item according to the expansion word and expansion word described in the final similarity calculation of corresponding keyword
Weight, including:
Obtain the original weight of the corresponding keyword of the expansion word;
By the product of the original weight and the expansion word and the final similarity of corresponding keyword, as the extension
The lexical item weight of word.
The embodiment of the present invention additionally provides a kind of community's question and answer searching system, including:
Keyword extracting unit, for extracting at least one keyword from question sentence;
Keyword expansion unit, expansion word for obtaining each keyword that the keyword extracting unit obtains and
The final similarity of each expansion word and corresponding keyword;
Relationship analysis unit, for analyzing the interdependent pass in the question sentence between each two lexical item with grammar association
System;
Weights determination unit, for according to the weight for being in advance the dependence setting that the relationship analysis unit is analyzed
Weight is spent, determines the distance weighting for reflecting tightness degree between each two lexical item in the question sentence;
Degree of association determination unit, the distance weighting for being determined according to the weights determination unit determine every in the question sentence
The degree of association between two lexical items;
First weight determining unit, the degree of association for being determined according to the degree of association determination unit determine in the question sentence
The lexical item weight of each lexical item;
First retrieval unit, the lexical item power of each lexical item in the question sentence for being determined according to first weight determining unit
Retrieval and the relevant question and answer pair of the question sentence again;
Second weight determining unit, the expansion word for being obtained according to the keyword expansion unit extensions and corresponding key
The lexical item weight of expansion word described in the final similarity calculation of word;
Second retrieval unit, the lexical item weight retrieval of the expansion word for being determined according to second weight determining unit and
The relevant question and answer pair of question sentence;
Retrieval result display unit, it is all for retrieving first retrieval unit and second retrieval unit
Question and answer according to preset rules to being ranked up display.
Optionally, the keyword expansion unit, including:
Hownet expansion module, at least one expansion word for obtaining each keyword respectively using Hownet HowNet, and
It is 1 that each expansion word, which is defined, with the initial similarity of corresponding keyword;
Word woods expansion module, at least one expansion word for obtaining each keyword respectively using Chinese thesaurus, and
It is 1 that each expansion word, which is defined, with the initial similarity of corresponding keyword;
Model extension module obtains each respectively for text depth representing model word2vec of the utilization after trained
The initial similarity of at least one expansion word and each expansion word and corresponding keyword of keyword;
Similarity calculation module expands for merging the Hownet expansion module, institute's predicate woods expansion module and the model
The identical expansion word that exhibition module is got, calculates separately the final similarity of each expansion word and corresponding keyword after merging
SR, wherein SR=Ssum/ 3, SsumFor the sum of corresponding all initial similarities of the expansion word.
Optionally, the weights determination unit, including:
First weight computation module, for calculating separately the distance between the first lexical item and each second lexical item weight D, institute
It is any one lexical item in the question sentence to state the first lexical item, and second lexical item is that there are described interdependent with first lexical item
The lexical item of relationship;
Wherein,Y be in advance between first lexical item and second lexical item dependence setting
Importance weight, be worth on the basis of α;
Second weight computation module, for calculating separately the distance between first lexical item and each third lexical item weight
Dis, the third lexical item are any one lexical item in addition to first word in the question sentence, and Dis is first lexical item
Corresponding at least one the first weight computation module meter of existing at least one dependence between the third lexical item
The sum of obtained distance weighting D.
Optionally, the degree of association determination unit, specifically for calculating lexical item t in the question sentence according to the following equationiWith
Lexical item tjBetween degree of association wrel(i,j):
wrel(i,j)=λ Dep (ti,tj)+(1-λ)Closepmi(ti,tj);
tiIndicate i-th of lexical item in the question sentence, tjIndicate j-th of lexical item in the question sentence, i=1,2 ... n, j
=1,2 ... n, n are the lexical item sum in the question sentence;
λ is regulatory factor;
B is a constant more than 1;
Dis(ti,tj) it is lexical item tiWith lexical item tjThe distance between weight;
Lexical item t is concentrated for question sentenceiWith lexical item tjThe probability occurred jointly, Nd(ti,tj) it is question sentence
Concentrate lexical item tiWith lexical item tjThe number of the question sentence occurred simultaneously, NDThe question sentence sum concentrated for question sentence;
WithLexical item t is indicated respectivelyiWith lexical item tjEach leisure question sentence concentrates appearance
Probability, Nd(ti) it is that question sentence is concentrated comprising lexical item tiQuestion sentence sum, Nd(tj) it is that question sentence is concentrated comprising lexical item tjQuestion sentence it is total
Number, NDThe question sentence sum concentrated for question sentence.
Optionally, first weight determining unit, specifically for calculating each lexical item in the question sentence according to the following equation
The weight matrix of final weight composition
Wherein, α is given constant;
E is that incidence matrix M is carried out the random matrix after orthogonal transformation, and the incidence matrix M is every two in the question sentence
The symmetrical matrix that the degree of association between a lexical item is formed;
For the weight matrix of the original weight composition of each lexical item in the question sentence.
Optionally, second weight determining unit, including:
Original Weight Acquisition module, the original weight for obtaining the corresponding keyword of the expansion word;
Second weight determination module, the original weight for obtaining the original Weight Acquisition module and the expansion word
With the product of the final similarity of corresponding keyword, the lexical item weight as the expansion word.
Question and answer provided in an embodiment of the present invention are to search method and community's question and answer searching system, by being different in question sentence
The importance weight of dependence setting, it may be determined that the association tight ness rating in question sentence between lexical item, it can be further according to the degree of association
The lexical item weight for determining each lexical item in question sentence is can be found that by merging the lexical item weight that importance weight obtains in question sentence
Important lexical item, to obtain solving existing community's question and answer searching system to retrieval result with the more relevant question and answer of question sentence
Not the shortcomings that not accounting for that question sentence is complicated and clause is tediously long and can not finding question sentence important lexical item, and then improve retrieval result
The purpose of accuracy.
Description of the drawings
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technology description to be briefly described, it should be apparent that, the accompanying drawings in the following description is the present invention
Some embodiments for those of ordinary skill in the art without creative efforts, can also basis
These attached drawings obtain other attached drawings.
Fig. 1 is flow diagram of the question and answer of the embodiment of the present invention to search method;
Fig. 2 is expansion word of the embodiment of the present invention and the acquisition methods block diagram of similarity;
Fig. 3 is question sentence retrieving block diagram of the embodiment of the present invention;
Fig. 4 is dependence schematic diagram of the embodiment of the present invention;
Fig. 5 is the composition schematic diagram of community of embodiment of the present invention question and answer searching system.
Specific implementation mode
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art
The every other embodiment obtained without making creative work, shall fall within the protection scope of the present invention.
When the problem that user is proposed in the form of question sentence, by the way that the question sentence is input to community's question and answer searching system
In, community's question and answer searching system can retrieve the problem same or similar with the question semanteme, while retrieve and the question sentence
The semantic same or similar corresponding answer of problem, that is, retrieve question and answer pair, and according to the degree of approximation with question sentence, by question and answer into
Row sequencing display.Lower mask body introduces the search method of question and answer pair.
It is flow diagram of the question and answer provided in an embodiment of the present invention to search method referring to Fig. 1, this method includes:
Step 101:At least one keyword is extracted from question sentence, and obtains the expansion word of the keyword and each expansion
Open up the final similarity of word and the keyword.
The acquisition methods block diagram of expansion word and similarity shown in Figure 2.
First, the keyword in the question sentence is extracted according to part of speech and deactivated vocabulary.Specifically, it removes to belong in the question sentence and stop
With the stop words in vocabulary, then extraction belongs to the lexical item of specific part of speech from the lexical item after removal stop words, such as default specific
Part of speech is respectively nouns and adjectives (can certainly be other types of part of speech), from each lexical item after removal stop words
It is the lexical item of noun and part of speech is adjectival part of speech to extract part of speech, using the noun lexical item extracted and adjective lexical item as
Keyword.The purpose of extraction keyword is that the quantity by limiting keyword, which can play, prevents lexical item from propagating through mostly and influencing
The problem of recall precision.
Then, conventional Extension and the word2vec extensions of text depth representing model are carried out to each keyword.Wherein, described
Conventional Extension includes:Each keyword is extended using Hownet HowNet, to obtain an expansion word item collection, and, make
Each keyword is extended with Chinese thesaurus, to obtain an expansion word item collection;The word2vec extensions refer to making
Each keyword is extended with the word2vec models after trained, to obtain an expansion word item collection and extension lexical item
Concentrate each expansion word respectively with the initial similarity of corresponding keyword.Since conventional Extension method cannot obtain similarity,
For the expansion word obtained using conventional Extension method, the initial similarity for defining its keyword that is corresponding to it is 1.
Finally, it by merging the expansion word and its initial similarity that three kinds of extended methods obtain, obtains finally extending lexical item
With similarity collection.Specifically, for an expansion word, above-mentioned three kinds of extended methods can may obtain the expansion word, it is also possible to
Two of which extended method, which can obtain the expansion word or only one of which extended method, can obtain the expansion word, at this point, will
Identical expansion word merges into an expansion word, will be at the beginning of corresponding three or two or one for the expansion word after merging
Beginning similarity is added, and the sum of will add up divided by 3 after final similarity of the result as the expansion word and corresponding keyword.
To sum up, the specific expansion word for obtaining the keyword in the following manner of step 101 and each expansion word and institute
The final similarity of keyword is stated, including:
Obtain at least one expansion word of each keyword respectively using Hownet HowNet, and define each expansion word with it is right
It is 1 to answer the initial similarity of keyword;Obtain at least one expansion word of each keyword respectively using Chinese thesaurus, and
It is 1 that each expansion word, which is defined, with the initial similarity of corresponding keyword;Utilize the text depth representing model after trained
Word2vec obtains at least one expansion word of each keyword and the initial phase of each expansion word and corresponding keyword respectively
Like degree;Merge the identical expansion word got, each expansion word calculated separately after merging is final similar with corresponding keyword
Spend SR, wherein SR=Ssum/ 3, SsumFor the sum of corresponding all initial similarities of the expansion word.
In order to more easily understand the concrete methods of realizing of step 101, it is exemplified below:
Assuming that the keyword extracted from question sentence includes " screen " word.
For keyword " screen ", HowNet, Chinese thesaurus and trained word2vec is used to carry out respectively same
Adopted word extension, respectively obtains the extension lexical item and its similarity of respective method.
Assuming that including using the HowNet expansion words being extended to " screen ":
Expansion word | Define similarity |
Screen | 1 |
Display screen | 1 |
Display | 1 |
Assuming that including using the expansion word that Chinese thesaurus is extended " screen ":
Expansion word | Define similarity |
Screen | 1 |
Display screen | 1 |
Assuming that the expansion word and its similarity that are extended to " screen " using the word2vec models after trained
Including:
Merge expansion word " screen ", the final similarity with " screen " is (1+1+0.788869)/3;
Merge expansion word " display screen ", the final similarity with " screen " is (1+1+0.775589)/3;
Merge expansion word " display ", the final similarity with " screen " is (1+0.654054)/3;
Merge expansion word " touch screen ", the final similarity with " screen " is 0.649287/3.
For other keywords in question sentence, expansion word and expansion word and corresponding key are obtained also according to aforesaid way
The final similarity of word.
Since the prior art is mainly using HowNet extensions and Chinese thesaurus extension, but obtained using existing way
Expansion word is not often inconsistent with the keywords semantics in question sentence, and the embodiment of the present invention merges word2vec synonym extended modes
Into existing synonym extended method, solve the problems, such as that expansion word is not inconsistent with the corresponding keyword meaning of a word.
Step 102:Analyze the dependence between each two lexical item with grammar association in the question sentence.
With reference to question sentence retrieving block diagram shown in Fig. 3, understand step 102 to step 107.
Step 102 for ease of understanding now illustrates, it is assumed that question sentence is:" how iphone5s passes through the side such as missing mode
Formula getting back mobile phone ", the result that grammatical relation analysis is carried out to the question sentence are:
Dependence schematic diagram shown in Figure 4, above-mentioned question sentence by " iphone5s ", " how ", " passing through ", " lose
Lose ", " pattern ", " etc. ", " mode ", " giving for change ", " mobile phone " these lexical items composition, wherein root be directed toward lexical item " giving for change " be
The starting point of syntactic analysis, the dependence between lexical item are respectively:
1, it is nominal subject relationship (nsubj) " to give " dependence between " iphone5s " for change;
2, " give for change " and " how " between dependence be adverbial modified relationship (advmod);
3, it is premodification relationship (prep) " to give " dependence between " loss " for change;
4, it is direct object relationship (dobj) " to give " dependence between " mobile phone " for change;
5, the dependence between " loss " and " passing through " is example relationship (case);
6, the dependence between " loss " and " pattern " is direct object relationship (dobj);
7, " pattern " and " etc. " between dependence be etc. relationship (etc);
8, the dependence between " pattern " and " mode " is direct object relationship (dobj).
Step 103:According to being in advance the importance weight of dependence setting, determines and reflect every two in the question sentence
The distance weighting of tightness degree between a lexical item.
In embodiments of the present invention, step 103 is realized as steps described below:
Before executing each step of the embodiment of the present invention, according to the significance level of different dependences, examined in community's question and answer
It is provided with different importance weights for each dependence in advance in cable system, such as can be to " subject relationship " and " object
Larger weight is arranged in relationship ", because this two kinds of relationships occupy major part in sentence structure and ingredient, such as can be arranged
The importance weight of " subject relationship " is 5, and the importance weight of " object relationship " is 4, and the important of other dependences is arranged
Spend weight.Wherein, the importance weight of dependence of the setting without importance is 1.
After obtaining the dependence of question sentence to the question sentence progress interdependent syntactic analysis of question sentence by step 102, next, root
According to the importance weight being arranged for dependence, the different distance weight of different dependences in question sentence, the distance power are calculated
The tightness degree of dependence between lexical item is reflected again, and distance weighting is smaller, then tightness degree is higher, conversely, distance weighting is got over
Greatly, then tightness degree is lower.
Such as:If the importance weight of subject relationship " subj " is " 5 ", the distance for two lexical items which is related to is weighed
Weight isIf the importance weight of direct object relationship " dobj " be " 4 ", two lexical items which is related to away from
It is from weightWherein, it is worth on the basis of α, α can be obtained by parameter regulation.Difference can be obtained in this manner
The corresponding different distance weight of dependence.
Next, between ignoring lexical item it is interdependent with by interdependent points relationship, can be according to the every of dependence
The distance between two lexical items weight D structure weighted-graph G=(V, E, W), in this manner, between any two lexical item
It is connection.
Wherein, n lexical item v in question sentencei, composition vertex set V={ vi| i=1 ..., n };
Any two lexical item (the lexical item v of specific dependenceiWith lexical item vj) be connected the side to be formed, composition side collection E=<
vi,vj>i|i≠j,1≤i≤n,1≤j≤n};
Any two lexical item (the lexical item v of specific dependenceiWith lexical item vj) corresponding distance weighting D, composition weights collection W
=<vi,vj>|vi∈V,vj∈V,<vi,vj>∈E}。
Based on weighted-graph, the distance weighting Dis of each two lexical item in question sentence can be calculated in the following manner, below
It illustrates:
Such as:For question sentence " how iphone5s passes through the modes such as missing mode getting back mobile phone ", need to calculate each word
8 distance weightings of (such as " iphone5s ") and other 8 lexical items, totally 64 distance weightings.Due to being based on importance weight
Can obtain a lexical item (being defined as the first lexical item) and with its another lexical item (defining the second lexical item) with dependence
The distance between weight D, and then the first lexical item and each word in addition to the first lexical item can be further calculated based on distance weighting D
The distance between item (defining third lexical item) weight Dis.
Such as:For question sentence " how iphone5s passes through the modes such as missing mode getting back mobile phone ", it is assumed that first word
Be " giving for change " when, respectively from the second lexical item " iphone5s ", " how ", " loss ", " mobile phone " there are different interdependent passes
System, according to formulaFour distance weighting D of " giving for change " and this four the second lexical items can be calculated, for example, when meter
When calculating " giving for change " and the distance weighting of " iphone5s ", y is " giving for change " and nominal subject relationship possessed by " iphone5s "
Importance weight, other three distance weighting D are equally according to said method calculated.
Based on this four distance weighting D, the distance between " giving for change " and other 8 lexical items weight Dis are respectively:
1, " give for change " with " iphone5s ", " how ", " loss ", the distance between " mobile phone " weight Dis be exactly that its is interdependent
The distance weighting D of relationship;The distance between " giving for change " and " passing through " weight Dis=A+B, A is between " giving for change " and " loss "
Distance weighting D, B are the distance between " loss " and " passing through " weight D;
2, the distance between " pattern " weight Dis=A+C, A " is given " for change to weigh for the distance between " giving for change " and " loss "
Weight D, C are the distance between " loss " and " pattern " weight D;
3, " give for change " with " etc. " the distance between weight Dis=A+C+E, A be the distance between " giving for change " and " loss " power
Weight D, C are the distance between " losss " and " pattern " weight D, E for " pattern " and " etc. " the distance between weight D;
4, it is the distance between " giving for change " and " loss " " to give " the distance between " mode " weight Dis=A+C+F, A for change
Weight D, C are the distance between " loss " and " pattern " weight D, and F is the distance between " pattern " and " mode " weight D.
When the first lexical item is other lexical items (such as " how "), also according to the above method calculate the first lexical item with it is other
The distance between each lexical item weight Dis, details are not described herein.
To sum up, the specific step 103 in the following manner:
First, the distance between the first lexical item and each second lexical item weight D are calculated separately, first lexical item is described
Any one lexical item in question sentence, second lexical item are that there are the lexical items of the dependence with first lexical item;Wherein,Y be in advance between first lexical item and second lexical item dependence setting importance weight, α
On the basis of be worth;Then, the distance between first lexical item and each third lexical item weight Dis, the third word are calculated separately
Item is any one lexical item in addition to first word in the question sentence, and Dis is first lexical item and the third lexical item
Between the sum of the corresponding at least one distance weighting D of existing at least one dependence.
Step 104:The degree of association in the question sentence between each two lexical item is determined according to the distance weighting.
By the corresponding distance weighting Dis of the calculated each two lexical item of step 103, it is n × n's that can build a dimension
Incidence matrix M, due to having ignored the interdependent direction between lexical item, matrix M is symmetrical matrix, the element that the i-th row jth arranges in M
MijIndicate lexical item t in question sentenceiWith lexical item tjThe degree of association.
Lexical item t in the question sentence is calculated according to the following equationiWith lexical item tjBetween degree of association wrel(i,j):
wrel(i,j)=λ Dep (ti,tj)+(1-λ)Closepmi(ti,tj);
tiIndicate i-th of lexical item in the question sentence, tjIndicate j-th of lexical item in the question sentence, i=1,2 ... n, j
=1,2 ... n, n are the lexical item sum in the question sentence;
λ is regulatory factor;
B is a constant more than 1;
Dis(ti,tj) it is lexical item tiWith lexical item tjThe distance between weight;
Lexical item t is concentrated for question sentenceiWith lexical item tjThe probability occurred jointly, Nd(ti,tj) it is question sentence
Concentrate lexical item tiWith lexical item tjThe number of the question sentence occurred simultaneously, NDThe question sentence sum concentrated for question sentence;
WithLexical item t is indicated respectivelyiWith lexical item tjEach leisure question sentence concentrates appearance
Probability, Nd(ti) it is that question sentence is concentrated comprising lexical item tiQuestion sentence sum, Nd(tj) it is that question sentence is concentrated comprising lexical item tjQuestion sentence it is total
Number, NDThe question sentence sum concentrated for question sentence.
Step 105:The lexical item weight of each lexical item in the question sentence is determined according to the degree of association, and according to the question sentence
In each lexical item the retrieval of lexical item weight and the relevant question and answer pair of the question sentence.
In embodiments of the present invention, the specific lexical item weight for determining each lexical item in question sentence in the following manner of step 105,
The weight matrix of each lexical item final weight composition in the question sentence is calculated according to the following equation
Wherein, α is given constant;E is that incidence matrix M is carried out the random matrix after orthogonal transformation, the incidence matrix M
The symmetrical matrix that the degree of association between each two lexical item in the question sentence is formed;For each original power of lexical item in the question sentence
The weight matrix reassembled into.
Wherein, original weightIt is to be determined by the retrieval model of community's question and answer searching system, the retrieval model
For:Vector space model (Vector space model, VSM) or probabilistic model okapi BM25 or language model
(Language Model, LM) or other models.The retrieval model includes but not limited to above-mentioned model.
It should be noted that incidence matrix M is converted to obtain by random matrix E using orthogonal transformation, to ensure the pass
The corresponding equation groups of connection matrix M certainly exist analytic solutions, i.e. E=D-1M(D-1For an orthogonal matrix).For every in matrix E
One element, value all [0,1) between, and the sum of each row element of E be equal to 1.Therefore, E must there are one characteristic values etc.
It is the corresponding solution of equations vectors of E in 1, and equal to the feature vector corresponding to 1 this characteristic value, thus E is corresponded to
Equation group certainly exist analytic solutions.Incidence matrix is converted into after a random matrix E for M gusts, so that it may to be somebody's turn to do by solving
The corresponding solution of equations of matrix analyses solution to obtain lexical item weight to be calculatedBut the weight and lexical item which obtains
Original valueIt is not related, and original weightTo be obtained by original retrieval model, as VSM, okapi BM25,
LM models etc. utilizeAnd more accurate lexical item weight can be calculated in incidence matrix M
Step 106:It is weighed with the lexical item of expansion word described in the final similarity calculation of corresponding keyword according to the expansion word
Weight, and according to the retrieval of the lexical item weight of the expansion word and the relevant question and answer pair of the question sentence.
In embodiments of the present invention, for each expansion word, first, the retrieval model by community's question and answer searching system is true
Then the original weight of fixed keyword corresponding with the expansion word calculates extension lexical item weight:Extend lexical item weight=key word item
The similarity of original weight * extensions lexical item and key word item;Then, retrieval model obtains retrieval knot using extension lexical item weight
Fruit, the retrieval result may include not only question and answer to that can also include the marking value to question and answer pair.
Wherein, the retrieval model is:Vector space model (Vector space model, VSM) or probabilistic model
Okapi BM25 or language model (Language Model, LM) or other models.The retrieval model includes but not limited to
Above-mentioned model.
To sum up, the specific lexical item weight for calculating the expansion word in the following manner of step 106:
Obtain the original weight of the corresponding keyword of the expansion word;By the original weight and the expansion word with it is corresponding
The product of the final similarity of keyword, the lexical item weight as the expansion word.
Step 107:By all question and answer that step 105 and step 106 retrieve to being ranked up display according to preset rules.
Step 105 and step 106 obtain question and answer to while, can also further obtain the marking to each question and answer pair
Value, the embodiment of the present invention can show the question and answer that step 105 and step 106 obtain to being ranked up according to marking value, that is, give a mark
It is worth high question and answer in preceding display, marking is worth low question and answer in rear display.
It should be noted that step 101 and step 106 can be executed in any position before step 107, step 101
In preceding step 106 after, for example step 106 is moved to after step 101 and is executed, or by step 101 move to step 106 it
Preceding execution etc. is not limited this present invention.
Question and answer provided in an embodiment of the present invention are important by being arranged for dependence different in question sentence to search method
Spend weight, it may be determined that the association tight ness rating in question sentence between lexical item can further determine that each lexical item in question sentence according to the degree of association
Lexical item weight, the important lexical item being can be found that in question sentence by merging the obtained lexical item weight of importance weight, to obtain
With the more relevant question and answer of question sentence to retrieval result, solving existing community's question and answer searching system, not account for question sentence structure multiple
Miscellaneous and clause is tediously long and the shortcomings that can not find question sentence important lexical item, and then improve the purpose of retrieval result accuracy.
It is the composition schematic diagram of question and answer searching system in community's provided in an embodiment of the present invention referring to Fig. 5, including:
Keyword extracting unit 501, for extracting at least one keyword from question sentence;
Keyword expansion unit 502, the extension of each keyword for obtaining the acquisition of the keyword extracting unit 501
The final similarity of word and each expansion word and corresponding keyword;
Relationship analysis unit 503, it is interdependent between each two lexical item with grammar association in the question sentence for analyzing
Relationship;
Weights determination unit 504, for being set according to the dependence obtained in advance for the relationship analysis unit 503 analysis
Fixed importance weight determines the distance weighting for reflecting tightness degree between each two lexical item in the question sentence;
Degree of association determination unit 505, distance weighting for being determined according to the weights determination unit 504 determine described in ask
The degree of association in sentence between each two lexical item;
First weight determining unit 506, described in the degree of association determination for being determined according to the degree of association determination unit 505
The lexical item weight of each lexical item in question sentence;
First retrieval unit 507, each lexical item in the question sentence for being determined according to first weight determining unit 506
Lexical item weight is retrieved and the relevant question and answer pair of the question sentence;
Second weight determining unit 508, for according to the keyword expansion unit 501 obtained expansion word of extension with it is right
Answer the lexical item weight of expansion word described in the final similarity calculation of keyword;
Second retrieval unit 509, the lexical item weight of the expansion word for being determined according to second weight determining unit 508
Retrieval and the relevant question and answer pair of the question sentence;
Retrieval result display unit 510, for examining first retrieval unit 507 and second retrieval unit 509
All question and answer that rope goes out according to preset rules to being ranked up display.
In embodiments of the present invention, the keyword expansion unit 502, including:
Hownet expansion module, at least one expansion word for obtaining each keyword respectively using Hownet HowNet, and
It is 1 that each expansion word, which is defined, with the initial similarity of corresponding keyword;
Word woods expansion module, at least one expansion word for obtaining each keyword respectively using Chinese thesaurus, and
It is 1 that each expansion word, which is defined, with the initial similarity of corresponding keyword;
Model extension module obtains each respectively for text depth representing model word2vec of the utilization after trained
The initial similarity of at least one expansion word and each expansion word and corresponding keyword of keyword;
Similarity calculation module expands for merging the Hownet expansion module, institute's predicate woods expansion module and the model
The identical expansion word that exhibition module is got, calculates separately the final similarity of each expansion word and corresponding keyword after merging
SR, wherein SR=Ssum/ 3, SsumFor the sum of corresponding all initial similarities of the expansion word.
In embodiments of the present invention, the weights determination unit 504, including:
First weight computation module, for calculating separately the distance between the first lexical item and each second lexical item weight D, institute
It is any one lexical item in the question sentence to state the first lexical item, and second lexical item is that there are described interdependent with first lexical item
The lexical item of relationship;
Wherein,Y be that dependence between first lexical item and second lexical item is arranged in advance
Importance weight is worth on the basis of α;
Second weight computation module, for calculating separately the distance between first lexical item and each third lexical item weight
Dis, the third lexical item are any one lexical item in addition to first word in the question sentence, and Dis is first lexical item
Corresponding at least one the first weight computation module meter of existing at least one dependence between the third lexical item
The sum of obtained distance weighting D.
In embodiments of the present invention, the degree of association determination unit 505, specifically for calculating described ask according to the following equation
Lexical item t in sentenceiWith lexical item tjBetween degree of association wrel(i,j):
wrel(i,j)=λ Dep (ti,tj)+(1-λ)Closepmi(ti,tj);
tiIndicate i-th of lexical item in the question sentence, tjIndicate j-th of lexical item in the question sentence, i=1,2 ... n, j
=1,2 ... n, n are the lexical item sum in the question sentence;
λ is regulatory factor;
B is a constant more than 1;
Dis(ti,tj) it is lexical item tiWith lexical item tjThe distance between weight;
Lexical item t is concentrated for question sentenceiWith lexical item tjThe probability occurred jointly, Nd(ti,tj) it is question sentence
Concentrate lexical item tiWith lexical item tjThe number of the question sentence occurred simultaneously, NDThe question sentence sum concentrated for question sentence;
WithLexical item t is indicated respectivelyiWith lexical item tjEach leisure question sentence concentrates appearance
Probability, Nd(ti) it is that question sentence is concentrated comprising lexical item tiQuestion sentence sum, Nd(tj) it is that question sentence is concentrated comprising lexical item tjQuestion sentence it is total
Number, NDThe question sentence sum concentrated for question sentence.
In embodiments of the present invention, first weight determining unit 506 is specifically used for described in calculating according to the following equation
The weight matrix of each lexical item final weight composition in question sentence
Wherein, α is given constant;
E is that incidence matrix M is carried out the random matrix after orthogonal transformation, and the incidence matrix M is every two in the question sentence
The symmetrical matrix that the degree of association between a lexical item is formed;
For the weight matrix of the original weight composition of each lexical item in the question sentence.
In embodiments of the present invention, second weight determining unit 508, including:
Original Weight Acquisition module, the original weight for obtaining the corresponding keyword of the expansion word;
Second weight determination module, the original weight for obtaining the original Weight Acquisition module and the expansion word
With the product of the final similarity of corresponding keyword, the lexical item weight as the expansion word.
Question and answer searching system in community's provided in an embodiment of the present invention passes through the weight for dependence setting different in question sentence
Spend weight, it may be determined that the association tight ness rating in question sentence between lexical item can further determine that each word in question sentence according to the degree of association
The lexical item weight of item, the important lexical item being can be found that in question sentence by merging the lexical item weight that importance weight obtains, to
It has arrived with the more relevant question and answer of question sentence to retrieval result, has solved existing community's question and answer searching system and do not account for question sentence structure
Complicated and clause is tediously long and the shortcomings that can not find question sentence important lexical item, and then improve the purpose of retrieval result accuracy.
As seen through the above description of the embodiments, those skilled in the art can be understood that above-mentioned implementation
All or part of step in example method can add the mode of required general hardware platform to realize by software.Based on such
Understand, substantially the part that contributes to existing technology can be in the form of software products in other words for technical scheme of the present invention
It embodies, which can be stored in a storage medium, such as ROM/RAM, magnetic disc, CD, including several
Instruction is used so that a computer equipment (can be the network communications such as personal computer, server, or Media Gateway
Equipment, etc.) execute method described in certain parts of each embodiment of the present invention or embodiment.
It should be noted that for system disclosed in embodiment, since it is corresponded to the methods disclosed in the examples,
So description is fairly simple, reference may be made to the description of the method.
It should also be noted that, herein, relational terms such as first and second and the like are used merely to one
Entity or operation are distinguished with another entity or operation, without necessarily requiring or implying between these entities or operation
There are any actual relationship or orders.Moreover, the terms "include", "comprise" or its any other variant are intended to contain
Lid non-exclusive inclusion, so that the process, method, article or equipment including a series of elements is not only wanted including those
Element, but also include other elements that are not explicitly listed, or further include for this process, method, article or equipment
Intrinsic element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that
There is also other identical elements in process, method, article or equipment including the element.
The foregoing description of the disclosed embodiments enables those skilled in the art to implement or use the present invention.
Various modifications to these embodiments will be apparent to those skilled in the art, as defined herein
General Principle can be realized in other embodiments without departing from the spirit or scope of the present invention.Therefore, of the invention
It is not intended to be limited to the embodiments shown herein, and is to fit to and the principles and novel features disclosed herein phase one
The widest range caused.
Claims (12)
1. a kind of question and answer are to search method, which is characterized in that including:
At least one keyword is extracted from question sentence, and obtain each keyword expansion word and each expansion word and it is corresponding close
The final similarity of keyword;
Analyze the dependence between each two lexical item with grammar association in the question sentence;
According to the importance weight for being in advance dependence setting, determine tight between each two lexical item in the reflection question sentence
The distance weighting of close degree;
The degree of association in the question sentence between each two lexical item is determined according to the distance weighting;
The lexical item weight of each lexical item in the question sentence is determined according to the degree of association, and according to each lexical item in the question sentence
Lexical item weight is retrieved and the relevant question and answer pair of the question sentence;
According to the lexical item weight of the expansion word and expansion word described in the final similarity calculation of corresponding keyword, and according to described
The retrieval of lexical item weight and the relevant question and answer pair of the question sentence of expansion word;
By all question and answer retrieved to being ranked up display according to preset rules.
2. according to the method described in claim 1, it is characterized in that, the expansion word for obtaining each keyword and each expansion
The final similarity of word and corresponding keyword is opened up, including:
It obtains at least one expansion word of each keyword respectively using Hownet HowNet, and defines each expansion word and closed with corresponding
The initial similarity of keyword is 1;
It obtains at least one expansion word of each keyword respectively using Chinese thesaurus, and defines each expansion word and closed with corresponding
The initial similarity of keyword is 1;
Using the text depth representing model word2vec after trained, at least one expansion word of each keyword is obtained respectively
And the initial similarity of each expansion word and corresponding keyword;
Merge the identical expansion word got, calculates separately the final similarity of each expansion word and corresponding keyword after merging
SR, wherein SR=Ssum/ 3, SsumFor the sum of corresponding all initial similarities of the expansion word.
3. according to the method described in claim 1, it is characterized in that, the basis is the important of dependence setting in advance
Weight is spent, determines the distance weighting for reflecting tightness degree between each two lexical item in the question sentence, including:
The distance between the first lexical item and each second lexical item weight D are calculated separately, first lexical item is in the question sentence
Any one lexical item, second lexical item are that there are the lexical items of the dependence with first lexical item;
Wherein,Y be in advance between first lexical item and second lexical item dependence setting it is important
Weight is spent, is worth on the basis of α;
The distance between first lexical item and each third lexical item weight Dis are calculated separately, the third lexical item is described asks
Any one lexical item in sentence in addition to first lexical item, Dis are existing between first lexical item and the third lexical item
The sum of corresponding at least one distance weighting D of at least one dependence.
4. according to the method described in claim 3, it is characterized in that, described determine in the question sentence often according to the distance weighting
The degree of association between two lexical items, including:
Lexical item t in the question sentence is calculated according to the following equationiWith lexical item tjBetween degree of association wrel(i,j):
wrel(i,j)=λ Dep (ti,tj)+(1-λ)Closepmi(ti,tj);
Wherein,
tiIndicate i-th of lexical item in the question sentence, tjIndicate j-th of lexical item in the question sentence, i=1,2 ... n, j=1,
2 ... n, n are the lexical item sum in the question sentence;
λ is regulatory factor;
B is a constant more than 1;
Dis(ti,tj) it is lexical item tiWith lexical item tjThe distance between weight;
Lexical item t is concentrated for question sentenceiWith lexical item tjThe probability occurred jointly, Nd(ti,tj) it is that question sentence is concentrated
Lexical item tiWith lexical item tjThe number of the question sentence occurred simultaneously, NDThe question sentence sum concentrated for question sentence;
WithLexical item t is indicated respectivelyiWith lexical item tjEach leisure question sentence concentrates the probability occurred, Nd
(ti) it is that question sentence is concentrated comprising lexical item tiQuestion sentence sum, Nd(tj) it is that question sentence is concentrated comprising lexical item tjQuestion sentence sum, NDFor
The question sentence sum that question sentence is concentrated.
5. method according to any one of claims 1 to 4, which is characterized in that described according to described in degree of association determination
The lexical item weight of each lexical item in question sentence, including:
The weight matrix W of each lexical item final weight composition in the question sentence is calculated according to the following equationq *:
Wherein, α is given constant;
E is that incidence matrix M is carried out the random matrix after orthogonal transformation, and the incidence matrix M is each two word in the question sentence
The symmetrical matrix that the degree of association between is formed;
Wq 0For the weight matrix of the original weight composition of each lexical item in the question sentence.
6. according to the method described in claim 1, it is characterized in that, described according to the final of the expansion word and corresponding keyword
The lexical item weight of expansion word described in similarity calculation, including:
Obtain the original weight of the corresponding keyword of the expansion word;
By the product of the original weight and the expansion word and the final similarity of corresponding keyword, as the expansion word
Lexical item weight.
7. a kind of community's question and answer searching system, which is characterized in that including:
Keyword extracting unit, for extracting at least one keyword from question sentence;
Keyword expansion unit, expansion word for obtaining each keyword that the keyword extracting unit obtains and each
The final similarity of expansion word and corresponding keyword;
Relationship analysis unit, for analyzing the dependence in the question sentence between each two lexical item with grammar association;
Weights determination unit, for according to the importance for being in advance the dependence setting that the relationship analysis unit is analyzed
Weight determines the distance weighting for reflecting tightness degree between each two lexical item in the question sentence;
Degree of association determination unit, the distance weighting for being determined according to the weights determination unit determine each two in the question sentence
The degree of association between lexical item;
First weight determining unit, the degree of association for being determined according to the degree of association determination unit determine each in the question sentence
The lexical item weight of lexical item;
First retrieval unit, the lexical item weight inspection of each lexical item in the question sentence for being determined according to first weight determining unit
Rope and the relevant question and answer pair of the question sentence;
Second weight determining unit, the expansion word for being obtained according to the keyword expansion unit extensions and corresponding keyword
The lexical item weight of expansion word described in final similarity calculation;
Second retrieval unit, the lexical item weight of the expansion word for being determined according to second weight determining unit retrieve with it is described
The relevant question and answer pair of question sentence;
Retrieval result display unit, all question and answer for retrieving first retrieval unit and second retrieval unit
To being ranked up display according to preset rules.
8. system according to claim 7, which is characterized in that the keyword expansion unit, including:
Hownet expansion module, at least one expansion word for obtaining each keyword respectively using Hownet HowNet, and define
Each expansion word is 1 with the initial similarity of corresponding keyword;
Word woods expansion module, at least one expansion word for obtaining each keyword respectively using Chinese thesaurus, and define
Each expansion word is 1 with the initial similarity of corresponding keyword;
Model extension module, it is each crucial for using the text depth representing model word2vec after trained, obtaining respectively
The initial similarity of at least one expansion word and each expansion word and corresponding keyword of word;
Similarity calculation module, for merging the Hownet expansion module, institute's predicate woods expansion module and the model extension mould
The identical expansion word that block is got calculates separately the final similarity S of each expansion word and corresponding keyword after mergingR,
In, SR=Ssum/ 3, SsumFor the sum of corresponding all initial similarities of the expansion word.
9. system according to claim 7, which is characterized in that the weights determination unit, including:
First weight computation module, for calculating separately the distance between the first lexical item and each second lexical item weight D, described
One lexical item is any one lexical item in the question sentence, and second lexical item is that there are the dependences with first lexical item
Lexical item;
Wherein,Y be in advance between first lexical item and second lexical item dependence setting it is important
Weight is spent, is worth on the basis of α;
Second weight computation module, for calculating separately the distance between first lexical item and each third lexical item weight Dis,
The third lexical item is any one lexical item in addition to first lexical item in the question sentence, and Dis is first lexical item and institute
Corresponding at least one first weight computation module of existing at least one dependence between third lexical item is stated to calculate
The sum of distance weighting D arrived.
10. system according to claim 9, which is characterized in that the degree of association determination unit is specifically used for according to following
Formula calculates lexical item t in the question sentenceiWith lexical item tjBetween degree of association wrel(i,j):
wrel(i,j)=λ Dep (ti,tj)+(1-λ)Closepmi(ti,tj);
Wherein,
tiIndicate i-th of lexical item in the question sentence, tjIndicate j-th of lexical item in the question sentence, i=1,2 ... n, j=1,
2 ... n, n are the lexical item sum in the question sentence;
λ is regulatory factor;
B is a constant more than 1;
Dis(ti,tj) it is lexical item tiWith lexical item tjThe distance between weight;
Lexical item t is concentrated for question sentenceiWith lexical item tjThe probability occurred jointly, Nd(ti,tj) it is that question sentence is concentrated
Lexical item tiWith lexical item tjThe number of the question sentence occurred simultaneously, NDThe question sentence sum concentrated for question sentence;
WithLexical item t is indicated respectivelyiWith lexical item tjEach leisure question sentence concentrates the probability occurred, Nd
(ti) it is that question sentence is concentrated comprising lexical item tiQuestion sentence sum, Nd(tj) it is that question sentence is concentrated comprising lexical item tjQuestion sentence sum, NDFor
The question sentence sum that question sentence is concentrated.
11. according to claim 7 to 10 any one of them system, which is characterized in that first weight determining unit, specifically
Weight matrix W for calculating each lexical item final weight composition in the question sentence according to the following equationq *:
Wherein, α is given constant;
E is that incidence matrix M is carried out the random matrix after orthogonal transformation, and the incidence matrix M is each two word in the question sentence
The symmetrical matrix that the degree of association between is formed;
Wq 0For the weight matrix of the original weight composition of each lexical item in the question sentence.
12. system according to claim 7, which is characterized in that second weight determining unit, including:
Original Weight Acquisition module, the original weight for obtaining the corresponding keyword of the expansion word;
Second weight determination module, original weight and the expansion word for obtaining the original Weight Acquisition module with it is right
The product for answering the final similarity of keyword, the lexical item weight as the expansion word.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610082304.6A CN105786794B (en) | 2016-02-05 | 2016-02-05 | A kind of question and answer are to search method and community's question and answer searching system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610082304.6A CN105786794B (en) | 2016-02-05 | 2016-02-05 | A kind of question and answer are to search method and community's question and answer searching system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105786794A CN105786794A (en) | 2016-07-20 |
CN105786794B true CN105786794B (en) | 2018-09-04 |
Family
ID=56403313
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610082304.6A Active CN105786794B (en) | 2016-02-05 | 2016-02-05 | A kind of question and answer are to search method and community's question and answer searching system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105786794B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106294656B (en) * | 2016-08-04 | 2019-03-19 | 武汉大学 | A kind of method of map locating keyword to relevant issues |
CN110019701B (en) * | 2017-09-18 | 2021-12-31 | 京东方科技集团股份有限公司 | Method for question answering service, question answering service system and storage medium |
CN108932326B (en) * | 2018-06-29 | 2021-02-19 | 北京百度网讯科技有限公司 | Instance extension method, device, equipment and medium |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2007099812A1 (en) * | 2006-03-01 | 2007-09-07 | Nec Corporation | Question answering device, question answering method, and question answering program |
US8543565B2 (en) * | 2007-09-07 | 2013-09-24 | At&T Intellectual Property Ii, L.P. | System and method using a discriminative learning approach for question answering |
CN103440287A (en) * | 2013-08-14 | 2013-12-11 | 广东工业大学 | Web question-answering retrieval system based on product information structuring |
CN104573028A (en) * | 2015-01-14 | 2015-04-29 | 百度在线网络技术(北京)有限公司 | Intelligent question-answer implementing method and system |
CN104899188A (en) * | 2015-03-11 | 2015-09-09 | 浙江大学 | Problem similarity calculation method based on subjects and focuses of problems |
WO2016027714A1 (en) * | 2014-08-21 | 2016-02-25 | 国立研究開発法人情報通信研究機構 | Question sentence generation device and computer program |
-
2016
- 2016-02-05 CN CN201610082304.6A patent/CN105786794B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2007099812A1 (en) * | 2006-03-01 | 2007-09-07 | Nec Corporation | Question answering device, question answering method, and question answering program |
US8543565B2 (en) * | 2007-09-07 | 2013-09-24 | At&T Intellectual Property Ii, L.P. | System and method using a discriminative learning approach for question answering |
CN103440287A (en) * | 2013-08-14 | 2013-12-11 | 广东工业大学 | Web question-answering retrieval system based on product information structuring |
WO2016027714A1 (en) * | 2014-08-21 | 2016-02-25 | 国立研究開発法人情報通信研究機構 | Question sentence generation device and computer program |
CN104573028A (en) * | 2015-01-14 | 2015-04-29 | 百度在线网络技术(北京)有限公司 | Intelligent question-answer implementing method and system |
CN104899188A (en) * | 2015-03-11 | 2015-09-09 | 浙江大学 | Problem similarity calculation method based on subjects and focuses of problems |
Also Published As
Publication number | Publication date |
---|---|
CN105786794A (en) | 2016-07-20 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110442777B (en) | BERT-based pseudo-correlation feedback model information retrieval method and system | |
CN111475729B (en) | Search content recommendation method and device | |
CN103136352B (en) | Text retrieval system based on double-deck semantic analysis | |
US9710547B2 (en) | Natural language semantic search system and method using weighted global semantic representations | |
CN110188168A (en) | Semantic relation recognition methods and device | |
US20110078205A1 (en) | Method and system for finding appropriate semantic web ontology terms from words | |
CN110020189A (en) | A kind of article recommended method based on Chinese Similarity measures | |
CN109783806B (en) | Text matching method utilizing semantic parsing structure | |
CN102253982A (en) | Query suggestion method based on query semantics and click-through data | |
CN104462399B (en) | The processing method and processing device of search result | |
US20190266286A1 (en) | Method and system for a semantic search engine using an underlying knowledge base | |
WO2020060718A1 (en) | Intelligent search platforms | |
CN111831786A (en) | Full-text database accurate and efficient retrieval method for perfecting subject term | |
CN108416008A (en) | A kind of BIM product database semantic retrieving methods based on natural language processing | |
CN105786794B (en) | A kind of question and answer are to search method and community's question and answer searching system | |
CN101571852A (en) | Dictionary generating device and information retrieving device | |
CN112036178A (en) | Distribution network entity related semantic search method | |
Raman et al. | Performance comparison of various information retrieval models used in search engines | |
US20190012388A1 (en) | Method and system for a semantic search engine using an underlying knowledge base | |
Juan | An effective similarity measurement for FAQ question answering system | |
CN111737413A (en) | Feedback model information retrieval method, system and medium based on concept net semantics | |
Bunescu et al. | A utility-driven approach to question ranking in social QA | |
Yang et al. | A novel ontology-based semantic retrieval model for food safety domain | |
Foley et al. | Integrating wordnet for multiple sense embeddings in vector semantics | |
KR102363131B1 (en) | Multi-dimensional knowledge searching method and system for expert systems |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |