CN101593206A - Searching method and device based on answer in the question and answer interaction platform - Google Patents

Abstract

Embodiment of the present invention provides a kind of searching method and device based on answer in the question and answer interaction platform, this method and device belong to network communication field, and this method comprises: the pairing expansion word of keyword that finds out described problem according to the keyword of described problem from the keyword expansion tabulation of storage; Calculate the number of times of key in the answer that each searches out, and calculate the number of times of the pairing expansion word of described keyword in the answer that each searches out; Confirm the priority of each answer according to result calculated, after the order of determining this answer according to this priority, export.The present invention also provides a kind of searcher based on answer in the question and answer interaction platform, and this method and device have answer accuracy height, the measured advantage of matter.

Description

Searching method and device based on answer in the question and answer interaction platform

Technical field

The present invention relates to network communication field, relate in particular to a kind of searching method and device based on answer in the question and answer interaction platform.

Background technology

Development along with the Internet (internet), increasing user selects by the internet Query Information, in order to allow the user can easily inquire the information of wanting, prior art provides a kind of question and answer interaction platform technology, the question and answer interaction platform is as a kind of open field question answering system, the information that can provide it to need to the user fast, simultaneously, because the complex nature of the problem, and the otherness between the answer, cause in this platform information quality uneven, make the user can't find its needed information, in order to address the above problem, prior art provides a kind of searching method based on answer in the question and answer interaction platform again, and this method is once searched for the answer that is tentatively met this problem earlier according to this problem after receiving the problem that needs to answer, then the answer that tentatively meets this problem is carried out the calculating of correlativity, and offered the user after according to the value of correlativity this answer that tentatively meets this problem being sorted.The above-mentioned concrete grammar that carries out the calculating of correlativity is, this problem is carried out finding out after the word segmentation processing keyword of this problem, then word segmentation processing is carried out in the answer that tentatively meets this problem, the keyword that will carry out answer after the word segmentation processing and this problem is compared and is obtained the value of correlativity, keyword as this problem is one, the value that then occurs a correlativity in answer is 1, as secondary occurs, and then the value of correlativity is 2.

In realizing process of the present invention, there are the following problems for the technical scheme that prior art provided:

The value of the correlativity in the scheme of prior art is only relevant with the number of times that keyword occurs, and does not have other correlation computations, and just the same as answer and problem so its accuracy is not high, then this relevance values is very high, but answer does not provide effective information.

Summary of the invention

Embodiment of the present invention provides a kind of searching method and device based on answer in the question and answer interaction platform, and this device and method has answer accuracy height, the measured advantage of matter.

The specific embodiment of the present invention provides a kind of searching method based on answer in the question and answer interaction platform, and this method comprises:

The problem of receive to need answering, and after described problem carried out word segmentation processing, extract the keyword of described problem;

From the keyword expansion tabulation of storage, find out the pairing expansion word of keyword of described problem according to the keyword of described problem; Calculate the number of times of keyword in the answer that each searches out of described problem, and calculate the number of times of the pairing expansion word of described keyword in the answer that each searches out;

Confirm the priority of each answer according to result calculated, after the order of determining this answer according to this priority, export.

The specific embodiment of the invention also provides a kind of searcher based on answer in the question and answer interaction platform, and this device comprises:

Receiving element is used to receive the problem that needs are answered;

The word segmentation processing unit after being used for described problem carried out word segmentation processing, extracts the keyword of described problem;

Query unit is used for finding out from the keyword expansion tabulation of storage according to the keyword of described problem the pairing expansion word of keyword of described problem;

Statistic unit is used for calculating the number of times of the keyword of described problem in each answer that searches out, and calculates the number of times of the pairing expansion word of described keyword in the answer that each searches out;

The priority output unit is used for confirming according to result calculated the priority of each answer, exports after the order of determining this answer according to this priority.

By the above-mentioned technical scheme that provides as can be seen, the technical scheme of the embodiment of the invention is because to the equal calculation times of the expansion word of keyword and keyword, and the number of times of the expansion word of the number of times of the keyword that calculates and keyword combined the priority of judging answer, and answer is exported, thereby has the high advantage of accuracy according to the order of priority.And this method can also be judged the priority of answer by the result of calculation of vector, thereby has further improved accuracy.

Description of drawings

A kind of process flow diagram that Fig. 1 provides for the specific embodiment of the invention based on the searching method of answer in the question and answer interaction platform;

Fig. 2 is L of the present invention ₁Norm (W, W _e) value and L ₁(W, W _e) graph of relation;

Fig. 3 is IRad of the present invention (W, W _e) value and A (W, W _e) graph of relation;

Fig. 4 is Cos of the present invention (W, W _e) and vector (W, W _e) graph of relation;

Fig. 5 is REL of the present invention (W, W _e) and vector (W, W _e) graph of relation;

A kind of structural drawing that Fig. 6 provides for the specific embodiment of the invention based on the searcher of answer in the question and answer interaction platform.

Embodiment

Embodiment of the present invention provides a kind of searching method based on answer in the question and answer interaction platform, and this method comprises the steps: as shown in Figure 1

Step 11, the problem of receive to need answering, and after this problem carried out word segmentation processing, extract the keyword of this problem;

Wherein, extract in this step this problem keyword concrete operations can for, after this problem carried out word segmentation processing, remove the stop words in this answer, this stop words is the speech that does not have practical significance, as insignificant function word, for example:, as,, etc., can certainly be other speech, for example: be, etc.Other remaining speech are the keyword of this problem.Certainly also can set up the stop words tabulation according to actual conditions in actual conditions, the user is provided with the scope of stop words as required voluntarily.

Step 12, from the keyword expansion tabulation of storage, find out the pairing expansion word of keyword of this problem according to the keyword of this problem;

The method for building up of the keyword expansion tabulation in this step can adopt any one in the following manner, following method can certainly be combined.

Mode A, user set up the keyword expansion tabulation according to the needs of oneself;

Illustrate with a real example below, as keyword is rogue software, then the user can be provided with the expansion word of rogue software correspondence and is, computer, network, program, technology, company etc., certainly the user also can be provided with other expansion word, as antivirus software, advertisement etc., the present invention only describes as example, is not limited to above-mentioned explanation in actual applications.

Mode B, all data that will store are carried out the keyword that word segmentation processing obtains each data, calculate the level and smooth tolerance between any two keywords in the keyword of each data, exceed setting value as the level and smooth tolerance of this calculating and then should be set at keyword and expansion word by any two keywords, and store in the keyword expansion tabulation.

Optionally, in order to reduce calculated amount, the user can limit the scope of all data of storage, as temporal qualification, in 1 year, in one month, also can be the qualification on the region, as the answer of regional, the answer of United States region, also can answer as English for answering the qualification of language, Chinese is answered.Certainly in actual conditions, in order to reduce calculated amount, can also set, when surpassing set point number, just calculate level and smooth tolerance as the number of times that in all data, occurs when keyword to the keyword of this answer.

Realize among this mode B calculating the smoothness amount specific algorithm can for, the probability calculation that occurs in cliction up and down according to the expansion word of this keyword and this keyword goes out level and smooth tolerance.Concrete grammar can for:

$\mathrm{REL}(W,W_e)=\mathrm{\&alpha;}\&times;\mathrm{IRad}(W,W_e)+\mathrm{\&beta;}\&times;\sqrt{\mathrm{Cos}(W,W_e)\&times;{\mathrm{<figure-callout\; id="1"\; label="L"\; filenames="A2009100879590013H2.png,A2009100879590014H1.png"\; state="\{\{state\}\}">L</figure-callout>}}_1\mathrm{norm}(W,W_e)}---\left(1\right)$

Wherein, α, β are weighting factor, and the user can adjust voluntarily.

L ₁norm(W，W _e)＝(2-L ₁(W，W _e)) ^0.5 (2)

$L_1(W,W_e)=\underset{y\&Element;Y}{\mathrm{\&Sigma;}}|W\left(y\right)-W_e\left(y\right)|---\left(3\right)$

Wherein, L ₁Norm (W, W _e) value and L ₁(W, W _e) graph of a relation as shown in Figure 2, as shown in Figure 2, if L ₁(W, W _e) be 0.5 when choosing expansion word, the speech that obtains is many, but accuracy is relatively low.

$\mathrm{IRad}(W,W_e)={10}^{-0.6A(W,W_e)}---\left(4\right)$

Wherein, measure IRad (W, W as a result _e) value and A (W, W _e) relation curve as shown in Figure 3, as shown in Figure 3, as A (W, W _e) be 0.5 when choosing expansion word, the speech that obtains is fewer, but accuracy is than higher.

$A(W,W_e)=2\log 2+\underset{y\&Element;Y}{\mathrm{\&Sigma;}}\{W\left(y\right)\log \frac{W\left(y\right)}{W\left(y\right)+W_e\left(y\right)}+W_e\left(y\right)\log \frac{W_e\left(y\right)}{W\left(y\right)+W_e\left(y\right)}\}---\left(5\right)$

$\mathrm{Cos}(W,W_e)=\frac{\underset{y\&Element;Y}{\mathrm{\&Sigma;}}{F}_{W,y}{F}_{w_e,y}}{\sqrt{\underset{y\&Element;X}{\mathrm{\&Sigma;}}{\left(F_{W,y}\right)}^2}\sqrt{\underset{y\&Element;X_e}{\mathrm{\&Sigma;}}{\left(F_{W_e,y}\right)}^2}}---\left(6\right)$

Cos (W, W _e) and vector (W, W _e) graph of relation as shown in Figure 4, as shown in Figure 4, as (W, W _e) be 0.5 when choosing expansion word, the speech that obtains is many, but accuracy is relatively low.

Wherein, W can represent any one keyword in all data, W _eCan represent any one keyword except that W in all data; W (y) can be the y of cliction up and down of W all probability that occur in the cliction up and down at W; The y of cliction up and down of We (y) is at W _eAll probability that occur in the cliction up and down; Wherein all of W up and down cliction can for, if W occurs in five documents, then its all cliction up and down can be the set of the up and down cliction of W in five documents; F _{W, y}Can be the frequency of speech W appearance in the specific data set,

Can be speech W _eThe frequency that occurs in the specific data set, X is the set of words of speech W co-occurrence, X _eCan be speech W and expansion word W _eCo-occurrence word set.

Wherein the implication of cliction belongs to existing proven technique up and down.

Optionally, α, β are preferred 0.65,0.35, when α, β preferred 0.65,0.35 the time, REL (W, W _e) and vector (W, W _e) graph of a relation as shown in Figure 5, curve as shown in Figure 5 is smoother, can guarantee to get access to an amount of expansion word under the accuracy rate condition, thereby improve the accuracy of search, and improve the quality of answer.

Step 13, calculate the number of times of keyword in the answer that each searches out of this problem, and calculate the number of times of the pairing expansion word of this keyword in the answer that each searches out;

Optionally, after step 13, this method can also comprise, calculates the vector of keyword in the answer that each searches out of this problem, and calculates the vector of the pairing expansion word of this keyword in the answer that each searches out;

The concrete grammar of compute vector can for, according to keyword, the expansion word of keyword set and keyword set calculates vector; Be specifically as follows:

$\mathrm{Rel}(Q,A)=\frac{\underset{y\&Element;\mathrm{SQ}\&cap;A,x\&Element;{\mathrm{SQ}}_e\&cap;A}{\mathrm{\&Sigma;}}({\mathrm{Wgt}}_{y,q}{\mathrm{Wgt}}_{y,a}+\mathrm{\&lambda;}{\mathrm{Wgt}}_{x,q}{\mathrm{Wgt}}_{x,a})}{\sqrt{\underset{y\&Element;\mathrm{SQ}}{\mathrm{\&Sigma;}}{{\mathrm{Wgt}}_{y,\mathrm{<figure-callout\; id="2"\; label="q"\; filenames="A2009100879590013H2.png"\; state="\{\{state\}\}">q</figure-callout>}}}^2}\sqrt{\underset{y\&Element;A}{\mathrm{\&Sigma;}}{{\mathrm{Wgt}}_{y,a}}^2}\sqrt{\mathrm{\&kappa;}\underset{y\&Element;{\mathrm{SQ}}_e}{\mathrm{\&Sigma;}}{{\mathrm{Wgt}}_{y,q}}^2}}---\left(7\right)$

Wherein, Wgt _{I, q}The weighted value of expression keyword i in problem, Wgt _{I, a}The weighted value of expression keyword i in answer, i are the keyword numberings; When calculating correlativity here, expansion word is carried out certain discount handle, wherein κ and λ are corresponding discount factors, satisfy 0＜κ, λ＜1, and SQ is the keyword set of problem; SQ _eBe the set of SQ expansion word, the set of this expansion word can obtain according to aforesaid way A or B, satisfies Q _e={ SQ, SQ _e, A is the keyword set in answering;

Wgt＝tf*idf*senmatic*relavity (8)

In the formula (8), tf is the frequency that vocabulary occurs, and idf is contrary document frequency, and senmatic is the semantic space calibration factor between the keyword, sets according to the importance of different parts of speech; As noun be 1, verb is 0.8, and other numeral can certainly be set, senmatic can set up on their own as required; Relavity is the semantic dependency metric between the vocabulary, and when being keyword, the value of relavity is 1, when being the expansion word of keyword, relavity can be calculated by above-mentioned formula (1), during calculating, this keyword be equivalent to formula (1) in W, expansion word is equivalent to We.Wherein the concrete computing method of contrary document frequency idf are: $\mathrm{idf}=1+\log \frac{\mathrm{docnum}}{\mathrm{df}+1}---\left(8\right)$

Wherein, docnum is a total number of documents; Df is current vocabulary document frequency.

Step 14, confirm the preferential utmost point of each answer,, and answer is exported by this order in the order of determining this answer according to this priority according to this result calculated.

Realize this step method can for, if what calculate is occurrence number, then carry out Determination of priority according to the result of calculation of number of times, if calculate be vector then carry out Determination of priority according to the result of calculation of vector sum number of times.

The method that the specific embodiment of the invention provides, because to the equal calculation times of the expansion word of keyword and keyword, and the number of times of the expansion word of the number of times of the keyword that calculates and keyword combined the priority of judging answer, and answer is exported, thereby has the high advantage of accuracy according to the order of priority.And this method can also be judged the priority of answer by the result of calculation of vector, thereby has further improved accuracy.

The specific embodiment of the invention provides an embodiment, and present embodiment provides a kind of searcher based on answer in the question and answer interaction platform, and this installs as shown in Figure 6, comprising:

Receiving element 61 is used to receive the problem that needs are answered;

Word segmentation processing unit 62 after being used for this problem carried out word segmentation processing, extracts the keyword of this problem;

Query unit 63 is used for finding out from the keyword expansion tabulation of storage according to the keyword of this problem the pairing expansion word of keyword of described problem;

Statistic unit 64 is used for calculating the number of times of the keyword of this problem in each answer that searches out, and calculates the number of times of the pairing expansion word of this keyword in the answer that each searches out;

Priority output unit 65 is used for confirming according to result calculated the priority of each answer, exports after the order of determining this answer according to this priority.

The present invention improves word segmentation processing unit 62, query unit 63, statistic unit 64 and priority output unit 55 to obtain technical scheme of the present invention.

Optionally, computing unit 64 can also be used for calculating the vector of the keyword of this problem in each answer that searches out, and calculates the vector of the pairing expansion word of described keyword in the answer that each searches out.The concrete computing method of vector can not given unnecessary details here referring to the associated description among the method embodiment.

Optionally, this device can also comprise:

Storage unit 65 is used to store the keyword expansion tabulation; Setting up described keyword expansion tabulation is specially, all data of storage are carried out the keyword that word segmentation processing obtains each data, calculate the level and smooth tolerance between any two keywords in the keyword of each data, exceed setting value as the level and smooth tolerance of this calculating and then should be set at keyword and expansion word by any two keywords, and store in the keyword expansion tabulation.Smoothly the computing method of tolerance can be referring to the description among the method embodiment.

The device that the specific embodiment of the invention provides, because the equal number of computations of expansion word of 64 pairs of keywords of computing unit and keyword, and export after by priority output unit 65 quantity of the expansion word of the quantity of the keyword that calculates and keyword being combined the priority of judging answer, thereby has the high advantage of accuracy.

One of ordinary skill in the art will appreciate that all or part of step that realizes in the foregoing description method is to instruct relevant hardware to finish by program, described program can be stored in the computer read/write memory medium, this program is when carrying out, comprise the said method step, described storage medium comprises: ROM/RAM, disk, CD etc.

In sum, the technical scheme that the specific embodiment of the invention provides has the high advantage of accuracy.

The above; only be the preferable embodiment of the present invention; but protection scope of the present invention is not limited thereto; anyly be familiar with those skilled in the art in the technical scope that the embodiment of the invention discloses; the variation that can expect easily or replacement all should be encompassed within protection scope of the present invention.Therefore, protection scope of the present invention should be as the criterion with the protection domain of claim.

Claims (7)

1, a kind of searching method based on answer in the question and answer interaction platform is characterized in that, described method comprises:

The problem of receive to need answering, and after described problem carried out word segmentation processing, extract the keyword of described problem;

From the keyword expansion tabulation of storage, find out the pairing expansion word of keyword of described problem according to the keyword of described problem; Calculate the number of times of keyword in the answer that each searches out of described problem, and calculate the number of times of the pairing expansion word of described keyword in the answer that each searches out;

Confirm the priority of each answer according to result calculated, after the order of determining this answer according to this priority, export.

2, method according to claim 1 is characterized in that, describedly confirms also to comprise before the priority of each answer according to result calculated:

Calculate the vector of keyword in the answer that each searches out of described problem, and calculate the vector of the pairing expansion word of described keyword in the answer that each searches out;

Wherein, the computing method of vector are specially: according to described keyword, the set of the expansion word of described keyword set and described keyword calculates vector.

3, method according to claim 1 is characterized in that, the method for building up of the keyword expansion of described storage tabulation is,

All data of storage are carried out the keyword that word segmentation processing obtains each data, level and smooth tolerance in the keyword of described each data of calculating between any two keywords, exceed setting value as the level and smooth tolerance of calculating and then should be set at keyword and expansion word by any two keywords, and store in the described keyword expansion tabulation;

The computing method of described level and smooth tolerance are specially: the probability calculation that occurs in cliction up and down according to the expansion word of described keyword and described keyword goes out described level and smooth tolerance.

4, method according to claim 3 is characterized in that, described method also comprised before calculating level and smooth tolerance:

The occurrence number of keyword in all answers of statistics answer, when occurrence number reaches set point number, the calculating of smoothly measuring.

5, a kind of searcher based on answer in the question and answer interaction platform is characterized in that, described device comprises:

Receiving element is used to receive the problem that needs are answered;

The word segmentation processing unit after being used for described problem carried out word segmentation processing, extracts the keyword of described problem;

Query unit is used for finding out from the keyword expansion tabulation of storage according to the keyword of described problem the pairing expansion word of keyword of described problem;

Statistic unit is used for calculating the number of times of the keyword of described problem in each answer that searches out, and calculates the number of times of the pairing expansion word of described keyword in the answer that each searches out;

The priority output unit is used for confirming according to result calculated the priority of each answer, exports after the order of determining this answer according to this priority.

6, device according to claim 5, it is characterized in that, described computing unit also is used for calculating the vector of the keyword of described problem in each answer that searches out, and calculate the vector of the pairing expansion word of described keyword in the answer that each searches out, described vector is according to described keyword, and the set of the expansion word of described keyword set and described keyword calculates.

7, device according to claim 5 is characterized in that, described device also comprises:

Storage unit is used to store the keyword expansion tabulation; Setting up described keyword expansion tabulation is specially, all data of storage are carried out the keyword that word segmentation processing obtains each data, level and smooth tolerance in the keyword of described each data of calculating between any two keywords, exceed setting value as the level and smooth tolerance of calculating and then should be set at keyword and expansion word by any two keywords, and store in the described keyword expansion tabulation; Described level and smooth tolerance draws according to the probability calculation that the expansion word of described keyword and described keyword occurs in cliction up and down.

Images