US 9,317,590 B2Granted

Search method, search device and storage medium

AssigneeTencent Technology (Shenzhen) Company Limited

InventorsYanxiong Lu, Yuekui Yang, Liang Wang, Feng Jiao

Life of this patent

Estimating the term…

Loading patent…

Abstract

Disclosed are a search method, a search device and a storage medium. The method comprises obtaining all relevant documents of information to be sought; calculating a relevancy between each relevant document and the information to be sought based on a word matching algorithm and a semantics matching algorithm; performing sequencing processing on all the relevant documents according to the relevancy obtained through calculation, and displaying a sequencing result. Further disclosed is a search device. The present invention comprehensively considers matching between words, and matching of the semantics relationship between words, obtains an accurate relevancy calculation result, provides an ideal search result to a user, and improves satisfaction of the user.

Claims7

Dependent claims

A search method, comprising:

obtaining relevant documents of information to be searched;
calculating a correlation between each obtained relevant document and the information to be searched, based on a word matching algorithm and a semantic matching algorithm;
and
sorting the obtained relevant documents according to the calculated correlation, and displaying a sorting result;
wherein said calculating a correlation between each obtained relevant document and the information to be searched based on a word matching algorithm and a semantic matching algorithm comprises:
vectorizing the information to be searched, to obtain m vectors ti, wherein m and i are both positive integers, and 1≦i≦m;
vectorizing each obtained relevant document, to obtain n vectors dj corresponding to each relevant document, wherein n and j are both positive integers, and 1≦j≦n;
calculating a relevancy score S1 of each relevant document to the information to be searched based on the word matching algorithm, and calculating a relevancy score S2 of each relevant document to the information to be searched based on the semantic matching algorithm;
calculating the correlation S between each relevant document and the information to be searched, according to the formula of S=α×S1+(1−α)×S2, wherein α is a preset weight, and 0<α<1;
wherein the formula of the word matching algorithm is as follows:
S
1
=
∑
i
=
1
m
⁢
⁢
(
qtf
i
k
3
+
qtf
i
)
×
(
(
k
1
+
1
)
×
tf
i
k
×
(
1
-
b
+
b
⁢
⁢
l
avdl
)
+
tf
i
)
⁢
log
⁡
(
w
i
)
wherein k1, k3, k and b are constants, qtfi is the word frequency of the ith vector ti in the information to be searched, tfi is the word frequency of the vector ti in a corresponding relevant document, l is the length of the corresponding relevant document, avdl is the average length of all obtained relevant documents, and wi is the weight of the vector ti.

62 more elements

A search method, comprising:

obtaining relevant documents of information to be searched;
calculating a correlation between each obtained relevant document and the information to be searched, based on a word matching algorithm and a semantic matching algorithm;
and
sorting the obtained relevant documents according to the calculated correlation, and displaying a sorting result;
wherein said calculating a correlation between each obtained relevant document and the information to be searched based on a word matching algorithm and a semantic matching algorithm comprises:
vectorizing the information to be searched, to obtain m vectors ti, wherein m and i are both positive integers, and 1≦i≦m;
vectorizing each obtained relevant document, to obtain n vectors dj corresponding to each relevant document, wherein n and j are both positive integers, and 1≦j≦n;
calculating a relevancy score S1 of each relevant document to the information to be searched based on the word matching algorithm, and calculating a relevancy score S2 of each relevant document to the information to be searched based on the semantic matching algorithm;
calculating the correlation S between each relevant document and the information to be searched, according to the formula of S=α×S1+(1−α)×S2, wherein α is a preset weight, and 0<α<1;
wherein the formula of the semantic matching algorithm is as follows:
S
2
=
∑
i
=
1
m
⁢
⁢
∑
j
n
⁢
⁢
1
k
3
+
1
×
(
k
1
+
1
k
×
(
1
-
b
+
b
⁢
⁢
l
avdl
)
)
⁢
mi
⁡
(
t
i
,
d
j
)
wherein k1, k3, k and b are constants, l is the length of a corresponding relevant document, avdl is the average length of all the obtained relevant documents, and mi(ti,dj) is the mutual information between the vector ti and the vector dj.

58 more elements

A search device, comprising:

a search module which is configured to obtain relevant documents of information to be searched;
a calculation module which is configured to calculate a correlation between each relevant document obtained by the search module and the information to be searched, based on a word matching algorithm and a semantic matching algorithm;
a sort module which is configured to sort the relevant documents obtained by the search module according to the correlation calculated by the calculation module;
and
a display module which is configured to display a sorting result obtained by the sort module;
wherein the calculation module comprises:
a first vectorization unit which is configured to vectorize the information to be searched, to obtain m vectors ti, wherein m and i are both positive integers, and 1≦i≦m;
a second vectorization unit which is configured to vectorize each relevant document obtained by the search module, to obtain n vectors dj corresponding to each relevant document, wherein n and j are both positive integers, and 1≦j≦n;
a word matching calculation unit which is configured to calculate a relevancy score S1 of the relevant document processed by the second vectorization unit to the information to be searched based on the word matching algorithm;
a semantic matching calculation unit which is configured to calculate a relevancy score S2 of the relevant document processed by the second vectorization unit to the information to be searched based on the semantic matching algorithm;
and
a correlation calculation unit which is configured to calculate the correlation S between the relevant document and the information to be searched, according to the formula of S=α×S1+(1−α)×S2, wherein α is a preset weight, and 0<α<1;
wherein the formula of the word matching algorithm is:
S
1
=
∑
i
=
1
m
⁢
⁢
(
qtf
i
k
3
+
qtf
i
)
×
(
(
k
1
+
1
)
×
tf
i
k
×
(
1
-
b
+
b
⁢
⁢
l
avdl
)
+
tf
i
)
⁢
log
⁡
(
w
i
)
the formula of the semantic matching algorithm is as follows:
S
2
=
∑
i
=
1
m
⁢
⁢
∑
j
n
⁢
⁢
1
k
3
+
1
×
(
k
1
+
1
k
×
(
1
-
b
+
b
⁢
⁢
l
avdl
)
)
⁢
mi
⁡
(
t
i
,
d
j
)
wherein k1, k3, k and b are constants, qtfi is the word frequency of the ith vector ti in the information to be searched, tfi is the word frequency of the vector ti in a corresponding relevant document, l is the length of the corresponding relevant document, avdl is the average length of all relevant documents obtained by the search module, wi is the weight of the vector ti, and mi(ti,dj) is the mutual information between the vector ti and the vector dj.

116 more elements

Dependencies are inferred from the claim text.

Facts

Application: 14/347,776Filed 6 Dec 2012
Priority: CN 2012 1 003152313 Feb 2012
Published: US 2014/0358914 A14 Dec 2014
Granted: 19 Apr 2016
Term ends: Not estimatedThe source records don’t support an estimate.

Classification5 codes

G06F 16/3344
Query execution: using natural language analysis
G06F 16/338
Querying: presentation of query results
G06F 16/36
Of unstructured textual data: creation of semantic tools, e.g. ontology or thesauri
G06F 16/93
Details of database functions independent of the retrieved data types: document management systems