A METHOD AND SYSTEM FOR SEARCHING BY USING FUZZY
RELATIONAL PRODUCTS
TECHNICAL FIELD
The present invention relates to a search algorithm for increasing search
efficiency of an Internet search engine, and in more detail, a search method and
apparatus for making a similar relationship regarding Internet search classifications by
using fuzzy 0 products.
BACKGROUND OF THE INVENTION
"Fuzzy" is defined as 'knowledge of degree' because it deals the degree of truth.
That is, fuzzy is the knowledge that overcomes the limits of Boolean logic by the
traditional computer. Fuzzy handles the Boolean logic of '0' and '1' specifically.
The idea of fuzzy logic was introduced by Dr. Lotfi Zadeh of the University of
California at Berkeley in the 1960's. He studied the problem how a computer can
understand natural language and perceived the truth that it could not express various
words specifically by using the absolute expression of '0' or '14 The study, which was
started based on the question "How pretty is my wife?", was called "computation with
words" because a human's natural expression is computerized. That is, upon analysis if
we indicate "She is pretty" as "1" and "She isn't pretty" as "0", then "She is slightly
pretty" may be "0.2" and "She is somewhat pretty" may be "0.5" and "She is very
pretty" may be "0.8". We call "0.2", "0.5", "0.8" as membership grades. Also, because
they are distinguished from the absolute expression, we call them soft computing. Of
course, people would have difficulty to express sentences of daily life as '0' or '1'
absolutely.
Fuzzy logic comprises '0' and '1' in the extreme case. Also, fuzzy logic
comprises all values between O'(false) and 'l'(true). Therefore, fuzzy logic indicates
knowledge by vagueness with the results comprising minute errors. Approximate
reasoning originated from fuzzy logic. Also, fuzzy logic uses alpha-cut(-cut) in order to
set a permissible limit of the errors. Because fuzzy logic is like the process of the human
thinking and comprises many philosophical elements, it can be defined as the logic of
emancipation to exclude some extreme thinking like Buddhistic theory. That is, the
fuzzy logic is even more similar to the way of oriental thinking.
The fuzzy logic seems similar to the working method of our brain. We receive
data, calculate the amount of partial truth, organize the high-level truths in order, and
elevate the result as the reaction of the exercise nerve when it exceeds a threshold. This
process can occur in the mechanical brain, the nervous system, and the expert system.
The present invention relates to the search algorithm of a search engine.
Hereinafter, we will explain the search method, along with the advantages and
disadvantages of the prior search engines.
Search engines are classified into the classification item search, the index
search, the meta search that uses a plurality of search engines at once, and etc. The
classification item search is the search method of constructing a directory database, and
the index search is the search method of indexing documents. In the classification item
search, the accuracy rate is high but the reproducibility is low. To the contrary in the
index search, the accuracy rate is low but the reproducibility is high.
A representative search engine using the classification item search method is
Yahoo, and a representative search engine using the index search method is AltaVista.
Simmani, a real Korean search engine, uses both the classification item search method
and the index search method.
We will compare the classification item search method with the index search
method by referring to the following table.
As described above, the two search methods have sides that supplement each
other. Almost all prior search engines provide two search methods together. That is, if
the user is not satisfied with the search result, then he can select the other search method
by using the search method option. Therefore, the prior search engines are still
somewhat troublesomeness and cannot unify the merits of each search method.
In the prior index search method, more time is required in order for the user to
select a desired website from voluminous search results. Also, in the prior index search
method, if the desired website is in another classification, then the user cannot readily
find this website.
Therefore, one object of the present invention is to provide a method and
system for setting up a relationship among each classification, multiplying the
classification of the document or website, and increasing the search efficiency of the
classification item search by using fuzzy relational products. A second object of the
present invention is to provide a search method and system by using fuzzy relational
products for making a search easy by increasing classifications of the vague search
word. A third object of the present invention is to provide a search method and system
by using fuzzy relational products for managing the search classification system
efficiently by providing shared classification items and fluidity of classification levels.
A fourth object of the present invention is to provide a search method and system by
using fuzzy relational products for changing the classification system variously by
setting up the permitted limit of the fuzzy logic. A final object of the present invention
is to provide a search method and system by using fuzzy relational products for solving
particularity among the search classifications by providing compatibility. In this
instance particularity means that one search classification item is both a subordinate
concept and a superordinate concept of another classification item simultaneously.
SUMMARY OF THE INVENTION
To accomplish the objects of the present invention according to one preferred
embodiment of the present invention, the present invention is provided for inputting a
classification item and a search word from a user, calculating the fuzzy degree
according to the degree of the search word about each of the search results, which
belong to the item classification, wherein the classification item is regarded as a fuzzy
set having the search word as an element, calculating an inclusive relationship between
the classification items of a basis classification system as the degree between a
predetermined truth value and a predetermined false value by applying fuzzy relational
products to the basis classification system by using the classification item and the fuzzy
degree, calculating a similar expanded classification system according to the degree,
extracting the first search result of the search word from a database about the
classification item in the basis classification system, extracting the second search result
of the search word from database about the similar expanded classification item in the
similar expanded classification system and displaying the first search result and the
second search result.
The formula corresponding to the fuzzy relational products is (R"1 < R)jra =
→ Rkm), wherein k is an index of the search word, j and m are
classification fuzzy sets having the search word as an element, R is a matrix in which
the classification item is a column and the search word is a row and has a fuzzy value
between 0 and 1 displaying the relationship between each classification and the search
word, R"1 is a matrix in which the classification item is a row and the search word is a
column, Nk is the total number of search words, Rj "1 is the degree that search word k is
included within classification item j, Rkm is the degree that search word k is included
within classification item m, and the arrow symbol " →" is an implication operator.
The basis classification system has an inclusive relationship of high and low
concepts between a plurality of classification items. A similar expanded classification
system changes according to membership function data and alpha-cut, which are
applied to fuzzy relational products.
To accomplish the objects of the present invention according to another
preferred embodiment of the present invention, the present invention is provided for
inputting a search word by the user, extracting at least one classification item list
corresponding to the search word and search result corresponding to the classification
item list from a database, selecting the classification item, which has the highest fuzzy
degree according to the degree of the search word, in the search result, calculating
inclusive relationship between the classification items of basis classification system as
the degree between a predetermined truth value and a predetermined false value by
applying fuzzy relational products to the basis classification system by using the
classification item and the fuzzy degree, calculating similar expanded classification
system according to the degree, extracting similar expanded classification item
corresponding to the classification item by using the similar expanded classification
system, extracting search results of the search word from a database about the
classification item and the similar expanded classification item and displaying the
search result.
The formula corresponding to the fuzzy relational products is (R"1 < R)jm =
l/NkS (Rjk"1 -* Rkm), wherein k is an index of the search word, j and m are
classification fuzzy sets having the search word as an element, R is a matrix in which
the classification item is a column and the search word is a row and has a fuzzy value
between 0 and 1 displaying a relationship between each classification and the search
word, R is a matrix in which the classification item is a row and the search word is a
column, Nk is the total number of search words, Rjk"1 is the degree that the search word
k is included within classification item j, Rkm is the degree that the search word k is
included within classification item m, and the arrow symbol "→" is an implication
operator.
The basis classification system has an inclusive relationship of high and low
concepts between a plurality of classification items. The similar expanded classification
system changes according to the membership function data and alpha-cut, which are
applied to fuzzy relational products.
To accomplish the objects of the present invention according to still another
preferred embodiment of the present invention, the present invention is provided for
receiving a classification item and a search word from a user, calculating the fuzzy
degree according to the degree of the search word about each of the search results,
which belong to the classification item, wherein the classification item is regarded as a
fuzzy set having the search word as an element, calculating an inclusive relationship
between the classification items of the basis classification system as a degree between a
predetermined truth value and a predetermined false value by applying fuzzy relational
products to the basis classification system by using the classification item and the fuzzy
degree, calculating a similar expanded classification system according to the degree,
extracting the first search result of the search word from a database about the
classification item in the basis classification system, extracting the second search result
of the search word from the database about the similar expanded classification item in
the similar expanded classification system and displaying the first search result and the
second search result.
To accomplish the objects of the present invention according to still another
preferred embodiment of the present invention, the present invention is provided for
receiving a search word from a user, extracting at least one classification item list
corresponding to the search word and search result corresponding to the classification
item list from a database, selecting the classification item, which has highest fuzzy
degree according to the degree of the search word, in the search result, calculating an
inclusive relationship between the classification items of the basis classification system
as a degree between a predetermined truth value and a predetermined false value by
applying fuzzy relational products to the basis classification system by using the
classification item and the fuzzy degree, calculating a similar expanded classification
system according to the degree, extracting a similar expanded classification item
corresponding to the classification item by using the similar expanded classification
system, extracting search results of the search word from a database about the
classification item and the similar expanded classification item and displaying the
search results.
BRIEF DESCRIPTIONS OF THE DRAWINGS
The above objects and other advantages of the present invention will become
more apparent by detailed descriptions of the preferred embodiments thereof with
reference to the attached drawings, in which:
FIG. 1 is a schematic diagram of a search system for making similar
relationships among search classifications by using fuzzy relational products in the
present invention.
FIG. 2 is a functional block diagram of making similar relationships among
search classifications by using fuzzy relational products in the present invention.
FIG. 3 is a diagram of result lists created by an implication operator in the
present invention.
FIG. 4 is a diagram of applying fuzzy relational products to a search
classification item in the present invention.
FIG. 5a and FIG. 5b are the results according to alpha-cut(α-cut) in the present
invention.
FIG. 6 is an example illustrating real classification and expanded classification
according to the fuzzy relational method in the present invention.
FIG. 7 is a diagram illustrating a classification system of the expanded
classification according to the fuzzy relational method in the present invention.
FIG. 8 is a flowchart illustrating the process of making a similar relationship
among search classifications and searching by using fuzzy relational products in the
present invention.
<A key of numerical references for the major parts of the drawings>
101: search server
103: fuzzy search module
107: database server
109: search information database
111: network
113: user computer
115: manager computer
EMBODIMENT
Hereinafter, preferred embodiments of the present invention will be described
in more detail with reference to the accompanying drawings, but it is understood that
the present invention should not be limited to the following embodiments.
FIG. 1 is a schematic diagram of a search system for making similar
relationships among search classifications by using fuzzy relational products in the
present invention. Referring to FIG. 1, a search server 101 comprises a fuzzy search
module 103 that provides a search according to the similar relationship among search
classifications by using fuzzy relational products. We will explain the search method
according to fuzzy relational products between FIG. 3 and FIG. 8.
The database server 107, which is coupled with a search server 101, comprises
a search information database 109. The search information database 109 comprises a
search word table 121, classification item table 123, fuzzy degree table 125, etc.
Keywords are stored on the search word table 121, and the classification items of the
basis classification system are stored on the classification item table 123. Fuzzy degrees
are stored on the fuzzy degree table 125. The fuzzy degree is calculated for each search
result belonging to each classification item, which is regarded as a fuzzy set comprising
a search word as an element, according to the frequency of the search word.
Manager 119 and user 115 connect to the search server 101 through network
111 by using the manager computer 117 and the user computer 113 equipping a web
browser.
Fuzzy logic was suggested as a possibility (the logic value between 0 and 1) in
order to apply the logic of vague reality about the crisp logic value, which is fixed as [0,
1] by Zadeh. The vagueness is caused by not only changeability of the logical subject
but also unreality of the crisp logic value. Fuzzy logic can apply the unreal human logic
to real life by breaking the crisp concept of logic value and becoming a logic value
between 0 and 1.
Fuzzy logic was introduced by Zadeh, developed by many researchers
especially mathematicians, and applied to real life in Japan. Fuzzy logic is used in order
to improve the resolution of a camera and applied to a mechanical brain of a washing
machine, a refrigerator, and an air conditioner. That is, it need not be asserted that fuzzy
logic is the seed of the mechanical brain algorithm.
Also, fuzzy logic can be applied to the relational database, the object-oriented
database, and etc. Fuzzy logic can be used in order to set up the vagueness of a
relationship at the relational database and set up the obscurity of the class hierarchy
structure. Fuzzy logic may be widely applied to setting up the relationship of objects.
The relationship of objects is usually described as the inclusive relationship. And the
relational products describe it well.
Each classification is regarded as a fuzzy set having a search word as an
element at the predetermined basis classification system. A fuzzy degree is decided
according to the frequency of a search word in documents or the summary of web sites
corresponding to each classification. And the fuzzy search module 103 displays the
inclusive relationship as a value between 0 and 1 by using the fuzzy degree and
applying fuzzy relational products to each classification.
If the fuzzy search module 103 knows the classification of a document, then it
looks for the classification in the classification system display window. Also if the fuzzy
search module 103 cannot search for the desired content in the classification, then
expansion occurs into similar classifications.
If the classification is not correct, then the fuzzy search module 103 selects the
classification of high degree in the classifications, which are related to the search word,
and displays it. Also if the fuzzy search module 103 cannot search for the desired
content in the classification, then it expands to similar classifications.
FIG. 2 is a functional block diagram for making a similar relationship among
search classifications by using fuzzy relational products in the present invention.
Referring to FIG. 2, the search condition, which is received by the user 115, is
transmitted to the database process and fuzzy search unit 207. The search condition can
be a classification item or a search word. The database process and fuzzy search unit
207 searches the search word about the classification item on the search information
database 109 and displays the search result 203.
The database process and fuzzy search unit 207 extracts search words and
classification items and then calculates fuzzy degree information between each search
word and each classification item. Also, the database process and fuzzy search unit 207
calculates the expanded similar classification item by applying the classification item to
fuzzy relational products. And the database process and fuzzy search unit 207 searches
the search word about each expanded similar classification item and displays the search
result 203.
FIG. 3 is a diagram of result lists created by an implication operator in the
present invention. Referring to FIG. 3 and regarding the general concept of a set,
whether the element x is comprised in set A(x£ΞA) is displayed as 0 or 1 according to
the crisp logic. However, if A is a fuzzy set, then the logical value of xEA can be a
value between 0 and 1.
Also, the crisp implication operator has values as follows.
If fuzzy logic applies a fuzzy set, fuzzy power set, and crisp implication
operator, then a binary function is needed, which outputs the result in the range of [0,1]
and the domain of [0,l]x[0,l]. We name the function implication operator and can
display various types according to each environment. Hundreds of kinds of known
implication operators exist.
The representative functions are as follows.
KD. The Kleene-Diense systems and operator
a→b = (l-a)vb = max(l-a,b)
Implication operator is the function for setting logical value of fuzzy
implication. From this, we can display an inclusive relationship of all elements, which
are comprised in the universal set, as shown in Table 301(Degree of membership).
FIG. 4 is a diagram of applying fuzzy relational products to search a
classification item in the present invention. We will examine first about fuzzy relational
products in order to explain how these fuzzy relational products are applied to the
search classification item. Referring to FIG. 4, fuzzy relational products were developed
for the first time by W. Bandler and LJ. Kohout. Some fuzzy relational products were
applied to establish a relationship between each of the patients by observing Parkinson's
disease patients' mental symptoms in London hospitals in order to handle the possibility
in the actual world.
The formula that works by fuzzy relational products is as follows.
< R
(1)(t))
jra -
→ R
(1)km(t)) formula 1)
Usually, search classification items become involved as an inclusive
relationship of high and low concepts. In the crisp environment, some classification
item has a classification of a subordinate concept like book classification of the library.
Therefore, high position classification item includes each low rank classification item
that the classification item of a subordinate concept includes. But the high and low
inclusive relationship can appear to be more or less different according to the subjective
viewpoint of the actual search subject.
When we look from the viewpoint of an ordinary person, we think of the
Internet as that which is obviously seen on the web pages that are searched through a
web browser. Contrarily from the expert's viewpoint, the Internet comprises
terminology that not only indicates the world wide web (WWW) but also
communication protocols such as FTP, Gopher, News, Email and so on. We embody
fluid classification system that changes according to the search environment and search
subject by applying fuzzy relational products in order to process vagueness of inclusive
relationships, uncertainty, and variability.
First, there is need to establish a relationship between a classification item and
search words. A keyword (which is related to the classification item) besides the
classification item itself does not exist in the typical search engine. In the present
invention, some search words of each classification are related to a degree that has a
value between 0 and 1.
As an example we are going to explain the search classification method in
fuzzy relational products by using formula 1.
Supposition)
1) Kl, K2, K3, K4, K5: search word
2) CI, C2, C3, C4, C5: fuzzy set that has the search word as an element
3) R: matrix in which classification is a column and search word is a row and
displays a relationship between each classification and search word. It has a
fuzzy value between 0 and 1.
4) R"1: matrix in which classification is a row and search word is a column.
The lower formula is calculated by the above supposition and formula 1.
(R-1 < R)jm = l/Nk 1 → Rkm) formula 2)
Nk: total number of search words
Rjk "1: degree that search word k is included within classification item j
R m: degree that search word k is included within classification item m
Arrow symbol "→" : implication operator
The formula 2 indicates the degree that the m times classification set comprises
search words of the j times classification set.
Example) (R"1 < R)23 = 1, (R'1 < R)i3 = 0.8
Then the relationship between each classification can be set up like 401, 403
and 405 by fuzzy relational products. Alpha-cut (α-cut) is a stand value to change the
final result to a crisp value. That is, if alpha-cut(α-cut) is 1, then the relational products
value which is less than or equal to 0.8 becomes 0, and the relational products value
which is over 0.8 becomes 1. If alpha-cut(α-cut) is 0.8, then relational products value
which is below 0.8 becomes 0, and the relational products value which is greater than or
equal to 0.8 becomes 1.
FIG. 5a and FIG. 5b illustrate the results according to alpha-cut(α-cut) in the
present invention. Referring to FIG. 5a and FIG. 5b, the relationship between search
classification items is created by the final value, which is made by the fuzzy relational
products at FIG. 4. The schematic diagram displaying relationship between search
classification items is same with 501 when alpha-cut is 1. Also, the schematic diagram
displaying a relationship between search classification items is the same with 503 when
alpha-cut is 0.8.
When alpha-cut is 1 like 501, classification item CI is a subordinate item of all
classification items, and C3 is a subordinate classification item of C2, and C5 is a
subordinate classification item of C4.
When alpha-cut is 0.8 like 503 C4 is the highest classification item, and CI and
C3 are the lowest classification items. The classification relationship of alpha-cut 0.8
expands comprising the classification relationship of alpha-cut 1.
We can find some particularity being different from the general search
classification in 501 and 503.
1) Share of classification: CI is composed simultaneously of subordinate
classification items of C2, C3, C4 and C5 in 501. The search method of the present
invention introduces a share concept and multi-inheritance instead of an exclusive
concept of general search classification. That is, the sites that are classified by CI share
C2, C3, C4, and C5 as a high position classification item.
2) Fluidity of classification : CI has fluidity that can be situated in several
classification hierarchies in 501. CI belongs in the third classification hierarchy when
the search path is C2 → C3 → CI. However, CI belongs in the second classification
hierarchy when search path is C2 → CI.
3) Representation of more than two about one meaning: We can find another
particularity in 503. CI, C3 are expressed differently as a search classification having
the same meaning. That is, CI and C3 are equal for all classification relationships.
4) Compatibility of classification: C2 and C5 to each other are both a
subordinate concept and a superordinate concept in 503 at the same time. However, C2
and C5 do not represent the same meaning as the relationship of CI and C3. C5 is a
subordinate classification of CI and C3, but C2 is not. This situation is similarly
explained for the relationship of CI, C3, C5.
5) Subjectivity of classification system: The classification system can be
different according to alpha-cut in 501 and 503. The alpha-cut can change according to
the search environment and search subject.
FIG. 6 is an example illustrating real classification and expanded classification
according to the fuzzy relational method in the present invention. FIG. 7 is a diagram
illustrating the classification system of the expanded classification according to the
fuzzy relational method in the present invention.
Referring to FIG. 6 and FIG. 7, the inclusive relationship between search
classifications A, B, C, D, E, F, G is marked with a solid line in 601. Furthermore, the
expanded classification relationship, which is created by fuzzy relational products, is
marked with dotted line in 601. That is, the expanded classification relationship, which
is created by fuzzy relational products, among search classifications A, B, C, D, E, F, G
can be expressed the same with 701.
FIG. 8 is a flowchart illustrating the process of making similar relationships
among search classifications and searching by using fuzzy relational products in the
present invention. Referring to FIG. 8, if the fuzzy search module 103 receives a
classification item and search word from user 115 S801, then the fuzzy search module
103 extracts the subordinate classification item corresponding to the classification item
from a search information database 109 S803. And the fuzzy search module 103
extracts the similar expanded classification item of the classification item and the
similar expanded classification item of the subordinate classification item according to
fuzzy relational products S805. If the fuzzy search module 103 extracts the search
classification item and the similar expanded classification item, then finds the search
word about each classification item S807.
In more detail, the search order by the search system, which supports fuzzy
relational products, is as follows. The user expands the search tree until he looks for the
desired search classification in the user interface screen of the tree view type. Otherwise,
the user inputs a search classification item in the search word input part and pushes the
return key; thereafter the fuzzy search module 103 unfolds the classification items
automatically in the user interface screen of the tree view type. Whenever the user
selects a classification item in the tree view, the summary explanation of sites, which
belong directly to the classification, can be displayed on another window.
If the user selects the summary explanation window, he then connects with the
site automatically. Also the fuzzy search module 103 modifies the statistics data of the
selection count.
If the desired site does not exist in the classification or the subordinate
classification, then the user selects the option for expanding to the similar classification.
Then the fuzzy search module 103 extracts the similar expanded classification item by
using fuzzy relational products and searches the search word in the similar expanded
classification item.
INDUSTRIAL APPLICABILITY
As described above, according to the present invention, the present invention
can provide a method and system for setting up the relationship among each
classification, multiplying the classification of a document or website and increasing the
search efficiency of the classification item search by using fuzzy relational products.
Also, the present invention can provide a search method and system by using
fuzzy relational products for making the search easy by increasing the number of
classifications of the vague search word.
Also, the present invention can provide a search method and system by using
fuzzy relational products for managing the search classification system efficiently by
providing for sharing of classifications and fluidity of classification levels.
Also, the present invention can provide a search method and system by using
fuzzy relational products for changing the classification system variously by setting up
the permitted limit of the fuzzy logic.
Also, the present invention can provide a search method and system by using
fuzzy relational products for solving the particularity among the search classifications
by providing compatibility. The particularity means that one search classification item is
a subordinate concept and a superordinate concept of another classification item
simultaneously.
Also, the present invention can provide a search method and system by using
fuzzy relational products for being utilized as a basis module of an intelligent
classification way, as well as a search engine.
Also, the present invention can provide a search method and system by using
fuzzy relational products for providing a search way uniting the advantages of a
classification item search and the index search through the excellent graphical user
interface of a tree view type.