# US 8,543,908 B2 · Aiding in creating, extending, and verifying accessibility metadata

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 8,543,908 B2
- Document ID: US8543908B2
- Status: Granted
- Kind: B2
- Application number: 12/825,317
- Filed: 2010-06-28
- Published: 2011-01-20
- Granted: 2013-09-24
- Earliest claimed: 2009-07-16
- Assignees: International Business Machines Corporation
- Applicants: Shinya Kawanaka; Masatomo Kobayashi; Daisuke Sato
- Inventors: Shinya Kawanaka; Masatomo Kobayashi; Daisuke Sato
- Web page: https://patentgrep.com/patent/US8543908B2

## CPC

- G06F 16/957: Browsing optimisation, e.g. caching or content distillation
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

Degrees of similarity between pages from the viewpoint of metadata creation and verification can be estimated based on degrees of similarity between pages in terms of structure, style (appearance), and accessibility error, and, based on the estimation, representative pages appropriate for efficiently adding metadata can be recommended, or target pages appropriate for efficiently verifying the metadata can be recommended.

## Claims

### Claim 1

A program product for aiding in correcting defects included in structured text, wherein the program product comprises a non-transitory storage medium storing a program that, when executed by a computer, causes the computer to perform:
calculating a degree of defect for each of the defects;
calculating a degree of similarity for each of the defects;
calculating, based on the degree of defect and the degree of similarity, a degree of contribution indicating a degree to which correction of each of the defects influences correction of remaining defects, wherein the structured text defined by a Document Object Model (DOM), and the degree of contribution cont(d) is calculated by an equation (1):
[
Expression
⁢
⁢
1
]
cont
⁡
(
d
)
=
∑
d
′
∈
D
⁢
⁢
psim
⁡
(
dd
′
)
·
es
⁡
(
d
′
)
(
Equation
⁢
⁢
1
)
where
cont(d) is the degree of correction contribution indicating the degree to which the correction of each of the defects in a page denoted by DOM d′ influences defects in a page denoted by DOM d,
psim(d,d′) is the degree of similarity between each of the defects included in the page denoted by DOM d and each of the defects included in the page denoted by DOM d′,
es(d′) is the degree of defect of the page denoted by DOM d′, and
D is a set of DOM in target HTML text; and
selecting a defect having a high degree of contribution as a higher-priority defect.

### Claim 2 (depends on 1)

The program product according to claim 1, wherein the program that, when executed by the computer, causes the computer to perform:
recalculating the degree of defect for each of the remaining defects on the assumption that the selected defect has been corrected;
recalculating the degree of contribution for the remaining defects based on the recalculated degree of defect and the degree of similarity; and
selecting another defect having a high degree of contribution among the remaining defects as a next higher-priority defect.

### Claim 3 (depends on 2)

The program product according to claim 2, wherein a correction order of the defects is determined by recursively repeating the recalculating and the selecting.

### Claim 4 (depends on 1)

The program product according to claim 1, wherein the defects are defects in each page included in the structured text.

### Claim 5 (depends on 1)

The program product according to claim 1, wherein the degree of similarity psim(d,d′) is calculated by an equation (2):
[
Expression
⁢
⁢
2
]
psim
⁡
(
d
,
d
′
)
=
∑
p
∈
P
⁡
(
d
)
⋂
P
⁡
(
d
′
)
⁢
⁢
s
⁡
(
p
)
·
k
⁡
(
d
,
d
′
,
p
)
∑
p
∈
P
⁡
(
d
)
⋃
P
⁡
(
d
′
)
⁢
s
⁡
(
p
)
(
Equation
⁢
⁢
2
)
where
P(d) and P(d′) are sets of the defects included in the page denoted by DOM d and the page denoted by DOM d′, respectively,
s(p) is a degree of defect due to a defect p, and
k(d,d′,p) is a degree of similarity indicating to which degree the defect p is similar between the page denoted by DOM d and the page denoted by DOM d′.

### Claim 6 (depends on 5)

The program product according to claim 5, wherein
on a condition that the defect p is a defect having a location identifiable in DOM, the degree of similarity k(d,d′,p) is calculated by an equation (3):
k(d,d′,p)=ssim(elems(d,p),elems(d′,p)) (Equation 3)
where
elems(d,p) and elems(d′,p) are element sets identified by the defect p in the page denoted by DOM d and the page denoted by DOM d′, respectively,
s=s(elems(d,p)) and s′=s(elems(d′,p)) are styles of the element sets elems(d,p) and elems(d′,p), respectively, and
ssim(s,s′) is a degree of style similarity indicating a degree of similarity between the styles.

### Claim 7 (depends on 6)

The program product according to claim 6, wherein
the degree of style similarity ssim(s,s′) is calculated by an equation (4):
⁢
[
Expression
⁢
⁢
4
]
ssim
⁡
(
s
,
s
′
)
=
∑
p
∈
s
⋂
s
′
⁢
Wa
⁢
∑
v
⁢
⁢
s
⁡
(
a
,
v
)
-
s
′
⁡
(
a
,
v
)
(
∑
v
⁢
⁢
s
⁡
(
a
,
v
)
2
)
·
(
∑
v
⁢
⁢
s
′
⁡
(
a
,
v
)
2
)
⁢
⁢
⁢
where
⁢
⁢
⁢
∑
a
⁢
Wa
=
1
(
Equation
⁢
⁢
4
)
where
the style s is a function s(a,v) of a property a and a value v of the style, and
Wa is a weight indicating a degree to which the property a influences the degree of style similarity.

### Claim 8 (depends on 5)

The program product according to claim 5, wherein
if the defect p is a defect not having a location identifiable in DOM,
the degree of similarity k(d,d′,p) is calculated by an equation (5):
k(d,d′,p)=dsim(d,d′) (Equation 5)
where
dsim(d,d′) is a degree of similarity between DOM d and DOM d′.

### Claim 9

A program product for aiding in correcting defects included in structured text, wherein the program product comprises a non-transitory storage medium storing a program that, when executed by a computer, causes the computer to perform:
receiving an input of a correction for a defect;
calculating a degree of similarity between the defect and remaining defects;
calculating, based on the degree of similarity, a degree of confidence indicating a degree as to whether or not the correction is applicable to the remaining defects, wherein the structured text is HTML text defined by Document Object Model (DOM), and the degree of confidence conf(m,d,d′) is calculated by an equation (1):
conf(m,d,d′)=ssim(elems(d,m),elems(cr,m))·dsim(d,d′) (Equation 1)
where
conf(m,d,d′) is a degree of confidence indicating a degree as to whether or not a correction m for a defect included in a page denoted by DOM d is also applicable to a defect in a page denoted by DOM d′,
dsim(d,d′) is a degree of similarity between the page denoted by DOM d and the page denoted by DOM d′,
elems(d,m) and elems(d′,m) are element sets identified by the correction m in the page denoted by DOM d and the page denoted by DOM d′, respectively,
s=s(elems(d,m)) and s′=s(elems(d′,m)) are styles of the element sets elems(d,m) and elems(d′,m), respectively, and
ssim(s,s′) is a degree of similarity between the styles; and
selecting the correction as a correction candidate for defects having the degree of confidence equal to or higher than a threshold.

### Claim 10 (depends on 9)

The program product according to claim 9, wherein the remaining defects are defects across pages included in the structured text.

### Claim 11 (depends on 9)

The program product according to claim 9, wherein the remaining defects are defects within a page included in the structured text.

### Claim 12

A program product for aiding in verifying corrections included in structured text, wherein the program product comprises a non-transitory storage medium storing a program that, when executed by a computer, causes the computer to perform:
calculating a degree of verification indicating a degree to which each of the corrections have been verified;
calculating a degree of similarity for the corrections;
calculating, based on the degree of similarity, a degree of verification contribution indicating a degree to which verification of a correction influences verification of the remaining corrections, wherein the structured text is HTML text defined by a Document Object Model (DOM), and the degree of verification contribution vcond(d) is calculated by an equation (1):
⁢
[
Expression
⁢
⁢
7
]
vcond
⁡
(
d
)
=
∑
d
′
∈
D
⁢
⁢
{
Σconf
⁡
(
m
,
d
′
,
d
)
·
(
1
-
v
⁡
(
d
′
,
m
)
)
}
(
Equation
⁢
⁢
7
)
where
D is a set of target DOM,
v(d′,m) is a degree of verification indicating a degree to which a correction m included in a page denoted by DOM d′ has been verified, and
conf(m,d′,d) is a degree of confidence indicating a degree as to whether or not the correction m included in the page denoted by DOM d′ is also applicable to a defect in a page denoted by DOM d; and
selecting a correction having a high degree of verification contribution as a higher-priority correction.

### Claim 13 (depends on 12)

The program product according to claim 12, wherein the program that, when executed by the computer, causes the computer to perform:
recalculating the degree of verification for each remaining correction on the assumption that the selected correction has been verified;
recalculating the degree of verification contribution for the remaining corrections based on the recalculated degree of verification and the degree of similarity; and
selecting another correction having a high degree of verification contribution among the remaining corrections as a next higher-priority correction.

### Claim 14 (depends on 13)

The program product according to claim 13, wherein a verification order of the corrections is determined by recursively repeating the recalculating and the selecting.

### Claim 15 (depends on 13)

The program product according to claim 13, wherein
the structured text is HTML text defined by a Document Object Model (DOM), and on the assumption that a correction m included in a page denoted by DOM d′ has been verified,
the recalculated degree of verification v′(d,m) is calculated by an equation (2):
v′(d,m)=v(d,m)+conf(m,d,d)N·v(d,m) (Equation 2)
where
v(d,m) is a degree of verification indicating a degree to which the correction m included in a page denoted by DOM d has been verified, and
conf(m,d,d′) is a degree of confidence indicating a degree as to whether or not the correction m included in the page denoted by DOM d is also applicable to a defect in the page denoted by DOM d′.

### Claim 16

A method implemented by a computer system, the method aiding in correcting defects included in structured text, comprising:
calculating a degree of defect for each of the defects;
calculating a degree of similarity for each of the defects;
calculating, based on the degree of defect and the degree of similarity, a degree of contribution indicating a degree to which correction of each of the defects influences correction of remaining defects, wherein the structured text defined by a Document Object Model (DOM), and the degree of contribution cont(d) is calculated by an equation (1):
[
Expression
⁢
⁢
1
]
cont
⁡
(
d
)
=
∑
d
′
∈
D
⁢
⁢
psim
⁡
(
dd
′
)
·
es
⁡
(
d
′
)
(
Equation
⁢
⁢
1
)
where
cont(d) is the degree of correction contribution indicating the degree to which the correction of each of the defects in a page denoted by DOM d′ influences defects in a page denoted by DOM d,
psim(d,d′) is the degree of similarity between each of the defects included in the page denoted by DOM d and each of the defects included in the page denoted by DOM d′,
es(d′) is the degree of defect of the page denoted by DOM d′, and
D is a set of DOM in target HTML text; and
selecting a defect having a high degree of contribution as a higher-priority defect.

### Claim 17

A method implemented by a computer system, the method aiding in correcting defects included in structured text, comprising:
receiving an input of a correction for a defect;
calculating a degree of similarity between the defect and remaining defects;
calculating, based on the degree of similarity, a degree of confidence indicating a degree as to whether or not the correction is applicable to the remaining defects, wherein the structured text is HTML text defined by Document Object Model (DOM), and the degree of confidence conf(m,d,d′) is calculated by an equation (1):
conf(m,d,d′)=ssim(elems(d,m),elems(d′,m))·dsim(d,d′) (Equation 1)
where
conf(m,d,d′) is a degree of confidence indicating a degree as to whether or not a correction m for a defect included in a page denoted by DOM d is also applicable to a defect in a page denoted by DOM d′,
dsim(d,d′) is a degree of similarity between the page denoted by DOM d and the page denoted by DOM d′,
elems(d,m) and elems(d′,m) are element sets identified by the correction m in the page denoted by DOM d and the page denoted by DOM d′, respectively,
s=s(elems(d,m)) and s′=s(elems(d′,m)) are styles of the element sets elems(d,m) and elems(d′,m), respectively, and
ssim(s,s′) is a degree of similarity between the styles; and
selecting the correction as a correction candidate for defects having the degree of confidence equal to or higher than a threshold.

### Claim 18

A method implemented by a computer system, the method aiding in verifying corrections included in structured text, comprising:
calculating a degree of verification indicating a degree to which each of the corrections have been verified;
calculating a degree of similarity for the corrections;
calculating, based on the degree of similarity, a degree of verification contribution indicating a degree to which verification of a correction influences verification of the remaining corrections, wherein the structured text is HTML text defined by a Document Object Model (DOM), and the degree of verification contribution vcond(d) is calculated by an equation (1):
⁢
[
Expression
⁢
⁢
7
]
vcond
⁡
(
d
)
=
∑
d
′
∈
D
⁢
⁢
{
Σconf
⁡
(
m
,
d
′
,
d
)
·
(
1
-
v
⁡
(
d
′
,
m
)
)
}
(
Equation
⁢
⁢
7
)
where
D is a set of target DOM,
v(d′,m) is a degree of verification indicating a degree to which a correction m included in a page denoted by DOM d′ has been verified, and
conf(m,d′,d) is a degree of confidence indicating a degree as to whether or not the correction m included in the page denoted by DOM d′ is also applicable to a defect in a page denoted by DOM d; and
selecting a correction having a high degree of verification contribution as a higher-priority correction.

### Claim 19

A computer system for aiding in correcting defects included in structured text, comprising:
storage; and
a processing unit coupled to the storage and that performs:
calculating a degree of defect for each of the defects;
calculating a degree of similarity for each of the defects;
calculating, based on the degree of defect and the degree of similarity, a degree of contribution indicating a degree to which correction of each of the defects influences correction of remaining defects, wherein the structured text defined by a Document Object Model (DOM), and the degree of contribution cont(d) is calculated by an equation (1):
[
Expression
⁢
⁢
1
]
cont
⁡
(
d
)
=
∑
d
′
∈
D
⁢
⁢
psim
⁡
(
dd
′
)
·
es
⁡
(
d
′
)
(
Equation
⁢
⁢
1
)
where
cont(d) is the degree of correction contribution indicating the degree to which the correction of each of the defects in a page denoted by DOM d′ influences defects in a page denoted by DOM d,
psim(d,d′) is the degree of similarity between each of the defects included in the page denoted by DOM d and each of the defects included in the page denoted by DOM d′, es(d′) is the degree of defect of the page denoted by DOM d′, and
D is a set of DOM in target HTML text; and
selecting a defect having a high degree of contribution as a higher-priority defect.

### Claim 20

A computer system for aiding in correcting defects included in structured text, comprising:
storage; and
a processing unit coupled to the storage and that performs:
receiving an input of a correction for a defect;
calculating a degree of similarity between the defect and remaining defects;
calculating, based on the degree of similarity, a degree of confidence indicating a degree as to whether or not the correction is applicable to the remaining defects, wherein the structured text is HTML text defined by Document Object Model (DOM), and the degree of confidence conf(m,d,d′) is calculated by an equation (1):
conf(m,d,d′)=ssim(elems(d,m),elems(d′,m))·dsim(d,d′) (Equation 1)
where
conf(m,d,d′) is a degree of confidence indicating a degree as to whether or not a correction m for a defect included in a page denoted by DOM d is also applicable to a defect in a page denoted by DOM d′,
dsim(d,d′) is a degree of similarity between the page denoted by DOM d and the page denoted by DOM d′,
elems(d,m) and elems(d′,m) are element sets identified by the correction m in the page denoted by DOM d and the page denoted by DOM d′, respectively,
s=s(elems(d,m)) and s′=s(elems(d′,m)) are styles of the element sets elems(d,m) and elems(d′,m), respectively, and
ssim(s,s′) is a degree of similarity between the styles; and
selecting the correction as a correction candidate for defects having the degree of confidence equal to or higher than a threshold.

### Claim 21

A computer system for aiding in verifying corrections included in structured text, comprising:
storage; and
a processing unit coupled to the storage and that performs:
calculating a degree of verification indicating a degree to which each of the corrections have been verified;
calculating a degree of similarity for the corrections;
calculating, based on the degree of similarity, a degree of verification contribution indicating a degree to which verification of a correction influences verification of the remaining corrections, wherein the structured text is HTML text defined by a Document Object Model (DOM), and the degree of verification contribution vcond(d) is calculated by an equation (1):
⁢
[
Expression
⁢
⁢
7
]
vcond
⁡
(
d
)
=
∑
d
′
∈
D
⁢
⁢
{
Σconf
⁡
(
m
,
d
′
,
d
)
·
(
1
-
v
⁡
(
d
′
,
m
)
)
}
(
Equation
⁢
⁢
7
)
where
D is a set of target DOM,
v(d′,m) is a degree of verification indicating a degree to which a correction m included in a page denoted by DOM d′ has been verified, and
conf(m,d′,d) is a degree of confidence indicating a degree as to whether or not the correction m included in the page denoted by DOM d′ is also applicable to a defect in a page denoted by DOM d; and
selecting a correction having a high degree of verification contribution as a higher-priority correction.