Information Technology Reference
In-Depth Information
Given product class
C
with instances
I
and corresponding reviews
R
,
opine's goal is to find a set of (feature, opinions) tuples
{
(
f, o
i
, ...o
j
)
}
such
that
f
O
,where:
a)
F
is the set of product class features in
R
.
b)
O
is the set of opinion phrases in
R
.
c)
f
is a feature of a particular product instance.
d)
o
is an opinion about
f
in a particular sentence.
d) the opinions associated with
f
are ranked based on opinion strength.
∈
F
and
o
i
, ...o
j
∈
Input: product class C, reviews R
.
Output: set of [feature, ranked opinion list] tuples
R'
parseReviews(R);
E
←
findExplicitFeatures(R', C);
O
←
findOpinions(R', E);
CO
←
clusterOpinions(O);
I
←
findImplicitFeatures(CO, E);
RO
←
rankOpinions(CO);
{
(
f
,
o
i
, ...o
j
)...
}←
outputTuples(RO, I
∪
E);
Fig. 2.1. OPINE Overview.
←
The steps of our solution are outlined in Figure 2.1 above. opine parses the
reviews using MINIPAR [3] and applies a simple pronoun-resolution module
to the parsed review data. opine then uses the data to find
explicit
prod-
uct features. opine's
Feature Assessor
and its use of Web
Point-wise Mutual
Information
(PMI) statistics are vital for the extraction of high-quality fea-
tures (see 2.3.3). opine then identifies
opinion phrases
associated with explicit
features and finds their polarity. opine's novel use of relaxation labeling tech-
niques for determining the semantic orientation of potential opinion words
in the context of given features and sentences leads to high precision and
recall on the tasks of
opinion phrase extraction
and
opinion phrase polarity
extraction
(see 2.3.5).
Opinion phrases refer to
properties
, which are sometimes
implicit
(
e.g.
,
“tiny phone” refers to the size of the phone). In order to extract implicit
properties, opine first clusters opinion phrases (
e.g.
,
tiny
and
small
will be
placed in the same cluster), automatically labels the clusters with property
names (
e.g.
,
Size
) and uses them to extract
implicit
features (
e.g.
,
PhoneSize
).
The final component of our system is the ranking of opinions which refer to the
same property based on their strength (
e.g.
,
fantastic >
(
almost, great
)
>
good
). Finally, opine outputs a set of (feature, ranked opinions) tuples for
each identified feature.
2.3.1 The KnowItAll System
opine is built on top of KnowItAll, a Web-based, domain-independent infor-
mation extraction system [4]. Given a set of relations of interest, KnowItAll
Search WWH ::
Custom Search