Web Content Classification based on Topic and Sentiment Analysis
of Text
Shuhua Liu and Thomas Forss
Arcada University of Applied Sciences, Jan-Magnus Janssonin aukio 1, 00560 Helsinki, Finland
Keywords: Web Content Classification, Text Summarization, Topic Similarity, Sentiment Analysis, Online Safety
Solutions.
Abstract: Automatic classification of web content has been studied extensively, using different learning methods and
tools, investigating different datasets to serve different purposes. Most of the studies have made use of the
content and structural features of web pages. However, previous experience has shown that certain groups
of web pages, such as those that contain hatred and violence, are much harder to classify with good
accuracy when both content and structural features are already taken into consideration. In this study we
present a new approach for automatically classifying web pages into pre-defined topic categories. We apply
text summarization and sentiment analysis techniques to extract topic and sentiment indicators of web
pages. We then build classifiers based on combined topic and sentiment features. A large amount of
experiments were carried out. Our results suggest that incorporating the sentiment dimension can indeed
bring much added value to web content classification. Topic similarity based classifiers solely did not
perform well, but when topic similarity and sentiment features are combined, the classification model
performance is significantly improved for many web categories. Our study offers valuable insights and
inputs to the development of web detection systems and Internet safety solutions.
1 INTRODUCTION
Web content classification, also known as web
content categorization, is the process of assigning
one or more predefined category labels to a web
page. Classification models are built through
training and validation using a set of labeled data,
and are then applied to label new web pages, or in
other words, to detect if a new webpage falls into
certain predefined categories.
Automatic classification of web pages has been
studied extensively, using different learning methods
and tools, investigating different datasets to serve
different purposes (Qi and Davidson, 2007).
However, practical experience has shown that
certain groups of web pages, such as those
containing hatred and violence, are much harder to
classify with good accuracy even when both content
and structural features are already taken into
consideration. There is a great need for better
content detection systems that can accurately
identify excessively offensive and harmful websites.
Hate and violence related web pages often carry
strong negative sentiment while their topics may
vary a lot. In the meantime, advanced developments
in computing methodologies and technology have
brought us many new and better means for text
content analysis such as topic extraction, topic
modeling and sentiment analysis. In our research we
set out to explore the effectiveness of combined
topic and sentiment features for improving
automatic classification of web content. We present
a new approach for automatically classifying web
pages into pre-defined topic categories, which first
applies text summarization and sentiment analysis
techniques to extract topic and sentiment indicators
of web pages, and then builds classifiers based on
the extracted topic and sentiment features. Our
results offer valuable insights and inputs to the
development of web detection systems and Internet
safety solutions.
The rest of the paper is organized as follows: in
Section 2, we give an overview of related research in
web content classification and Internet safety and
security solutions. In Section 3, we describe our
approach and explain the methods and techniques
used in topic extraction and sentiment analysis. In
Section 4, we describe our data and three sets of
300
Liu S. and Forss T..
Web Content Classification based on Topic and Sentiment Analysis of Text.
DOI: 10.5220/0005101803000307
In Proceedings of the International Conference on Knowledge Discovery and Information Retrieval (KDIR-2014), pages 300-307
ISBN: 978-989-758-048-2
Copyright
c
2014 SCITEPRESS (Science and Technology Publications, Lda.)
experiments including solely sentiment-based
classifiers on three web categories, classifiers based
on combined topic and sentiment features while
extending our work first to eight web categories,
then further to all web categories. Our results are
presented and analyzed. Section 5 concludes the
paper.
2 RELATED RESEARCH
Earliest studies on web classification already
appeared in the late 1990s soon after the web was
invented. Chakrabarti et al (1998) studied hypertext
categorization using hyperlinks. Cohen (2002)
combined anchor extraction with link analysis to
improve web page classifiers. The method exploits
link structure within a site as well as page structure
within hub pages, and it brought substantial
improvement on the accuracy of a bag-of-words
classifier, reducing error rate by about half on
average (Cohen, 2002).
Dumais and Chen (2000) explored the use of
hierarchical structure for classifying a large,
heterogeneous collection of web content. They
applied SVM classifiers in the context of
hierarchical classification and found small
advantages in accuracy for hierarchical models over
flat (non-hierarchical) models. They also found the
same accuracy using a sequential Boolean decision
rule and a multiplicative decision rule, with higher
efficiency. Yu et al (2004) presented a framework
for web page classification without negative
examples, called Positive Example Based Learning
(PEBL), to address issues related with the
acquisition of negative training examples and to
avoid bias. They found that given the same set of
positive examples, their Mapping-Convergence
algorithm outperforms one-class SVMs, and was
almost as accurate as the traditional SVMs.
There is a huge amount of research on text
classification in general. However, web content
classification differs from general text categorization
due to its special structure, meta-data and its
dynamics. Shen et al (2004, 2007) studied web-page
classification based on text summarization. They
gave empirical evidence that web-page summaries
created manually by human editors can indeed
improve the performance of web-page classification
algorithms. They proposed sentence-based
summarization methods and showed that their
summarization-based classification algorithm
achieves an approximately 8.8% improvement as
compared to pure-text-based classification
algorithm, and an ensemble classifier using the
improved summarization algorithm achieves about
12.9% improvement over pure-text-based methods.
Our approach differs in that we take a word-based
instead of a sentence-based approach.
In recent years, there have been many studies on
text classification techniques for social media
analysis (e.g. customer reviews, twitter), sentiment
analysis, etc. For example, an interesting study by
Zhang et al (2013) investigated the classification of
short texts using an information path to deal with the
less informative word co-occurrences and sparseness
with such texts. Their method makes use of ordered
subsets in short texts, which is termed “information
path”. They found the classification based on each
subset to result in higher overall accuracy than
classifying the entire data set directly.
Related with online safety solutions, Hammami
et al (2003) developed a web filtering system
WebGuard that focuses on automatically detecting
and filtering adult content on the Web. It combines
the textual content, image content, and URL of a
web page to construct its feature vector, and classify
a web page into two classes: Suspect and Normal.
The suspect URLs are stored in a database, which is
constantly and automatically updated in order to
reflect the highly dynamic evolution of the Web.
Last et al (2003) and Elovici et al (2005)
developed systems for anomaly detection and
terrorist detection on the Web using content-based
methods. Web content is used as the audit
information provided to the detection system to
identify abnormal activities. The system learns the
normal behavior by applying an unsupervised
clustering algorithm to the content of web pages
accessed by a normal group of users and computes
their typical interests. The content models of normal
behavior are then used in real-time to reveal
deviation from normal behavior at a specific location
on the web (Last et al, 2003). They system can thus
monitor the traffic emanating from the monitored
group of users, issues an alarm if the access
information is not within the typical interests of the
group, and tracks down suspected terrorists by
analyzing the content of information they access
(Elovici et al, 2005).
In more recent years, Calado et al (2006) studied
link-based similarity measures as well as a
combination with text-based similarity metrics for
the classification of web documents for Internet
safety and anti-terrorism applications (Calado et al.
2006). Qi and Davidson (2007) presented a survey
of features and algorithms in the space of web
content classification.
WebContentClassificationbasedonTopicandSentimentAnalysisofText
301
3 WEB CONTENT
CLASSIFICATION BASED ON
TOPIC AND SENTIMENT
ANALYSIS
Our approach to web content classification is
illustrated in Figure 1. Exploring the textual
information, we apply word weighting, text
summarization and sentiment analysis techniques to
extract topic features, content similarity features and
sentiment indicators of web pages to build
classifiers.
In this study we only take into consideration the
page attributes that are text-related. Our focus is on
added value to web classification that can be gained
from textual content analysis. We should point out
that structural features and hyperlink information
capture the design elements of web pages that may
also serve as effective indicators of their content
nature and category (Cohen, 2002). They contain
very useful information for web classification. In
addition, analysis of images contained in a web page
would provide another source of useful information
for web classification (Chen et al, 2006; Kludas,
2007). However, these topics are dealt with in other
projects.
3.1 Topic Extraction and Text
Summarization
The Topic Extraction step takes web textual
information as input and generates a set of topic
terms. We start with extracting topics from each web
page and then each of the collections of web pages
belonging to same categories. The extracted topics
hopefully give a good representation of the core
content of a web page or a web category.
Text summarization tools have the capability to
distil the most important content from text
documents. However, most of the text
summarization systems are concerned with sentence
extraction targeted for human users (Radev et al,
2004a, 2004b). The summarization approach that
Shen et al (2004, 2007) took is also sentence based.
To help web content classification, we believe
simple term extraction could already be a
sufficiently effective and more efficient approach, as
the extracted content (terms) are only used as cues
for classifying the content instead of presenting it to
human users. Thus, in this study, we used the time-
tested tf-idf weighting method (Salton and Buckley,
1988) to extract topic terms from web pages and
their collections.
3.1.1 Topic Extraction for Individual Pages
For each webpage, we make use of its different
content attributes (full page or meta-content) as
input. By applying different compression rates, we
obtained different sets of topic words (for example
top 30, top 50, top 20%, 35%, 50%, 100%).
3.1.2 Topic Extraction for Web Categories
The topic terms of a web category are obtained
through summarization of the collection of all web
pages in the same category. For each collection, we
apply the Centroid method of the MEAD
summarization tool (Radev et al, 2004a; 2004b) to
create summaries of it. Through this we try to
extract topics that are a good representation of a
specific web category. The Centroid method has
been a benchmarking text summarization method.
Given a document or a collection of documents to be
summarized, it creates a cluster and all sentences in
the cluster are represented using a tf-idf weighted
vector space model. A pseudo sentence, which is the
average of all sentences in the cluster, is then
calculated. This pseudo sentence is regarded as the
centroid of the document (cluster); it is a set of the
most important/informative words of the whole
cluster, and thus can be regarded as the best
representation of the entire document collection. By
applying different compression rates, different sets
Figure 1: Web content classification based on topic and sentiment analysis.
KDIR2014-InternationalConferenceonKnowledgeDiscoveryandInformationRetrieval
302
of topic terms can be obtained for each category. In
our case, we try to match up the number of extracted
terms for each web category with the number of
extracted terms for each web page.
3.2 Sentiment Feature Extraction
Sentiment analysis is the process of automatic
extraction and assessment of sentiment-related
information from text. Sentiment analysis has been
applied widely in extracting opinions from product
reviews, discovering an affective dimension of the
social web (Pang and Lee, 2008; Thelwall et al,
2010; Liu, 2012).
Sentiment analysis methods generally fall into
two categories: (1) the lexical approach –
unsupervised, use direct indicators of sentiment, i.e.
sentiment bearing words; (2) the learning approach –
supervised, classification based algorithms, exploit
indirect indicators of sentiment that can reflect genre
or topic specific sentiment patterns. Performance of
supervised methods and unsupervised methods vary
depending on text types (Thelwall et al, 2012).
SentiStrength (Thelwall et al, 2010, 2012) takes
a lexical approach to sentiment analysis, making use
of a combination of sentiment lexical resources,
semantic rules, heuristic rules and additional rules. It
contains an EmotionLookupTable of 2,310
sentiment words and word stems taken from the
Linguistic Inquiry and Word Count (LIWC)
program (Pennebaker et al, 2003), the General
Inquirer list of sentiment terms (Stone et al, 1966)
and ad-hoc additions made during testing of the
system. The SentiStrength algorithm has been tested
on several social web data sets such as MySpace,
Twitter, YouTube, Digg, Runners World, BBC
Forums. It was found to be robust enough to be
applied to a wide variety of social web contexts.
While most opinion mining algorithms attempt to
identify the polarity of sentiment in text – positive,
negative or neutral, SentiStrength gives sentiment
measurement to both positive and negative
directions with the strength of the sentiment
expressed on different scales. To help web content
classification, we use sentiment features to get a
grasp of the sentiment tone of a web page. This is
different from the sentiment of opinions concerning
a specific entity, as in traditional opinion mining
literature.
As a starting point, we apply an unsupervised
approach with the original SentiStrength system. For
each web page, sentiment features are extracted by
using the key topic terms obtained from the topic
extraction process as input to SentiStrength. This
gives sentiment strength value for each web page in
the range of -5 to +5, with -5 indicating strong
negative sentiment and +5 indicating strong positive
sentiment. We tested with different selections of key
topic terms: top 20%, 30%, 50%, and 100% (which
represents the raw text after stop-word removing).
3.3 Page Vs. Category Topic Similarity
We use topic similarity to measure the content
similarity between a web page and a web category.
There are several different approaches for text
similarity analysis: (1) Lexical/word based similarity
analysis making use of hierarchical lexical resources
such as WordNet – words are considered similar if
they are synonyms, antonyms, used in the same way,
used in the same context, or one is a type of another;
(2) the vector space model and Cosine similarity
analysis; (3) corpus based word semantic similarity
analysis by SVD (Singular Value Decomposition)
supported Latent Semantic Analysis methods
(Landauer et al 1998; Laudauer and Dumais, 2008);
(4) explicit semantic analysis: using external
resources such as Wikipedia concepts to define
vector space (Gabrilovich and Markovich 2007); (5)
Language model based similarity measures; (6)
graph based similarity analysis.
Our web page-category similarity is simply
implemented as the cosine similarity between topic
terms of a web page and topic terms of each web
category. The Cosine similarity measure is generic
and robust. We consider it as a good starting choice
for our purpose.
4 DATA AND EXPERIMENTS
Our dataset is a collection of over 165,000 web
pages in 20 categories, with each web page only
labeled with one class (instead of multi-class). Each
webpage is represented by a total of 31 attributes
including full page, URL, Title and other meta-
content, plus structural info and link information.
Taking into account the missing entries for
different attributes, we selected a subset of the
content features as the raw data for our study. In our
first round of experiments (Section 4.1) we utilized
full-page free text content. In our second and third
round of experiments (Section 4.2 and Section 4.3),
we used mainly the textual meta-content of web
pages including URL words, title words, and several
meta-description words (CobraMetaDescription,
CobraMetaKeywords, TagTextA and
TagTextMetaContent).
WebContentClassificationbasedonTopicandSentimentAnalysisofText
303
In the following, we report three sets of
experiments and results of our study. In our first set
of experiments, we focused on the three most
problematic web categories i.e. Hate, Violence and
Racism, and investigated classification models
solely based on web pages’ sentiment features. In
our second set of experiments, we extend our
modeling work into eight web categories (including
Hate, Violence and Racism), developed
classification models based on combined topic
similarity and sentiment features of the web pages.
In our third set of experiments, we expand our work
further to include twenty web categories and built
classification models based on combined topic
similarity and sentiment features.
4.1 Sentiment based Classifier for
Detecting Hate, Violence and
Racism Web Content
To build classifiers for Hate, Violence and Racism
web content, we sampled three datasets from the full
database. The datasets contain training data with
balanced positive and negative examples for the
three web categories. Each dataset makes maximal
use of the positive examples available. Negative
samples are distributed evenly in the other 19 web
categories.
Through some quick exploration, we found
negative sentiment strength a better discriminator of
web content than positive sentiment strength at least
for the three web categories Hate, Violence and
Racism. Thus, in our first set of experiments we only
used negative sentiment strength value as data for
learning and prediction. Here strong negative
sentiment strength (e.g. -3, -4, -5) represents
something with “bad sentiment" instead of "lack of
sentiment". Lack of sentiment is when the sentiment
strength is on the weak side (e.g. -1, -2).
Corresponding to the six sets of topic words for
each web page, six sentiment features are obtained.
Features for learning include a number of negative
sentiment strength values of each web page, based
on the different sets of topic terms (top 30, top 50,
top 20%, 35%, 50% and 100%). Eventually we
found that 30% cut-off seems to be a good level, as
the average sentiment strength for each of the web
categories starts to have larger variations and would
not be at its best as a distinctive indicator for
different web categories.
We built a classification model using the
NäiveBayes (NB) method with cross validation, as
three binary classifiers: c = 1, belongs to the
category, (Violence, Hate, Jew-Racism), c = 0 (does
not belong to the category). NB Classifier is a
simple but highly effective text classification
algorithm that has been shown to perform very well
on language data. It uses the joint probabilities of
features and categories to estimate the probabilities
of categories given a document. Support Vector
Machines (SVM) is another most commonly used
algorithm in classification and foundation for
building highly effective classifiers to achieve
impressive accuracy in text classification. We
experimented with both NB and SVM methods, and
found that they achieved similar results, while SVM
training takes much longer.
We tested with different combinations of the
sentiment features. The best results show fairly good
precision and recall levels for all three categories.
Table 1: Sentiment based NB classifiers.
Model Performance
Category Precision Recall
Hate 71.38% 77.16%
Jew-Racism 63.29% 72.79%
Violence 81.91% 73.92%
4.2 Combining Topic Similarity with
Sentiment Analysis in Web Content
Classification
Following our first set of experiments, we try to find
ways to further improve the classification
performance, while in the meantime extend our
study from 3 to 8 web categories. Eight datasets
were sampled from the full database, in the same
way as described earlier. Each dataset contains
training data with balanced positive and negative
examples for a particular web category. In this round
of experiments we made use of combined metadata
of web pages instead of full page content as the raw
data.
4.2.1 Topic Similarity based Classifier
To derive topic similarity features for each web
page, we first summarize the data collection of each
web category (as described in 3.1.2) and then
calculate the cosine similarity between the topic
vectors of each web page and each web category (as
described in Section 3.3).
For each web category under study, we built
NäiveBayes classification models (with cross
validation) using only topic similarity features. It
turns out that the results were very disappointing for
most categories, low on both precision and recall
measures. We thus conclude that our topic similarity
KDIR2014-InternationalConferenceonKnowledgeDiscoveryandInformationRetrieval
304
based classifiers solely do not perform well.
Next we seek to improve the classification
performance through combined use of topic
similarity and sentiment features. The results are
very encouraging and the classification performance
is significantly improved for most categories.
4.2.2 Extracting New Sentiment Features
For each web page, using the new raw data (meta
content of web pages), we extract topic terms based
on the tf-idf term weighting method and then the
sentiment features using the original SentiSrength as
well as the customized algorithms we defined. We
tried different ways to customize the SentiStrength
algorithm: (1) Counts of the amount of positives and
negative sentiment words in a web page; (2)
NewScale: sum of word SentiStrength value
weighted by word frequency, normalized on total
word counts, value between 5 and -5; (3) update the
EmotionLookupTable. We found only few novel
terms comparing with the original
EmotionLookupTable, so we did not pursue it
further as the effect would be minor. Figure 2 gives
an overview of sentiment strength variations
corresponding to changes of compression rate when
extracting topic terms for each web collection.
Figure 2: Sentiment strength value (NewScale) for each
web category as compression rate changes.
Each category seems to keep their relative sentiment
position rather consistently (with Violence being a
bit different). We tested the new sentiment features
by building NB classifiers for a few web categories.
We found the classifiers to not necessarily perform
better than the earlier sentiment based classifier. The
performance varies from category to category, some
slightly better, some not.
4.2.3 Classification using Combined
Features
Next, we built NäiveBayes models (with cross
validation) for eight web categories, using combined
topic similarity features and sentiment features (as
listed in Table 2). The results are very encouraging
and the model performances are significantly
improved for almost all categories when compared
with solely sentiment based or solely topic similarity
based classifiers, as is shown in Table 3.
Table 2: List of combined features for each web page.
Features of Web Page for Classification Models
Page-Category topic
similarity
Sentiment features
Sim1
to
Sim8
Topic similarity
between a web page
and web category #1
to web category #8
Pos1
to
Pos5
Counts of +1 to +5
SentiStrength values
Note:
In the experiments for
Section 4.3, there will be 20
similarity features Sim1 to
Sim20 as we work on all the
20 web categories.
Neg1
to
Neg5
Counts of -1 to -5
SentiStrength values
NS New scale
Table 3: Classifiers making use of combined sentiment
and topic similarity features (8 web categories).
Model Performance (combined features)
Category Precision Recall
C1: Cults 75.80% 90.55%
C2: Occults 87.08% 91.84%
C3: Racism 98.26% 96.30%
C4: Racist 69.96% 91.82%
C5: JewRel 64.43% 96.28%
C6: Religion 67.01% 92.81%
C7: Violence 93.69% 82.75%
C8: Unknown 89.59% 93.31%
The best performing models are for Category C3
(RacWh) and C7 (Violence), for which both of the
precision levels are over 93% and the recall levels
are also very good. For web category C5 (JewRel)
and C8 (Unknown), the models achieve very good
recall levels (over 96%), but precision is lower. The
C2 model performance also reaches a good level in
terms of a good balance of precision and recall.
4.3 Extending to All Web Categories
Finally, in trying to gather more understanding about
the effect of combined topic and sentiment analysis
on classification performance, we extend the
application of our approach to the other 12 web
categories as well. The results are shown in Table 4.
From the table we can see that, the classification
WebContentClassificationbasedonTopicandSentimentAnalysisofText
305
models perform on a relatively good level only on
two web categories: C9 (Adult) and C15 (Marijuna),
for which the performance level is comparable to the
previous 8 web categories. The recall levels for all
the web categories are quite good, with category
C13 and C20 models achieving the highest recall
level of about 96%. However, the precision levels
are in general much lower than the previous 8 web
categories. This is interesting, and it seems to
confirm that our approach works better on detecting
negative web content than more neutral content.
Table 4: Classifiers making use of combined features
(additional 12 web categories).
Model Performance (combined features)
P: Precision R: Recall
Category
P
R
Categor
y
P
R
C9 (AD) 78.69% 80.24% C15 (M
J
74.90% 87.97%
C10 58.25% 93.48% C16 54.75% 87.56%
C11 57.42% 94.81% C17 62.81% 90.21%
C12 62.96% 86.14% C18 53.57% 93.23%
C13 57.38% 96.43% C19 57.61% 93.24%
C14 64.42% 90.65% C20 60.85% 95.85%
5 CONCLUSIONS
In this research we set out to develop sentiment-
aware web content detection system. We proposed a
new approach for automatically classifying web
pages into pre-defined topic categories. Word
weighting, text summarization and sentiment
analysis techniques are applied to extract topic and
sentiment indicators of web pages. NäiveBayes
classifiers are developed based on the extracted topic
similarity and sentiment features. A large amount of
experiments were carried out.
Overall, our experiment results are very
encouraging and suggest that incorporating the
sentiment dimension can indeed bring much added
value to web content classification. Topic similarity
based classifiers solely do not perform well, but
when topic similarity and sentiment features are
combined, the classification model performance is
significantly improved for many web categories
under study. Our results offer valuable insights and
inputs to the development of web detection systems
and Internet safety solutions.
The main contributions of this paper are: (1)
investigation of a new approach for web content
classification to serve online safety applications,
through an integrated use of term weighting, text
summarization and sentiment analysis methods; (2)
customization of SentiStrength algorithm; (3) large
amount of feature extraction and model developing
experiments contributes to better understanding of
text summarization, sentiment analysis methods, and
the learning models; (4) several sets of analytical
results that directly benefit the development of cyber
safety solutions.
Our future work would include the incorporation
of n-grams and multi-granularity topics,
probabilistic topic models (Blei et al, 2003; Blei,
2012; Lu et al, 2011b), revisiting topic-aware
sentiment lexicons (Lu et al, 2011a), word ontology,
structural features, and fine-tuning the models with
different learning methods. We will also look into
new topic similarity measures and refining language
processing during topic extraction. We believe there
is still much room for improvement and some of
these methods will hopefully help to enhance the
classification performance to a new level. Our goal
will be on improving precision and reducing false
positives.
ACKNOWLEDGEMENTS
This research is supported by the Tekes funded
DIGILE D2I research program, Arcada Research
Foundation, and our industry partner. We thank the
reviewers for their helpful comments.
REFERENCES
Blei, D, Ng, A., and Jordan, M. I. 2003. Latent dirichlet
allocation. Advances in neural information processing
systems. 601-608.
Blei, D. 2012. Probabilistic topic models.
Communications of the ACM, 55(4):77–84, 2012.
Calado, P., Cristo, M., Goncalves, M. A., de Moura, E. S.,
Ribeiro-Neto, B., and Ziviani, N. 2006. Link-based
similarity measures for the classification of web
documents. Journal of the American Society for
Information Science and Technology (57:2), 208-221.
Chakrabarti, S., B. Dom and P. Indyk. 1998. Enhanced
hypertext categorization using hyperlinks.
Proceedings of ACM SIGMOD 1998.
Chen, Z., Wu, O., Zhu, M., and Hu, W. 2006. A novel web
page filtering system by combining texts and images.
Proceedings of the 2006 IEEE/WIC/ACM
International Conference on Web Intelligence, 732–
735. Washington, DC IEEE Computer Society.
Cohen, W. 2002. Improving a page classifier with anchor
extraction and link analysis. In S. Becker, S. Thrun,
and K. Obermayer (Eds.), Advances in Neural
Information Processing Systems (Volume 15,
KDIR2014-InternationalConferenceonKnowledgeDiscoveryandInformationRetrieval
306
Cambridge, MA: MIT Press) 1481–1488.
Dumais, S. T., and Chen, H. 2000. Hierarchical
classification of web content. Proceedings of
SIGIR'00, 256-263.
Elovici, Y., Shapira, B., Last, M., Zaafrany, O., Friedman,
M., Schneider, M., and Kandel, A. 2005. Content-
based detection of terrorists browsing the web using
an advanced terror detection system (ATDS),
Intelligence and Security Informatics (Lecture Notes
in Computer Science Volume 3495, 2005), 244-255.
Gabrilovich, E., and Markovich, S. 2007. Computing
Semantic Relatedness using Wikipedia-based Explicit
Semantic Analysis. Proceedings of the 20th
International Joint Conference on Artificial
Intelligence (IJCAI’07), Hyderabad, India.
Hammami, M., Chahir, Y., and Chen, L. 2003.
WebGuard: web based adult content detection and
filtering system. Proceedings of the IEEE/WIC Inter.
Conf. on Web Intelligence (Oct. 2003), 574 – 578.
Kludas, J. 2007. Multimedia retrieval and classification for
web content, Proc. of the 1st BCS IRSG conference on
Future Directions in Information Access, British
Computer Society Swinton, UK ©2007.
Last, M., Shapira, B., Elovici, Y., Zaafrany, O., and
Kandel, A. 2003. Content-Based Methodology for
Anomaly Detection on the Web. Advances in Web
Intelligence, Lecture Notes in Computer Science (Vol.
2663, 2003), 113-123.
Landauer, T. K., Foltz, P. W., and Laham, D. 1998.
Introduction to Latent Semantic Analysis. Discourse
Processes (25), 259-284.
Landauer, T. K., and Dumais, S. T. 2008. Latent semantic
analysis. Scholarpedia 3(11): 4356.
Liu, B. 2012. Sentiment Analysis and Opinion Mining.
Synthesis Lectures on Human Language Technologies,
Morgan & Claypool Publishers 2012.
Lu, Y., M. Castellanos, U. Dayal, C. Zhai. 2011a.
"Automatic Construction of a Context-Aware
Sentiment Lexicon: An Optimization Approach", In
Proceedings of the 20th international conference on
World wide web (WWW'2011) Pages: 347-356.
Lu, Y., Q. Mei, C. Zhai. 2011b. "Investigating Task
Performance of Probabilistic Topic Models - An
Empirical Study of PLSA and LDA", Information
Retrieval, April 2011, Volume 14, Issue 2, pp 178-
203.
Pang, B., and Lee, L. 2008. Opinion mining and sentiment
analysis. Foundations and Trends in Information
Retrieval 2(1-2), 1-135, July 2008.
Pennebaker, J., Mehl, M., & Niederhoffer, K. 2003.
Psychological aspects of natural language use: Our
words, our selves. Annual review of psychology, 54(1),
547–577.
Qi, X., and Davidson, B. 2007. Web Page Classification:
Features and Algorithms. Technical Report LU-CSE-
07-010, Dept. of Computer Science and Engineering,
Lehigh University, Bethlehem, PA, 18015.
Radev, D., Allison, T., Blair-Goldensohn, S., Blitzer, J.,
Celebi, A., Dimitrov, S., and Zhang, Z. 2004a.
MEAD-a platform for multidocument multilingual text
summarization. Proeedings of the 4
th
LREC
Conference (Lisbon, Portugal, May 2004).
Radev, D., Jing, H., Styś, M., and Tam, D. 2004b.
Centroid-based summarization of multiple documents.
Information Processing and Management (40) 919–
938.
Salton, G., and Buckley, C. 1988. Term-weighting
approaches in automatic text retrieval. Information
Processing and Management, 24(5), 513-523.
Shen, D., Z. Chen, Q. Yang, H. Zeng, B. Zhang, Y. Lu, W.
Ma: Web-page classification through summarization.
SIGIR 2004: 242-249.
Shen, D., Qiang Yang, Zheng Chen: Noise reduction
through summarization for Web-page classification.
Information Processing and Management 43(6): 1735-
1747 (2007).
Stone, P. J., Dunphy, D. C., Smith, M. S., and Ogilvie, D.
M. 1966. The general inquirer: a computer approach
to content analysis. The MIT Press, Cambridge,
Massachusetts, 1966. 651.
Thelwall, M., Buckley, K., Paltoglou, G., Cai, D., and
Kappas, A. 2010. Sentiment strength detection in short
informal text. Journal of the American Society for
Information Science and Technology, 61(12), 2544–
2558.
Thelwall, M., Buckley, K., and Paltoglou, G. 2012.
Sentiment strength detection for the social Web.
Journal of the American Society for Information
Science and Technology, 63(1), 163-173.
Yu, H., Han, J., and Chang, K. C.-C. 2004. PEBL: Web
Page Classification without Negative Examples. IEEE
Trans. on Knowledge and Data Eng. (16:1), 70-81.
Zhang, S., Xiaoming Jin, Dou Shen, Bin Cao, Xuetao
Ding, Xiaochen Zhang: Short text classification by
detecting information path. CIKM 2013: 727-732.
WebContentClassificationbasedonTopicandSentimentAnalysisofText
307
