Citation Context

...ts: (unigrams + bigrams) and (unigrams + bigrams + trigrams). We found that although an increase in classification accuracy is observed for the movie reviews, this is not the case for any of the other datasets (see Table 2). We also examined POS-based n-grams in conjuction with a unigram model and observed a decrease in accuracy across all corpora. This indicates that the syntactic patterns represented by the POS n-gram features do contain information which is more discriminative than unigrams. The most promising results came from a POS-based stopwording approach proposed by Matsumoto et al. ([9]). This approach (which Matsumoto et al. refer to as “word subsequences”) consists of an n-gram model, where terms have Table 3: Most discriminative unigrams, bigrams and trigrams according to Information Gain Ratio for binary classification. Microblogs Blogs Microreviews Reviews 1 ! witherspoon great bad 2 <Url> joaquin boring worst 3 <Topic> reese best stupid witherspoon 4 amazing joaquin terrible boring phoenix 5 . . sharon the best the worst 6 ! ! ledger worst waste 7 ? heath n’t ridiculous ledger 8 ! ! ! heath love wasted 9 love johnny loved awful cash 10 <Topic> ! palestinians ? ? Table ...

Citation Context

... is that of supervised machine learning, for which, many different lexical, syntactic and knowledge-based feature representations have been explored in the literature (Dave et al., 2003; Gamon, 2004; =-=Matsumoto et al., 2005-=-; Ng et al., 2006). However, the use of syntactic features for opinion mining has achieved varied results. In our work, we show that by altering syntactic dependency relation triples in a particular w...

Citation Context

...ossible future work. 2 Related Work Supervised machine learning methods including Support Vector Machines (SVM) are often used in sentiment analysis and shown to be very promising (Pang et al., 2002; =-=Matsumoto et al., 2005-=-; Kudo and Matsumoto, 2004; Mullen and Collier, 2004; Gamon, 2004). One of the advantages of these methods is that a wide variety of features such as dependency trees and sequences of words can easily...

Citation Context

...lso, syntactic features such as the dependency relationship of words and subtrees have been shown to effectively improve the performances of sentiment analysis (Kudo and Matsumoto, 2004; Gamon, 2004; =-=Matsumoto et al., 2005-=-; Ng et al., 2006). While these features are usually employed by data-driven approaches, there are unsupervised approaches for sentiment analysis that make use of a set of terms that are semantically ...

Citation Context

... 3). An empirical evaluation on a widely used sentiment corpus shows an improvement of 1.45 point in accuracy over the baseline resulting from a combination of bag-of-words and high-impact parse features (Section 4). 2 Related Work Our research builds on previous work in the field of sentiment classification and convolution kernels. For sentiment classification, the design of lexical and syntactic features is an important first step. Several approaches propose feature-based learning algorithms for this problem. Pang and Lee (2004) and Dave et al. (2003) represent a document as a bag-of-words; Matsumoto et al., (2005) extract frequently occurring connected subtrees from dependency parsing; Joshi and Penstein-Rose (2009) use a transformation of dependency relation triples; Liu and Seneff (2009) extract adverb-adjective-noun relations from dependency parser output. Previous research has convincingly demonstrated a kernel’s ability to generate large feature sets, which is useful to quickly model new and not well understood linguistic phenomena in machine learning, and has led to improvements in various NLP tasks, including relation extraction (Bunescu and Mooney, 2005a; Bunescu and Mooney, 2005b; Zhang et al....

Citation Context

...thi belong to the Indo-Aryan subgroup of the Indo-European language family and are two widely spoken Indian languages with a speaker population of 450 million and 72 million respectively. 413syntax (=-=Matsumoto et al., 2005-=-; Nakagawa et al., 2010), semantic (Balamurali et al., 2011) and negation (Ikeda et al., 2008) have also been explored for this task. There has been research related to clustering and sentiment analys...