Automatically Extracting Polarity-Bearing Topics for Cross-Domain Sentiment Classification
He, Yulan and Lin, Chenghua and Alani, Harith

Article Structure

Abstract

Joint sentiment-topic (JST) model was previously proposed to detect sentiment and topic simultaneously from text.

Introduction

Given a piece of text, sentiment classification aims to determine whether the semantic orientation of the text is positive, negative or neutral.

Related Work

There has been significant amount of work on algorithms for domain adaptation in NLP.

Joint Sentiment-Topic (J ST) Model

Assume that we have a corpus with a collection of D documents denoted by C 2 {d1, d2, ..., d D}; each document in the corpus is a sequence of Nd words denoted by d = (2121,2122, ...,de), and each word in the document is an item from a vocabulary index with V distinct terms denoted by {1, 2, ..., V}.

Topics

sentiment classification

Appears in 20 sentences as: sentiment class (1) Sentiment Classification (1) sentiment classification (18) sentiment classifier (1)
In Automatically Extracting Polarity-Bearing Topics for Cross-Domain Sentiment Classification
  1. Furthermore, using feature augmentation and selection according to the information gain criteria for cross-domain sentiment classification , our proposed approach performs either better or comparably compared to previous approaches.
    Page 1, “Abstract”
  2. Given a piece of text, sentiment classification aims to determine whether the semantic orientation of the text is positive, negative or neutral.
    Page 1, “Introduction”
  3. With prior polarity words extracted from both the MPQA subjectivity lexicon1 and the appraisal lexiconz, the J ST model achieves a sentiment classification accuracy of 74% on the movie review data3 and 71% on the multi-domain sentiment dataset4.
    Page 1, “Introduction”
  4. The fact that the J ST model does not required any labeled documents for training makes it desirable for domain adaptation in sentiment classification .
    Page 1, “Introduction”
  5. proposed structural correspondence learning (SCL) for domain adaptation in sentiment classification .
    Page 2, “Related Work”
  6. These observations motivate us to explore polarity-bearing topics extracted by JST for cross-domain sentiment classification since grouping words from different domains but bearing similar sentiment has the effect of overcoming the data distribution difference of two domains.
    Page 4, “Joint Sentiment-Topic (J ST) Model”
  7. Output: A sentiment classifier for the target domain Pt 1: Merge D8 and Pt with document labels discarded, D: {(51331 3 ng Nfimfwl 3 ng Nt}
    Page 5, “Joint Sentiment-Topic (J ST) Model”
  8. Input: The target domain data, 13’5 = E X : 1 g n g N75, N75 > N5}, document sentiment classification threshold 7' Output: A labeled document pool 13 1: Train a J ST model parameterized by A on D75 2: for each document 533,5, 6 D75 do 3: Infer its sentiment class label from JST as ln = arg max, P(l|:cf,; A)
    Page 5, “Joint Sentiment-Topic (J ST) Model”
  9. 6.2 Supervised Sentiment Classification
    Page 5, “Joint Sentiment-Topic (J ST) Model”
  10. We performed 5-fold cross validation for the performance evaluation of supervised sentiment classification .
    Page 5, “Joint Sentiment-Topic (J ST) Model”
  11. Table 2: Supervised sentiment classification accuracy.
    Page 6, “Joint Sentiment-Topic (J ST) Model”

See all papers in Proc. ACL 2011 that mention sentiment classification.

See all papers in Proc. ACL that mention sentiment classification.

Back to top.

domain adaptation

Appears in 15 sentences as: Domain Adaptation (2) Domain adaptation (2) domain adaptation (11)
In Automatically Extracting Polarity-Bearing Topics for Cross-Domain Sentiment Classification
  1. The fact that the J ST model does not required any labeled documents for training makes it desirable for domain adaptation in sentiment classification.
    Page 1, “Introduction”
  2. We proceed with a review of related work on sentiment domain adaptation .
    Page 2, “Introduction”
  3. We subsequently show that words from different domains can indeed be grouped under the same polarity-bearing topic through an illustration of example topic words extracted by J ST before proposing a domain adaptation approach based on J ST. We verify our proposed approach by conducting experiments on both the movie review data
    Page 2, “Introduction”
  4. There has been significant amount of work on algorithms for domain adaptation in NLP.
    Page 2, “Related Work”
  5. for domain adaptation where a mixture model is defined to learn differences between domains.
    Page 2, “Related Work”
  6. proposed structural correspondence learning (SCL) for domain adaptation in sentiment classification.
    Page 2, “Related Work”
  7. There has also been research in exploring careful structuring of features for domain adaptation .
    Page 2, “Related Work”
  8. 5 Domain Adaptation using J ST
    Page 4, “Joint Sentiment-Topic (J ST) Model”
  9. Given input data cc and a class label 3/, labeled patterns of one domain can be drawn from the joint distribution P(:c, y) = Domain adaptation usually assume that data distribution are different in source and target domains, i.e., Ps(x) 75 Pt(:c).
    Page 4, “Joint Sentiment-Topic (J ST) Model”
  10. The task of domain adaptation is to predict the label corresponding to in the target domain.
    Page 4, “Joint Sentiment-Topic (J ST) Model”
  11. shows how to perform domain adaptation using the J ST model.
    Page 4, “Joint Sentiment-Topic (J ST) Model”

See all papers in Proc. ACL 2011 that mention domain adaptation.

See all papers in Proc. ACL that mention domain adaptation.

Back to top.

LDA

Appears in 12 sentences as: LDA (13)
In Automatically Extracting Polarity-Bearing Topics for Cross-Domain Sentiment Classification
  1. was extended from the latent Dirichlet allocation ( LDA ) model (?)
    Page 1, “Introduction”
  2. It is worth pointing out that the J ST model with single topic becomes the standard LDA model with only three sentiment topics.
    Page 6, “Joint Sentiment-Topic (J ST) Model”
  3. that the J ST model with word polarity priors incorporated performs significantly better than the LDA model without incorporating such prior information.
    Page 6, “Joint Sentiment-Topic (J ST) Model”
  4. For comparison purpose, we also run the LDA model and augmented the BOW features with the
    Page 6, “Joint Sentiment-Topic (J ST) Model”
  5. Baseline 82.53 79.96 81.32 83.61 85.82 LDA 83.76 84.32 85.62 85.4 87.68
    Page 6, “Joint Sentiment-Topic (J ST) Model”
  6. The best accuracy was obtained when the number of topics is set to 15 in the LDA model.
    Page 6, “Joint Sentiment-Topic (J ST) Model”
  7. We have performed significance test and found that LDA performs statistically significant better than Baseline according to a paired t-test with p < 0.005 for the Kitchen domain and with p < 0.001 for all the other domains. ]
    Page 6, “Joint Sentiment-Topic (J ST) Model”
  8. ST performs statistically significant better than both Baseline and LDA with p < 0.001.
    Page 6, “Joint Sentiment-Topic (J ST) Model”
  9. performs worse than either LDA or J ST feature augmentation.
    Page 6, “Joint Sentiment-Topic (J ST) Model”
  10. LDA results were generated from an ME classifier trained on document vectors augmented with topics generated from the LDA model.
    Page 7, “Joint Sentiment-Topic (J ST) Model”
  11. that LDA only improves slightly compared to the baseline with an error reduction of 11%.
    Page 7, “Joint Sentiment-Topic (J ST) Model”

See all papers in Proc. ACL 2011 that mention LDA.

See all papers in Proc. ACL that mention LDA.

Back to top.

in-domain

Appears in 8 sentences as: in-domain (8)
In Automatically Extracting Polarity-Bearing Topics for Cross-Domain Sentiment Classification
  1. We study the polarity-bearing topics extracted by J ST and show that by augmenting the original feature space with polarity-bearing topics, the in-domain supervised classifiers learned from augmented feature representation achieve the state-of-the-art performance of 95% on the movie review data and an average of 90% on the multi-domain sentiment dataset.
    Page 1, “Abstract”
  2. We study the polarity-bearing topics extracted by the JST model and show that by augmenting the original feature space with polarity-bearing topics, the performance of in-domain supervised classifiers learned from augmented feature representation improves substantially, reaching the state-of-the-art results of 95% on the movie review data and an average of 90% on the multi-domain sentiment dataset.
    Page 2, “Introduction”
  3. The adaptation loss is calculated with respect to the in-domain gold standard classification result.
    Page 7, “Joint Sentiment-Topic (J ST) Model”
  4. For example, the in-domain goal standard for the Book domain is 79.96%.
    Page 7, “Joint Sentiment-Topic (J ST) Model”
  5. Table 3: Adaptation loss with respect to the in-domain gold standard.
    Page 7, “Joint Sentiment-Topic (J ST) Model”
  6. The thick horizontal lines are in-domain sentiment classification accuracies.
    Page 7, “Joint Sentiment-Topic (J ST) Model”
  7. It is worth noting that our in-domain results are slightly different from those reported in (?
    Page 7, “Joint Sentiment-Topic (J ST) Model”
  8. In this paper, we have studied polarity-bearing topics generated from the J ST model and shown that by augmenting the original feature space with polarity-bearing topics, the in-domain supervised classifiers learned from augmented feature representation achieve the state-of-the-art performance on both the movie review data and the multi-domain sentiment dataset.
    Page 8, “Joint Sentiment-Topic (J ST) Model”

See all papers in Proc. ACL 2011 that mention in-domain.

See all papers in Proc. ACL that mention in-domain.

Back to top.

feature space

Appears in 5 sentences as: feature space (5)
In Automatically Extracting Polarity-Bearing Topics for Cross-Domain Sentiment Classification
  1. We study the polarity-bearing topics extracted by J ST and show that by augmenting the original feature space with polarity-bearing topics, the in-domain supervised classifiers learned from augmented feature representation achieve the state-of-the-art performance of 95% on the movie review data and an average of 90% on the multi-domain sentiment dataset.
    Page 1, “Abstract”
  2. We study the polarity-bearing topics extracted by the JST model and show that by augmenting the original feature space with polarity-bearing topics, the performance of in-domain supervised classifiers learned from augmented feature representation improves substantially, reaching the state-of-the-art results of 95% on the movie review data and an average of 90% on the multi-domain sentiment dataset.
    Page 2, “Introduction”
  3. proposed a kemel-mapping function which maps both source and target domains data to a high-dimensional feature space so that data points from the same domain are twice as similar as those from different domains.
    Page 2, “Related Work”
  4. In this paper, we have studied polarity-bearing topics generated from the J ST model and shown that by augmenting the original feature space with polarity-bearing topics, the in-domain supervised classifiers learned from augmented feature representation achieve the state-of-the-art performance on both the movie review data and the multi-domain sentiment dataset.
    Page 8, “Joint Sentiment-Topic (J ST) Model”
  5. First, polarity-bearing topics generated by the J ST model were simply added into the original feature space of documents, it is worth investigating attaching different weight to each topic
    Page 8, “Joint Sentiment-Topic (J ST) Model”

See all papers in Proc. ACL 2011 that mention feature space.

See all papers in Proc. ACL that mention feature space.

Back to top.

sentiment lexicon

Appears in 4 sentences as: sentiment lexicon (4)
In Automatically Extracting Polarity-Bearing Topics for Cross-Domain Sentiment Classification
  1. The previously proposed J ST model uses the sentiment prior information in the Gibbs sampling inference step that a sentiment label will only be sampled if the current word token has no prior sentiment as defined in a sentiment lexicon .
    Page 2, “Introduction”
  2. For each word 21) E {1, ..., V}, if w is found in the sentiment lexicon , for each I E {1, ..., S}, the element Alw is updated as follows
    Page 3, “Joint Sentiment-Topic (J ST) Model”
  3. where the function 8 returns the prior sentiment label of w in a sentiment lexicon , i.e.
    Page 3, “Joint Sentiment-Topic (J ST) Model”
  4. The MPQA subjectivity lexicon is used as a sentiment lexicon in our experiments.
    Page 5, “Joint Sentiment-Topic (J ST) Model”

See all papers in Proc. ACL 2011 that mention sentiment lexicon.

See all papers in Proc. ACL that mention sentiment lexicon.

Back to top.

Gibbs sampling

Appears in 3 sentences as: Gibbs sampling (3)
In Automatically Extracting Polarity-Bearing Topics for Cross-Domain Sentiment Classification
  1. The previously proposed J ST model uses the sentiment prior information in the Gibbs sampling inference step that a sentiment label will only be sampled if the current word token has no prior sentiment as defined in a sentiment lexicon.
    Page 2, “Introduction”
  2. Gibbs sampling was used to estimate the posterior distribution by sequentially sampling each variable of interest, 2,; and It here, from the distribution over
    Page 3, “Joint Sentiment-Topic (J ST) Model”
  3. In our experiment, a was updated every 25 iterations during the Gibbs sampling procedure.
    Page 5, “Joint Sentiment-Topic (J ST) Model”

See all papers in Proc. ACL 2011 that mention Gibbs sampling.

See all papers in Proc. ACL that mention Gibbs sampling.

Back to top.

PDF available from www.aclweb.org.
Send us an email