Aspect Extraction through Semi-Supervised Modeling
Mukherjee, Arjun and Liu, Bing

Article Structure

Abstract

Aspect extraction is a central problem in sentiment analysis.

Introduction

Aspect-based sentiment analysis is one of the main frameworks for sentiment analysis (Hu and Liu, 2004; Pang and Lee, 2008; Liu, 2012).

Related Work

There are many existing works on aspect extraction.

Proposed Seeded Models

The standard LDA and existing aspect and sentiment models (ASMs) are mostly governed by the phenomenon called “higher-order co-occurrence” (Heinrich, 2009), i.e., based on how often terms co-occur in different contexts].

Experiments

This section evaluates the proposed models.

Conclusion

This paper studied the issue of using seeds to discover aspects in an opinion corpus.

Topics

topic models

Appears in 8 sentences as: topic modeling (2) Topic models (1) topic models (5)
In Aspect Extraction through Semi-Supervised Modeling
  1. Current methods either extract aspects without categorizing them, or extract and categorize them using unsupervised topic modeling .
    Page 1, “Abstract”
  2. The second type uses statistical topic models to extract aspects and group them at the same time in an unsupervised manner.
    Page 1, “Introduction”
  3. Our models are related to topic models in general (Blei et al., 2003) and joint models of aspects and sentiments in sentiment analysis in specific (e.g., Zhao et al., 2010).
    Page 2, “Introduction”
  4. In recent years, topic models have been used to perform extraction and grouping at the same time.
    Page 2, “Related Work”
  5. Aspect and sentiment extraction using topic modeling come in two flavors: discovering aspect words sentiment wise (i.e., discovering positive and negative aspect words and/or sentiments for each aspect without separating aspect and sentiment terms) (Lin and He, 2009; Brody and Elhadad, 2010, Jo and Oh, 2011) and separately discovering both aspects and sentiments (e.g., Mei et al., 2007; Zhao et al., 2010).
    Page 2, “Related Work”
  6. (2009) stated that one reason is that the objective function of topic models does not always correlate well with human judgments.
    Page 2, “Related Work”
  7. Setting the number of topics/aspects in topic models is often tricky as it is difficult to know the
    Page 6, “Experiments”
  8. Topic models are often evaluated quantitatively using perplexity and likelihood on held-out test data (Blei et al., 2003).
    Page 7, “Experiments”

See all papers in Proc. ACL 2012 that mention topic models.

See all papers in Proc. ACL that mention topic models.

Back to top.

proposed models

Appears in 7 sentences as: proposed models (7)
In Aspect Extraction through Semi-Supervised Modeling
  1. Our experimental results show that the two proposed models are indeed able to perform the task effectively.
    Page 1, “Abstract”
  2. The proposed models are evaluated using a large number of hotel reviews.
    Page 2, “Introduction”
  3. Experimental results show that the proposed models outperform the two baselines by large margins.
    Page 2, “Introduction”
  4. We will show in Section 4 that the proposed models outperform it by a large margin.
    Page 3, “Related Work”
  5. This section evaluates the proposed models .
    Page 5, “Experiments”
  6. Even with this noisy automatically-labeled data, the proposed models can produce good results.
    Page 6, “Experiments”
  7. However, it is important to note that the proposed models are flexible and do not need to have seeds for every aspect/topic.
    Page 6, “Experiments”

See all papers in Proc. ACL 2012 that mention proposed models.

See all papers in Proc. ACL that mention proposed models.

Back to top.

jointly model

Appears in 4 sentences as: Joint modeling (1) joint models (1) jointly model (2)
In Aspect Extraction through Semi-Supervised Modeling
  1. Our models also jointly model both aspects and aspect specific sentiments.
    Page 1, “Introduction”
  2. Our models are related to topic models in general (Blei et al., 2003) and joint models of aspects and sentiments in sentiment analysis in specific (e.g., Zhao et al., 2010).
    Page 2, “Introduction”
  3. First of all, we jointly model aspect and sentiment, while DF-LDA is only for topics/aspects.
    Page 2, “Introduction”
  4. Joint modeling ensures clear separation of aspects from sentiments producing better results.
    Page 2, “Introduction”

See all papers in Proc. ACL 2012 that mention jointly model.

See all papers in Proc. ACL that mention jointly model.

Back to top.

co-occurrence

Appears in 3 sentences as: co-occurrence (2) co-occurrence” (1)
In Aspect Extraction through Semi-Supervised Modeling
  1. The standard LDA and existing aspect and sentiment models (ASMs) are mostly governed by the phenomenon called “higher-order co-occurrence” (Heinrich, 2009), i.e., based on how often terms co-occur in different contexts].
    Page 3, “Proposed Seeded Models”
  2. W1 co-occumng With W2 Wthh in turn co-occurs With W3 denotes a second-order co-occurrence between W1 and W3.
    Page 3, “Proposed Seeded Models”
  3. Thus, having only one seed per seed set will result in sampling that single word whenever that seed set is chosen which will not have the effect of correlating seed words so as to pull other words based on co-occurrence with constrained seed words.
    Page 7, “Experiments”

See all papers in Proc. ACL 2012 that mention co-occurrence.

See all papers in Proc. ACL that mention co-occurrence.

Back to top.

Gibbs sampling

Appears in 3 sentences as: Gibbs sampler (1) Gibbs sampling (2)
In Aspect Extraction through Semi-Supervised Modeling
  1. (2011) relied on user feedback during Gibbs sampling iterations.
    Page 2, “Related Work”
  2. We employ collapsed Gibbs sampling (Griffiths and Steyvers, 2004) for posterior inference.
    Page 4, “Proposed Seeded Models”
  3. The variations in the results are due to the random initialization of the Gibbs sampler .
    Page 8, “Experiments”

See all papers in Proc. ACL 2012 that mention Gibbs sampling.

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

Back to top.

labeled data

Appears in 3 sentences as: labeled data (4)
In Aspect Extraction through Semi-Supervised Modeling
  1. We adopt this method as well but with no use of manually labeled data in training.
    Page 2, “Related Work”
  2. Note that unlike traditional Max-Ent training, we do not need manually labeled data for training (see Section 4 for details).
    Page 5, “Proposed Seeded Models”
  3. Since ME-LDA used manually labeled training data for Max-Ent, we again randomly sampled 1000 terms from our corpus appearing at least 20 times and labeled them as aspect terms or sentiment terms, so this labeled data clearly has less noise than our automatically labeled data .
    Page 6, “Experiments”

See all papers in Proc. ACL 2012 that mention labeled data.

See all papers in Proc. ACL that mention labeled data.

Back to top.

LDA

Appears in 3 sentences as: LDA (3)
In Aspect Extraction through Semi-Supervised Modeling
  1. Existing works are based on two basic models, pLSA (Hofmann, 1999) and LDA (Blei et al., 2003).
    Page 2, “Related Work”
  2. The standard LDA and existing aspect and sentiment models (ASMs) are mostly governed by the phenomenon called “higher-order co-occurrence” (Heinrich, 2009), i.e., based on how often terms co-occur in different contexts].
    Page 3, “Proposed Seeded Models”
  3. DF-LDA adds constraints to LDA .
    Page 5, “Experiments”

See all papers in Proc. ACL 2012 that mention LDA.

See all papers in Proc. ACL that mention LDA.

Back to top.

semi-supervised

Appears in 3 sentences as: Semi-Supervised (1) semi-supervised (2)
In Aspect Extraction through Semi-Supervised Modeling
  1. Semi-Supervised Modeling
    Page 1, “Introduction”
  2. With seeds, our models are thus semi-supervised and need a different formulation.
    Page 2, “Introduction”
  3. In (Lu and Zhai, 2008), a semi-supervised model was proposed.
    Page 2, “Related Work”

See all papers in Proc. ACL 2012 that mention semi-supervised.

See all papers in Proc. ACL that mention semi-supervised.

Back to top.

sentiment analysis

Appears in 3 sentences as: sentiment analysis (4)
In Aspect Extraction through Semi-Supervised Modeling
  1. Aspect extraction is a central problem in sentiment analysis .
    Page 1, “Abstract”
  2. Aspect-based sentiment analysis is one of the main frameworks for sentiment analysis (Hu and Liu, 2004; Pang and Lee, 2008; Liu, 2012).
    Page 1, “Introduction”
  3. Our models are related to topic models in general (Blei et al., 2003) and joint models of aspects and sentiments in sentiment analysis in specific (e.g., Zhao et al., 2010).
    Page 2, “Introduction”

See all papers in Proc. ACL 2012 that mention sentiment analysis.

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

Back to top.

sentiment lexicon

Appears in 3 sentences as: sentiment lexicon (3)
In Aspect Extraction through Semi-Supervised Modeling
  1. To learn the Max-Ent parameters 2 of MESAS, we used the sentiment lexicon 4 of (Hu and Liu, 2004) to automatically generate training data (no manual labeling).
    Page 5, “Experiments”
  2. Of those 1000 terms if they appeared in the sentiment lexicon , they were treated as sentiment terms, else aspect terms.
    Page 6, “Experiments”
  3. Clearly, labeling words not in the sentiment lexicon as aspect terms may not always be correct.
    Page 6, “Experiments”

See all papers in Proc. ACL 2012 that mention sentiment lexicon.

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

Back to top.

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