Learning with Compositional Semantics as Structural Inference for Subsentential Sentiment Analysis

By Yejin Choi, Claire Cardie. Conference on Empirical Methods in Natural Language Processing, 2008

365 citations recorded by the source corpus.

Determining the polarity of a sentiment-bearing expression requires more than a simple bag-of-words approach. In particular, words or constituents within the expression can interact with each other to yield a particular overall polarity. In this paper, we view such subsentential interactions in light of compositional semantics, and present a novel learning-based approach that incorporates structural inference motivated by compositional semantics into the learning procedure. Our experiments show that (1) simple heuristics based on compositional semantics can perform better than learning-based methods that do not incorporate compositional semantics (accuracy of 89.7% vs. 89.1%), but (2) a method that integrates compositional semantics into learning performs better than all other alternatives (90.7%). We also find that "content-word negators", not widely employed in previous work, play an important role in determining expression-level polarity. Finally, in contrast to conventional wisdom, we find that expression-level classification accuracy uniformly decreases as additional, potentially disambiguating, context is considered.

Subfields: Artificial Intelligence

View Paper: https://www.semanticscholar.org/paper/b35e340a4356eb8953b04c17375ffb631da68916

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

Multi-entity Sentiment Scoring (2009, background): “One of the most recent developments in the area - compositional semantics has shown potential for sentence- and expression-level analysis in both logic-oriented [11],[9] and machine learning-oriented [3] paradigms.”
MPQA 3.0: An Entity/Event-Level Sentiment Corpus (2015, background): “, 2007), phrase-level (Choi and Cardie, 2008), aspect-level (Hu and Liu, 2004; Titov and McDonald, 2008), etc.”
United we Stand: Improving Sentiment Analysis by Joining Machine Learning and Rule Based Methods (2010, methodology): “In (Choi and Cardie, 2008) a combined method is presented performing expression level polarity classiﬁ-cation which integrates inference-rules inspired by compositional semantics into learning.”
Sentiment Analysis of Short Informal Texts (2014, background): “A conventional way of addressing negation in sentiment analysis is to reverse the polarity of a word, i.e. change a word’s sentiment score from s to − s (Kennedy & Inkpen, 2005; Choi & Cardie, 2008).”
Semi-Supervised Recursive Autoencoders for Predicting Sentiment Distributions (2011, background): “The work of (Polanyi and Zaenen, 2006; Choi and Cardie, 2008) focuses on manually constructing several lexica and rules for both polar words and related content-word negators, such as “prevent cancer”, where prevent reverses the negative polarity of cancer.”
Compositional Matrix-Space Models for Sentiment Analysis (2011, background): “Choi and Cardie (2008), for example, propose an algorithm for phrase-based sentiment analysis that learns proper assignments of intermediate sentiment analysis decision variables given the a priori (i.e., out of context) polarity of the words in the phrase and the (correct) phrase-level polarity.”
A review of natural language processing techniques for opinion mining systems (2017, background): “Intensification and diminution are increasing and decreasing the strength of sentiment, respectively, which are also useful for opinion mining [40, 65, 41, 66].”
Long Short-Term Memory Over Recursive Structures (2015, methodology): “Following (Socher et al., 2013), we also use the classification accuracy to measure the performances.”
Survey on Aspect-Level Sentiment Analysis (2016, methodology): “Instead of a binary sentiment classifier, as is used in the above two methods [50], [63], a Support Vector Regression model is employed in [20] to find the sentiment score for an aspect.”
Multi-Level Structured Models for Document-Level Sentiment Classification (2010, background): “(2010), Clarke et al. (2010)) and other fields (e.g., Felzenszwalb et al. (2008), Yu and Joachims (2009)) have analyzed joint multilevel models (i.”
Beneath the Tip of the Iceberg: Current Challenges and New Directions in Sentiment Analysis Research (2020, background): “These works include approaches that use named-entity recognition, Kim & Hovy [94], parsing and ranking candidates, Kim&Hovy [95], semantic role labeling, Wiegand & Ruppenhofer [224], structured prediction using CRFs, Choi et al. [24], multi-tasking, Yang & Cardie [233], amongst others.”
Learning Sentence Embeddings with Auxiliary Tasks for Cross-Domain Sentiment Classification (2016, background): “With the growing need of correctly identifying the sentiments expressed in subjective texts such as product reviews, sentiment classification has received continuous attention in the NLP community for over a decade (Pang et al., 2002; Pang and Lee, 2004; Hu and Liu, 2004; Choi and Cardie, 2008; Naka”
Lexicon-Based Methods for Sentiment Analysis (2011, methodology): “We use the same Mechanical Turk judgments as in the previous section, with six Turkers per word or pair.”
Dependency Tree-based Sentiment Classification using CRFs with Hidden Variables (2010, methodology): “Choi and Cardie (2008) categorized polarity reversing words into two categories: function-word negators such as not and content-word negators such as eliminate.”
Challenges in Sentiment Analysis (2017, methodology): “Earlier works on negation handling employ simple heuristics such as ﬂipping the polarity of the words in a negator’s scope (Kennedy & Inkpen, 2005; Choi & Cardie, 2008) or changing the degree of sentiment of the modiﬁed word by a ﬁxed constant (Taboada, Brooke, Toﬁloski, Voll, & Stede, 2011).”