SentiMerge: Combining Sentiment Lexicons in a Bayesian Framework

Abstract

Many approaches to sentiment analysis rely on a lexicon that labels words with a prior polarity. This is particularly true for languages other than English, where labelled training data is not easily available. Existing efforts to produce such lexicons exist, and to avoid duplicated effort, a principled way to combine multiple resources is required. In this paper, we introduce a Bayesian probabilistic model, which can simultaneously combine polarity scores from several data sources and estimate the quality of each source. We apply this algorithm to a set of four German sentiment lexicons, to produce the SentiMerge lexicon, which we make publically available. In a simple classification task, we show that this lexicon outperforms each of the underlying resources, as well as a majority vote model.

Untitled

“…These values are "encoded" into two three-dimensional vectors, which are then summed and added to (1, 1, 1) (not shown) to form the parameters of a Dirichlet over the latent representation of the word's polarity value. such lexica has emerged (Emerson and Declerck, 2014;Altrabsheh et al, 2017), borrowing ideas from crowdsourcing (Raykar et al, 2010;Hovy et al, 2013). However, this is a non-trivial task, because lexica can use binary, categorical, or continuous scales to quantify polarity-in addition to different interpretations for each-and thus cannot easily be combined.…”

“…Our representation is particularly efficacious for datasets from domains that are not well-supported by standard sentiment lexica. 1 The existing research that is most closely related to our work is SentiMerge (Emerson and Declerck, 2014), a Bayesian approach for aligning sentiment lexica with different continuous scales. SentiMerge consists of two steps: (i) aligning the lexica via rescaling, and (ii) combining the rescaled lexica using a Gaussian distribution.…”

DebugSL: An Interactive Tool for Debugging Sentiment Lexicons

“…A tag may be a single value, e.g., positive, or it may be a distribution.The large number of SLs and methods to generate them renders errors and disagreements inevitable (Liu, 2015;Feldman, 2013). Numerous works raise the issue of polarity disagreements between SLs and its negative impact on SA tasks (Potts, 2011;Emerson and Declerck, 2014;Liu, 2015). Schneider and gives examples of SLs that disagree on up to 78% of their annotations and shows that the accuracy of an SA task can improve by 8.5% by correcting a modest number of inconsistencies in an SL.…”

Evaluating the Morphological Compositionality of Polarity

“…Our data mainly consists of lemmas sampled out of the PolArt polarity dictionary (Klenner et al, 2009), a manually curated resource with around 8000 entries, found to be of high precision by (Emerson and Declerck, 2014). Most entries are subjective words but PolArt also includes a few intensifiers (INT) 2 , and shifters (SHI).…”