In Ordinal Classification tasks, items have to be assigned to classes that have a relative ordering, such as “positive”, “neutral”, “negative” in sentiment analysis. Remarkably, the most popular evaluation metrics for ordinal classification tasks either ignore relevant information (for instance, precision/recall on each of the classes ignores their relative ordering) or assume additional information (for instance, Mean Average Error assumes absolute distances between classes). In this paper we propose a new metric for Ordinal Classification, Closeness Evaluation Measure, that is rooted on Measurement Theory and Information Theory. Our theoretical analysis and experimental results over both synthetic data and data from NLP shared tasks indicate that the proposed metric captures quality aspects from different traditional tasks simultaneously. In addition, it generalizes some popular classification (nominal scale) and error minimization (interval scale) metrics, depending on the measurement scale in which it is instantiated.

This paper presents SentiSense, a concept-based affective lexicon. It is intended to be used in sentiment analysis-related tasks, specially in polarity and intensity classification and emotion identification. SentiSense attaches emotional meanings to concepts from the WordNet lexical database, instead of terms, thus allowing to address the word ambiguity problem using one of the many WordNet-based word sense disambiguation algorithms. SentiSense consists of 5,496 words and 2,190 synsets labeled with an emotion from a set of 14 emotional categories, which are related by an antonym relationship. SentiSense has been developed semi-automatically using several semantic relations between synsets in WordNet. SentiSense is endowed with a set of tools that allow users to visualize the lexicon and some statistics about the distribution of synsets and emotions in SentiSense, as well as to easily expand the lexicon. SentiSense is available for research purposes.