Linguistic-based Approach for Recognizing Implicit Language in Hate Speech: Exploratory Insights

Abstract

Language, in all its forms, is one of the most comprehensive ways to characterize human societies. By means of the analysis of regular components, either at phonetic, morphological, syntactic or semantic level, human language provides valuable information that can be translated into knowledge in order to represent behavioral patterns. For instance, in web texts, such as the ones posted on Twitter or Facebook, it is quite frequent to find linguistic expressions, such as the following one: “Don’t come here. If you are afraid for your life and you have no place to go, don’t pick this country.” This text could denote an explicit description of the current immigration phenomenon and, likewise, it could connote an implicit content of mockery, aggressiveness, or even hate. Both interpretations are possible, but only one of them is more likely according to the author profiling. This fact stresses out the underlying problem that it is faced in this investigation: Many of our utterances entail two communicative dimensions. The explicit dimension (literal use of language), and the implicit dimension (figurative use of language). Both dimensions are supposed to communicate information thought consciously. In this respect, the most challenging issue for this approach relies on the recognition of the correct communicative dimension profiled by the author in a web text. In this context, this article focuses on analyzing textual information, mainly extracted from Twitter, in order to set a computational framework to differentiate between explicit and implicit language. In particular, we are interested in recognizing figurative uses regarding irony and sarcasm, in order to apply the findings to better understand and prevent social problems related to hate speech.