Compilation and evaluation of the Spanish SatiCorpus 2021 for satire identification using linguistic features and transformers

Abstract

Satirical content on social media is hard to distinguish from real news, misinformation, hoaxes or propaganda when there are no clues as to which medium these news were originally written in. It is important, therefore, to provide Information Retrieval systems with mechanisms to identify which results are legitimate and which ones are misleading. Our contribution for satire identification is twofold. On the one hand, we release the Spanish SatiCorpus 2021, a balanced dataset that contains satirical and non-satirical documents. On the other hand, we conduct an extensive evaluation of this dataset with linguistic features and embedding-based features. All feature sets are evaluated separately and combined using different strategies. Our best result is achieved with a combination of the linguistic features and BERT with an accuracy of 97.405%. Besides, we compare our proposal with existing datasets in Spanish regarding satire and irony.

UMUTeam at SemEval-2022 Task 5: Combining image and textual embeddings for multi-modal automatic misogyny identification

“…The LF included low-level linguistic categories concerning phonetics and syntax's, and high-level features related to semantics and pragmatics, including features proper from figurative language (del Pilar Salas-Zárate et al, 2020). Moreover, these kinds of features have proven to be effective for performing other automatic classification tasks such as irony and satire identification (García-Díaz and Valencia-García, 2022). As some of the dictionaries of UMUTextStats are not translated to English, we select a subset of language-independent linguistic features, based on linguistic metrics, Part-of-Speech features and the usage of social media jargon.…”

Smart Analysis of Economics Sentiment in Spanish Based on Linguistic Features and Transformers

“…For compiling the dataset, we relied on the UMUCorpus-Classifier tool [47], developed by our research group. This tool crawls data from Twitter, a social network in which users can send and receive micro-blogging posts of less than 280 characters.…”

“…Following the spirit of our previous works [50], we evaluated several feature sets for conducting opinion mining in the financial domain. Specifically, linguistic features along with contextual and non-contextual pre-trained embeddings were evaluated.…”

Satiric Content Detection Through Linguistic Features

“…Further works try to describe, in more detail, the characteristic of satiric content. In particular, the authors of [4] affirm that satire is characterized by: controversial or sensitive issues, aggressive language -negative emotions and tone -for entertainment purposes, a shorter form with respect to true news even if the words that can be found are more complex, a language not so clear since they are not written by professional journalists, based on imagination and figurative language such as metaphors, similes, personifications, idioms, etc. Moreover, in [5], three language dimensions are considered to better understand satires: the use of first-person singular, which is a proclamation of one's ownership of statements, the more negative words reflect negative emotions, and the use of exclusive words which emphasize a cognitive complexity.…”