Exploring the fine-grained analysis and automatic detection of irony on Twitter

Abstract

To push the state of the art in text mining applications, research in natural language processing has increasingly been investigating automatic irony detection, but manually annotated irony corpora are scarce. We present the construction of a manually annotated irony corpus based on a fine-grained annotation scheme that allows for identification of different types of irony. We conduct a series of binary classification experiments for automatic irony recognition using a support vector machine (SVM) that exploits a varied feature set and compare this method to a deep learning approach that is based on an LSTM network and (pre-trained) word embeddings. Evaluation on a held-out corpus shows that the SVM model outperforms the neural network approach and benefits from combining lexical, semantic and syntactic information sources. A qualitative analysis of the classification output reveals that the classifier performance may be further enhanced by integrating implicit sentiment information and context-and user-based features.

An Analysis of Figurative Language in Maroon 5 – Nobody’s Love Song Lyrics

“…Besides that, the irony was also included in verse one in line one and line three. Hee et al (2018) argue that irony is a kind of sentence that communicates subtle references. Figurative language, in Irony terminology, typically states something by reversing the sense of the intention that has taken place (Green, 2018;Hee et al, 2018).…”

A transformer-based approach to irony and sarcasm detection

“…In a follow-up study [21], a knowledge-based k-NN classifier was fed with a feature set that captures a wide range of linguistic phenomena (e.g., structural, emotional). Significant results were achieved in [91], were a combination of lexical, semantic and syntactic features passed through an SVM classifier that outperformed LSTM deep neural network approaches. Apart from local content, several approaches claimed that global context may be essential to capture FL phenomena.…”

“…• Several approaches search words on large dictionaries which demand large computational times and can be considered as impractical [76,87]. • Many studies exhaustively preprocess the input texts, including stemming, tagging, emoji processing, etc., that tend to be time-consuming especially in large datasets [52,91]. • Many approaches attempt to create datasets using social media API's to automatically collect data rather than exploiting their system on benchmark datasets, with proven quality.…”

“…• Many approaches attempt to create datasets using social media API's to automatically collect data rather than exploiting their system on benchmark datasets, with proven quality. To this end, it is impossible to be compared and evaluated [52,57,91].…”

Dataset Analysis and Augmentation for Emoji-Sensitive Irony Detection

“…Previous work on irony detection relied on hand-crafted features such as punctuation and smiles (Veale and Hao, 2010) or lexical features, such as gap between rare and common words, intensity of adverbs and adjectives, sentiments, and sentence structure (Barbieri and Saggion, 2014 14 (8.9%) 10 (6.9%) 14 (8.9%) 9 (6.2%) 11 (7.0%) 9 (6.2%) 11 (7.0%) 9 (6.2%) 11 (7.0%) 9 (6.2%) 9 (5.5%) 8 (5.5%) 8 (4.8%) for the same. Machine learning algorithms such as SVMs informed with sentiment features have shown good performance gains in irony detection (Van Hee, 2017, 2018. Some neural network-based methods have been conducted.…”