Context-aware Convolutional Neural Networks for Twitter Sentiment Analysis in Italian

Castellucci, Giuseppe; Croce, Danilo; Basili, Roberto

doi:10.4000/books.aaccademia.2001

Cited by 17 publications

(27 citation statements)

References 2 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…In particular, two binary SVM classifiers (one per subtask) are designed to adopt specific combinations of different kernel functions, each operating over a taskspecific tweet representation. This work extends the modeling proposed in (Castellucci et al, 2014) that was proved to be beneficial within the Irony Detection subtask within SENTIPOLC 2014. The UNITOR system here presented ranked 1 st and 2 nd in the Sarcasm Detection subtask, while it ranked 6 th and 7 th within the Irony Detection subtask.…”

Section: Introductionsupporting

confidence: 58%

“…The aim of this set of features is to capture irony by defining a set of irony-specific features inspired by the work of (Castellucci et al, 2014). Word Space Vector (WS)i sa250-dimensional vector representation of the average semantic meaning of a tweet according to a Word space model.…”

Section: Irony-specific Featuresmentioning

confidence: 99%

“…For each dimension representing a lemma/part-of-speech pair, its weight is computed as the cosine similarity between the word embedding vector of the considered word and the vector obtained from the linear combination of all the other words in the message (WS) 5 . This vector aims at capturing how much odd is the occurrence of a given word in a sentence aiming at capturing its unconventional uses: it should be an indicator of potential ironic mechanisms, as suggested in (Castellucci et al, 2014). Irony Specific BOW(Adjective, Noun, Verb) (ISBOW-A), (ISBOW-S), (ISBOW-V) are three BoW vectors that use the same weighting scheme specified in ISBOW.…”

Section: Irony-specific Featuresmentioning

confidence: 99%

See 2 more Smart Citations

Detecting Hate Speech for Italian Language in Social Media

Santucci¹,

Spina²,

Milani³

et al. 2018

EVALITA Evaluation of NLP and Speech Tools for Italian

View full text Add to dashboard Cite

show abstract

Section: Introductionsupporting

confidence: 58%

Section: Irony-specific Featuresmentioning

confidence: 99%

Section: Irony-specific Featuresmentioning

confidence: 99%

See 1 more Smart Citation

Detecting Hate Speech for Italian Language in Social Media

Santucci¹,

Spina²,

Milani³

et al. 2018

EVALITA Evaluation of NLP and Speech Tools for Italian

View full text Add to dashboard Cite

show abstract

“…48,49 Distributional polarity vectors capture the main usage of words but not their ironic or metaphorical senses. It should be interesting to verify if an approach similar to the one suggested by Castellucci et al 50 could be beneficial. In that work, deviation from standard semantic usages of words provides effective information on the irony detection task.…”

Section: Discussionmentioning

confidence: 99%

Dynamic polarity lexicon acquisition for advanced Social Media analytics

Basili

Croce

Castellucci

2017

International Journal of Engineering Business Management

Self Cite

View full text Add to dashboard Cite

Social media analytics tool aims at eliciting information and knowledge about individuals and communities, as this emerges from the dynamics of interpersonal communications in the social networks. Sentiment analysis (SA) is a core component of this process as it focuses onto the subjective levels of this knowledge, including the agreement/rejection, the perception, and the expectations by which individual users socially evolve in the network. Analyzing user sentiments thus corresponds to recognize subjective opinions and preferences in the texts they produce in social contexts, gather collective evidence across one or more communities, and trace some inferences about the underlying social phenomena. Automatic SA is a complex process, often enabled by hand-coded dictionaries, called polarity lexicons, that are intended to capture the a priori emotional aspects of words or multiword expressions. The development of such resources is an expensive, and, mainly, language and task-dependent process. Resulting polarity lexicons may be inadequate at fully covering Social Media phenomena, which are intended to capture global communities. In the area of SA over Social Media, this article presents an unsupervised and language independent method for inducing large-scale polarity lexicons from a specific but representative medium, that is, Twitter. The model is based on a novel use of Distributional Lexical Semantics methodologies as these are applied to Twitter. Given a set of heuristically annotated messages, the proposed methodology transfers the known sentiment information of subjective sentences to individual words. The resulting lexical resource is a large-scale polarity lexicon whose effectiveness is measured with respect to different SA tasks in English, Italian, and Arabic. Comparison of our method with different Distributional Lexical Semantics paradigms confirms the beneficial impact of our method in the design of very accurate SA systems in several natural languages.

show abstract

“…The UniPI team adopted a deep learning method that required the modeling of individual tweets through both WE and CNN (system named UniPI.2.c by Attardi et al [ 78 ]). The Unitor team took a similar approach to the UniPI team, using an extended representation of tweets with additional features taken from the Distributional Polarity Lexicons in combination with a CNN (systems named Unitor.1.u and Unitor.2.u by Castellucci et al [ 79 ]). The ItaliaNLP team used a SVM learning algorithm paired to an LSTM network based on specific linguistic and semantic feature engineering and existing external resources, such as lexicons specific for sentiment analysis tasks (system named ItaliaNLP.1.c by Cimino and Dell’Orletta [ 80 ]).…”

Section: Background and Related Workmentioning

confidence: 99%

An Effective BERT-Based Pipeline for Twitter Sentiment Analysis: A Case Study in Italian

Pota

Ventura

Catelli

et al. 2020

Sensors

102

View full text Add to dashboard Cite

Over the last decade industrial and academic communities have increased their focus on sentiment analysis techniques, especially applied to tweets. State-of-the-art results have been recently achieved using language models trained from scratch on corpora made up exclusively of tweets, in order to better handle the Twitter jargon. This work aims to introduce a different approach for Twitter sentiment analysis based on two steps. Firstly, the tweet jargon, including emojis and emoticons, is transformed into plain text, exploiting procedures that are language-independent or easily applicable to different languages. Secondly, the resulting tweets are classified using the language model BERT, but pre-trained on plain text, instead of tweets, for two reasons: (1) pre-trained models on plain text are easily available in many languages, avoiding resource- and time-consuming model training directly on tweets from scratch; (2) available plain text corpora are larger than tweet-only ones, therefore allowing better performance. A case study describing the application of the approach to Italian is presented, with a comparison with other Italian existing solutions. The results obtained show the effectiveness of the approach and indicate that, thanks to its general basis from a methodological perspective, it can also be promising for other languages.

show abstract

Context-aware Convolutional Neural Networks for Twitter Sentiment Analysis in Italian

Cited by 17 publications

References 2 publications

Detecting Hate Speech for Italian Language in Social Media

Detecting Hate Speech for Italian Language in Social Media

Dynamic polarity lexicon acquisition for advanced Social Media analytics

An Effective BERT-Based Pipeline for Twitter Sentiment Analysis: A Case Study in Italian

Contact Info

Product

Resources

About