In this contribution, we examine the proficiency of Large Language Models (LLMs) in solving the linguistic game “La Ghigliottina,” the final game of the popular Italian TV quiz show “L’Eredità”. This game is particularly challenging as it requires LLMs to engage in semantic inference reasoning for identifying the solutions of the game. Our experiment draws inspiration from Ghigliottin-AI, a task of EVALITA 2020, an evaluation campaign focusing on Natural Language Processing (NLP) and speech tools designed for the Italian language. To benchmark our experiment, we use the results of the most successful artificial player in this task, namely Il Mago della Ghigliottina. The paper describes the experimental setting and the results which show that LLMs perform poorly.

This paper describes how idiom-related language resources, collected through a crowdsourcing experiment carried out by means of Dodiom, a Game-with-a-purpose, have been analysed by language experts. The paper focuses on the criteria adopted for the data annotation and evaluation process. The main scope of this project is, indeed, the evaluation of the quality of the linguistic data obtained through a crowdsourcing project, namely to assess if the data provided and evaluated by the players who joined the game are actually considered of good quality by the language experts. Finally, results of the annotation and evaluation processes as well as future work are presented.

In this paper, we present a web service platform for disinformation detection in hotel reviews written in English. The platform relies on a hybrid approach of computational stylometry techniques, machine learning and linguistic rules written using COGITO, Expert System Corp.’s semantic intelligence software thanks to which it is possible to analyze texts and extract all their characteristics. We carried out a research experiment on the Deceptive Opinion Spam corpus, a balanced corpus composed of 1,600 hotel reviews of 20 Chicago hotels split into four datasets: positive truthful, negative truthful, positive deceptive and negative deceptive reviews. We investigated four different classifiers and we detected that Simple Logistic is the most performing algorithm for this type of classification.

In this paper, we describe UniOr_ExpSys team participation in TRAC-2 (Trolling, Aggression and Cyberbullying) shared task, a workshop organized as part of LREC 2020. TRAC-2 shared task is organized in two sub-tasks: Aggression Identification (a 3-way classification between “Overtly Aggressive”, “Covertly Aggressive” and “Non-aggressive” text data) and Misogynistic Aggression Identification (a binary classifier for classifying the texts as “gendered” or “non-gendered”). Our approach is based on linguistic rules, stylistic features extraction through stylometric analysis and Sequential Minimal Optimization algorithm in building the two classifiers.

In this paper we describe the system submitted to the ELEXIS Monolingual Word Sense Alignment Task. We test different systems,which are two types of LSTMs and a system based on a pretrained Bidirectional Encoder Representations from Transformers (BERT)model, to solve the task. LSTM models use fastText pre-trained word vectors features with different settings. For training the models,we did not combine external data with the dataset provided for the task. We select a sub-set of languages among the proposed ones,namely a set of Romance languages, i.e., Italian, Spanish, Portuguese, together with English and Dutch. The Siamese LSTM withattention and PoS tagging (LSTM-A) performed better than the other two systems, achieving a 5-Class Accuracy score of 0.844 in theOverall Results, ranking the first position among five teams.