Licia Capra

Salvatore Greco, Ke Zhou, Licia Capra et al.

AI regulations are expected to prohibit machine learning models from using sensitive attributes during training. However, the latest Natural Language Processing (NLP) classifiers, which rely on deep learning, operate as black-box systems, complicating the detection and remediation of such misuse. Traditional bias mitigation methods in NLP aim for comparable performance across different groups based on attributes like gender or race but fail to address the underlying issue of reliance on protected attributes. To partly fix that, we introduce NLPGuard, a framework for mitigating the reliance on protected attributes in NLP classifiers. NLPGuard takes an unlabeled dataset, an existing NLP classifier, and its training data as input, producing a modified training dataset that significantly reduces dependence on protected attributes without compromising accuracy. NLPGuard is applied to three classification tasks: identifying toxic language, sentiment analysis, and occupation classification. Our evaluation shows that current NLP classifiers heavily depend on protected attributes, with up to $23\%$ of the most predictive words associated with these attributes. However, NLPGuard effectively reduces this reliance by up to $79\%$, while slightly improving accuracy.

Marzieh Saeidi, Alessandro Venerandi, Licia Capra et al.

In this paper, we investigate whether text from a Community Question Answering (QA) platform can be used to predict and describe real-world attributes. We experiment with predicting a wide range of 62 demographic attributes for neighbourhoods of London. We use the text from QA platform of Yahoo! Answers and compare our results to the ones obtained from Twitter microblogs. Outcomes show that the correlation between the predicted demographic attributes using text from Yahoo! Answers discussions and the observed demographic attributes can reach an average Pearson correlation coefficient of \r{ho} = 0.54, slightly higher than the predictions obtained using Twitter data. Our qualitative analysis indicates that there is semantic relatedness between the highest correlated terms extracted from both datasets and their relative demographic attributes. Furthermore, the correlations highlight the different natures of the information contained in Yahoo! Answers and Twitter. While the former seems to offer a more encyclopedic content, the latter provides information related to the current sociocultural aspects or phenomena.

Giovanni Quattrone, Emilio Ferrara, Pasquale De Meo et al.

Social (or folksonomic) tagging has become a very popular way to describe content within Web 2.0 websites. Unlike taxonomies, which overimpose a hierarchical categorisation of content, folksonomies enable end-users to freely create and choose the categories (in this case, tags) that best describe some content. However, as tags are informally defined, continually changing, and ungoverned, social tagging has often been criticised for lowering, rather than increasing, the efficiency of searching, due to the number of synonyms, homonyms, polysemy, as well as the heterogeneity of users and the noise they introduce. To address this issue, a variety of approaches have been proposed that recommend users what tags to use, both when labelling and when looking for resources. As we illustrate in this paper, real world folksonomies are characterized by power law distributions of tags, over which commonly used similarity metrics, including the Jaccard coefficient and the cosine similarity, fail to compute. We thus propose a novel metric, specifically developed to capture similarity in large-scale folksonomies, that is based on a mutual reinforcement principle: that is, two tags are deemed similar if they have been associated to similar resources, and vice-versa two resources are deemed similar if they have been labelled by similar tags. We offer an efficient realisation of this similarity metric, and assess its quality experimentally, by comparing it against cosine similarity, on three large-scale datasets, namely Bibsonomy, MovieLens and CiteULike.