This paper investigates the appropriate responses that Conversational Agent systems (CAs) should employ when subjected to sexual harassment by users. Previous studies indicate that conventional CAs often respond neutrally or evade such requests. Enhancing the responsiveness of CAs to offensive speech is crucial, as users might carry over these interactions into their social interactions. To address this issue, we selected evaluators to compare a series of responses to sexual harassment from four commercial CAs (Amazon Alexa, Apple Siri, Google Home, and Microsoft Cortana) with alternative responses we realized based on insights from psychological and sociological studies. Focusing on CAs with a female voice, given their increased likelihood of encountering offensive language, we conducted two experiments involving 22 evaluators (11 females and 11 males). In the initial experiment, participants assessed the responses in a textual format, while the second experiment involved the evaluation of responses generated with a synthetic voice exhibiting three different intonations (angry, neutral, and assertive). Results from the first experiment revealed a general preference for the responses we formulated. For the most voted replies, female evaluators exhibited a tendency towards responses with an assertive intent, emphasizing the sexually harassing nature of the request. Conversely, male evaluators leaned towards a more neutral response, aligning with prior findings that highlight gender-based differences in the perception of sexual harassment. The second experiment underscored a preference for assertive responses. The study’s outcomes highlight the need to develop new, educational responses from CAs to instances of sexual harassment, aiming to discourage harmful behavior.

The increasing popularity of natural language processing has led to a race to improve machine learning models that often leaves aside the core study object, the language itself. In this study, we present classification models designed to detect stereotypes related to immigrants, along with both quantitative and qualitative analyses, shedding light on linguistic distinctions in how humans and various models perceive stereotypes. Given the subjective nature of this task, one of the models incorporates the judgments of all annotators by utilizing soft labels. Through a comparative analysis of BERT-based models using both hard and soft labels, along with predictions from GPT-4, we gain a clearer understanding of the linguistic challenges posed by texts containing stereotypes. Our dataset comprises Spanish Twitter posts collected as responses to immigrant-related hoaxes, annotated with binary values indicating the presence of stereotypes, implicitness, and the requirement for conversational context to understand the stereotype. Our findings suggest that both model prediction confidence and inter-annotator agreement are higher for explicit stereotypes, while stereotypes conveyed through irony and other figures of speech prove more challenging to detect than other implicit stereotypes.

In this paper, we focus on the topics of misinformation and racial hoaxes from a perspective derived from both social psychology and computational linguistics. In particular, we consider the specific case of anti-immigrant feeling as a first case study for addressing racial stereotypes. We describe the first corpus-based study for multilingual racial stereotype identification in social media conversational threads. Our contributions are: (i) a multilingual corpus of racial hoaxes, (ii) a set of common guidelines for the annotation of racial stereotypes in social media texts, and a multi-layered, fine-grained scheme, psychologically grounded on the work by Fiske, including not only stereotype presence, but also contextuality, implicitness, and forms of discredit, (iii) a multilingual dataset in Italian, Spanish, and French annotated following the aforementioned guidelines, and cross-lingual comparative analyses taking into account racial hoaxes and stereotypes in online discussions. The analysis and results show the usefulness of our methodology and resources, shedding light on how racial hoaxes are spread, and enable the identification of negative stereotypes that reinforce them.

The growth of social media has brought with it a massive channel for spreading and reinforcing stereotypes. This issue becomes critical when the affected targets are minority groups such as women, the LGBT+ community and immigrants. Although from the perspective of computational linguistics, the detection of this kind of stereotypes is steadily improving, most stereotypes are expressed implicitly and identifying them automatically remains a challenge. One of the problems we found for tackling this issue is the lack of an operationalised definition of implicit stereotypes that would allow us to annotate consistently new corpora by characterising the different forms in which stereotypes appear. In this paper, we present thirteen criteria for annotating implicitness which were elaborated to facilitate the subjective task of identifying the presence of stereotypes. We also present NewsCom-Implicitness, a corpus of 1,911 sentences, of which 426 comprise explicit and implicit racial stereotypes. An experiment was carried out to evaluate the applicability of these criteria. The results indicate that different criteria obtain different inter-annotator agreement values and that there is a greater agreement when more criteria can be identified in one sentence.

In this paper we present InferES - an original corpus for Natural Language Inference (NLI) in European Spanish. We propose, implement, and analyze a variety of corpus-creating strategies utilizing expert linguists and crowd workers. The objectives behind InferES are to provide high-quality data, and at the same time to facilitate the systematic evaluation of automated systems. Specifically, we focus on measuring and improving the performance of machine learning systems on negation-based adversarial examples and their ability to generalize across out-of-distribution topics. We train two transformer models on InferES (8,055 gold examples) in a variety of scenarios. Our best model obtains 72.8% accuracy, leaving a lot of room for improvement. The “hypothesis-only” baseline performs only 2%-5% higher than majority, indicating much fewer annotation artifacts than prior work. We show that models trained on InferES generalize very well across topics (both in- and out-of-distribution) and perform moderately well on negation-based adversarial examples.

This paper presents the main sources of disagreement found during the annotation of the Spanish SFU Review Corpus with negation (SFU ReviewSP -NEG). Negation detection is a challenge in most of the task related to NLP, so the availability of corpora annotated with this phenomenon is essential in order to advance in tasks related to this area. A thorough analysis of the problems found during the annotation could help in the study of this phenomenon.

This paper presents the ADN-Classifier, an Automatic classification system of Spanish Deverbal Nominalizations aimed at identifying its semantic denotation (i.e. event, result, underspecified, or lexicalized). The classifier can be used for NLP tasks such as coreference resolution or paraphrase detection. To our knowledge, the ADN-Classifier is the first effort in acquisition of denotations for nominalizations using Machine Learning. We compare the results of the classifier when using a decreasing number of Knowledge Sources, namely (1) the complete nominal lexicon (AnCora-Nom) that includes sense distictions, (2) the nominal lexicon (AnCora-Nom) removing the sense-specific information, (3) nominalizations’ context information obtained from a treebank corpus (AnCora-Es) and (4) the combination of the previous linguistic resources. In a realistic scenario, that is, without sense distinction, the best results achieved are those taking into account the information declared in the lexicon (89.40% accuracy). This shows that the lexicon contains crucial information (such as argument structure) that corpus-derived features cannot substitute for.

This paper presents AnCora, a multilingual corpus annotated at different linguistic levels consisting of 500,000 words in Catalan (AnCora-Ca) and in Spanish (AnCora-Es). At present AnCora is the largest multilayer annotated corpus of these languages freely available from http://clic.ub.edu/ancora. The two corpora consist mainly of newspaper texts annotated at different levels of linguistic description: morphological (PoS and lemmas), syntactic (constituents and functions), and semantic (argument structures, thematic roles, semantic verb classes, named entities, and WordNet nominal senses). All resulting layers are independent of each other, thus making easier the data management. The annotation was performed manually, semiautomatically, or fully automatically, depending on the encoded linguistic information. The development of these basic resources constituted a primary objective, since there was a lack of such resources for these languages. A second goal was the definition of a consistent methodology that can be followed in further annotations. The current versions of AnCora have been used in several international evaluation competitions

In this paper we present two large-scale verbal lexicons, AnCora-Verb-Ca for Catalan and AnCora-Verb-Es for Spanish, which are the basis for the semantic annotation with arguments and thematic roles of AnCora corpora. In AnCora-Verb lexicons, the mapping between syntactic functions, arguments and thematic roles of each verbal predicate it is established taking into account the verbal semantic class and the diatheses alternations in which the predicate can participate. Each verbal predicate is related to one or more semantic classes basically differentiated according to the four event classes -accomplishments, achievements, states and activities-, and on the diatheses alternations in which a verb can occur. AnCora-Verb-Es contains a total of 1,965 different verbs corresponding to 3,671 senses and AnCora-Verb-Ca contains 2,151 verbs and 4,513 senses. These figures correspond to the total of 500,000 words contained in each corpus, AnCora-Ca and AnCora-Es. The lexicons and the annotated corpora constitute the richest linguistic resources of this kind freely available for Spanish and Catalan. The big amount of linguistic information contained in both resources should be of great interest for computational applications and linguistic studies. Currently, a consulting interface for these lexicons is available at (http://clic.ub.edu/ancora/).