Aligning song lyrics with sung audio is challenging, especially for languages and music styles where annotated datasets are scarce. We address this gap by presenting the first dataset of Italian opera arias annotated with lyrics and time-stamps per word. The dataset comprises of 24 arias drawn from well-known operas of the 18th to 20th centuries with a total audio duration of nearly two hours. We benchmark both music alignment models and speech forced alignment models and show that existing methods face significant challenges on this dataset, with performance dropping by 45% compared to other datasets. Multilingual and speech-based models exhibit relatively better performance on this dataset. We also evaluate few-shot fine-tuning of these models on the new dataset and find that, while it yields only marginal overall improvement, it produces localized gains on specific arias, suggesting that limited exposure helps the model adapt to some patterns but cannot fully overcome differences in language or musical style.

Large Language Models (LLMs) face significant challenges when queried about long-tail knowledge, i.e., information that is rarely encountered during their training process. These difficulties arise due to the inherent sparsity of such data. Furthermore, LLMs often lack the ability to verify or ground their responses in authoritative sources, which can lead to plausible yet inaccurate outputs when addressing infrequent subject matter. Our work aims to investigate these phenomena by introducing KE-MHISTO, a multilingual benchmark for Entity Linking and Question Answering in the domain of historical music knowledge, available in both Italian and English. We demonstrate that KE-MHISTO provides significantly broader coverage of long-tail knowledge compared to existing alternatives. Moreover, it poses substantial challenges for state-of-the-art models. Our experiments reveal that smaller, multilingual models can achieve performance comparable to significantly larger counterparts, highlighting the potential of efficient, language-aware approaches for long-tail knowledge extraction. KE-MHISTO is available at: https://github.com/polifonia-project/KE-MHISTO.

We present a clear distinction between the phenomena of comparisons and similes along with a fine-grained annotation guideline that facilitates the structural annotation and assessment of the two classes, with three major contributions: 1) a publicly available annotated data set of 100 comparative statements; 2) theoretically grounded annotation guidelines for human annotators; and 3) results of machine learning experiments to establish how the–often subtle–distinction between the two phenomena can be automated.

In this paper, we present an evaluation of two different approaches to the free-form Question Answering (QA) task. The main difference between the two approaches is that one is based on latent representations of knowledge, and the other uses explicit knowledge representation. For the evaluation, we developed DynaKnowledge, a new benchmark composed of questions concerning Wikipedia low-frequency entities. We wanted to ensure, on the one hand, that the questions are answerable and, on the other, that the models can provide information about very specific facts. The evaluation that we conducted highlights that the proposed benchmark is particularly challenging. The best model answers correctly only on 50% of the questions. Analysing the results, we also found that ChatGPT shows low reliance on low-frequency entity questions, manifesting a popularity bias. On the other hand, a simpler model based on explicit knowledge is less affected by this bias. With this paper, we want to provide a living benchmark for open-form QA to test knowledge and latent representation models on a dynamic benchmark.