Large language models (LLMs) demonstrate great performance in natural language processing and understanding tasks. However, much work remains to enhance their interpretability. Annotated datasets with explanations could be key to addressing this issue, as they enable the development of models that provide human-like explanations for their decisions. In this paper, we introduce the SimilEx dataset, the first Italian dataset reporting human evaluations of similarity between pairs of sentences. For a subset of these pairs, the annotators also provided explanations in natural language for the scores assigned. The SimilEx dataset is valuable for exploring the variability in similarity perception between sentences and for training LLMs to offer human-like explanations for their predictions.

State-of-the-art Large Language Models (LLMs) demonstrate exceptional proficiency across diverse tasks, yet systematic evaluations of their linguistic abilities remain limited. This paper addresses this gap by proposing a new evaluation framework leveraging the potentialities of Controllable Text Generation. Our approach evaluates the models’ capacity to generate sentences that adhere to specific linguistic constraints and their ability to recognize the linguistic properties of their own generated sentences, also in terms of consistency with the specified constraints. We tested our approach on six Italian LLMs using various linguistic constraints.

Large Language Models (LLMs) undergo extensive evaluation against various benchmarks collected in established leaderboards to assess their performance across multiple tasks. However, to the best of our knowledge, there is a lack of comprehensive studies evaluating these models’ linguistic abilities independent of specific tasks. In this paper, we introduce a novel evaluation methodology designed to test LLMs’ sentence generation abilities under specific linguistic constraints. Drawing on the ‘linguistic profiling’ approach, we rigorously investigate the extent to which five LLMs of varying sizes, tested in both zero- and few-shot scenarios, effectively adhere to (morpho)syntactic constraints. Our findings shed light on the linguistic proficiency of LLMs, revealing both their capabilities and limitations in generating linguistically-constrained sentences.

In this paper, we explore the impact of augmenting pre-trained Encoder-Decoder models, specifically T5, with linguistic knowledge for the prediction of a target task. In particular, we investigate whether fine-tuning a T5 model on an intermediate task that predicts structural linguistic properties of sentences modifies its performance in the target task of predicting sentence-level complexity. Our study encompasses diverse experiments conducted on Italian and English datasets, employing both monolingual and multilingual T5 models at various sizes. Results obtained for both languages and in cross-lingual configurations show that linguistically motivated intermediate fine-tuning has generally a positive impact on target task performance, especially when applied to smaller models and in scenarios with limited data availability.

Several studies in the literature on the interpretation of Neural Language Models (NLM) focus on the linguistic generalization abilities of pre-trained models. However, little attention is paid to how the linguistic knowledge of the models changes during the fine-tuning steps. In this paper, we contribute to this line of research by showing to what extent a wide range of linguistic phenomena are forgotten across 50 epochs of fine-tuning, and how the preserved linguistic knowledge is correlated with the resolution of the fine-tuning task. To this end, we considered a quite understudied task where linguistic information plays the main role, i.e. the prediction of the evolution of written language competence of native language learners. In addition, we investigate whether it is possible to predict the fine-tuned NLM accuracy across the 50 epochs solely relying on the assessed linguistic competence. Our results are encouraging and show a high relationship between the model’s linguistic competence and its ability to solve a linguistically-based downstream task.

This paper describes the process of acquisition, cleaning, interpretation, coding and linguistic annotation of a collection of parliamentary debates from the Senate of the Italian Republic covering the COVID-19 period and a former period for reference and comparison according to the CLARIN ParlaMint guidelines and prescriptions. The corpus contains 1199 sessions and 79,373 speeches, for a total of about 31 million words and was encoded according to the ParlaCLARIN TEI XML format, as well as in CoNLL-UD format. It includes extensive metadata about the speakers, the sessions, the political parties and Parliamentary groups. As required by the ParlaMint initiative, the corpus was also linguistically annotated for sentences, tokens, POS tags, lemmas and dependency syntax according to the universal dependencies guidelines. Named entity classification was also included. All linguistic annotation was performed automatically using state-of-the-art NLP technology with no manual revision. The Italian dataset is freely available as part of the larger ParlaMint 2.1 corpus deposited and archived in CLARIN repository together with all other national corpora. It is also available for direct analysis and inspection via various CLARIN services and has already been used both for research and educational purposes.

We report the results of the SemEval 2022 Task 3, PreTENS, on evaluation the acceptability of simple sentences containing constructions whose two arguments are presupposed to be or not to be in an ordered taxonomic relation. The task featured two sub-tasks articulated as: (i) binary prediction task and (ii) regression task, predicting the acceptability in a continuous scale. The sentences were artificially generated in three languages (English, Italian and French). 21 systems, with 8 system papers were submitted for the task, all based on various types of fine-tuned transformer systems, often with ensemble methods and various data augmentation techniques. The best systems reached an F1-macro score of 94.49 (sub-task1) and a Spearman correlation coefficient of 0.80 (sub-task2), with interesting variations in specific constructions and/or languages.

This paper presents an investigation aimed at studying how the linguistic structure of a sentence affects the perplexity of two of the most popular Neural Language Models (NLMs), BERT and GPT-2. We first compare the sentence-level likelihood computed with BERT and the GPT-2’s perplexity showing that the two metrics are correlated. In addition, we exploit linguistic features capturing a wide set of morpho-syntactic and syntactic phenomena showing how they contribute to predict the perplexity of the two NLMs.

In this paper we present an NLP-based approach for tracking the evolution of written language competence in L2 Spanish learners using a wide range of linguistic features automatically extracted from students’ written productions. Beyond reporting classification results for different scenarios, we explore the connection between the most predictive features and the teaching curriculum, finding that our set of linguistic features often reflect the explicit instructions that students receive during each course.

In this paper we investigate the linguistic knowledge learned by a Neural Language Model (NLM) before and after a fine-tuning process and how this knowledge affects its predictions during several classification problems. We use a wide set of probing tasks, each of which corresponds to a distinct sentence-level feature extracted from different levels of linguistic annotation. We show that BERT is able to encode a wide range of linguistic characteristics, but it tends to lose this information when trained on specific downstream tasks. We also find that BERT’s capacity to encode different kind of linguistic properties has a positive influence on its predictions: the more it stores readable linguistic information of a sentence, the higher will be its capacity of predicting the expected label assigned to that sentence.

“Voices of the Great War” is the first large corpus of Italian historical texts dating back to the period of First World War. This corpus differs from other existing resources in several respects. First, from the linguistic point of view it gives account of the wide range of varieties in which Italian was articulated in that period, namely from a diastratic (educated vs. uneducated writers), diaphasic (low/informal vs. high/formal registers) and diatopic (regional varieties, dialects) points of view. From the historical perspective, through a collection of texts belonging to different genres it represents different views on the war and the various styles of narrating war events and experiences. The final corpus is balanced along various dimensions, corresponding to the textual genre, the language variety used, the author type and the typology of conveyed contents. The corpus is fully annotated with lemmas, part-of-speech, terminology, and named entities. Significant corpus samples representative of the different “voices” have also been enriched with meta-linguistic and syntactic information. The layer of syntactic annotation forms the first nucleus of an Italian historical treebank complying with the Universal Dependencies standard. The paper illustrates the final resource, the methodology and tools used to build it, and the Web Interface for navigating it.

In this paper, we introduce Profiling–UD, a new text analysis tool inspired to the principles of linguistic profiling that can support language variation research from different perspectives. It allows the extraction of more than 130 features, spanning across different levels of linguistic description. Beyond the large number of features that can be monitored, a main novelty of Profiling–UD is that it has been specifically devised to be multilingual since it is based on the Universal Dependencies framework. In the second part of the paper, we demonstrate the effectiveness of these features in a number of theoretical and applicative studies in which they were successfully used for text and author profiling.

In this paper, we present a crowdsourcing-based approach to model the human perception of sentence complexity. We collect a large corpus of sentences rated with judgments of complexity for two typologically-different languages, Italian and English. We test our approach in two experimental scenarios aimed to investigate the contribution of a wide set of lexical, morpho-syntactic and syntactic phenomena in predicting i) the degree of agreement among annotators independently from the assigned judgment and ii) the perception of sentence complexity.

Detection and correction of errors and inconsistencies in “gold treebanks” are becoming more and more central topics of corpus annotation. The paper illustrates a new incremental method for enhancing treebanks, with particular emphasis on the extension of error patterns across different textual genres and registers. Impact and role of corrections have been assessed in a dependency parsing experiment carried out with four different parsers, whose results are promising. For both evaluation datasets, the performance of parsers increases, in terms of the standard LAS and UAS measures and of a more focused measure taking into account only relations involved in error patterns, and at the level of individual dependencies.

In this paper, we present the CItA corpus (Corpus Italiano di Apprendenti L1), a collection of essays written by Italian L1 learners collected during the first and second year of lower secondary school. The corpus was built in the framework of an interdisciplinary study jointly carried out by computational linguistics and experimental pedagogists and aimed at tracking the development of written language competence over the years and students’ background information.

In this paper, we present T2K^2, a suite of tools for automatically extracting domain―specific knowledge from collections of Italian and English texts. T2K^2 (Text―To―Knowledge v2) relies on a battery of tools for Natural Language Processing (NLP), statistical text analysis and machine learning which are dynamically integrated to provide an accurate and incremental representation of the content of vast repositories of unstructured documents. Extracted knowledge ranges from domain―specific entities and named entities to the relations connecting them and can be used for indexing document collections with respect to different information types. T2K^2 also includes “linguistic profiling” functionalities aimed at supporting the user in constructing the acquisition corpus, e.g. in selecting texts belonging to the same genre or characterized by the same degree of specialization or in monitoring the “added value” of newly inserted documents. T2K^2 is a web application which can be accessed from any browser through a personal account which has been tested in a wide range of domains.

The paper describes the design and the results of a manual annotation methodology devoted to enrich the ISST--TANL Corpus, derived from the Italian Syntactic--Semantic Treebank (ISST), with Semantic Frames information. The main issues encountered in applying the English FrameNet annotation criteria to a corpus of Italian language are discussed together with the choice of anchoring the semantic annotation layer to the underlying dependency syntactic structure. The results of a case study aimed at extending and specialising this methodology for the annotation of a corpus of legislative texts are also discussed.

In this paper, we present a novel approach to multi-word terminology extraction combining a well-known automatic term recognition approach, the C--NC value method, with a contrastive ranking technique, aimed at refining obtained results either by filtering noise due to common words or by discerning between semantically different types of terms within heterogeneous terminologies. Differently from other contrastive methods proposed in the literature that focus on single terms to overcome the multi-word terms' sparsity problem, the proposed contrastive function is able to handle variation in low frequency events by directly operating on pre-selected multi-word terms. This methodology has been tested in two case studies carried out in the History of Art and Legal domains. Evaluation of achieved results showed that the proposed two--stage approach improves significantly multi--word term extraction results. In particular, for what concerns the legal domain it provides an answer to a well-known problem in the semi--automatic construction of legal ontologies, namely that of singling out law terms from terms of the specific domain being regulated.

This paper reports on the design and construction of a bio-event annotated corpus which was developed with a specific view to the acquisition of semantic frames from biomedical corpora. We describe the adopted annotation scheme and the annotation process, which is supported by a dedicated annotation tool. The annotated corpus contains 677 abstracts of biomedical research articles.