This thesis is concerned with the role played by software tools in the analysis and dissemination of linguistic corpora and their contribution to a more widespread adoption of corpora in different fields. Chapter 1 contains an overview of some of the most relevant corpus analysis tools available today, presenting their most interesting features and some of their drawbacks. Chapter 2 begins with an explanation of the reasons why none of the available tools appear to satisfy the requirements of the user community and then continues with technical overview of the current status of the new system developed as part of this work. This presentation is followed by highlights of features that make the system appealing to users and corpus builders (i.e. scholars willing to make their corpora available to the public). The chapter concludes with an indication of future directions for the projects and information on the current availability of the software. Chapter 3 describes the design of an experiment devised to evaluate the usability of the new system in comparison to another corpus tool. Usage of the tool was tested in the context of a documentation task performed on a real assignment during a translation class in a master's degree course. In chapter 4 the findings of the experiment are presented on two levels of analysis: firstly a discussion on how participants interacted with and evaluated the two corpus tools in terms of interface and interaction design, usability and perceived ease of use. Then an analysis follows of how users interacted with corpora to complete the task and what kind of queries they submitted. Finally, some general conclusions are drawn and areas for future work are outlined.

The focus of this study is the relationship among three different manuscripts (Modena, Bibl. Estense, MS α.R.4.4; Firenze, Bibl. Laurenziana MS Rediano 9; and London, BL, MS Harley, 2253) and the poetry they transmit. The aim of this research is to show the ways that the Bible was used in the transmission of the lyric poetry in the three literatures that they represent: Occitan (primarily through Marcabru’s songs), Italian (through the love poetry of Guittone d’Arezzo), and Middle English (through the Harley love lyrics and the MS.’s primary scribe), in a medieval European context.

The way humans learn the meaning of words is a fundamental question in many different disciplines and, from a computational perspective, an answer to this question could lead to important advances in artificial intelligence. While the details of the learning process are still an open question, what we do know is that humans make use of the very rich perceptual input present in the communicative setups in which learning takes place. In this work, we will present three models of human learning from naturalistic multi-modal input. We will start by introducing a model that assumes a purely predictive learner existing in a non-communicative setup and show that such a computational learner when tested displays comparable learning behaviour to human learners on a novel word learning setup. We will then relax some of the learning assumptions and present a model that, instead of exposing the computational learner to a passive environment, such as the text corpora traditionally used in semantic learning experiments, it exposes the learner to communicative episodes, simulated in our experiments by corpora capturing multi-modal interactions between children and their caregivers, allowing the learner to make use of information beyond words and passive percept during learning. Finally, we will present on-going work towards interactive learning between two agents.

This thesis addresses the cognitive foundations of categorization and acquisition of intonational categories in native (L1) and second language (L2). It focuses on the link between the processing of intonational categories and the and pragmatic functions of language. The thesis reports two behavioral psychoacoustic experiments that studied the disambiguation of sentence-modality (statement vs. question) signaled by sentence-final Boundary Tones by manipulating lexical and linguistic status of the underlying segmental information. A third ERPs experiment studied with ERPs the association of specific Pitch Accents with the discourse status of a referent in German and how different processing-correlates of PA violation are processed in L1 and L2 speakers. In all experiments, specific attention has been devoted to individual differences both at the theoretical and empirical level. I showed that perceivers can display variability in processing as a function of biographic factors, in the quantity and quality of training in a second language, and in the presence of variables related to the construct of Theory-of-Mind (ToM). I support the view that the processing of intonational categories, modulated by Fundamental Frequency contours, links with the processing of segmental information, the semantic access at word-level, and the decoding of the information structure within the discourse model. The study of processing of pitch contours is a highly multidisciplinary discipline, but the different theoretical perspectives are not always considered within specific research. I propose to approach the study of pitch processing by trying to integrate the different theoretical and empirical approaches with the aim to use the available knowledge. This broader perspective considers the auditory categorization process, the integration of the sound-domain information with higher-order linguistic structure, and the modeling of individual variability of the perceivers. I support the view that the presence of individual traits that favor the efficient decoding of the interlocutor’s perspective and intentions correlates with a more efficient processing of the discourse information structure. I propose that this is observable through the manipulation of the associated intonational categories. I think that the adoption of a multidisciplinary perspective, centered on the processing of intonational categories, and the approach developed in this thesis is relevant to develop further the study of specific populations known to display less efficient processing of the pragmatic aspects of discourse, such as individuals with Autism Spectrum Conditions.

This thesis focuses on the neology in minority languages, especially in the Ladin spoken in the Fassa Valley (Trentino-Alto Adige, Italy) and Aranes spoken in the Aran Valley (Catalonia, Spain), as a means of evaluation of the vitality of these languages. The real grade of Ladin and Aranes is definitely endangered as UNESCO Atlas of the World's Languages in Danger demonstrated in 2003, but I think that the rule of the neology in Ladin is enough important to improve this grade. For this reason, I will analyse the lexicon related to Information Technology (IT) as a semantic field rich in neologisms. In the first part I will examine two corpora: one, based upon the language as used in websites and databases and the other, taken from questionnaires carried out in elementary schools and middle schools. In the second part, I will provide concrete examples of how minority languages create new words and form neologisms (i.e. with innovative or conservative lexicon or morphological structures, etc.). This will produce an overview of neology in Ladin and Aranes. However, I will specify the rule of norms, use and contact throughout this process.

An actual trend in the computational linguistics and natural language processing is the implementation of multilingual utilities for different tasks, like information retrival, summarization of documents in different languages or machine translation, tasks in which the resolution of anaphoric references plays a crucial role. This dissertation presents a proposal of annotation scheme for the creation of corpus resources for linguistic based multilingual anaphora resolution. This scheme has been implemented for the annotation of English and Italian data. Inter-annotator agreement studies show that the annotation scheme is relaiable. The annotated corpora have been used for the anaphora resolution task, and the results have been compared with well known corpora. Finally hand annotated linguistic features have been used to help in the anaphora resolution process. The results show that our multilingual annotation scheme proposal has been utilized to produce data useful to build anaphora resolution systems for languages with different grammatical and typological features, like English and Italian.

Scopo della tesi è analizzare testi in sincronia (ma anche in diacronia) e di diversa natura (discorsi e interviste politici, racconti letterari, testi pratici, articoli di giornale) attraverso la grammatica delle valenze, o meglio attraverso l'analisi di omissioni e aumenti argomentali in contesti dati. L'intento è quello di mostrare come omissioni e aumenti argomentali contribuiscano a creare nei contesti considerati modificazioni semantiche ed effetti comunicativi differenti (ma anche ricorrenti), rilevanti per la stessa comprensione dei testi. La tesi è articolata in quattro capitoli che sviluppano i seguenti argomenti: a) la teoria della valenza di Tesnière e la sua ricezione; b) l'applicazione della grammatica delle valenze ai testi; c) le omissioni argomentali (aspetti teorici e analisi applicata ai testi); d) gli aumenti argomentali di verbi e nomi atmosferici (aspetti teorici e analisi applicata ai testi). Il tipo di analisi svolta mette a disposizione sia dati utili a livello teorico, sia materiale linguistico autentico commentato spendibile anche in ambito didattico.

Although being one very simple statement, the distributional hypothesis - namely, words that occur in similar contexts are semantically similar - has been granted the role of main assumption in many computational linguistic techniques. This is mostly due to the fact that it allows to easily and automatically construct a representation of word meaning from a large textual input. Among the computational linguistic techniques that are corpus-based and adopt the distributional hypothesis, Distributional semantic models (DSMs) have been shown to be a very effective method in many semantic-related tasks. DSMs approximate word meaning by vectors that keep track of the patterns of co-occurrence of words in the processed corpora. In addition, DSMs have been shown to be a very plausible computational model for human concept cognition, since they are able to simulate several psychological phenomena. Despite their success, one of their strongest limitations is that they entirely represent word meaning in terms of connections with other words. Cognitive scientists have argued that, in this way, DSMs neglect that humans rely also on non-verbal experiences and have access to rich sources of perceptual knowledge when they learn the meaning of words. In this work, the lack of perceptual grounding of distributional models is addressed by exploiting computer vision techniques that automatically identify discrete "visual words" in images, so that the distributional representation of a word can be extended to also encompass its co-occurrence with the visual words of images it is associated with. A flexible architecture to integrate text- and image-based distributional information is introduced and tested on a set of empirical evaluations, showing that an integrated model is superior to a purely text-based approach, and it provides somewhat complementary semantic information with respect to the latter.

Svariati contributi comparsi negli ultimi decenni hanno avviato la parziale rilettura del pensiero linguistico di Aristotele. Su queste premesse, lo studio si propone, a partire dall’analisi dei testi più significativi, di ricostruire una teoria della metafora coerente con il resto del sistema filosofico aristotelico. Aristotele concepisce la metafora come un fatto di lingua, e ne delinea le principali caratteristiche e funzioni all’interno della comunicazione. Per Aristotele, tuttavia, la metafora è anche il segno del processo mentale che l’ha prodotta. Il pensiero metaforico, che soggiace alla metafora intesa semplicemente come tropo, è un’attività cognitiva che si fonda sulla capacità umana di cogliere la somiglianza. A sua volta, il vedere ciò che è simile è una capacità che precede il linguaggio, ma tuttavia si connette inevitabilmente ad esso sul piano sia analogico che logico, nel momento del concepimento di un giudizio. Il processo metaforico è dunque uno strumento di conoscenza che, procedendo dal pensiero al linguaggio, permette all’uomo di cogliere le relazioni tra gli enti, mettendolo a sua volta in relazione con il mondo. / Several studies have recently started a partial reinterpretation of Aristotle’s linguistics. Moving from these premises, this work tries to rebuild Aristotle’s theory of metaphor, in conformity with his philosophy and the analysis of his most relevant papers. Aristotle conceives metaphor a fact of language, and defines metaphor most important features and functions in relationship with communication. Nevertheless Aristotle means metaphor as a sign of the psychical process that produces it. Metaphorical thought, which is in our mind and which we can understand looking through the trope of metaphor, is a cognitive process, based on the human capability of catching similarity. Seeing what is similar is a capability that precedes utterance, but nevertheless it is connected to the language in an analogical and logical way every time we make an assertion. Metaphorical action is a cognitive appliance that, proceeding from thought to utterance, makes man capable of understanding relationships between things, and brings man himself in relationship with the world.

In recent years, Neural Language Models (NLMs) have consistently demonstrated increasing linguistic abilities. However, the extent to which such networks can actually learn grammar remains an object of investigation, and experimental results are often inconclusive. Notably, the mainstream evaluation framework in which NLMs are tested seems largely based on Generative Grammar and nativist principles, and a shared constructionist approach on the matter has not yet emerged: this is at odds with the fact that usage-based theories are actually better suited to inspect the behaviour of such models. The main contribution of this thesis is the introduction of CALaMo, a novel framework for evaluating Neural Language Models’ linguistic abilities, using a constructionist approach. We especially aim at formalizing the relationship between the computational modelling phase and the underlying linguistic theory, thus allowing a more refined and informed discussion of settings and results. We focus on two specific areas that, we believe, are currently not easily tractable within the mainstream evaluation framework. The first scenario deals with language acquisition from child-directed data. Our main experimental result shows how it is possible to follow schematization paths during the acquisition process of the model, and how this relates to core hypotheses in constructionist theories. The second scenario deconstructs the mainstream view of the Neural Model as an average idealized speaker by proposing a way to simulate and analyze a population of artificial individuals. We show how the amount of “shared linguistic knowledge” across speakers is highly dependent on the specific linguistic background of each individual. Overall, we believe our framework opens the path for future discussion on the role of computational modelling in usage-based linguistic theory and vice versa, and provides a new formal methodology to both fields of study.