Computational methods applied to undeciphered scripts from the Aegean and Cyprus: Linear A and Cypro-Minoan

The study of ancient, undeciphered scripts presents unique challenges, that depend both on the nature of the problem and on the peculiarities of each writing system. In this thesis, I present two computational approaches that are tailored to two different tasks and writing systems. The first of these methods is aimed at the decipherment of the Linear A afraction signs, in order to discover their numerical values. This is achieved with a combination of constraint programming, ad-hoc metrics and paleographic considerations. The second main contribution of this thesis regards the creation of an unsupervised deep learning model which uses drawings of signs from ancient writing system to learn to distinguish different graphemes in the vector space. This system, which is based on techniques used in the field of computer vision, is adapted to the study of ancient writing systems by incorporating information about sequences in the model, mirroring what is often done in natural language processing. In order to develop this model, the Cypriot Greek Syllabary is used as a target, since this is a deciphered writing system. Finally, this unsupervised model is adapted to the undeciphered Cypro-Minoan and it is used to answer open questions about this script. In particular, by reconstructing multiple allographs that are not agreed upon by paleographers, it supports the idea that Cypro-Minoan is a single script and not a collection of three script like it was proposed in the literature. These results on two different tasks shows that computational methods can be applied to undeciphered scripts, despite the relatively low amount of available data, paving the way for further advancement in paleography using these methods.

Predire la borsa valori con language model neurali

Nonostante lo scetticismo di molti studiosi circa la possibilità di prevedere l'andamento della borsa valori, esistono svariate teorie ipotizzanti la possibilità di utilizzare le informazioni conosciute per predirne i movimenti futuri. L’avvento dell’intelligenza artificiale nella seconda parte dello scorso secolo ha permesso di ottenere risultati rivoluzionari in svariati ambiti, tanto che oggi tale disciplina trova ampio impiego nella nostra vita quotidiana in molteplici forme. In particolare, grazie al machine learning, è stato possibile sviluppare sistemi intelligenti che apprendono grazie ai dati, riuscendo a modellare problemi complessi. Visto il successo di questi sistemi, essi sono stati applicati anche all’arduo compito di predire la borsa valori, dapprima utilizzando i dati storici finanziari della borsa come fonte di conoscenza, e poi, con la messa a punto di tecniche di elaborazione del linguaggio naturale umano (NLP), anche utilizzando dati in linguaggio naturale, come il testo di notizie finanziarie o l’opinione degli investitori. Questo elaborato ha l’obiettivo di fornire una panoramica sull’utilizzo delle tecniche di machine learning nel campo della predizione del mercato azionario, partendo dalle tecniche più elementari per arrivare ai complessi modelli neurali che oggi rappresentano lo stato dell’arte. Vengono inoltre formalizzati il funzionamento e le tecniche che si utilizzano per addestrare e valutare i modelli di machine learning, per poi effettuare un esperimento in cui a partire da dati finanziari e soprattutto testuali si tenterà di predire correttamente la variazione del valore dell’indice di borsa S&P 500 utilizzando un language model basato su una rete neurale.

Vocabulary Transfer and Knowledge Distillation for Language Model Compression

Natural Language Processing (NLP) has seen tremendous improvements over the last few years. Transformer architectures achieved impressive results in almost any NLP task, such as Text Classification, Machine Translation, and Language Generation. As time went by, transformers continued to improve thanks to larger corpora and bigger networks, reaching hundreds of billions of parameters. Training and deploying such large models has become prohibitively expensive, such that only big high tech companies can afford to train those models. Therefore, a lot of research has been dedicated to reducing a model’s size. In this thesis, we investigate the effects of Vocabulary Transfer and Knowledge Distillation for compressing large Language Models. The goal is to combine these two methodologies to further compress models without significant loss of performance. In particular, we designed different combination strategies and conducted a series of experiments on different vertical domains (medical, legal, news) and downstream tasks (Text Classification and Named Entity Recognition). Four different methods involving Vocabulary Transfer (VIPI) with and without a Masked Language Modelling (MLM) step and with and without Knowledge Distillation are compared against a baseline that assigns random vectors to new elements of the vocabulary. Results indicate that VIPI effectively transfers information of the original vocabulary and that MLM is beneficial. It is also noted that both vocabulary transfer and knowledge distillation are orthogonal to one another and may be applied jointly. The application of knowledge distillation first before subsequently applying vocabulary transfer is recommended. Finally, model performance due to vocabulary transfer does not always show a consistent trend as the vocabulary size is reduced. Hence, the choice of vocabulary size should be empirically selected by evaluation on the downstream task similar to hyperparameter tuning.

NLG-Metricverse: An End-to-End Library for Evaluating Natural Language Generation

Driven by recent deep learning breakthroughs, natural language generation (NLG) models have been at the center of steady progress in the last few years. However, since our ability to generate human-indistinguishable artificial text lags behind our capacity to assess it, it is paramount to develop and apply even better automatic evaluation metrics. To facilitate researchers to judge the effectiveness of their models broadly, we suggest NLG-Metricverse—an end-to-end open-source library for NLG evaluation based on Python. This framework provides a living collection of NLG metrics in a unified and easy- to-use environment, supplying tools to efficiently apply, analyze, compare, and visualize them. This includes (i) the extensive support of heterogeneous automatic metrics with n-arity management, (ii) the meta-evaluation upon individual performance, metric-metric and metric-human correlations, (iii) graphical interpretations for helping humans better gain score intuitions, (iv) formal categorization and convenient documentation to accelerate metrics understanding. NLG-Metricverse aims to increase the comparability and replicability of NLG research, hopefully stimulating new contributions in the area.

Bio-QA-GNN: Reasoning with Language Models and Knowledge Graphs for Interpretable Biomedical Question Answering

Nowadays the idea of injecting world or domain-specific structured knowledge into pre-trained language models (PLMs) is becoming an increasingly popular approach for solving problems such as biases, hallucinations, huge architectural sizes, and explainability lack—critical for real-world natural language processing applications in sensitive fields like bioinformatics. One recent work that has garnered much attention in Neuro-symbolic AI is QA-GNN, an end-to-end model for multiple-choice open-domain question answering (MCOQA) tasks via interpretable text-graph reasoning. Unlike previous publications, QA-GNN mutually informs PLMs and graph neural networks (GNNs) on top of relevant facts retrieved from knowledge graphs (KGs). However, taking a more holistic view, existing PLM+KG contributions mainly consider commonsense benchmarks and ignore or shallowly analyze performances on biomedical datasets. This thesis start from a propose of a deep investigation of QA-GNN for biomedicine, comparing existing or brand-new PLMs, KGs, edge-aware GNNs, preprocessing techniques, and initialization strategies. By combining the insights emerged in DISI's research, we introduce Bio-QA-GNN that include a KG. Working with this part has led to an improvement in state-of-the-art of MCOQA model on biomedical/clinical text, largely outperforming the original one (+3.63\% accuracy on MedQA). Our findings also contribute to a better understanding of the explanation degree allowed by joint text-graph reasoning architectures and their effectiveness on different medical subjects and reasoning types. Codes, models, datasets, and demos to reproduce the results are freely available at: \url{https://github.com/disi-unibo-nlp/bio-qagnn}.

SAB(IO): A BIOLOGICALLY PLAUSIBLE CONNECTIONIST APPROACH TO AUTOMATIC TEXT SUMMARIZATION

An implementation of a computational tool to generate new summaries from new source texts is presented, by means of the connectionist approach (artificial neural networks). Among other contributions that this work intends to bring to natural language processing research, the use of a more biologically plausible connectionist architecture and training for automatic summarization is emphasized. The choice relies on the expectation that it may bring an increase in computational efficiency when compared to the sa-called biologically implausible algorithms.

Developing and implementing an open and non-proprietary device description for FOUNDATION fieldbus based on software standards

Support for interoperability and interchangeability of software components which are part of a fieldbus automation system relies on the definition of open architectures, most of them involving proprietary technologies. Concurrently, standard, open and non-proprietary technologies, such as XML, SOAP, Web Services and the like, have greatly evolved and been diffused in the computing area. This article presents a FOUNDATION fieldbus (TM) device description technology named Open-EDD, based on XML and other related technologies (XLST, DOM using Xerces implementation, OO, XMIL Schema), proposing an open and nonproprietary alternative to the EDD (Electronic Device Description). This initial proposal includes defining Open-EDDML as the programming language of the technology in the FOUNDATION fieldbus (TM) protocol, implementing a compiler and a parser, and finally, integrating and testing the new technology using field devices and a commercial fieldbus configurator. This study attests that this new technology is feasible and can be applied to other configurators or HMI applications used in fieldbus automation systems. (c) 2008 Elsevier B.V. All rights reserved.

Frequency Domain Connectivity Identification: An Application of Partial Directed Coherence in fMRI

Functional magnetic resonance imaging (fMRI) has become an important tool in Neuroscience due to its noninvasive and high spatial resolution properties compared to other methods like PET or EEG. Characterization of the neural connectivity has been the aim of several cognitive researches, as the interactions among cortical areas lie at the heart of many brain dysfunctions and mental disorders. Several methods like correlation analysis, structural equation modeling, and dynamic causal models have been proposed to quantify connectivity strength. An important concept related to connectivity modeling is Granger causality, which is one of the most popular definitions for the measure of directional dependence between time series. In this article, we propose the application of the partial directed coherence (PDC) for the connectivity analysis of multisubject fMRI data using multivariate bootstrap. PDC is a frequency domain counterpart of Granger causality and has become a very prominent tool in EEG studies. The achieved frequency decomposition of connectivity is useful in separating interactions from neural modules from those originating in scanner noise, breath, and heart beating. Real fMRI dataset of six subjects executing a language processing protocol was used for the analysis of connectivity. Hum Brain Mapp 30:452-461, 2009. (C) 2007 Wiley-Liss, Inc.

Using Reverberation to Improve Range and Elevation Discrimination for Small Array Sound Source Localization

Sound source localization (SSL) is an essential task in many applications involving speech capture and enhancement. As such, speaker localization with microphone arrays has received significant research attention. Nevertheless, existing SSL algorithms for small arrays still have two significant limitations: lack of range resolution, and accuracy degradation with increasing reverberation. The latter is natural and expected, given that strong reflections can have amplitudes similar to that of the direct signal, but different directions of arrival. Therefore, correctly modeling the room and compensating for the reflections should reduce the degradation due to reverberation. In this paper, we show a stronger result. If modeled correctly, early reflections can be used to provide more information about the source location than would have been available in an anechoic scenario. The modeling not only compensates for the reverberation, but also significantly increases resolution for range and elevation. Thus, we show that under certain conditions and limitations, reverberation can be used to improve SSL performance. Prior attempts to compensate for reverberation tried to model the room impulse response (RIR). However, RIRs change quickly with speaker position, and are nearly impossible to track accurately. Instead, we build a 3-D model of the room, which we use to predict early reflections, which are then incorporated into the SSL estimation. Simulation results with real and synthetic data show that even a simplistic room model is sufficient to produce significant improvements in range and elevation estimation, tasks which would be very difficult when relying only on direct path signal components.

Refining specifications to logic programs

The refinement calculus provides a framework for the stepwise development of imperative programs from specifications. In this paper we study a refinement calculus for deriving logic programs. Dealing with logic programs rather than imperative programs has the dual advantages that, due to the expressive power of logic programs, the final program is closer to the original specification, and each refinement step can achieve more. Together these reduce the overall number of derivation steps. We present a logic programming language extended with specification constructs (including general predicates, assertions, and types and invariants) to form a wide-spectrum language. General predicates allow non-executable properties to be included in specifications. Assertions, types and invariants make assumptions about the intended inputs of a procedure explicit, and can be used during refinement to optimize the constructed logic program. We provide a semantics for the extended logic programming language and derive a set of refinement laws. Finally we apply these to an example derivation.

A refinement calculus for logic programs

Existing refinement calculi provide frameworks for the stepwise development of imperative programs from specifications. This paper presents a refinement calculus for deriving logic programs. The calculus contains a wide-spectrum logic programming language, including executable constructs such as sequential conjunction, disjunction, and existential quantification, as well as specification constructs such as general predicates, assumptions and universal quantification. A declarative semantics is defined for this wide-spectrum language based on executions. Executions are partial functions from states to states, where a state is represented as a set of bindings. The semantics is used to define the meaning of programs and specifications, including parameters and recursion. To complete the calculus, a notion of correctness-preserving refinement over programs in the wide-spectrum language is defined and refinement laws for developing programs are introduced. The refinement calculus is illustrated using example derivations and prototype tool support is discussed.

Agents as multi-threaded logical objects

In this paper we describe a distributed object oriented logic programming language in which an object is a collection of threads deductively accessing and updating a shared logic program. The key features of the language, such as static and dynamic object methods and multiple inheritance, are illustrated through a series of small examples. We show how we can implement object servers, allowing remote spawning of objects, which we can use as staging posts for mobile agents. We give as an example an information gathering mobile agent that can be queried about the information it has so far gathered whilst it is gathering new information. Finally we define a class of co-operative reasoning agents that can do resource bounded inference for full first order predicate logic, handling multiple queries and information updates concurrently. We believe that the combination of the concurrent OO and the LP programming paradigms produces a powerful tool for quickly implementing rational multi-agent applications on the internet.

iMap : um mecanismo de inferência para mapas conceituais

Mapas Conceituais são representações gráficas do conhecimento de uma pessoa num dado momento e área de conhecimento. Por sua natureza investigativa, são utilizados como ferramentas de apoio em abordagens pedagógicas que objetivam promover a aprendizagem significativa. No entanto, o processo de avaliação de um mapa tende a ser custoso pois acarreta uma pesada carga de processamento cognitivo por parte do avaliador, já que este precisa mapear os conceitos e relações em busca de nuances de conhecimento alí presentes. Essa pesquisa tem por objetivo aumentar o nível de abstração nas interações entre o avaliador e os mapas conceituais fornecendo uma camada intermediária de inteligência computacional que favoreça a comunicação por meio de perguntas e respostas em linguagem natural, fornecendo ao avaliador ferramentas que lhe permita examinar o conteúdo do mapa conceitual sem exigir deste o mapeamento visual dos conceitos e relações presentes nos mapas avaliados. Uma ferramenta é prototipada e uma prova de conceito apresentada. A análise da arquitetura proposta permitiu definir uma arquitetura final com características que permitem potencializar o uso de mapas conceituais e facilitar diversas operações pedagógicas com estes. Essa pesquisa situa-se na área de investigação de sistemas de perguntas e resposta, aplicando técnicas de processamento de linguagem natural para análise da pergunta e interpretação do mapa conceitual e aplica técnica de inteligência artificial para inferir respostas às perguntas.

An SPSS R-Menu for Ordinal Factor Analysis

Exploratory factor analysis is a widely used statistical technique in the social sciences. It attempts to identify underlying factors that explain the pattern of correlations within a set of observed variables. A statistical software package is needed to perform the calcula- tions. However, there are some limitations with popular statistical software packages, like SPSS. The R programming language is a free software package for statistical and graphical computing. It o ers many packages written by contributors from all over the world and programming resources that allow it to overcome the dialog limitations of SPSS. This paper o ers an SPSS dialog written in the R programming language with the help of some packages, so that researchers with little or no knowledge in programming, or those who are accustomed to making their calculations based on statistical dialogs, have more options when applying factor analysis to their data and hence can adopt a better approach when dealing with ordinal, Likert-type data.

Concepção de uma pequena turbina eólica destinada ao segmento doméstico (simplificação e análise aerodinâmica das pás)

O trabalho que se apresenta incide sobre o estudo aerodinâmico das pás de uma turbina eólica de pequeno porte, com vista à simplificação geométrica, de forma a que estas sejam baratas e de fácil concepção. A teoria da quantidade de movimento do elemento de pá (BEMT), que é o modelo de referência para o projecto e análise aerodinâmica das pás das turbinas eólicas, foi utilizada neste trabalho de forma a projectar e analisar aerodinamicamente as pás da turbina. Sendo assim, desenvolveu-se um programa computacional em MATLAB, denominado de “Turbina”, de forma a implementar a teoria BEM. Introduzindo os dados dos parâmetros de projecto no programa (potência requerida, o número de pás, velocidade do vento, a TSR e o tipo de perfil alar), obtêm-se os parâmetros geométricos das pás (distribuição da corda ao longo da envergadura, o raio da pá e a distribuição da torção da pá), os parâmetros aerodinâmicos e de desempenho. Uma pá ideal foi calculada e de seguida foi modificada de forma a obter-se uma pá simples e menos carregada aerodinamicamente. Introduzidas as modificações na geometria da pá ideal, obtiveram-se duas configurações distintas. Uma configuração linear, onde a distribuição da corda e do ângulo de torção se tornam lineares, e outra configuração bi-linear, onde a distribuição da corda continua linear mas o ângulo de torção se torna bi-linear, isto é, a pá é composta por dois troços onde cada troço apresenta uma distribuição linear do ângulo de torção geométrica. As conclusões demonstram que a configuração bi-linear é uma boa alternativa a configuração ideal, apresentando uma redução do desempenho do rotor de 2.8% para um aumento do raio da pá em 1.41%, para se obter a mesma potência da configuração ideal. A análise aos perfis alares, utilizados neste trabalho, foi efectuada a partir dos programas comerciais ICEM e FLUENT. De forma a automatizar a análise de CFD, três programas foram desenvolvidos utilizando a linguagem de programação “C”. Os programas são denominados de “Malha2D”, “Calcula_Coeficientes” e “Plot_Graficos”. Finalmente, um estudo paramétrico foi feito de forma a avaliar a influências das variáveis de projecto no desempenho geral da turbina.