A supervised machine learning method for word sense disambiguation of Portuguese nouns

Dissertação de Mestrado, Ciências da Linguagem, Faculdade de Ciências Humanas e Sociais, Universidade do Algarve, 2010

Développement d’un système d’appariement pour l’e-recrutement

Ce mémoire tente de répondre à une problématique très importante dans le domaine de recrutement : l’appariement entre offre d’emploi et candidats. Dans notre cas nous disposons de milliers d’offres d’emploi et de millions de profils ramassés sur les sites dédiés et fournis par un industriel spécialisé dans le recrutement. Les offres d’emploi et les profils de candidats sur les réseaux sociaux professionnels sont généralement destinés à des lecteurs humains qui sont les recruteurs et les chercheurs d’emploi. Chercher à effectuer une sélection automatique de profils pour une offre d’emploi se heurte donc à certaines difficultés que nous avons cherché à résoudre dans le présent mémoire. Nous avons utilisé des techniques de traitement automatique de la langue naturelle pour extraire automatiquement les informations pertinentes dans une offre d’emploi afin de construite une requête qui nous permettrait d’interroger notre base de données de profils. Pour valider notre modèle d’extraction de métier, de compétences et de d’expérience, nous avons évalué ces trois différentes tâches séparément en nous basant sur une référence cent offres d’emploi canadiennes que nous avons manuellement annotée. Et pour valider notre outil d’appariement nous avons fait évaluer le résultat de l’appariement de dix offres d’emploi canadiennes par un expert en recrutement.

Développement d’un système d’appariement pour l’e-recrutement

Ce mémoire tente de répondre à une problématique très importante dans le domaine de recrutement : l’appariement entre offre d’emploi et candidats. Dans notre cas nous disposons de milliers d’offres d’emploi et de millions de profils ramassés sur les sites dédiés et fournis par un industriel spécialisé dans le recrutement. Les offres d’emploi et les profils de candidats sur les réseaux sociaux professionnels sont généralement destinés à des lecteurs humains qui sont les recruteurs et les chercheurs d’emploi. Chercher à effectuer une sélection automatique de profils pour une offre d’emploi se heurte donc à certaines difficultés que nous avons cherché à résoudre dans le présent mémoire. Nous avons utilisé des techniques de traitement automatique de la langue naturelle pour extraire automatiquement les informations pertinentes dans une offre d’emploi afin de construite une requête qui nous permettrait d’interroger notre base de données de profils. Pour valider notre modèle d’extraction de métier, de compétences et de d’expérience, nous avons évalué ces trois différentes tâches séparément en nous basant sur une référence cent offres d’emploi canadiennes que nous avons manuellement annotée. Et pour valider notre outil d’appariement nous avons fait évaluer le résultat de l’appariement de dix offres d’emploi canadiennes par un expert en recrutement.

La poésie à l’école secondaire québécoise (1963-2016) - Une histoire à travers les textes

Afin de comprendre et d’expliquer le malaise actuel vis-à-vis de l’enseignement de la poésie au secondaire, la thèse remonte à la publication du Rapport Parent (1963-1966), document fondateur de l’école québécoise moderne, et à la création du ministère de l’Éducation du Québec (1964) pour suivre, à travers les documents, l’histoire de l’enseignement de la poésie au secondaire au cours des 50 dernières années. Cette période se divise en trois grandes étapes : une révolution en éducation marquée par la publication du Rapport Parent et de deux programmes; le retour aux sources amorcé par L’école québécoise qui verra deux programmes et deux séries de manuels scolaires; la réforme articulée autour de L’école, tout un programme dont les programmes et les manuels sont actuellement en vigueur. C’est à travers ces trois types de documents (documents d’orientation, programmes d’enseignement et manuels scolaires) que je retracerai le parcours de l’enseignement de la poésie à l’école secondaire avec comme toile de fond les grands courants qui ont marqué l’éducation et l’enseignement du français au Québec depuis les années 1960. L’enseignement de la poésie ne peut se détacher du contexte dans lequel il se fait. Ce contexte comprend diverses facettes qui lui sont internes et d’autres qui lui sont externes. Les facettes externes dans ce travail sont la vision de la culture et de l’enseignement ainsi que la place faite à la littérature et à la poésie. L’idée de culture a beaucoup évolué depuis 50 ans. Le Rapport Parent abandonne la vision traditionnelle de la culture au profit d’une vision plus large qui englobe l’ensemble des activités humaines. Nuancée et actualisée, cette vision large de la culture restera la référence. La littérature qui détenait une place importante dans la culture traditionnelle perd ainsi de l’influence. Sa place et son importance, malgré les discours qui la valorisent, vont diminuer jusqu’à ne devenir qu’un adjuvant à l’enseignement de la technique de la langue et des types de textes. L’évolution de l’enseignement transformera la classe traditionnelle bien ordonnée sous l’autorité du maitre en lieu qui doit permettre à l’élève de construire ses connaissances. Les éléments reliés directement à l’enseignement de la poésie comprennent les programmes, les manuels, l’enseignement de la lecture et de l’écriture et le travail du professeur et de l’élève. Les programmes subiront d’importantes modifications. Le programme-cadre de 1969 marque la fin des programmes à visées littéraires. Le seul vrai but des programmes depuis, c’est de parfaire l’apprentissage technique de la langue. Le programme de 1980 marque l’abandon des genres littéraires comme base de l’enseignement des textes au profit de la notion de discours. Comme le discours poétique n’y apparait pas, la poésie peine à se trouver une place. Les programmes de 1995 et de la réforme optent pour les types de textes plutôt que les discours, mais la poésie n’entre pas dans cette typologie. La place de la poésie en lecture et en écriture est variable selon les époques. Sur un même pied que les autres types de textes dans le programme-cadre, elle apparait seulement en lecture dans le programme de 1980. Les programmes subséquents lui redonneront une place en écriture. Depuis 1980, elle reste un genre marginal qui souvent ne sert qu’à illustrer des notions utiles aux discours ou types de textes utilisés par le programme. La relation professeur/élève s’est considérablement modifiée au cours de cette période. De maitre, le professeur est devenu guide et médiateur. L’élève évolue en sens inverse : de passif récepteur de la connaissance, il devient responsable de construire ses connaissances. Le professeur, compte tenu du peu de place qu’occupe la poésie dans les programmes et du fait qu’elle ne fait pas partie de la typologie des textes adoptée par ceux-ci, ne peut compter que sur ses connaissances pour en assurer un enseignement valable. Les manuels accompagnent les programmes sauf pour le programme-cadre. Depuis 1980, les manuels de français ont une visée autoéducative parce qu’ils s’adressent directement à l’étudiant. Ils se présentent comme des préparations de cours que l’élève n’a qu’à suivre pour acquérir les connaissances au programme. La place de la poésie en lecture et en écriture y est variable selon les époques et les manuels.

OntoSPARES: da linguagem natural às ontologias. Contributos para a classificação automática de dados históricos (séc. XVI-XVIII)

O processamento de linguagem natural e as ontologias são ferramentas cuja interação permite uma melhor compreensão dos dados armazenados. Este trabalho, ao associar estas duas áreas aos elementos disponíveis numa base de dados prosopográfica, tornou possível identificar e classificar relacionamentos entre setores de ocupação na forma como eram designados na época, setores de atividade num formato mais próximo do de hoje e o estatuto social que essas incumbências tinham na sociedade coeva. Os dados utilizados são sobretudo de membros do Santo Ofício – do século XVI ao século XVIII. Para atingir este objetivo utilizaram-se algumas descrições textuais de ocorrências da época e outras pouco estruturadas, disponíveis no repositório SPARES. A aplicação de processamento de linguagem natural (remoção de stopwords e aplicação de stemming), conjugada com a construção de duas ontologias, tornou possível classificar esses dados, permitindo consultas mais eficazes. Ao contribuir para a classificação automática de dados históricos, propõem-se metodologias que podem ser aplicadas em dados de qualquer outra área do conhecimento, especialmente as que lidam com as variáveis de tempo e espaço de forma mais intensa; Abstract: OntoSPARES: from natural language to ontologies Contributions to the automatic classification of historical data (16th-18th centuries) The interaction between the natural language processing and ontologies are tools allowing a better understanding of the data stored. This work, by combining these two areas to the elements available in a prosopographic database, has made possible to identify and classify relationships between occupations of many individuals (in general Holy Office members of the 16th-18th centuries). To achieve this goal the data used was gathered in SPARES repository, including some textual descriptions of the time occurrences. They are all few structured. The application of natural language processing (stopwords removal and stemming application), combined with the construction of two ontologies, made possible to classify those data, allowing a more effective search. By contributing to the automatic classification of historical data, this thesis proposes methodologies that can be applied to data from any other field of knowledge, specially data dealing with time and space variables.

Methods for integrating machine learning and constrained optimization

In the framework of industrial problems, the application of Constrained Optimization is known to have overall very good modeling capability and performance and stands as one of the most powerful, explored, and exploited tool to address prescriptive tasks. The number of applications is huge, ranging from logistics to transportation, packing, production, telecommunication, scheduling, and much more. The main reason behind this success is to be found in the remarkable effort put in the last decades by the OR community to develop realistic models and devise exact or approximate methods to solve the largest variety of constrained or combinatorial optimization problems, together with the spread of computational power and easily accessible OR software and resources. On the other hand, the technological advancements lead to a data wealth never seen before and increasingly push towards methods able to extract useful knowledge from them; among the data-driven methods, Machine Learning techniques appear to be one of the most promising, thanks to its successes in domains like Image Recognition, Natural Language Processes and playing games, but also the amount of research involved. The purpose of the present research is to study how Machine Learning and Constrained Optimization can be used together to achieve systems able to leverage the strengths of both methods: this would open the way to exploiting decades of research on resolution techniques for COPs and constructing models able to adapt and learn from available data. In the first part of this work, we survey the existing techniques and classify them according to the type, method, or scope of the integration; subsequently, we introduce a novel and general algorithm devised to inject knowledge into learning models through constraints, Moving Target. In the last part of the thesis, two applications stemming from real-world projects and done in collaboration with Optit will be presented.

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.

Ethics in the flesh: formalizing moral values in embodied cognition

Values are beliefs or principles that are deemed significant or desirable within a specific society or culture, serving as the fundamental underpinnings for ethical and socio-behavioral norms. The objective of this research is to explore the domain encompassing moral, cultural, and individual values. To achieve this, we employ an ontological approach to formally represent the semantic relations within the value domain. The theoretical framework employed adopts Fillmore’s frame semantics, treating values as semantic frames. A value situation is thus characterized by the co-occurrence of specific semantic roles fulfilled within a given event or circumstance. Given the intricate semantics of values as abstract entities with high social capital, our investigation extends to two interconnected domains. The first domain is embodied cognition, specifically image schemas, which are cognitive patterns derived from sensorimotor experiences that shape our conceptualization of entities in the world. The second domain pertains to emotions, which are inherently intertwined with the realm of values. Consequently, our approach endeavors to formalize the semantics of values within an embodied cognition framework, recognizing values as emotional-laden semantic frames. The primary ontologies proposed in this work are: (i) ValueNet, an ontology network dedicated to the domain of values; (ii) ISAAC, the Image Schema Abstraction And Cognition ontology; and (iii) EmoNet, an ontology for theories of emotions. The knowledge formalization adheres to established modeling practices, including the reuse of semantic web resources such as WordNet, VerbNet, FrameNet, DBpedia, and alignment to foundational ontologies like DOLCE, as well as the utilization of Ontology Design Patterns. These ontological resources are operationalized through the development of a fully explainable frame-based detector capable of identifying values, emotions, and image schemas generating knowledge graphs from from natural language, leveraging the semantic dependencies of a sentence, and allowing non trivial higher layer knowledge inferences.

Distributed argumentation technology: advancing risk analysis and regulatory compliance of distributed ledger technologies for transaction and management of securities

Distributed argumentation technology is a computational approach incorporating argumentation reasoning mechanisms within multi-agent systems. For the formal foundations of distributed argumentation technology, in this thesis we conduct a principle-based analysis of structured argumentation as well as abstract multi-agent and abstract bipolar argumentation. The results of the principle-based approach of these theories provide an overview and guideline for further applications of the theories. Moreover, in this thesis we explore distributed argumentation technology using distributed ledgers. We envision an Intelligent Human-input-based Blockchain Oracle (IHiBO), an artificial intelligence tool for storing argumentation reasoning. We propose a decentralized and secure architecture for conducting decision-making, addressing key concerns of trust, transparency, and immutability. We model fund management with agent argumentation in IHiBO and analyze its compliance with European fund management legal frameworks. We illustrate how bipolar argumentation balances pros and cons in legal reasoning in a legal divorce case, and how the strength of arguments in natural language can be represented in structured arguments. Finally, we discuss how distributed argumentation technology can be used to advance risk management, regulatory compliance of distributed ledgers for financial securities, and dialogue techniques.

Advancing digital healthcare research and machine learning-aided knowledge extraction amidst the COVID-19 pandemic

The rapid progression of biomedical research coupled with the explosion of scientific literature has generated an exigent need for efficient and reliable systems of knowledge extraction. This dissertation contends with this challenge through a concentrated investigation of digital health, Artificial Intelligence, and specifically Machine Learning and Natural Language Processing's (NLP) potential to expedite systematic literature reviews and refine the knowledge extraction process. The surge of COVID-19 complicated the efforts of scientists, policymakers, and medical professionals in identifying pertinent articles and assessing their scientific validity. This thesis presents a substantial solution in the form of the COKE Project, an initiative that interlaces machine reading with the rigorous protocols of Evidence-Based Medicine to streamline knowledge extraction. In the framework of the COKE (“COVID-19 Knowledge Extraction framework for next-generation discovery science”) Project, this thesis aims to underscore the capacity of machine reading to create knowledge graphs from scientific texts. The project is remarkable for its innovative use of NLP techniques such as a BERT + bi-LSTM language model. This combination is employed to detect and categorize elements within medical abstracts, thereby enhancing the systematic literature review process. The COKE project's outcomes show that NLP, when used in a judiciously structured manner, can significantly reduce the time and effort required to produce medical guidelines. These findings are particularly salient during times of medical emergency, like the COVID-19 pandemic, when quick and accurate research results are critical.

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.

Quality and Aspect based Argument Generation

Natural Language Processing has always been one of the most popular topics in Artificial Intelligence. Argument-related research in NLP, such as argument detection, argument mining and argument generation, has been popular, especially in recent years. In our daily lives, we use arguments to express ourselves. The quality of arguments heavily impacts the effectiveness of our communications with others. In professional fields, such as legislation and academic areas, arguments of good quality play an even more critical role. Therefore, argument generation with good quality is a challenging research task that is also of great importance in NLP. The aim of this work is to investigate the automatic generation of arguments with good quality, according to the given topic, stance and aspect (control codes). To achieve this goal, a module based on BERT [17] which could judge an argument's quality is constructed. This module is used to assess the quality of the generated arguments. Another module based on GPT-2 [19] is implemented to generate arguments. Stances and aspects are also used as guidance when generating arguments. After combining all these models and techniques, the ranks of the generated arguments could be acquired to evaluate the final performance. This dissertation describes the architecture and experimental setup, analyzes the results of our experimentation, and discusses future directions.

SynBA: A contextualized Synonim-Based adversarial Attack for text classification

With the advent of high-performance computing devices, deep neural networks have gained a lot of popularity in solving many Natural Language Processing tasks. However, they are also vulnerable to adversarial attacks, which are able to modify the input text in order to mislead the target model. Adversarial attacks are a serious threat to the security of deep neural networks, and they can be used to craft adversarial examples that steer the model towards a wrong decision. In this dissertation, we propose SynBA, a novel contextualized synonym-based adversarial attack for text classification. SynBA is based on the idea of replacing words in the input text with their synonyms, which are selected according to the context of the sentence. We show that SynBA successfully generates adversarial examples that are able to fool the target model with a high success rate. We demonstrate three advantages of this proposed approach: (1) effective - it outperforms state-of-the-art attacks by semantic similarity and perturbation rate, (2) utility-preserving - it preserves semantic content, grammaticality, and correct types classified by humans, and (3) efficient - it performs attacks faster than other methods.

Hierarchical Multi-Label Text Classification in a Low-Resource Setting

In this thesis we address a multi-label hierarchical text classification problem in a low-resource setting and explore different approaches to identify the best one for our case. The goal is to train a model that classifies English school exercises according to a hierarchical taxonomy with few labeled data. The experiments made in this work employ different machine learning models and text representation techniques: CatBoost with tf-idf features, classifiers based on pre-trained models (mBERT, LASER), and SetFit, a framework for few-shot text classification. SetFit proved to be the most promising approach, achieving better performance when during training only a few labeled examples per class are available. However, this thesis does not consider all the hierarchical taxonomy, but only the first two levels: to address classification with the classes at the third level further experiments should be carried out, exploring methods for zero-shot text classification, data augmentation, and strategies to exploit the hierarchical structure of the taxonomy during training.

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.