Entity extraction e text classification come strumenti per la sentiment analysis di testi giornalistici

In this thesis we are going to talk about technologies which allow us to approach sentiment analysis on newspapers articles. The final goal of this work is to help social scholars to do content analysis on big corpora of texts in a faster way thanks to the support of automatic text classification.

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.

Metodi di previsione di indici di borsa con tecniche di text & opinion mining applicate a twitter

Il problema relativo alla predizione, la ricerca di pattern predittivi all‘interno dei dati, è stato studiato ampiamente. Molte metodologie robuste ed efficienti sono state sviluppate, procedimenti che si basano sull‘analisi di informazioni numeriche strutturate. Quella testuale, d‘altro canto, è una tipologia di informazione fortemente destrutturata. Quindi, una immediata conclusione, porterebbe a pensare che per l‘analisi predittiva su dati testuali sia necessario sviluppare metodi completamente diversi da quelli ben noti dalle tecniche di data mining. Un problema di predizione può essere risolto utilizzando invece gli stessi metodi : dati testuali e documenti possono essere trasformati in valori numerici, considerando per esempio l‘assenza o la presenza di termini, rendendo di fatto possibile una utilizzazione efficiente delle tecniche già sviluppate. Il text mining abilita la congiunzione di concetti da campi di applicazione estremamente eterogenei. Con l‘immensa quantità di dati testuali presenti, basti pensare, sul World Wide Web, ed in continua crescita a causa dell‘utilizzo pervasivo di smartphones e computers, i campi di applicazione delle analisi di tipo testuale divengono innumerevoli. L‘avvento e la diffusione dei social networks e della pratica di micro blogging abilita le persone alla condivisione di opinioni e stati d‘animo, creando un corpus testuale di dimensioni incalcolabili aggiornato giornalmente. Le nuove tecniche di Sentiment Analysis, o Opinion Mining, si occupano di analizzare lo stato emotivo o la tipologia di opinione espressa all‘interno di un documento testuale. Esse sono discipline attraverso le quali, per esempio, estrarre indicatori dello stato d‘animo di un individuo, oppure di un insieme di individui, creando una rappresentazione dello stato emotivo sociale. L‘andamento dello stato emotivo sociale può condizionare macroscopicamente l‘evolvere di eventi globali? Studi in campo di Economia e Finanza Comportamentale assicurano un legame fra stato emotivo, capacità nel prendere decisioni ed indicatori economici. Grazie alle tecniche disponibili ed alla mole di dati testuali continuamente aggiornati riguardanti lo stato d‘animo di milioni di individui diviene possibile analizzare tali correlazioni. In questo studio viene costruito un sistema per la previsione delle variazioni di indici di borsa, basandosi su dati testuali estratti dalla piattaforma di microblogging Twitter, sotto forma di tweets pubblici; tale sistema include tecniche di miglioramento della previsione basate sullo studio di similarità dei testi, categorizzandone il contributo effettivo alla previsione.

Generazione automatica di una knowledge base con applicazione ai Sustainable Development Goals

Questa tesi di laurea compie uno studio sull’ utilizzo di tecniche di web crawling, web scraping e Natural Language Processing per costruire automaticamente un dataset di documenti e una knowledge base di coppie verbo-oggetto utilizzabile per la classificazione di testi. Dopo una breve introduzione sulle tecniche utilizzate verrà presentato il metodo di generazione, prima in forma teorica e generalizzabile a qualunque classificazione basata su un insieme di argomenti, e poi in modo specifico attraverso un caso di studio: il software SDG Detector. In particolare quest ultimo riguarda l’applicazione pratica del metodo esposto per costruire una raccolta di informazioni utili alla classificazione di documenti in base alla presenza di uno o più Sustainable Development Goals. La parte relativa alla classificazione è curata dal co-autore di questa applicazione, la presente invece si concentra su un’analisi di correttezza e performance basata sull’espansione del dataset e della derivante base di conoscenza.

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.

Meta-learning for informed model searches in Automated Machine Learning

Day by day, machine learning is changing our lives in ways we could not have imagined just 5 years ago. ML expertise is more and more requested and needed, though just a limited number of ML engineers are available on the job market, and their knowledge is always limited by an inherent characteristic of theirs: they are humans. This thesis explores the possibilities offered by meta-learning, a new field in ML that takes learning a level higher: models are trained on other models' training data, starting from features of the dataset they were trained on, inference times, obtained performances, to try to understand the relationship between a good model and the way it was obtained. The so-called metamodel was trained on data collected by OpenML, the largest ML metadata platform that's publicly available today. Datasets were analyzed to obtain meta-features that describe them, which were then tied to model performances in a regression task. The obtained metamodel predicts the expected performances of a given model type (e.g., a random forest) on a given ML task (e.g., classification on the UCI census dataset). This research was then integrated into a custom-made AutoML framework, to show how meta-learning is not an end in itself, but it can be used to further progress our ML research. Encoding ML engineering expertise in a model allows better, faster, and more impactful ML applications across the whole world, while reducing the cost that is inevitably tied to human engineers.

Frame-driven Extraction of Linked Data and Ontologies from Text

Ontology design and population -core aspects of semantic technologies- re- cently have become fields of great interest due to the increasing need of domain-specific knowledge bases that can boost the use of Semantic Web. For building such knowledge resources, the state of the art tools for ontology design require a lot of human work. Producing meaningful schemas and populating them with domain-specific data is in fact a very difficult and time-consuming task. Even more if the task consists in modelling knowledge at a web scale. The primary aim of this work is to investigate a novel and flexible method- ology for automatically learning ontology from textual data, lightening the human workload required for conceptualizing domain-specific knowledge and populating an extracted schema with real data, speeding up the whole ontology production process. Here computational linguistics plays a fundamental role, from automati- cally identifying facts from natural language and extracting frame of relations among recognized entities, to producing linked data with which extending existing knowledge bases or creating new ones. In the state of the art, automatic ontology learning systems are mainly based on plain-pipelined linguistics classifiers performing tasks such as Named Entity recognition, Entity resolution, Taxonomy and Relation extraction [11]. These approaches present some weaknesses, specially in capturing struc- tures through which the meaning of complex concepts is expressed [24]. Humans, in fact, tend to organize knowledge in well-defined patterns, which include participant entities and meaningful relations linking entities with each other. In literature, these structures have been called Semantic Frames by Fill- 6 Introduction more [20], or more recently as Knowledge Patterns [23]. Some NLP studies has recently shown the possibility of performing more accurate deep parsing with the ability of logically understanding the structure of discourse [7]. In this work, some of these technologies have been investigated and em- ployed to produce accurate ontology schemas. The long-term goal is to collect large amounts of semantically structured information from the web of crowds, through an automated process, in order to identify and investigate the cognitive patterns used by human to organize their knowledge.

New markov chain based methods for single and cross-domain sentiment classification

Nowadays communication is switching from a centralized scenario, where communication media like newspapers, radio, TV programs produce information and people are just consumers, to a completely different decentralized scenario, where everyone is potentially an information producer through the use of social networks, blogs, forums that allow a real-time worldwide information exchange. These new instruments, as a result of their widespread diffusion, have started playing an important socio-economic role. They are the most used communication media and, as a consequence, they constitute the main source of information enterprises, political parties and other organizations can rely on. Analyzing data stored in servers all over the world is feasible by means of Text Mining techniques like Sentiment Analysis, which aims to extract opinions from huge amount of unstructured texts. This could lead to determine, for instance, the user satisfaction degree about products, services, politicians and so on. In this context, this dissertation presents new Document Sentiment Classification methods based on the mathematical theory of Markov Chains. All these approaches bank on a Markov Chain based model, which is language independent and whose killing features are simplicity and generality, which make it interesting with respect to previous sophisticated techniques. Every discussed technique has been tested in both Single-Domain and Cross-Domain Sentiment Classification areas, comparing performance with those of other two previous works. The performed analysis shows that some of the examined algorithms produce results comparable with the best methods in literature, with reference to both single-domain and cross-domain tasks, in $2$-classes (i.e. positive and negative) Document Sentiment Classification. However, there is still room for improvement, because this work also shows the way to walk in order to enhance performance, that is, a good novel feature selection process would be enough to outperform the state of the art. Furthermore, since some of the proposed approaches show promising results in $2$-classes Single-Domain Sentiment Classification, another future work will regard validating these results also in tasks with more than $2$ classes.

A text mining approach to materiality assessment

Worldwide companies currently make a significant effort in performing the materiality analysis, whose aim is to explain corporate sustainability in an annual report. Materiality reflects what are the most important social, economic and environmental issues for a company and its stakeholders. Many studies and standards have been proposed to establish what are the main steps to follow to identify the specific topics to be included in a sustainability report. However, few existing quantitative and structured approaches help understanding how to deal with the identified topics and how to prioritise them to effectively show the most valuable ones. Moreover, the use of traditional approaches involves a long-lasting and complex procedure where a lot of people have to be reached and interviewed and several companies' reports have to be read to extrapolate the material topics to be discussed in the sustainability report. This dissertation aims to propose an automated mechanism to gather stakeholders and the company's opinions identifying relevant issues. To accomplish this purpose, text mining techniques are exploited to analyse textual documents written by either a stakeholder or the reporting company. It is then extracted a measure of how much a document deals with some defined topics. This kind of information is finally manipulated to prioritise topics based on how the author's opinion matters. The entire work is based upon a real case study in the domain of telecommunications.

Edge-aware Graph Attention Networks: Joint Reasoning On Text and Knowledge Graphs for Biomedical Question Answering

Much of the real-world dataset, including textual data, can be represented using graph structures. The use of graphs to represent textual data has many advantages, mainly related to maintaining a more significant amount of information, such as the relationships between words and their types. In recent years, many neural network architectures have been proposed to deal with tasks on graphs. Many of them consider only node features, ignoring or not giving the proper relevance to relationships between them. However, in many node classification tasks, they play a fundamental role. This thesis aims to analyze the main GNNs, evaluate their advantages and disadvantages, propose an innovative solution considered as an extension of GAT, and apply them to a case study in the biomedical field. We propose the reference GNNs, implemented with methodologies later analyzed, and then applied to a question answering system in the biomedical field as a replacement for the pre-existing GNN. We attempt to obtain better results by using models that can accept as input both node and edge features. As shown later, our proposed models can beat the original solution and define the state-of-the-art for the task under analysis.

Automatic classification of stellar orbits in axisymmetric potentials

Within the classification of orbits in axisymmetric stellar systems, we present a new algorithm able to automatically classify the orbits according to their nature. The algorithm involves the application of the correlation integral method to the surface of section of the orbit; fitting the cumulative distribution function built with the consequents in the surface of section of the orbit, we can obtain the value of its logarithmic slope m which is directly related to the orbit’s nature: for slopes m ≈ 1 we expect the orbit to be regular, for slopes m ≈ 2 we expect it to be chaotic. With this method we have a fast and reliable way to classify orbits and, furthermore, we provide an analytical expression of the probability that an orbit is regular or chaotic given the logarithmic slope m of its correlation integral. Although this method works statistically well, the underlying algorithm can fail in some cases, misclassifying individual orbits under some peculiar circumstances. The performance of the algorithm benefits from a rich sampling of the traces of the SoS, which can be obtained with long numerical integration of orbits. Finally we note that the algorithm does not differentiate between the subtypes of regular orbits: resonantly trapped and untrapped orbits. Such distinction would be a useful feature, which we leave for future work. Since the result of the analysis is a probability linked to a Gaussian distribution, for the very definition of distribution, some orbits even if they have a certain nature are classified as belonging to the opposite class and create the probabilistic tails of the distribution. So while the method produces fair statistical results, it lacks in absolute classification precision.