1. Αποθετήριο Παραδοτέων Υποτρόφων Δράσεων ΕΣΠΑ Τριτοβάθμιας Εκπαίδευσης 2017-2023
2. Όλο το Αποθετήριο
3. Υποψήφιοι διδάκτορες

Τύπος:	Διδακτορική διατριβή
Τίτλος:	Natural language processing and information extraction
Εναλλακτικός τίτλος:	Επεξεργασία φυσικής γλώσσας και εξαγωγή πληροφοριών από κείμενα
Συγγραφέας:	[EL] Στυλιανού, Νικόλαος[EN] Stylianou, Nikolaos
Επιβλέπων διατριβής:	[EL] Βλαχάβας, Ιωάννης[EN] Vlahavas, Ioannis
Συμβουλευτική επιτροπή:	[EL] Βακάλη, Αθηνά[EN] Vakali, Athina [EL] Τσουμάκας, Γρηγόριος[EN] Tsoumakas, Grigorios
Μέλος εξεταστικής επιτροπής:	[EL] Βασιλειάδης, Νικόλαος[EN] Bassiliades, Nick [EL] Τέφας, Αναστάσιος[EN] Tefas, Anastasios [EL] Κουμπαράκης, Μανόλης[EN] Koubarakis, Manolis [EL] Αλέτρας, Νικόλαος[EN] Aletras, Nikolaos
Ημερομηνία:	14/10/2021
Περίληψη:	This thesis presents original research in the subject of Machine Learning and more specifically in the fields of Natural Language Processing and Information Extraction. We focus on the following research problems which concern specific tasks in Natural Language Processing and Information Extraction: a) improving clinical decision making through Biomedical Entity Recognition, b) advancing Biomedical Argumentation Mining, c) efficient Language Modeling with distant contextual information and d) deploying Natural Language Processing applications in the real world. First, we present a series of novel architectures for refined Biomedical Entity Recognition, with specific focus in Evidence-Based Medicine entities. These semantically rich entities, which are more descriptive than generic biomedical entity types, offer useful insights in the treatment formulation process and are harder for Machine Learning models to identify. The incrementally proposed changes to the Deep Neural Network architectures intend to improve the performance in all biomedical entity categories and provide more efficient solutions. We also explore the use of Evidence-Based Medicine entities towards the extraction of conclusions from medical publications. To further enhance the clinical practice, we expand our approach of conclusion extraction to create argumentative structures of inference from medical publications. We present a novel approach to fully handle Biomedical Argumentation Mining that incorporates Evidence-Based Medicine entities. The final system has increased performance in all Argumentation Mining sub-tasks and creates argument graphs with higher level information. Capitalizing on the importance of semantic entities, we present two methodologies to incorporate coreferent information in Language Modeling. Combining Coreference Resolution with Language Modeling, we introduce new architectures to efficiently use coreferent mentions and create latent entity representations. The resulting language models have better performance in Language Modeling and in downstream tasks, with minimum added complexity. Finally, we complete a number of case studies for the implementation of Natural Language Processing techniques in the real world and investigate their added value. We present a real-time Sentiment Analysis platform with location inference techniques, a framework for the transformation of large document repositories to Knowledge Graphs and an open-source Machine Learning platform that makes research contributions more accessible to the industry.
Γλώσσα:	Αγγλικά
Τόπος δημοσίευσης:	Θεσσαλονίκη, Ελλάδα
Σελίδες:	214
Θεματική κατηγορία:	[EL] Τεχνητή νοημοσύνη[EN] Artificial Intelligence
Λέξεις-κλειδιά:	Επεξεργασία φυσικής γλώσσας; βαθιά μάθηση; Μηχανική μάθηση; τεχνητή νοημοσύνη; machine learning; natural language processing; Εξαγωγή πληροφοριών
Κάτοχος πνευματικών δικαιωμάτων:	© 2021 Nikolaos Stylianou
Σημειώσεις:	This research is co-financed by Greece and the European Union (European Social Fund - ESF) through the Operational Programme “Human Resources Development, Education and Lifelong Learning” in the context of the project “Strengthening Human Resources Research Potential via Doctorate Research” (MIS-5000432), implemented by the State Scholarships Foundation (IKY).
Εμφανίζεται στις συλλογές:	Υποψήφιοι διδάκτορες