Eugeniu Costetchi is a skillful Software Engineer with over twelve years of experience, proficient in a wide range of technologies and Python, Java and C++ programming languages; furthermore, he is a knowledgeable researcher in Semantic Web, Computational Linguistics and related domains. Eugeniu is a proficient English writer and speaker, an effective communicator and due to his open and friendly personality, an excellent team player.
Eugeniu has been passionate and curious about computers since his late child hood when he started self-schooling about computer hardware, software and programming. In high school, he won regional competitions and participated at the National Informatics Olympiads in Moldova in 1997, and in Romania in 1999 and 2000.
Almost a decade ago, in 2006, after finishing his Bachelor at Alexandru Ioan Cuza University (Iasi, Romania) in Business Information Systems, specialised in Service Oriented and Model Driven Architectures, Eugeniu worked for Inpro Systems Ltd. (Beijing, China) as a software Engineer and Team Leader for two years. There he made essential contributions to the design and development of ChainBuilder ESB project, a Java based messaging infrastructure and a set of adjacent tools.
At the end of 2008, Eugeniu enrolled to an academic Master in Informatio and Computer Science at the University of Luxembourg. There, for almost a year, he partook in two research projects, both related to Wireless Ad-Hoc Mesh Networks: the first one concerning agent behaviour strategies and their impact on the agent’s reputation within the network; the second investigated movement patterns on highways and on ski slopes for the purpose of simulating these mobility models and testing their impact on the topology of ad hoc networks.
In the final semester, to write his dissertation, he did an extended internship at the department of Knowledge Intensive Software Systems (KISS) of the Research Centre Henri Tudor (currently LIST). In his thesis Eugeniu explores problems and possible approaches to building a dialogue system for Automated Ontology Elicitation. Such dialogue system would interview a human expert on a particular topic/domain and automatically propose a conceptual model of that domain in the form of ontology. The thesis explores a range of questions from philosophical issues of knowledge representation and ontologies to pragmatic approaches on how to implement them with existing Semantic Web technologies and Ontology Engineering Methodologies. Eugeniu dives into what constitutes the nature of a question and how can it be theoretically grasped, explores the known models of interviewer-respondent interactions (specifically in classroom environment and in requirements elicitation sessions) and analyses various interviewing techniques to propose a dialogue management model for ontology elicitation constrained by a very simple interaction language.
While writing his thesis, Eugeniu became even more interested in the topics related to modelling, representing and interactively acquiring knowledge/ontologies; in the process of natural language understanding and the question how surface linguistic forms are connected through deeper linguistic meanings to pragmatic domain knowledge. This motivated him to apply for a research grant from the Luxembourgish Ministry of Research that he was granted. Therefore, after successfully defending his distinguished Master thesis, he continued this line of research into a joint PhD programme between the Research Centre Henri Tudor and the University of Bremen under the supervision of Prof. John Bateman and Dr. Eric Ras.
The first part of his PhD focused on deepening the understanding of Semantic Web Technologies and Ontology Building Methodologies. In 2011, after presenting his research project in the poster session of the European Future Technologies Conference and Exhibition (FET11, Budapest, Hungary), he participated at the Extended Semantic Web Conference Summer School (ESWC-SS-2011, Crete, Greece), where he learned from the leading world class experts about the philosophy, historical development, applications and existing technologies of Semantic Web. Later that year, to solidify the gained knowledge and having had made new connections, he visited Oxford, Leeds and Milton Keynes Universities, spending one month at each of them, attending lectures, seminars and project presentations and learning about on-going projects on ontology reasoning, ontology authoring with controlled natural language and ontology versioning and maping. He also participated at the European Summer School in Language, Logic and Information (ESSLLI-2011, Ljubljana, Slovenia) where he received an answer to many of his questions unanswered to that time, and where his concepts regarding connections between natural language and knowledge representation started to crystallize - it was at this moment that he opened up towards understanding of what natural language is and how it works.
Upon his return from the summer school, he implemented a first prototype of a dialogue system for ontology elicitation which still maintained the limitation of a small interaction language and an ontology metrics module to measure different parameters of the resulting ontologies. This dialogue system was properly executing the interview strategies; however the lack of natural language interaction was slowing down the process and making it unfit for use in practice. Fluent interaction, the linguistic competence and ability of the computer to understand natural language has become a key need and a major turning point in his research project taking Eugeniu slowly but surely into the domain of Computational Linguistics and specifically to the task of Natural Language Processing and Understanding.
After having done an extensive state of the art research on existing technologies in Natural Language Processing for English, as well as on major linguistic theories and grammar, Eugeniu realised that none of the existing ones satisfy the requirements of his dialogue systems and started to develop his own Natural Language Parser for English with Systemic Functional Grammar. This work would become his PhD thesis in the end.
By the summer of 2013, based on the newly developed parser named Parsimonious Vole, he published three papers that were presented at the International Conference on Dependency Linguistics (DepLing-2013, Prague, Czech Rep.), the Recent Advancements in Natural Language Processing conference (RANLP-2013, Hissar, Bulgaria) and the student workshop of the European Summer School in Logic, Language and Information (ESSLLI-2013, Dusseldorf, Germany) respectively. He showed how it can we derive rich systemic functional grammatical parses accounting for experiential and interactional aspects of language which could not be accounted for by any of the existing parses at the time, from dependency parses (generated with Stanford Parser) through graph pattern matching.
He spent the next year deepening and perfecting his parser but he still did not have a gold standard corpus for evaluating the accuracy of it. The opportunity came in early 2014, when he went on a short scientific mission to the Department of Psychology at the Tel-Aviv University (Israel) to help researchers there take advantage of existing NLP technologies to analyse texts of people with Obsessive Compulsive Disorders. During this scientific mission, with the help of other PhD students, he created a first gold standard corpus by hand annotating blog texts written by people diagnosed with OCD. According to evaluations carried out with this corpus, the parser performs with an accuracy of 92% for syntactic and interpersonal features and with an accuracy varying between 30 - 70% for various experiential (semantic) features.
In the fall of 2014 he attended the workshop and summer school in Systemic Functional Linguistics (3rd LinC-2014, Cardiff, United Kingdom) and in the summer of 2015 the International Systemic Functional Congress (42nd ISFC-2015, Aachen, Germany), where he received positive feedback from the SFL community on the value and importance of Parsimonious Vole parser. In January 2017 he gave a talk about the importance of cross-theoretic transformations and correspondences at the Cardiff Symposium for Systemic Functional Linguistics.