Back to search

Aristotle University Of Thessaloniki (AUTH)

Ένα σύστημα αυτόματης ανάλυσης κειμένων της Νέας Ελληνικής: μέθοδοι αναπαράστασης των κυρίων ονομάτων προσώπων

Abstract

dc:description

The purpose of this study had been twofold. Firstly, we wanted to construct a morphological electronic dictionary of persons’ proper names in Modern Greek; secondly, we wanted to provide a morphological description of them in order to achieve their automatic recognition and integration in automatic text analysis. Proper names are encountered with great frequency in texts and unless they are part of a database, such as the electronic dictionary that we have created, then the automatic analysis systems treats (faces) them as unknown words. The starting point and inspirational basis of our work has been the morphological electronic dictionary of simple verbs in Modern Greek (DelasGR) developed by Kyriacopoulou T. (1990). The theoretical background of our research specifically concerning the morphological description of proper names of persons was based on the transformational grammar by Z. S. Harris, and our methodological framework was defined by the model of finite state automata, which record the linguistic data in an economic and structured way both for the computer and the linguist. We searched for an adequate definition of the term ‘proper name’ and discussed various theories of linguistic categorization, thus distinguishing it from the entirety of named entities. Furthermore, we analyzed the presence of proper names in Modern Greek according to the morphological, syntactic and semantic features that govern this specific category of nouns. After having clearly and systematically defined the lexical and morpho-syntactic properties of persons’ proper names in Modern Greek, we examined the morphological problems and other problems concerning the ambiguities of names; finally we tackled the difficulties that arose when we tried to encode the names, so as they can be used by a computer. Thus, we created a dictionary of anthroponyms, and then all inflectional types (inflected forms) were automatically compiled and converted into electronic dictionaries compressed in an appropriate way to allow their use by a Natural Language Processing system. The morphological dictionary was used to identify lexical units that occur in texts collected from the internet. Later, we presented the first results of the automatic processing and recognition, which were very encouraging. Then, we turned to the context of proper names, in order to achieve automatic recognition and classification of those anthroponyms that appear in texts under study, but are not contained in prefabricated dictionaries. We described the methodological steps followed and the criteria for the recognition of the studied structures, which are the basis for the compilation of according finite state automata. To achieve automatic recognition we used the method of regular expressions and at the same time we used the semantic features that are already integrated in dictionaries. Having applied these methods of automatic recognition, we observed a significant improvement in the results drawn from our system, since the words that were recovered were multiple in relation to those recovered by the mere application of electronic dictionaries. Furthermore, we observed improvement in the classification of words exported, which became more accurate thanks to the use of attributes that belong to the context of the entries. The combined method we propose, through the use both of our dictionary and the automatic representation of Greek surnames and first names, presents remarkably high rates of retrieving results with quality and validity. Finally, and most importantly, the encouraging results prompting from this specific method can contribute to the onset of automatic analysis of all named entities.

Degree

thesis:*
Grantor dc:publisher
Aristotle University Of Thessaloniki (AUTH)
Year dc:date
2012

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Μαυρόπουλος, Αθανάσιος

Subjects

dc:subject × 11

Rights

Language dc:language
gre

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/32216

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Μαυρόπουλος, Αθανάσιος. Ένα σύστημα αυτόματης ανάλυσης κειμένων της Νέας Ελληνικής: μέθοδοι αναπαράστασης των κυρίων ονομάτων προσώπων. Aristotle University Of Thessaloniki (AUTH), 2012. http://hdl.handle.net/10442/hedi/32216