Naslov
Comparison of Language Networks Measures for Legal Texts and Literature

Autori
Miličić, Tanja ; Meštrović, Ana

Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, sažetak, ostalo

Skup
7th International Conference on Information Technologies and Information Society (ITIS2015)

Mjesto i datum
Novo Mesto, Slovenija, 04.11.2015. - 06.11.2015

Vrsta sudjelovanja
Predavanje

Vrsta recenzije
Međunarodna recenzija

Ključne riječi
Language networks

Sažetak
In the last decade we have witnessed tremendous advances in understanding networked systems across a number of disciplines. One of the reasons for this lies in the discovery that for each system there exists a common set of fundamental laws and principles, despite their diversity. Inspired by complexity theory, it is recently acknowledged that human language can be modeled as a complex network and that it shares a number of non trivial statistical patterns such as small world phenomenon, disassortative mixing, power low degree distribution, etc. As the network model of any other type of real-world system, linguistic networks consist of a set of nodes that represent a linguistic unit (e.g. word) and a set of edges representing the pairwise relations between them. Various linguistic networks have already been analyzed, such as word co-occurrence network, syntactic, syllables or semantic networks. In this experiment co-occurrence language network measures from two different categories of texts are compared on a global and a local level. On a global level, we consider average values of a given measure, while comparison on a local level is performed via rank plots. Networks are constructed in a way that words represent nodes which are in turn joined by an edge if they are adjacent in an area between delimiters. All networks are generated as directed and weighted, where weight of a link between two nodes represents overall co-occurrence frequencies of the corresponding words. Our dataset consists of eight texts divided into two categories of four legal texts and four short novellas both written in English. The reason for choosing this particular text types is their obvious structural and linguistic distinction. The aim of this experiment was to investigate how complex network measures operate in different structures of texts and which of them are sensitive to different text categories. The results of our measuring show that there is no uniform rule to differentiate mentioned styles of texts on a global level. However, local perspective rank plots of average node strength indicate that there are structural differences between legal texts and literature.

Izvorni jezik
Engleski

Znanstvena područja
Računarstvo, Informacijske i komunikacijske znanosti

POVEZANOST RADA