Adaptation of texts to context

diploma thesis

Luka Žontar (Author), Zoran Bosnić (Mentor)

Abstract

In this thesis we try to develop a methodology that can adapt texts to target publication types using summarization, natural language generation and paraphrasing. The solution is based on key text characteristics that describe different publication types. To examine types such as social media posts, newspaper articles, research articles and official statements, we use three distinct text evaluation metrics: length, text polarity and readability. While altering key text evaluation metrics, we mostly focus on length due to much research that was done in this field (either with summarization or natural language generation). Using paraphrasing we will try to adjust text readability and polarity that describes reader's negative or positive orientation towards the topic. The process of text adaptation will be implemented iteratively. The developed methodology will automatize writing articles that are based on existing articles. The more crucial contribution of this thesis is that we help to gain access of harder works to those that cannot understand the origin texts.

Keywords

text adaptation;context-aware;artificial intelligence;text summarization;natural language processing;computer and information science;diploma;

Data

Language:	English
Year of publishing:	2020
Typology:	2.11 - Undergraduate Thesis
Organization:	UL FRI - Faculty of Computer and Information Science
Publisher:	[L. Žontar]
UDC:	004.8:81'322.2(043.2)
COBISS:	28976643
Views:	1276
Downloads:	175
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	Slovenian
Secondary title:	Prilagoditev besedil kontekstu objave
Secondary abstract:	V tem diplomskem delu skušamo razviti metodologijo, ki bo znala s pomočjo povzemanja, generiranja naravnega jezika in parafraziranja prilagoditi be\-se\-di\-lo ciljnemu kontekstu objave. Za rešitev problema se najprej osredotočimo na ključne karakteristike, ki določajo obravnavane tipe objav. Za obravnavo štirih razlilčnih tipov, ki so objave na socialnih omrežjih, novice, znanstveni članki in uradne izjave, uporabimo tri karakteristike: dolžino, polarnost besedila in berljivost. Pri spreminjanju ključnih karakteristik se osredotočimo predvsem na dolžino, saj se veliko raziskav nanaša na prilagajanje teksta, bodisi s povzemanjem ali generiranjem naravnega jezika. S pomočjo parafraziranja bomo skušali prilagajati berljivost in polarnost teksta, ki priča o negativni oziroma pozitivni naravnanosti bralca k besedilu. Proces prilagajanja besedil bo potekal iterativno. Razvita metodologija bo avtomatizirala proces pisanja člankov, ki temeljijo na obstoječih delih. Še bolj pomembno pa je, da s tem omogočimo dostop zahtevnejših del tistim, ki jih v prvotni obliki ne razumejo.
Secondary keywords:	prilagajanje besedil;kontekstno-odvisen;umetna inteligenca;procesiranje naravnega jezika;računalništvo in informatika;univerzitetni študij;diplomske naloge;
Type (COBISS):	Bachelor thesis/paper
Study programme:	1000468
Embargo end date (OpenAIRE):	1970-01-01
Thesis comment:	Univ. v Ljubljani, Fak. za računalništvo in informatiko
Pages:	79 str.
ID:	12029426