Rimarju

Rhyming English words using Natural Language Processing has been done using multiple techniques over recent years, however not many studies have carried out the same process on different languages. The Maltese language and culture has multiple aspects in which rhyming words comes into play, from traditional poetry to ‘Għana’. In creating these artistic forms, having one flowing theme throughout is very important. This paper investigates the various techniques and use of phonetics to find words that rhyme, incorporating these phenomena with a topic modelling system which can rank and group words in terms of context. To achieve this, a rhyme detection technique is utilized, employing the use of the International Phonetic Alphabet (IPA). This approach is able to handle the challenges and complexities inherent when using a model with a Semitic base such as Maltese, which also incorporates influences from Romance languages. This method also takes into account the diverse pronunciations and dialects that may affect rhyme detection, as well as the inclusion of special characters and distinct sounds. The methodology continues by building a topic modelling system based on Latent Dirichlet Allocation (LDA), which is able to successfully detect latent topics and rank word pairs based on relevance to a given theme. In the evaluation, the rhyming system is tested in a subjective manner, having a number of participants evaluate different rhymes it produces to specific words, ranking them using a Likert scale; whilst the topic modelling system uses various graph and statistical techniques to compare different factors such as efficiency and quality of the results produced by the model. To test the quality of the model itself, the results of the Maltese topic modelling system are compared to the same model, but with an English dataset. This experiment highlights the effectiveness of the model when given all the resources and datasets needed. The system is designed to address the challenges of working with a low‐resource language and provide a tool for language learners, poets, and other artists, by creating an interactive and intuitive system that is accessible to anyone, with the addition of a dynamic web page which has been developed to serve as a user interface for the system, providing an easy‐to‐use platform for students, poets, and artists to find rhyming words and generate poetry. The system has limitations due to the scarcity of resources for Maltese, but it provides a foundation for future work in automatic poetry generation and other applications.

Grantor dc:publisher.institution

University of Malta

Year dc:date.issued

2023

• Maltese language -- Phonetics
• Maltese language -- Rhyme
• Natural language processing (Computer science)

Statement dc:rights

• info:eu-repo/semantics/restrictedAccess

Language dc:language.iso

en

Repository record dc:identifier.uri

https://www.um.edu.mt/library/oar/handle/123456789/115290

OAI identifier oai:identifier

oai:www.um.edu.mt:123456789/115290