Simplification of Finance Articles
Name
Kristin Petersel
Abstract
Many financial articles contain field-specific vocabulary, which can be difficult for the ordinary reader to understand. Based on articles from frequently consumed sources, the Master's thesis examines, how much complex vocabulary they contain and how successfully can language technology make the text easier for the reader. External readers marked complex words in the articles, which are simplified by two methods: translation of the text using Neurotõlge by the University of Tartu and synonymization of complicated words using the EstNLTK Wordnet synonym set and the pre-trained skip-gram model. The analysis showed that complex words accounted for nearly 2% of all words in the articles under consideration. In order to simplify them, the results of the translation method were generally more successful, as the synonym for a complex word was not always easier to understand or there was no synonym for the word in Wordnet.
Graduation Thesis language
Estonian
Graduation Thesis type
Master - Data Science
Supervisor(s)
Mark Fišel, Lehar Oha
Defence year
2022