PERFORMA MODEL NAMED ENTITY RECOGNITION MENGGUNAKAN METODE BLSTM–HMM PADA DOKUMEN TWITTER BERBAHASA INDONESIA

ADI, WIJAYA (2024) PERFORMA MODEL NAMED ENTITY RECOGNITION MENGGUNAKAN METODE BLSTM–HMM PADA DOKUMEN TWITTER BERBAHASA INDONESIA. Undergraduate Thesis thesis, Institut Teknologi Telkom Purwokerto.

	Text Cover.pdf Download (894kB)
	Text Abstrack.pdf Download (524kB)
	Text Abstrak.pdf Download (522kB)
	Text BAB I.pdf Download (474kB)
	Text BAB II.pdf Download (650kB)
	Text BAB III.pdf Download (739kB)
	Text BAB IV.pdf Restricted to Registered users only Download (748kB)
	Text BAB V.pdf Download (459kB)
	Text Daftar Pustaka.pdf Download (529kB)
	Text Lampiran.pdf Restricted to Registered users only Download (509kB)

Abstract

Twitter is a popular microblogging platform in Indonesia, providing text data in Indonesian language that can be utilized for Natural Language Processing (NLP) applications. Named Entity Recognition (NER) is a crucial subtask in Information Extraction (IE) aiming to identify named entities such as persons, locations, and organizations within text. Limitations in Indonesian NER research on the Twitter platform have prompted the urgency to develop optimized NER models. NER plays a significant role in IE and NLP in general, enabling automatic extraction of useful information from Twitter text. The objective of this research is to determine the F1 score performance value in the NER model using the BLSTM-HMM method, with FastText and POS tag features. Also, to ascertain whether the application of these methods can enhance the F1 score performance of the NER model. This study implements the Bidirectional Long Short-Term Memory (BLSTM) – Hidden Markov Model (HMM) method with FastText features and POS tags to enhance the performance of NER models on Indonesian-language Twitter document data. The developed model achieved an F1 score of 89.23%. The combination of the NER model with BLSTM-HMM and the addition of FastText and POS tag features can be an interesting alternative to enhance entity recognition performance. In this study, this combination successfully yielded positive results.

Actions (login required)

View Item