Nurfaizy, Muh. (2024) Ekspansi Kueri yang Terkontekstualisasi untuk Pemeringkatan Informasi Lowongan Pekerjaan Menggunakan Multilingual Bidirectional Encoders Representations From Transformers. Other thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 06111940000078-Undergraduate_Thesis.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
06111940000078-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 April 2026. Download (2MB) | Request a copy |
Abstract
Terdapat ribuan mahasiswa yang lulus dari program sarjana di setiap universitas di Indonesia setiap tahunnya. Ini menyebabkan informasi lowongan pekerjaan menjadi salah satu jenis informasi yang banyak diakses. Informasi lowongan pekerjaan ini sangat berlimpah di situs job portal dikarenakan kemudahan dalam mengunggah informasi bagi pencari tenaga kerja. Oleh karena itu, perlu adanya sistem pencarian informasi lowongan pekerjaan yang efektif sehingga dapat membantu user. Sistem tersebut dinamakan Information Retrieval (IR). Informasi yang paling relevan kemudian diperingkatkan berdasarkan nilai relevansinya dari nilai yang paling tinggi ke yang paling rendah. Salah satu metode IR yang banyak digunakan belakangan ini ialah Query Expansion (QE). Perkembangan Natural Language Processing (NLP) turut serta mendorong perkembangan IR. Dampaknya ialah word embedding marak diaplikasikan kedalam sistem IR. Penelitian ini mencoba mengaplikasikan Large Language Models (LLMs) multilingual Bidirectional Representations From Transformers (mBERT) yang merupakan salah satu Language Representasion Models (LRM) terbaru dengan metode Ekspansi Kueri. Dengan menerapkan Contextualized Query Expansion (CQE), hasil pemeringkatan informasi lowongan pekerjaan dengan ekspansi kueri memperoleh skor rata-rata evaluasi NDGC@25 dan NDCG@50 sebesar 0.8967 dan 0.8803. Skor evaluasi tersebut meningkat dibandingkan dengan evaluasi hasil pemeringkatan informasi lowongan pekerjaan sebelum ekspansi kueri yang memperoleh skor rata-rata NDCG@25 dan NDCG@50 sebesar 0.8735 dan 0.8661.
=============================================================================================================================
Thousands of students graduate from undergraduate programs at every university in Indonesia each year. This has led job vacancy information to become one of the most accessed types of information. Job vacancy information is abundant on job portal websites due to the ease of uploading information for job seekers. Therefore, there is a need for an effective job vacancy information retrieval system to assist users. This system is known as Information Retrieval (IR). The most relevant information is then ranked based on its relevance score, from the highest to the lowest. One widely used IR method in recent times is Query Expansion (QE). The development of Natural Language Processing (NLP) has also contributed to the advancement of IR. As a result, word embedding is widely applied in IR systems and has proven to be effective.This research attempts to apply multilingual Bidirectional Representations from Transformers (mBERT), which is one of the latest Language Representation Models (LRM), using the Query Expansion (QE) method. By implementing the Contextualized Query Expansion, the ranking results of job vacancy information achieved an average evaluation score of 0.8967 for NDGC@25 and 0.8803 for NDCG@50. This represents an improvement compared to the ranking results of job vacancy information without Query Expansion, which obtained average scores of 0.8735 and 0.8661 for NDCG@25 and NDCG@50.
| Item Type: | Thesis (Other) |
|---|---|
| Uncontrolled Keywords: | Lowongan Pekerjaan, Information Retrieval, mBERT, Contextualized Query Expansion, Job Vacancy |
| Subjects: | Q Science > QA Mathematics > QA336 Artificial Intelligence |
| Divisions: | Faculty of Science and Data Analytics (SCIENTICS) > Mathematics > 44201-(S1) Undergraduate Thesis |
| Depositing User: | Muh. Nurfaizy |
| Date Deposited: | 19 Feb 2024 04:07 |
| Last Modified: | 19 Feb 2024 04:07 |
| URI: | http://repository.its.ac.id/id/eprint/107459 |
Actions (login required)
 |
View Item |