DEVELOPMENT OF KNOWLEDGE GRAPH FOR LEGAL RULE SEARCH IN INDONESIAN LAWS AND REGULATIONS BASED ON TEXT SIMILARITY AND SEMATCH TRAVERSAL
The legal framework in Indonesia faces challenges in the form of several overlapping and inconsistent regulations, which hinder the ease of access and analysis of legal documents. This study aims to develop a query-based search system utilizing a Knowledge Graph (KG) to analyze inter-article relatio...
Saved in:
| Main Author: | |
|---|---|
| Format: | Theses |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/87713 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | The legal framework in Indonesia faces challenges in the form of several overlapping and inconsistent regulations, which hinder the ease of access and analysis of legal documents. This study aims to develop a query-based search system utilizing a Knowledge Graph (KG) to analyze inter-article relationships within legal documents, particularly in the health sector. The system is expected to support more efficient access to legal information.
The Knowledge Graph construction process involves converting legal texts into nodes and edges, as well as applying one of several text similarity methods, including TF-IDF, IndoBERT, Indo-LegalBERT, and SentenceBERT. Experiments were conducted to compare the performance of these text similarity methods to identify the best one, including analyzing similarity score thresholds to determine the optimal value. Additionally, the query-based search system was evaluated using the User Acceptance Testing (UAT) technique by legal experts to assess the relevance of search results to user needs.
The Knowledge Graph development demonstrated that the program successfully converted legal texts into nodes and edges that represent inter-article and inter-section relationships within legal documents. Experimental results showed that the TF-IDF method outperformed semantic similarity methods. A similarity threshold of 0.5 was selected as the optimal value to produce terminologically and contextually relevant search results. Based on the evaluation by legal experts, the system was able to generate accurate and article-focused search results.
This study concludes that the query-based search system utilizing a Knowledge Graph can help improve the ease of access to legal documents in Indonesia. The system is beneficial for legal analysts in supporting more efficient analysis and interpretation of legal documents. |
|---|