SIGAP: Indonesia's Government Uses Large Language Models to Revolutionize Public Complaint Handling Systems

Core Insights

The SIGAP (Public Complaint Information System) project in Indonesia was initiated by developer zaakiiirrr (Source: GitHub, published on June 15, 2026). It uses large language model (LLM) technology to convert unstructured public complaint texts into structured data, addressing the pain points of low efficiency and easy omissions in traditional manual processing, improving the response efficiency of government public services, and providing data support for decision-making.

Project Value

This project integrates AI with government services, aiming to optimize the complaint handling process, realize intelligent text understanding, information extraction and summarization, cover multi-scenario applications, and bring significant value to citizens, staff and decision-makers.

In the process of digital transformation, governments around the world face the problem of unstructured public complaint texts:

• The content of complaints is complex and inconsistent in format, leading to low efficiency in manual classification and processing;
• It is prone to omissions and misjudgments, making it difficult to cope with the growing number of complaints;
• The lack of intelligent tools to automatically understand and extract demands restricts the improvement of service quality.

The SIGAP project deeply integrates AI technology with government services, with core objectives:

• Use LLM technology to process unstructured complaint texts;
• Automatically extract key information and convert it into structured table data;
• Improve data processing efficiency and provide reliable support for government decision-making.

The SIGAP system integrates LLM and has the following capabilities:

1. Text Understanding: Deeply understand Indonesian semantics, identify entities, emotions and contextual correlations (different from traditional keyword matching);
2. Information Extraction: Automatically extract structured fields such as complaint category, involved department, geographical location, and urgency level;
3. Intelligent Summarization: Generate concise summaries to help staff quickly grasp key points and prioritize urgent matters.

Application Scenarios

• Urban management: Handle complaints about infrastructure damage, environmental sanitation, traffic order, etc.;
• Public services: Summarize feedback on people's livelihood issues such as medical care, education, and social security;
• Government supervision: Identify common problems in departmental services and room for improvement.

Value Manifestation

• Citizens: Demands are accurately understood, and processing progress is transparent and traceable;
• Staff: Reduce repetitive text processing and focus on problem solving;
• Decision-makers: Structured data supports policy formulation and optimizes resource allocation.

The SIGAP project faces the following technical challenges:

1. Language Diversity: Indonesia's multilingual/dialect environment requires handling language variations;
2. Text Quality: Complaint texts have spelling errors, grammatical irregularities, and colloquial expressions;
3. Privacy and Security: Ensure complaint data (including personal information) complies with data protection regulations;
4. Real-time Performance: Cope with a large number of concurrent complaints, need to optimize model inference and infrastructure scalability.

Insights

• Provide a reference case for global government digital transformation, showing the combination of AI and government needs;
• Verify the feasibility of LLM in low-resource language government scenarios;
• Become an intelligent bridge between the government and the public, providing experience for developing countries.

Outlook

• Integrate voice, image and other multimodal inputs to lower the threshold for citizens to provide feedback;
• Combine knowledge graph and reasoning technology to realize the full-process automation from recording complaints to intelligent triage and solution recommendation, moving towards smart governance.