AcuChatbot: A University Intelligent Q&A System Powered by Localized Large Language Models

AcuChatbot is an AI-driven intelligent Q&A system designed specifically for Acıbadem University, aiming to address the inefficiency in accessing information in higher education institutions. It integrates campus network data with locally deployed large language models (LLMs) to provide high-quality academic information services while ensuring data privacy. The system adopts a Retrieval-Augmented Generation (RAG) architecture to avoid model 'hallucinations', supports multi-turn dialogues, and covers various daily academic and administrative needs of students and faculty. It is an innovative practice for higher education institutions to use AI to improve service quality.

In higher education institutions, traditional methods for students and faculty to access academic information (such as course schedules, exam timetables, facility regulations, scholarship procedures, etc.)—like browsing official websites or sending email inquiries—suffer from low efficiency and scattered information. AcuChatbot is designed to solve these pain points and provide instant, accurate academic information services.

Data-Driven Knowledge Base

The system builds a knowledge base through website data crawling and parsing, structured knowledge organization, and real-time data synchronization to ensure answers are based on the latest official information.

Local LLM Integration

Unlike robots relying on cloud APIs, AcuChatbot uses locally deployed LLMs, bringing four major advantages:

• Data Privacy Protection: All queries and responses are processed locally; sensitive information is not transmitted to external servers, complying with regulations like GDPR;
• Low-Latency Response: Eliminates network latency and provides near-instant answers;
• Controllable Costs: No API call fees, making long-term use more economical;
• Customization Capability: The model can be fine-tuned according to the university’s needs to improve performance in academic fields.

Student Service Scenarios

• Course-related inquiries: Query syllabi, prerequisites, credits, and faculty contact information;
• Administrative process guidance: Course registration operations, leave/return to school applications, transcript acquisition;
• Campus life support: Library hours, canteen locations, dormitory regulations;
• Academic resource navigation: Online databases, writing centers, lab reservations.

Faculty and Staff Auxiliary Functions

• Quick query of school calendars, meeting schedules, and committee information;
• Access to research funding application guidelines and deadlines;
• Understand teaching evaluation processes and policy updates.

RAG Architecture

The system adopts a Retrieval-Augmented Generation (RAG) architecture with the following process:

1. Query Understanding: Analyze user intent and key needs;
2. Knowledge Retrieval: Retrieve relevant document fragments from the knowledge base;
3. Context Integration: Input the retrieved information as context into the model;
4. Answer Generation: Generate accurate and coherent answers based on the retrieved content. Advantages of this architecture: Answers are traceable (with source annotations) and effectively avoid model 'hallucinations'.

Multi-Turn Dialogue Management

• Remember dialogue history and understand references and omissions;
• Clarify and follow up on complex queries;
• Adjust answer directions based on user feedback.

Improve Service Efficiency

24/7 online availability, instant answers to common questions, freeing up human customer service resources to handle complex cases.

Promote Information Equity

Provide a stress-free, barrier-free information access channel for international students, new students, or introverted students, narrowing the information gap.

Localized Deployment Trend

Choosing local LLMs reflects the trend of AI applications in education: with tighter data privacy regulations and increased institutional awareness of data sovereignty, localized solutions will be more favored.

AcuChatbot faces the following challenges:

• Knowledge Base Maintenance Costs: Need to establish automated data synchronization and quality monitoring mechanisms;
• Complex Problem Handling: For scenarios like cross-departmental coordination or policy ambiguities, guide users to contact human services;
• Multilingual Support: International universities need to support queries in multiple languages;
• User Acceptance: Some users doubt the reliability of AI, so trust needs to be built through optimization and transparency.

AcuChatbot represents an innovative practice of higher education institutions using AI to improve service quality. It combines LLM capabilities with domain knowledge to provide practical value while ensuring privacy. For other universities, this is a worthy reference: AI is not a replacement for human services, but a powerful supplement—making information access more convenient and allowing human resources to focus on more valuable work.