Intent classification for university administrative services using a bidirectional recurrent neural network modified...

Area of Science:

• Artificial Intelligence
• Natural Language Processing
• Educational Technology

Background:

• Conversational agents are transforming university-student interactions by providing instant support.
• Multilingual NLU systems face challenges with separate models for intent classification (IC) and named entity recognition (NER), leading to higher costs and lower performance.
• Existing NLU methods require significant computational resources and memory.

Purpose of the Study:

• To develop a joint NLU model that integrates IC and NER for improved efficiency and accuracy in multilingual educational conversational agents.
• To address the limitations of conventional NLU approaches in complex, multilingual university environments.
• To enhance the student experience and optimize university support services through advanced AI.

Main Methods:

• A novel joint model combining IC and NER using a modified Bidirectional Long Short-Term Memory (BiLSTM) network.
• Optimization of the joint model using a newly developed Kepler optimization (DKO) algorithm.
• Implementation and evaluation of the model in Greek and English for university student assistance.

Main Results:

• The proposed joint NLU model outperformed state-of-the-art models in accuracy, precision, and recall.
• Demonstrated enhanced NLU accuracy with efficient resource utilization for Greek and English.
• The model effectively integrates deep learning with optimization techniques.

Conclusions:

• Joint NLU models show significant potential for cognitive assistants in educational contexts, improving student experience and university support.
• The developed model offers substantial gains in efficiency and accuracy for multilingual AI applications in higher education.
• This research contributes to merging AI and education through multimodal interaction and conversational interface development.