Deep Learning Based Amharic Chatbot for FAQs in Universities

2024-01-26 · Source: cs.AI updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Natural Language Processing, Software Development & Engineering · Depth: Advanced, short

Summary

A deep learning-based chatbot model has been developed to answer frequently asked questions (FAQs) in Amharic for university students. This system aims to reduce the time students and administrators spend on common inquiries. The chatbot employs natural language processing techniques including tokenization, normalization, stop word removal, and stemming to process Amharic input. It utilizes three machine learning algorithms for classification: Support Vector Machine (SVM), Multinomial Naïve Bayes, and a deep neural network. The deep learning model, implemented with TensorFlow, Keras, and NLTK, achieved the highest accuracy at 91.55% with a validation loss of 0.3548, using an Adam optimizer and SoftMax activation. The chatbot is integrated with Facebook Messenger and deployed on a Heroku server, providing 24-hour accessibility and effectively handling Amharic linguistic challenges like Fidel and morphological variations.

Key takeaway

For research scientists developing NLP solutions for low-resource languages, you should consider deep learning models over traditional machine learning for improved accuracy. The demonstrated success with Amharic, achieving 91.55% accuracy, highlights the potential for similar approaches in other complex linguistic contexts. Focus on robust preprocessing and leveraging frameworks like TensorFlow and Keras to overcome language-specific challenges and enhance model performance.

Key insights

A deep learning chatbot effectively answers Amharic university FAQs, achieving 91.55% accuracy.

Principles

• Deep learning outperforms traditional ML for Amharic NLP.
• Chatbots reduce administrative burden in educational settings.

Method

The method involves NLP preprocessing (tokenization, normalization, stop word removal, stemming) followed by classification using SVM, Naïve Bayes, or a deep neural network for response retrieval.

In practice

• Deploy chatbots on platforms like Facebook Messenger.
• Use Adam optimizer and SoftMax for deep learning text classification.
• Address language-specific challenges like morphological variation.

Topics

• Deep Learning
• Amharic Chatbot
• Natural Language Processing
• University FAQs
• TensorFlow

Best for: Research Scientist, AI Scientist, Machine Learning Engineer, NLP Engineer

Related on AIssential

• Natural Language Processing Specialization by DeepLearning.AI — The specialization teaches modern NLP, from foundational concepts to advanced attention models for real-world applications.
• Neural edit-tree lemmatization for spaCy — A new ML-based lemmatizer for spaCy achieves over 95% accuracy by learning rules, eliminating manual linguistic engineering.
• NLP Chat — Simplifying Chatbot Creation with Python — nlpchat simplifies chatbot creation by abstracting NLP complexities, using Sentence Transformers and Logistic Regression for intent classification.
• I spent 4 days training an AI model to understand my country. It barely works — Low-resource NLP projects face significant challenges in data collection, translation, and accurate labeling, often yielding modest baselines.
• Training Azerbaijani language models on Amazon SageMaker AI — Optimizing LLM training for low-resource, morphologically rich languages requires custom tokenization and GPU memory optimizations.
• TalkTag: Fine-Grained Morphosyntactic Error Annotation for Transcribed Speech — TalkTag automates fine-grained morphosyntactic error annotation in spoken transcripts using an LLM, even with scarce data.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.AI updates on arXiv.org.