Enhancing Freelancer Project Matching with a BERT-Powered Deep Learning Indonesian Chatbot

The rapid expansion of the freelancing sector has underscored the necessity for more efficient project matching systems. Traditional methods, which largely depend on keyword searches and manual reviews, are often slow and error-prone. This paper is developing and evaluating Indonesian chatbot powered by Bidirectional Encoder Representations from Transformers (BERT), aimed at enhancing freelancer project matching. The proposed chatbot utilizes advanced natural language processing (NLP) techniques to deliver personalized project recommendations based on users' skills and experiences. It leverages BERT’s bidirectional training approach for better context understanding and integrates Long Short-Term Memory (LSTM) networks to maintain context over extended interactions. The system’s adaptability allows for continuous learning and updating, accommodating new fields and trends. Data was collected using ChatGPT 4.0 and was processed to create a dataset focused on Information Technology, Business Management, and Graphic Design. The models, including Recurrent Neural Networks (RNN) and LSTM, demonstrated impressive performance with LSTM showing superior results in 97% accuracy and context retention in the chatbot. This research highlights the advantages of using BERT with deep learning models for accurate and efficient project matching, providing a scalable solution for the evolving freelancing landscape. Future work will explore expanding the system to other languages and domains to further enhance its applicability.