Paper Details
Authors: Md. Sajid Alam Chowdhury, Mostak Mahmud Chowdhury, Anik Mahmud Shanto, Hasan Murad, Udoy Das.
Venue: Proceedings of the Second Arabic Natural Language Processing Conference, pages 410-414, Bangkok, Thailand, 2024.
Status: Published · DOI: 10.18653/v1/2024.arabicnlp-1.36.
Abstract
In the financial industry, identifying user intent from text inputs is crucial for automated trading, sentiment analysis, and customer support. Intent detection is significant for finance, but limited studies have addressed low-resource languages such as Arabic compared with high-resource languages such as English. AraFinNLP 2024 introduced a shared task for detecting banking intents from queries in various Arabic dialects using the ArBanking77 dataset, which contains banking queries across 77 intent classes. We presented Dual-Phase-BERT, a hierarchical approach in which dialect detection is performed first, followed by banking intent detection. We trained and evaluated conventional machine learning, deep learning, and transformer-based models on the shared dataset. Dual-Phase-BERT ranked 7th among competitors with an F1 score of 0.801 on the test set.
BibTeX
@inproceedings{chowdhury-etal-2024-fired-arafinnlp,
title = {{F}ired{\_}from{\_}{NLP} at {A}ra{F}in{NLP} 2024: Dual-Phase-{BERT} - A Fine-Tuned Transformer-Based Model for Multi-Dialect Intent Detection in The Financial Domain for The {A}rabic Language},
author = {Chowdhury, Md. Sajid Alam and Chowdhury, Mostak Mahmud and Shanto, Anik Mahmud and Murad, Hasan and Das, Udoy},
booktitle = {Proceedings of the Second Arabic Natural Language Processing Conference},
pages = {410--414},
year = {2024},
publisher = {Association for Computational Linguistics},
doi = {10.18653/v1/2024.arabicnlp-1.36},
url = {https://aclanthology.org/2024.arabicnlp-1.36/}
}