Transformer-based Models for Long-Form Document Matching: Challenges and Empirical Analysis

Akshita Jha, Adithya Samavedhi, Vineeth Rakesh, Jaideep Chandrashekar, Chandan Reddy


Abstract
Recent advances in the area of long document matching have primarily focused on using transformer-based models for long document encoding and matching. There are two primary challenges associated with these models. Firstly, the performance gain provided by transformer-based models comes at a steep cost – both in terms of the required training time and the resource (memory and energy) consumption. The second major limitation is their inability to handle more than a pre-defined input token length at a time. In this work, we empirically demonstrate the effectiveness of simple neural models (such as feed-forward networks, and CNNs) and simple embeddings (like GloVe, and Paragraph Vector) over transformer-based models on the task of document matching. We show that simple models outperform the more complex BERT-based models while taking significantly less training time, energy, and memory. The simple models are also more robust to variations in document length and text perturbations.
Anthology ID:
2023.findings-eacl.178
Volume:
Findings of the Association for Computational Linguistics: EACL 2023
Month:
May
Year:
2023
Address:
Dubrovnik, Croatia
Editors:
Andreas Vlachos, Isabelle Augenstein
Venue:
Findings
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
2345–2355
Language:
URL:
https://aclanthology.org/2023.findings-eacl.178
DOI:
10.18653/v1/2023.findings-eacl.178
Bibkey:
Cite (ACL):
Akshita Jha, Adithya Samavedhi, Vineeth Rakesh, Jaideep Chandrashekar, and Chandan Reddy. 2023. Transformer-based Models for Long-Form Document Matching: Challenges and Empirical Analysis. In Findings of the Association for Computational Linguistics: EACL 2023, pages 2345–2355, Dubrovnik, Croatia. Association for Computational Linguistics.
Cite (Informal):
Transformer-based Models for Long-Form Document Matching: Challenges and Empirical Analysis (Jha et al., Findings 2023)
Copy Citation:
PDF:
https://aclanthology.org/2023.findings-eacl.178.pdf
Video:
 https://aclanthology.org/2023.findings-eacl.178.mp4