Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

End-to-End Dialog Neural Coreference Resolution: Balancing Efficiency and Accuracy in Large-Scale Systems

Zhang Dong, Songhang deng, Mingbang Wang, Le Dai, Jiyuan Li, Xingzu Liu, Ruilin Nong

TL;DR

This work targets scalable end-to-end neural coreference resolution for large-scale systems, addressing the tension between model size and accuracy. It introduces an End-to-End Neural Coreference Resolution framework that combines multi-layer neural architectures, contextual embeddings, and attention with optimization techniques like pruning and quantization to reduce overhead, formalizing representations such as $ \mathbf{C} = \mathcal{F}(\mathbf{E}, \mathbf{A}) $ and the affinity-based training objective $ \mathcal{L}_{CR} = - (\alpha \log P + \beta \log(1-P)) $. Empirical results on benchmarks including OntoNotes v5.0 demonstrate high F1 scores (e.g., $F1$ $=86.2$ with $\text{BERT-large}$ and $87.3$ with SpanBERT) along with strong precision/recall and rapid inference, outperforming baselines like Z-coref and Seq2seq. The work shows that context-aware embeddings, hierarchical attention, and hybrid link identification significantly boost accuracy while preserving scalability, making the approach practical for real-world, large-scale text processing tasks.

Abstract

Large-scale coreference resolution presents a significant challenge in natural language processing, necessitating a balance between efficiency and accuracy. In response to this challenge, we introduce an End-to-End Neural Coreference Resolution system tailored for large-scale applications. Our system efficiently identifies and resolves coreference links in text, ensuring minimal computational overhead without compromising on performance. By utilizing advanced neural network architectures, we incorporate various contextual embeddings and attention mechanisms, which enhance the quality of predictions for coreference pairs. Furthermore, we apply optimization strategies to accelerate processing speeds, making the system suitable for real-world deployment. Extensive evaluations conducted on benchmark datasets demonstrate that our model achieves improved accuracy compared to existing approaches, while effectively maintaining rapid inference times. Rigorous testing confirms the ability of our system to deliver precise coreference resolutions efficiently, thereby establishing a benchmark for future advancements in this field.

End-to-End Dialog Neural Coreference Resolution: Balancing Efficiency and Accuracy in Large-Scale Systems

TL;DR

This work targets scalable end-to-end neural coreference resolution for large-scale systems, addressing the tension between model size and accuracy. It introduces an End-to-End Neural Coreference Resolution framework that combines multi-layer neural architectures, contextual embeddings, and attention with optimization techniques like pruning and quantization to reduce overhead, formalizing representations such as and the affinity-based training objective . Empirical results on benchmarks including OntoNotes v5.0 demonstrate high F1 scores (e.g., with and with SpanBERT) along with strong precision/recall and rapid inference, outperforming baselines like Z-coref and Seq2seq. The work shows that context-aware embeddings, hierarchical attention, and hybrid link identification significantly boost accuracy while preserving scalability, making the approach practical for real-world, large-scale text processing tasks.

Abstract

Large-scale coreference resolution presents a significant challenge in natural language processing, necessitating a balance between efficiency and accuracy. In response to this challenge, we introduce an End-to-End Neural Coreference Resolution system tailored for large-scale applications. Our system efficiently identifies and resolves coreference links in text, ensuring minimal computational overhead without compromising on performance. By utilizing advanced neural network architectures, we incorporate various contextual embeddings and attention mechanisms, which enhance the quality of predictions for coreference pairs. Furthermore, we apply optimization strategies to accelerate processing speeds, making the system suitable for real-world deployment. Extensive evaluations conducted on benchmark datasets demonstrate that our model achieves improved accuracy compared to existing approaches, while effectively maintaining rapid inference times. Rigorous testing confirms the ability of our system to deliver precise coreference resolutions efficiently, thereby establishing a benchmark for future advancements in this field.

Paper Structure

This paper contains 22 sections, 8 equations, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Neural Coreference Resolution
  4. Efficient Large-Scale Systems
  5. Accuracy in NLP Tasks
  6. Methodology
  7. Neural Network Architectures
  8. Contextual Embeddings
  9. Coreference Links Resolution
  10. Experimental Setup
  11. Datasets
  12. Baselines
  13. Models
  14. Implements
  15. Experiments
  16. ...and 7 more sections

Figures (4)

  • Figure 1: End-to-End Neural Coreference Resolution system for sentences level extraction.
  • Figure 2: Comparison of different contextual embedding types on coreference resolution performance metrics.
  • Figure 3: Evaluation of different attention mechanisms used in coreference resolution, showcasing their impact on F1 score, precision, and recall metrics.
  • Figure 4: Coreference link identification techniques and their corresponding performance metrics.