High-quality parallel corpora serve as the fundamental backbone for advancements in Machine Translation (MT) research and the development of effective translation pedagogy. Despite this need, robust resources for the Arabic-English language pair remain significantly scarce. Furthermore, existing datasets are often limited by their reliance on simplistic one-to-one sentence mappings, which fail to capture the structural complexities inherent in natural language translation. To address this deficiency, this paper presents AlignAR, a novel generative sentence alignment method, alongside a comprehensive new Arabic–English dataset that juxtaposes simple legal documents with complex literary texts. Our evaluation demonstrates that “Easy” datasets lack the discriminatory power to fully assess alignment methods. By reducing one-to-one mappings within our “Hard” subset, we exposed the limitations of traditional alignment techniques when faced with structural divergence. In contrast, Large Language Model (LLM) based approaches demonstrated superior robustness and adaptability. Specifically, the proposed LLM-based approaches demonstrated better robustness, achieving an overall F1-score of 85.5%, a nearly 9% improvement over previous methods. This study underscores the importance of complex benchmarks and validates the efficacy of generative models in handling the intricacies of bitext alignment. The codes and datasets are available on Github.

Machine reading comprehension is a heavily-studied research and test field for evaluating new pre-trained language models (PrLMs) and fine-tuning strategies, and recent studies have enriched the pre-trained language models with syntactic, semantic and other linguistic information to improve the performance of the models. In this paper, we imitate the human reading process in connecting the anaphoric expressions and explicitly leverage the coreference information of the entities to enhance the word embeddings from the pre-trained language model, in order to highlight the coreference mentions of the entities that must be identified for coreference-intensive question answering in QUOREF, a relatively new dataset that is specifically designed to evaluate the coreference-related performance of a model. We use two strategies to fine-tune a pre-trained language model, namely, placing an additional encoder layer after a pre-trained language model to focus on the coreference mentions or constructing a relational graph convolutional network to model the coreference relations. We demonstrate that the explicit incorporation of coreference information in the fine-tuning stage performs better than the incorporation of the coreference information in pre-training a language model.