LATA: A Tool for LLM-Assisted Translation Annotation

TL;DR

This paper presents LATA, an innovative LLM-assisted tool that enhances the accuracy and efficiency of complex translation annotation tasks, especially for linguistically challenging language pairs like Arabic-English.

Contribution

It introduces a template-based Prompt Manager that uses large language models for precise sentence segmentation and alignment within a human-in-the-loop framework.

Findings

Improves annotation accuracy for structurally divergent language pairs

Reduces manual effort through automation and LLM assistance

Supports complex, multi-layered translation annotations

Abstract

The construction of high-quality parallel corpora for translation research has increasingly evolved from simple sentence alignment to complex, multi-layered annotation tasks. This methodological shift presents significant challenges for structurally divergent language pairs, such as Arabic--English, where standard automated tools frequently fail to capture deep linguistic shifts or semantic nuances. This paper introduces a novel, LLM-assisted interactive tool designed to reduce the gap between scalable automation and the rigorous precision required for expert human judgment. Unlike traditional statistical aligners, our system employs a template-based Prompt Manager that leverages large language models (LLMs) for sentence segmentation and alignment under strict JSON output constraints. In this tool, automated preprocessing integrates into a human-in-the-loop workflow, allowing researchers to refine alignments and apply custom translation technique annotations through a stand-off architecture. By leveraging LLM-assisted processing, the tool balances annotation efficiency with the linguistic precision required to analyze complex translation phenomena in specialized domains.