Edit Distances and Their Applications to Downstream Tasks in Research and Commercial Contexts

The tutorial describes the concept of edit distances applied to research and commercial contexts. We use Translation Edit Rate (TER), Levenshtein, Damerau-Levenshtein, Longest Common Subsequence and $n$-gram distances to demonstrate the frailty of statistical metrics when comparing text sequences. Our discussion disassembles them into their essential components. We discuss the centrality of four editing actions: insert, delete, replace and move words, and show their implementations in openly available packages and toolkits. The application of edit distances in downstream tasks often assumes that these accurately represent work done by post-editors and real errors that need to be corrected in MT output. We discuss how imperfect edit distances are in capturing the details of this error correction work and the implications for researchers and for commercial applications, of these uses of edit distances. In terms of commercial applications, we discuss their integration in computer-assisted translation tools and how the perception of the connection between edit distances and post-editor effort affects the definition of translator rates.

Are Large Language Models State-of-the-art Quality Estimators for Machine Translation of User-generated Content?

This paper investigates whether large language models (LLMs) are state-of-the-art quality estimators for machine translation of user-generated content (UGC) that contains emotional expressions, without the use of reference translations. To achieve this, we employ an existing emotion-related dataset with human-annotated errors and calculate quality evaluation scores based on the Multi-dimensional Quality Metrics. We compare the accuracy of several LLMs with that of our fine-tuned baseline models, under in-context learning and parameter-efficient fine-tuning (PEFT) scenarios. We find that PEFT of LLMs leads to better performance in score prediction with human interpretable explanations than fine-tuned models. However, a manual analysis of LLM outputs reveals that they still have problems such as refusal to reply to a prompt and unstable output while evaluating machine translation of UGC.