Sanghoon Kim, Dahyun Kim, Chanjun Park, Wonsung Lee, Wonho Song, Yunsu Kim, Hyeonwoo Kim, Yungi Kim, Hyeonju Lee, Jihoo Kim, Changbae Ahn, Seonghoon Yang, Sukyung Lee, Hyunbyung Park, Gyoungjin Gim, Mikyoung Cha, Hwalsuk Lee, Sunghun Kim. Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 6: Industry Track). 2024.

The success of Transformer architecture has seen increased interest in machine translation (MT). The translation quality of neural network-based MT transcends that of translations derived using statistical methods. This growth in MT research has entailed the development of accurate automatic evaluation metrics that allow us to track the performance of MT. However, automatically evaluating and comparing MT systems is a challenging task. Several studies have shown that traditional metrics (e.g., BLEU, TER) show poor performance in capturing semantic similarity between MT outputs and human reference translations. To date, to improve performance, various evaluation metrics have been proposed using the Transformer architecture. However, a systematic and comprehensive literature review on these metrics is still missing. Therefore, it is necessary to survey the existing automatic evaluation metrics of MT to enable both established and new researchers to quickly understand the trend of MT evaluation over the past few years. In this survey, we present the trend of automatic evaluation metrics. To better understand the developments in the field, we provide the taxonomy of the automatic evaluation metrics. Then, we explain the key contributions and shortcomings of the metrics. In addition, we select the representative metrics from the taxonomy, and conduct experiments to analyze related problems. Finally, we discuss the limitation of the current automatic metric studies through the experimentation and our suggestions for further research to improve the automatic evaluation metrics.

Pre-processing and post-processing are significant aspects of natural language processing (NLP) application software. Pre-processing in neural machine translation (NMT) includes subword tokenization to alleviate the problem of unknown words, parallel corpus filtering that only filters data suitable for training, and data augmentation to ensure that the corpus contains sufficient content. Post-processing includes automatic post editing and the application of various strategies during decoding in the translation process. Most recent NLP researches are based on the Pretrain-Finetuning Approach (PFA). However, when small and medium-sized organizations with insufficient hardware attempt to provide NLP services, throughput and memory problems often occur. These difficulties increase when utilizing PFA to process low-resource languages, as PFA requires large amounts of data, and the data for low-resource languages are often insufficient. Utilizing the current research premise that NMT model performance can be enhanced through various pre-processing and post-processing strategies without changing the model, we applied various decoding strategies to Korean–English NMT, which relies on a low-resource language pair. Through comparative experiments, we proved that translation performance could be enhanced without changes to the model. We experimentally examined how performance changed in response to beam size changes and n-gram blocking, and whether performance was enhanced when a length penalty was applied. The results showed that various decoding strategies enhance the performance and compare well with previous Korean–English NMT approaches. Therefore, the proposed methodology can improve the performance of NMT models, without the use of PFA; this presents a new perspective for improving machine translation performance.

Grammar error correction (GEC) refers to the proper correction of grammatical errors in a given sentence. Important factors to consider in GEC are not only the grammatical correction of the sentence, but also the recognition of a correct sentence in which no changes are required. However, GEC approaches in which deep learning recently started being used consider only the former aspect, which leads to overcorrection, whereby changes are made to a correct sentence unnecessarily. Because this bias is also reflected in performance metrics, conventional performance metrics consider only part of the important factors in GEC. This study proposes a new metric to consider both important aspects in GEC and to provide a new viewpoint for the GEC task. To the best of the authors knowledge, this study is the first to deal with comprehensively considering the correction performance and overcorrection problem in GEC. The experimental results demonstrate that the model performance ranking was reversed when evaluating the performance with the proposed metric compared to the General Language Understanding Evaluation benchmark [21], which only considers the correction performance. This indicates that the high performance of the correction does not result in less problems with the overcorrection and that the overcorrection problem should also be considered when evaluating the model performance. Moreover, we found that the copy mechanism [14] helps to alleviate the problem of overcorrection.

Translation of the languages of ancient times can serve as a source for the content of various digital media and can be helpful in various fields such as natural phenomena, medicine, and science. Owing to these needs, there has been a global movement to translate ancient languages, but expert minds are required for this purpose. It is difficult to train language experts, and more importantly, manual translation is a slow process. Consequently, the recovery of ancient characters using machine translation has been recently investigated, but there is currently no literature on the machine translation of ancient Korean. This paper proposes the first ancient Korean neural machine translation model using a Transformer. This model can improve the efficiency of a translator by quickly providing a draft translation for a number of untranslated ancient documents. Furthermore, a new subword tokenization method called the Share Vocabulary and Entity Restriction Byte Pair Encoding is proposed based on the characteristics of ancient Korean sentences. This proposed method yields an increase in the performance of the original conventional subword tokenization methods such as byte pair encoding by 5.25 BLEU points. In addition, various decoding strategies such as n-gram blocking and ensemble models further improve the performance by 2.89 BLEU points. The model has been made publicly available as a software application.

Patents provide inventors exclusive rights to their inventions by protecting their intellectual property rights. However, analyzing patent documents generally requires knowledge of various fields, considerable human labor, and expertise. Recent studies to alleviate this problem on patent analysis deal only with the analysis of claims and abstract parts, neglecting the descriptions that contain essential technical cores. Moreover, few studies use a deep learning approach to handle the entire patent analysis process, including preprocessing, summarization, and key-phrase generation. Therefore, we propose a novel multi-stage framework that can aid in analyzing patent documents by using the description part of the patent rather than abstracts or claims with deep learning. The framework comprises two stages: key-sentence extraction and key-phrase generation tasks. These stages are based on the T5 model structure, transformer-based architecture that uses a text-to-text approach. To further improve the framework’s performance, we employed two key factors: i) post-training the model with a patent-related raw corpus for encouraging the model’s comprehension of the patent domain, and ii) utilizing a text rank algorithm for efficient training based on the priority score of each sentence. We verified that our key-phrase generation method of the framework shows higher performance in both superficial and semantic evaluation than other extraction methods. In addition, we provided the validity and effectiveness of our methods through quantitative and qualitative analysis, demonstrating the practical functionality of our methods. We also provided a practical contribution to the patent analysis by releasing the framework as a demo system.