NLP datasets may still contain annotation errors, even when they are manually annotated. Researchers have attempted to develop methods to automatically reduce the adverse effect of errors in datasets. However, existing methods are time-consuming because they require many trained models to detect errors. This paper proposes a time-saving method that utilizes a tokenization technique called subword regularization to simulate multiple error detection models for detecting errors. Our proposed method, SubRegWeigh, can perform annotation weighting four to five times faster than the existing method. Additionally, SubRegWeigh improved performance in document classification and named entity recognition tasks. In experiments with pseudo-incorrect labels, SubRegWeigh clearly identifies pseudo-incorrect labels as annotation errors. Our code is available at https://github.com/4ldk/SubRegWeigh.

This paper investigates how LLMs encode inputs with typos. We hypothesize that specific neurons and attention heads recognize typos and fix them internally using local and global contexts. We introduce a method to identify typo neurons and typo heads that work actively when inputs contain typos. Our experimental results suggest the following: 1) LLMs can fix typos with local contexts when the typo neurons in either the early or late layers are activated, even if those in the other are not. 2) Typo neurons in the middle layers are the core of typo-fixing with global contexts. 3) Typo heads fix typos by widely considering the context not focusing on specific tokens. 4) Typo neurons and typo heads work not only for typo-fixing but also for understanding general contexts.

Combining different input modalities beyond text is a key challenge for natural language processing. Previous work has been inconclusive as to the true utility of images as a supplementary information source for text classification tasks, motivating this large-scale human study of labelling performance given text-only, images-only, or both text and images. To this end, we create a new dataset accompanied with a novel annotation method—Japanese Entity Labeling with Dynamic Annotation—to deepen our understanding of the effectiveness of images for multi-modal text classification. By performing careful comparative analysis of human performance and the performance of state-of-the-art multi-modal text classification models, we gain valuable insights into differences between human and model performance, and the conditions under which images are beneficial for text classification.

Chunpeng Ma, Aili Shen, Hiyori Yoshikawa, Tomoya Iwakura, Daniel Beck, Timothy Baldwin. Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume. 2021.

We propose a global optimization method under length constraint (GOLC) for neural text summarization models. GOLC increases the probabilities of generating summaries that have high evaluation scores, ROUGE in this paper, within a desired length. We compared GOLC with two optimization methods, a maximum log-likelihood and a minimum risk training, on CNN/Daily Mail and a Japanese single document summarization data set of The Mainichi Shimbun Newspapers. The experimental results show that a state-of-the-art neural summarization model optimized with GOLC generates fewer overlength summaries while maintaining the fastest processing speed; only 6.70% overlength summaries on CNN/Daily and 7.8% on long summary of Mainichi, compared to the approximately 20% to 50% on CNN/Daily Mail and 10% to 30% on Mainichi with the other optimization methods. We also demonstrate the importance of the generation of in-length summaries for post-editing with the dataset Mainich that is created with strict length constraints. The ex- perimental results show approximately 30% to 40% improved post-editing time by use of in-length summaries.

Taiki Watanabe, Akihiro Tamura, Takashi Ninomiya, Takuya Makino, Tomoya Iwakura. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). 2019.

This paper investigates learning methods for multi-class classification using labeled data for the target classification scheme and another labeled data for a similar but different classification scheme (support scheme). We show that if we have prior knowledge about the relation between support and target classification schemes in the form of a class correspondence table, we can use it to improve the model performance further than the simple multi-task learning approach. Instead of learning the individual classification layers for the support and target schemes, the proposed method converts the class label of each example on the support scheme into a set of candidate class labels on the target scheme via the class correspondence table, and then uses the candidate labels to learn the classification layer for the target scheme. We evaluate the proposed method on two tasks in NLP. The experimental results show that our method effectively learns the target schemes especially for the classes that have a tight connection to certain support classes.

The authors compared two methods for annotating a corpus for the named entity (NE) recognition task using non-expert annotators: (i) revising the results of an existing NE recognizer and (ii) manually annotating the NEs completely. The annotation time, degree of agreement, and performance were evaluated based on the gold standard. Because there were two annotators for one text for each method, two performances were evaluated: the average performance of both annotators and the performance when at least one annotator is correct. The experiments reveal that semi-automatic annotation is faster, achieves better agreement, and performs better on average. However, they also indicate that sometimes, fully manual annotation should be used for some texts whose document types are substantially different from the training data document types. In addition, the machine learning experiments using semi-automatic and fully manually annotated corpora as training data indicate that the F-measures could be better for some texts when manual instead of semi-automatic annotation was used. Finally, experiments using the annotated corpora for training as additional corpora show that (i) the NE recognition performance does not always correspond to the performance of the NE tag annotation and (ii) the system trained with the manually annotated corpus outperforms the system trained with the semi-automatically annotated corpus with respect to newswires, even though the existing NE recognizer was mainly trained with newswires.

In this paper, a technique for visualizing the changing rank and value of events simultaneously with time, as well as a trend analysis tool that uses the resulting visualization, are proposed. This technique facilitates analysts' understanding of the trends of both the entire data set and each event. It also uses colors, which are estimated automatically from the events' text labels, to represent the category to which events belong. Using the tool, the authors analyzed trends of topic data extracted from Twitter as use case, and could observe the trends and overview of Twitter. In addition, using the tool it is possible to present the different ways in which people who use different media think, using media-generated data.

We propose a named entity (NE) recognition method in which word chunks are repeatedly decomposed and concatenated. Our method identifies word chunks with a base chunker, such as a noun phrase chunker, and then recognizes NEs from the recognized word chunk sequences. By using word chunks, we can obtain features that cannot be obtained in word-sequence-based recognition methods, such as the first word of a word chunk, the last word of a word chunk, and so on. However, each word chunk may include a part of an NE or multiple NEs. To solve this problem, we use the following operators: SHIFT for separating the first word from a word chunk, POP for separating the last word from a word chunk, JOIN for concatenating two word chunks, and REDUCE for assigning an NE label to a word chunk. We evaluate our method on a Japanese NE recognition dataset that includes about 200,000 annotations of 191 types of NEs from over 8,500 news articles. The experimental results show that the training and processing speeds of our method are faster than those of a linear-chain structured perceptron and a semi-Markov perceptron, while maintaining high accuracy.