Mao-Chang Ku

This study introduces CorPilot, an innovative multi-agent framework that integrates Large Language Models (LLMs) to function as a form of collaborative AI, designed to automate and streamline corpus linguistic research. By structuring interactions between specialized agents within this agentic framework, CorPilot assists with tasks such as querying, annotation, semantic classification, and data analysis. Demonstrated through an empirical case study of Mandarin Chinese constructions, CorPilot replicates existing manual annotation results and reveals finer semantic distinctions and novel linguistic patterns previously unattainable through traditional methods. Our findings illustrate that CorPilot represents a significant methodological advancement, addressing challenges in scalability, replicability, and interpretability of corpus linguistics research. The framework's modular design also facilitates future extensions into various linguistic domains, holding considerable potential for theoretical and practical advancements in linguistics and computational humanities.

Hyperbolic embeddings such as the Poincaré model effectively represent lexical hierarchies with low distortion, yet their cross-lingual generalizability remains largely unexplored. This study investigates cross-lingual transfer by training 20-dimensional Poincaré embeddings exclusively on Open English WordNet (OEWN) hypernymy relations and evaluating on aligned Chinese Wordnet (CWN) synsets under a vocabulary-constrained transfer setting, where CWN-relevant synsets appear in OEWN training data but no Chinese-language supervision is used. We report robust statistical evidence based on the final 10 training checkpoints: Poincaré embeddings achieve 2.57× higher Mean Reciprocal Rank (MRR) than Euclidean embeddings on CWN (0.030 ± 0.001 vs 0.012 ± 0.000, p < 0.001, Cohen’s d = 34.48) and 5.61× higher on OEWN (0.016 ± 0.000 vs 0.003 ± 0.000, p < 0.001, d = 42.48). Furthermore, hierarchical filtering leveraging the radial dimension of hyperbolic space provides substantial additional gains: +74.6% MRR improvement on CWN and +25.8% on OEWN (both p < 0.001). The model achieves higher absolute performance on the zero-shot CWN test set (MRR = 0.052 ± 0.002) than on the in-domain OEWN test set (MRR = 0.020 ± 0.001). We attribute this to structural alignment: CWN’s broader branching factor (4.32 vs 1.10) and moderate depth naturally suit hyperbolic geometry’s capacity to compactly represent hierarchies. Our findings demonstrate that geometric properties learned from English hypernymy transfer robustly across languages when semantic structures align. We release the aligned CWN–OEWN hypernymy evaluation dataset and complete evaluation framework to facilitate future research on geometry-based cross-lingual semantic modeling.

Large language models (LLMs) have in recent years spurred research across various sectors, owing to their remarkable zero-shot or few-shot performance. This capability has become indispensable for individuals seeking to integrate these language models into their workflows effectively. In this paper, based on in-depth linguistic analyses, we explore the application of an LLM, specifically GPT-4, in generating Chinese language textbooks tailored for grade school students. This encompasses the creation of main lesson texts alongside accompanying Chinese character exercises. Experimental results suggest that the LLM-generated textbook lessons are a viable research direction. The initial outcomes demonstrate the ability of LLM to generate texts of satisfactory quality appropriate for a specified grade level. The contributions of this work include pioneering the quantitative analysis of Chinese language textbooks for native speakers in Taiwan and leveraging an LLM to automatically generate textbook content and accompanying Chinese character exercises targeted at native Chinese speakers, which is a novel approach facilitated by the development of prompts tailored to different language learning levels. The study also conducts quantitative and qualitative comparisons between machine-generated lessons and those developed by educational professionals in Taiwan.

This study examines the variation in causative affixes (pa-∅-, pa-ka-, and pa-pe-) in Paiwan using a corpus-based approach. Building on Tang's (295) identification of the ∅-, ka-, and pe- affixes in Paiwan causative constructions, we apply logistic regression analysis to data extracted from corpora. Our research suggests a cognitive distinction among the three causative subtypes (Verhagen & Kemmer). The regression model results support the theory of a direct/indirect causation dichotomy, offering a plausible explanation for the characteristics and lexical meanings of the affixes. Specifically, the affix pa-∅- is associated with "direct causation," typically used in events involving inanimate participants where the cause directly results in the state of the causee. Conversely, the affix pa-ka- is linked to "indirect causation," often found in contexts with animate participants and additional contributing forces. The affix pa-pe- occupies an intermediary position, showing a preference for intransitive effected predicates. Additionally, this study conducts a cross-linguistic comparison of the Paiwan causative affixes with the causative verbs doen and laten in Dutch, and shi and rang in Mandarin. These findings enhance our understanding of Paiwan causative constructions and offer insights into the universality and specificity of causative structures in linguistic typology.

Contextualized embeddings have proven to be powerful tools in various NLP tasks. However, their interpretability and how they encode lexical semantics remain challenging issues. In this paper, we tackle this problem by using definition modeling, a technique that aims to generate human-readable definitions for words, as a means to evaluate and understand high-dimensional semantic vectors. We introduce the Vec2Gloss model, which generates glosses from the contextualized embeddings of target words. The systematic gloss patterns provided by Chinese Wordnet enable us to examine the mechanism behind the model’s gloss generation. To delve deeper into this mechanism, we devise two dependency indices to measure the semantic and contextual dependencies of the generated glosses. These indices allow us to analyze the generated texts at both the gloss and token levels. Our results demonstrate that the proposed Vec2Gloss model enhances our understanding of lexical semantics in contextualized embeddings.

Abstract Constructions are direct form-meaning pairs with possible schematic slots. These slots are simultaneously constrained by the embedded construction itself and the sentential context. We propose that the constraint could be described by a conditional probability distribution. However, as this conditional probability is inevitably complex, we utilize language models to capture this distribution. Therefore, we build CxLM, a deep learning-based masked language model explicitly tuned to constructions’ schematic slots. We first compile a construction dataset consisting of over ten thousand constructions in Taiwan Mandarin. Next, an experiment is conducted on the dataset to examine to what extent a pretrained masked language model is aware of the constructions. We then fine-tune the model specifically to perform a cloze task on the opening slots. We find that the fine-tuned model predicts masked slots more accurately than baselines and generates both structurally and semantically plausible word samples. Finally, we release CxLM and its dataset as publicly available resources and hope to serve as new quantitative tools in studying construction grammar.

The rapid flow of information and the abundance of text data on the Internet have brought about the urgent demand for the construction of monitoring resources and techniques used for various purposes. To extract facets of information useful for particular domains from such large and dynamically growing corpora requires an unsupervised yet transparent ways of analyzing the textual data. This paper proposed a hybrid collocation analysis as a potential method to retrieve and summarize Taiwan-related topics posted on Weibo and PTT. By grouping collocates of 臺灣 ‘Taiwan’into clusters of topics via either word embeddings clustering or Latent Dirichlet allocation, lists of collocates can be converted to probability distributions such that distances and similarities can be defined and computed. With this method, we conduct a diachronic analysis of the similarity between Weibo and PTT, providing a way to pinpoint when and how the topic similarity between the two rises or falls. A fine-grained view on the grammatical behavior and political implications is attempted, too. This study thus sheds light on alternative explainable routes for future social media listening method on the understanding of cross-strait relationship.

This paper aims to investigate the variation between two Chinese causative auxiliaries shi ‘使’ and rang ‘讓’ from a corpus-based perspective. We conduct a logistic regression analysis to the Chinese data extracted from two corpora and propose a direct/indirect distinction (Verhagen and Kemmer 1997) between the two auxiliary verbs. The results retrieved by the regression model show that the theory of direct/indirect causation provides a reasonable account for the characteristics and lexical meanings of the verbs. We indicate that the verb shi is correlated with “direct causation” because it is typically used when inanimate participants are involved in the causing event, in which the force initiated by the cause inevitably and directly leads to the resulted stage of the causee. On the other hand, the verb rang should be classified as “indirect causation” because it is typically used in scenarios where animate participants are both involved, and some extra force besides the causer also plays a role in the effected event.

Words are conventionalized symbols that present the function by which meaning is attached to form. The Word Sense Disambiguation, which has been taken as one of the core semantic processing tasks in the pipe-lined NLP architecture, aims to assign proper word sense to lemma form in varied contexts based on a word-sense inventory such as WordNet. However, there are some theoretical assumptions unattested from a functional linguistic point of view. This paper proposes an alternative by introducing a novel task called word action disambiguation task (WAD) concentrated on the observable pairs between words and actions. The accompanying dataset, which was manually edited and compiled, is composed of 419 multiple-choice questions. We further verified the dataset through item evaluation with human rating data, and the semantic relations among the dataset were annotated automatically. A baseline performance with an accuracy of 38.64% was also provided with BERT models and 43.18% after incorporating paradigmatic knowledge with semantic graph. We expect the proposal of the WAD task and dataset would motivate computational models to incorporate more complex aspects of human language.