Try to Substitute: An Unsupervised Chinese Word Sense Disambiguation Method Based on HowNet

Abstract

Word sense disambiguation (WSD) is a fundamental natural language processing task. Unsupervised knowledge-based WSD only relies on a lexical knowledge base as the sense inventory and has wider practical use than supervised WSD that requires a mass of sense-annotated data. HowNet is the most widely used lexical knowledge base in Chinese WSD. Because of its uniqueness, however, most of existing unsupervised WSD methods cannot work for HowNetbased WSD, and the tailor-made methods have not obtained satisfying results. In this paper, we propose a new unsupervised method for HowNet-based Chinese WSD, which exploits the masked language model task of pre-trained language models. In experiments, considering existing evaluation dataset is small and out-of-date, we build a new and larger HowNet-based WSD dataset. Experimental results demonstrate that our model achieves significantly better performance than all the baseline methods. All the code and data of this paper are available at https://github.com/thunlp/SememeWSD.

Automatic Construction of Sememe Knowledge Bases via Dictionaries

“…When deep learning becomes the mainstream approach of NLP, the usefulness of HowNet is also proved in diverse tasks including word representation learning (Sun and Chen, 2016;Niu et al, 2017), language modeling (Gu et al, 2018), semantic composition (Qi et al, 2019a), sequence modeling (Qin et al, 2020), reverse dictionary , word sense disambiguation (Hou et al, 2020), textual adversarial attacking and backdoor attacking (Qi et al, 2021).…”

Leveraging Word-Formation Knowledge for Chinese Word Sense Disambiguation

“…However, these methods require copious sense- annotated datasets (Raganato et al, 2017), which are difficult to obtain in Chinese. Thus, previous Chinese WSD datasets (Niu et al, 2004;Jin et al, 2007;Agirre et al, 2009;Hou et al, 2020) are small in vocabulary size (less than 100 words except for Agirre et al, 2009), and it is uneasy to combine these datasets to enlarge their size, since they differ in format, sense inventory and construction guidelines.…”

“…Word sense disambiguation (WSD) aims to identify the sense of a polysemous word in a specific context, which benefits multiple downstream tasks (Hou et al, 2020). With copious senseannotated data (Raganato et al, 2017), neural WSD methods achieve superior performance by leveraging definitional and relational features in knowledge bases (KB) (Luo et al, 2018a;Huang et al, 2019;Bevilacqua and Navigli, 2020).…”

ALaSca: an Automated approach for Large-Scale Lexical Substitution

“…Furthermore, the resource annotation was carried out by considering the meaning of each target word occurrence, thereby providing different substitutes for the same lemma in different contexts. CoInCo Despite the effort expended to create the TWSI, the corpus covered only nouns, hence, Kremer et al [2014] proposed a similar annotation task starting from a set of sentences of the MASC corpus [Ide et al, 2008]. Substitutes were annotated through Amazon Mechanical Turk in this case too and the resulting dataset (Concept In Context, CoInCo) contains 15K tagged instances in 2474 sentences for 3874 distinct words with diverse part-of-speech tags.…”

“…An effective substitution system would provide two distinct sets of possible replacements, such as {smart, intelligent, bril-liant} and {shining, luminous}, respectively. The implicit disambiguation provided by substitution systems has shown itself to be useful in several fields, such as word sense induction [Başkaya et al, 2013;Amrami and Goldberg, 2018;Arefyev et al, 2019], text augmentation [Jia et al, 2019;Arefyev et al, 2020], word sense disambiguation [Hou et al, 2020] or text simplification [Bingel et al, 2018]. However, despite its possible uses, there is a lack of appropriate largescale resources for the task [Soler et al, 2019].…”