06月04日(Tue) 14:10〜16:10 A会場(-国際会議場3F メインホール)
| 演題番号 | 1A3-IOS-3a-1 |
|---|---|
| 題目 | Unsupervised Sense Clustering of Related Chinese Words |
| 著者 | Lee Chia-Ling(National Taiwan University) Liu Chao-Lin(National Chengchi University) |
| 時間 | 06月04日(Tue) 14:10〜14:30 |
| 概要 | Chinese words which share the same character may carry related but different meanings, e.g., "spend", "expend", "garden", "bloom". The semantics of these words form two clusters: {spend", "expend"} and {"garden", "bloom"}. In this paper, we aim at unsupervised clustering of a given set of such related Chinese words, where the quality of clustering results is to be judged based on the senses of the related words. Successfully differentiating these words not only represents an important step toward Chinese word sense disambiguation but also helps for computer assisted Chinese learning. Currently, we considered semantic context and tried to determine the clustering roughly. We employed two knowledge-based methods using ConceptNet5 and two corpus-based measures based on Chinese Wikipedia. Our experimental result shows it can be achieved 68.49% of Rand index in the best case using knowledge-based approach. |
| 論文 | PDFファイル |