The Construction of A Concept-based Chinese Knowledge Base with Semantic Composition Capability
建構概念為本且具語義結合性的中文知識庫

Academia Sinica Assistant Research Fellow Wei-Yun Ma

計畫主持人:中央研究院 資訊科學研究所馬偉雲教授

一套實用的中文知識庫所需要的實體數量往往以百萬、千萬計,且層出不窮,必須倚賴自動化的構建方式。同時,知識庫的推論機制,如因果關係、動作過程等也都必需一併考慮。在這個計畫中,我們的策略是以廣義知網現有的上下位架構為本,將實體一個一個地自動化掛載在適合的概念之下,能夠承襲相關概念屬性。此外,我們擬以深度學習的技術自動地從網際網路中擷取實體的知識,並且利用語義結合的機制來增強實體的語義表達能力。

For a practical Chinese knowledge base, our strategy is to base on E-HowNet’s hierarchical taxonomy and attach every newly discovered entity to an appropriate concept defined in EHowNet, so the entity can inherit relevant conceptual attributes. In addition, we aim to automatically extract entities’ knowledge from Internet using deep learning techniques and empower entities’ semantic representation ability with semantic composition mechanism.