Verb Temporality Analysis using Reichenbach's Tense System
André Kenji Horie, Kumiko Tanaka‐Ishii, Mitsuru Ishizuka · International Conference on Computational Linguistics · 2012
This paper presents the analysis process of verb temporality using Reichenbach’s tense system, a language-independent system which describes tense as relations among linguistic and extralinguistic temporal entities. Several difficulties arise from the deep analysis required for classification into Reichenbach’s categories. They regard establishing the logical sequence of clauses in the skeletal structure of the discourse, and modeling the behavior of temporal markers according to this sequence. A dependency clause anchoring algorithm is then proposed and compared to other anchoring methods, and sequential supervised learning is used for abstracting surrounding context in order to determine temporal marker behavior. Experimental results show that the proposed approach is able to better abstract verb temporality than statistical ones, suggesting that analytical interlingual translation can complement existing SMT techniques by providing an additional layer of semantic information.