Construct a Sense-Frame Aligned Predicate Lexicon for Chinese AMR Corpus

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 2962–2969 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 2962 Construct a Sense-Frame Aligned Predicate Lexicon for Chinese AMR Corpus Li Song1,2, Yuling Dai2, Yihuan Liu2, Bin Li2, Weiguang Qu3 1. Department of Chinese Language and Literature, Tsinghua University, Beijing 100084, China 2. School of Chinese Language and Literature, Nanjing Normal University, Nanjing 210097, China 3. School of Computer Science and Technology, Nanjing Normal University, Nanjing 210023, China songli1105@sina.com Abstract The study of predicate frame is an important topic for semantic analysis. Abstract Meaning Representation (AMR) is an emerging graph based semantic representation of a sentence. Since core semantic roles defined in the predicate lexicon compose the backbone in an AMR graph, the construction of the lexicon becomes the key issue. The existing lexicons blur senses and frames of predicates, which needs to be refined to meet the tasks like word sense disambiguation and event extraction. This paper introduces the on-going project on constructing a novel predicate lexicon for Chinese AMR corpus. The new lexicon includes 14,389 senses and 10,800 frames of 8,470 words. As some senses can be aligned to more than one frame, and vice versa, we found the alignment between senses is not just one frame per sense. Explicit analysis is given for multiple aligned relations, which proves the necessity of the proposed lexicon for AMR corpus, and supplies real data for linguistic theoretical studies. Keywords: predicate frame, semantic role, Abstract Meaning Representation, lexicography 1. Introduction Sentences are the basic units of language use. Semantic analysis on sentence-level focuses on semantic structures of the sentences. The backbones of semantic structures are the various semantic relations contained in the event frames of predicates, mainly the dominating-dominated relations between predicates and nominal components. Therefore, the study of semantic role frame of predicates is of great importance to semantic analysis. However, the way of defining semantic roles has always been controversial in the linguistic field. It is unavoidable to explore methods to describe semantic relations and de- termine the granularity of semantic role labels. Xue (2006a) pointed out that existing annotated resources define seman- tic roles through different levels of abstraction. At present, there are three general methods for defining, namely, pred- icate-general, frame-specific and predicate-specific, but none of them is perfect. Predicate-general labels are not suitable for representing core semantic roles that bear many dynamic problems. The frame-specific method removes the limit to the number of labels while causes complex labels and burdens annotators. The predicate-specific method can effectively annotate the dynamic core semantic roles, but is hard to handle non-core semantic roles. (See Section 2 for details.) Therefore, we argue that it is more appropriate to use predicate-specific labels for core semantic roles and predicate-general labels for non-core semantic roles. Eng- lish Proposition Bank (PropBank) (Palmer et al., 2005) and the emerging sentence semantic representation method, Abstract Meaning Representation (AMR) (Banarescu et al., 2013), both adopt such a method. Core …

Content truncated to 500 words