Cross-lingual Linking of Automatically Constructed Frames and FrameNet

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 6620–6625 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 6620 Cross-lingual Linking of Automatically Constructed Frames and FrameNet Ryohei Sasano Graduate School of Informatics, Nagoya University, Japan sasano@i.nagoya-u.ac.jp Abstract A semantic frame is a conceptual structure describing an event, relation, or object along with its participants. Several semantic frame resources have been manually elaborated, and there has been much interest in the possibility of applying semantic frames designed for a particular language to other languages, which has led to the development of cross-lingual frame knowledge. However, manually developing such cross-lingual lexical resources is labor-intensive. To support the development of such resources, this paper presents an attempt at automatic cross-lingual linking of automatically constructed frames and manually crafted frames. Specifically, we link automatically constructed example-based Japanese frames to English FrameNet by using cross-lingual word embeddings and a two-stage model that first extracts candidate FrameNet frames for each Japanese frame by taking only the frame-evoking words into account, then finds the best alignment of frames by also taking frame elements into account. Experiments using frame-annotated sentences in Japanese FrameNet indicate that our approach will facilitate the manual development of cross-lingual frame resources. Keywords: semantic frame, cross-lingual frame linking, FrameNet 1. Introduction A semantic frame is a conceptual structure describ- ing an event, relation, or object along with its partic- ipants. Semantic frames have been shown to be useful for many natural language processing applications such as recognizing textual entailment (Tatu and Moldovan, 2005), question answering (Shen and Lapata, 2007), and knowledge extraction (Søgaard et al., 2015). Thus, several semantic frame resources, such as FrameNet (Baker et al., 1998), VerbNet (Kipper et al., 2000), and PropBank (Palmer et al., 2005), have been manu- ally elaborated. In addition, various systems have been proposed for automatic construction of frame knowl- edge from raw corpora (Korhonen et al., 2006; Kawa- hara et al., 2014; QasemiZadeh et al., 2019; Yamada et al., 2021). Among them, FrameNet is a representative resource of manually crafted cognitive frames, which provides rich semantic representations of the core En- glish vocabulary based on Fillmore’s frame semantics (Fillmore, 1976) with more than 200K frame-annotated sentences and has been extended to languages other than English. Resources based on FrameNet have now been created for roughly a dozen languages (Baker et al., 2018). However, manually developing such lexical resources is labor-intensive. In particular, defining frames, which entails considering their relationship to the definition of frames designed for another language, is a laborious process and thus it is difficult to develop such resources on a large scale. For example, Japanese FrameNet (JFN) (Ohara, 2013), consisting of cognitive frames, lexical units, and frame-annotated sentences, has been developed for two decades, but its coverage is still lim- ited. Table 1 shows the statistics of FrameNet and JFN. The number of frame-annotated sentences in JFN is FrameNet JFN # of cognitive frames 1222 947 # of lexical units (LUs) 13572 4957 …

Content truncated to 500 words