研究発表

This paper aims to address two critical issues within Frame Semantics and Construction Grammar. First, Frame Semantics has taken a back seat compared to its closely related sister theory Construction Grammar (cf. Boas To Appear). Second, recent Construction Grammar studies have tended to focus on semantic frames rather than interactional frames in analyzing constructional meanings. Through an examination of various Japanese constructions, this study investigates whether interactional frames effectively describe constructional meanings. The paper argues that interactional frames are essential for adequately describing conventionalized pragmatic functions of constructions. As Boas (To Appear) points out, understanding the nature of meaning is inherently more complex than analyzing the form of constructions. Nevertheless, it is important to look into the meaning side of constructions and in particular to incorporate conventionalized pragmatic functions into constructinal analyses. This study claims that many conventionalized pragmatic functions of constructions are characterized by interactional frames (“how we conceptualize what is going on between the speaker and the hearer, or between the author and the reader”), rather than by semantic frames (“script-like conceptual structures that describe types of situations, objects, or events along with their participants and props”) (Fillmore 1982, Fillmore and Baker 2015, Ruppenhofer et al. 2016). Building upon Ohara (2024), this paper further highlights the need for developing an inventory of interactional frames. By analyzing sentence-level and discourse-level constructions from Present-day Japanese listed in Group Jammassy (2024), this study examines constructions’ conventionalized pragmatic functions, by comparing them with Croft (2022)’s comparative concepts pertaining to information structure (Croft 2022: 660). It demonstrates that pragmatic functions, such as those related to politeness and discourse patterns, are most effectively described using interactional frames. To summarize, this study bridges the gap between Frame Semantics and Construction Grammar by demonstrating how interactional frames effectively describe conventionalized pragmatic functions of constructions.

Theme Session “Cross-theoretical Perspectives on Frame-based Lexical and Constructional Analyses: Bridging Qualitative and Quantitative Studies.”

Motivated by Fillmore’s characterization of interactional frames (1982:117), the goal of this panel is to bring together theoretical linguists working in pragmatics with frame semanticists, creators of FrameNet-inspired lexica (Ruppenhofer et al., 2016, Boas 2009) and constructicons (Lyngfelt et al., 2018), as well as developers of real world applications. The panel will provide a forum to discuss the importance of the role of interactional frames in human language understanding as well as the challenges and consequences of representing Fillmore’s interactional frames in resources based on Frame Semantics and Construction Grammar.

This paper discusses the usefulness of crowdsourcing in linking Japanese FrameNet (JFN) and Kyoto University Case Frame (KCF), which differ in: type of frame knowledge (semantic frames vs. case frames); methodology of building (manual vs. automatic); and size (small vs. large). It proposes a method to crowdsource the assignment of a JFN semantic frame to KCF case frames. The method matches a KCF example sentence with a JFN annotated sentence, where both share the same meaning of a lemma. This paper argues that crowdsourcing frame disambiguation is effective in detecting multiword expressions (MWEs), missing frames and lexical units (LUs), and errors in frame assignment in Japanese FrameNet (JFN). It also proposes ways to alleviate the linguistic task of evaluating crowdsourced results, in deciding what kinds of lemmas to focus, where to find them, and what to do with them.

In this talk I will discuss FrameNet (FN), an online language resource called FrameNet (FN), and similar language resources in other languages such as Spanish, Japanese, German, Swedish and Brazilian Portuguese. FN documents the meanings and usages of contemporary English words in terms of semantic frames, that is, background knowledge structures that speakers can access about various experiences associated with the words. I will focus on two aspects of FN: (1) its relations to the theories of Frame Semantics (e.g., Fillmore & Baker 2015) and Construction Grammar (e.g., Fillmore et al. 2012); and (2) its relevance to cognitive linguists in the age of deep learning. After characterizing FN as a practical realization of Frame Semantics and Construction Grammar (e.g., Ruppenhoffer et al. 2016, Baker & Fillmore 2004), I will argue that the FN research and resource-building project has contributed extensively to advances in the two theories. Next, I will show that FN has garnered attention in the field of natural language processing (NLP) (e.g., Gildea & Jurafsky 2002) and it will continue to gain further prominence for enhancing NLP technologies. Finally, I will suggest that in the age of deep learning, collaboration between linguists and NLP researchers is both important and critical (cf. Bender 2020). Equally important and necessary is our continued qualitative description and analysis of linguistic data, since only linguists can develop hypotheses concerning speaker knowledge of language (cf. Fillmore 1992). Examining FrameNet data is a good place to start.