Detecting Multiword Expression Type Helps Lexical Complexity Assessment

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 4426–4435 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 4426 Detecting Multiword Expression Type Helps Lexical Complexity Assessment Ekaterina Kochmar,1 Sian Gooding,1 Matthew Shardlow2 1 University of Cambridge, Department of Computer Science and Technology 2 Manchester Metropolitan University {ek358, shg36}@cam.ac.uk, m.shardlow@mmu.ac.uk Abstract Multiword expressions (MWEs) represent lexemes that should be treated as single lexical units due to their idiosyncratic nature. Multiple NLP applications have been shown to benefit from MWE identification, however the research on lexical complexity of MWEs is still an under-explored area. In this work, we re-annotate the Complex Word Identification Shared Task 2018 dataset of Yimam et al. (2017), which provides complexity scores for a range of lexemes, with the types of MWEs. We release the MWE-annotated dataset with this paper, and we believe this dataset represents a valuable resource for the text simplification community. In addition, we investigate which types of expressions are most problematic for native and non-native readers. Finally, we show that a lexical complexity assessment system benefits from the information about MWE types. Keywords: Complex word identification, multiword expressions (MWE), text simplification 1. Introduction Complex word identification (CWI) is a well-established task, with applications in text complexity assessment and lexical simplification (Paetzold, 2015; Saggion, 2017). CWI is concerned with the identification of words in need of simplification and is often considered the first step in a lexical simplification pipeline (Shardlow, 2013). For exam- ple, a CWI system may identify appreciate as complex in: It made me appreciate freedom A lexical simplification system may then suggest replacing appreciate with value, making the new sentence easier to understand for the intended reader. Most research to date has focused on complexity at the level of individual words only, despite the fact that complexity often relates to whole chunks of text. Take the following sentence for example: Protesters used sledge hammers to tear apart the security wall In a traditional lexical simplification pipeline, a CWI com- ponent may identify the word sledge as complex, and a lex- ical simplifier may then try to replace sledge, for exam- ple, with sleigh. However in this sentence sledge occurs as part of an expression sledge hammers, therefore a system tasked with lexical complexity assessment should instead identify sledge hammers as a single lexical unit, assess its complexity as such and, if necessary, attempt to simplify it as a whole (for instance, to lump hammers). Sledge hammers is an example of a multiword expres- sion (MWE) – an expression which is made up of at least two words and which has idiosyncratic interpreta- tion that crosses word boundaries or spaces (Sag et al., 2002). Due to this distinctive nature, many areas in NLP, including parsing (Constant et al., 2017), machine trans- lation (Constant et al., 2017; Carpuat and Diab, 2010), keyphrase/index term extraction (Newman and Baldwin, 2012), and language acquisition research (Ellis et al., 2008), benefit from treating MWEs as single lexical units. …

Content truncated to 500 words