BenchLS: A Reliable Dataset for Lexical Simplification

To create our dataset we combined two resources: the LexMTurk (Horn et al., 2014) and LSeval (De Belder and Moens, 2012) datasets. The instances in both datasets, 929 in total, contain a sentence, a target complex word, and several candidate substitutions ranked according to their simplicity. The candidates in both datasets were suggested and ranked by English speakers from the U.S. To increase its reliability, we applied the following corrections over each instance of our dataset: Spelling Filtering: We discard any misspelled can- didates using Norvig’s algorithm. We trained our spelling model over the News Crawl corpus. Inflection Correction: We inflected all candidates to the tense of the target word using the Text Adorning module of LEXenstein (Paetzold and Specia, 2015; Burns, 2013). The resulting dataset – BenchLS – contains 929 instances, with an average of 7.37 candidate substitutions per complex word.