GeoQuery-LSFB: A French Belgian Sign Language Corpus with Procedural Semantic Annotations

The GeoQuery-LSFB corpus is a unique resource for French-Belgian Sign Language (LSFB). It contains machine-readable transcriptions of questions in LSFB and procedural semantic representations that describe the meaning of these questions in terms of primitive operations. The operations can be evaluated against a database containing facts about American Geography, yielding the answer to the LSFB question. GeoQuery-LSFB is an extension of GeoQuery (Zelle & Mooney, 1996), a well-known benchmark for semantic parsing. It originally consisted of 250 and 880 questions in English, paired with logical prolog queries that can be executed on a database about US geography (Zelle & Mooney, 1996). Over the years, different languages have been added to GeoQuery. First, Wong & Mooney (2006) translated the English sentences of the smaller version of GeoQuery (250 examples) into Spanish, Turkish and Japanese. Later, Lu & Ng (2011) translated all 880 English sentences of GeoQuery into Chinese, followed by Jones et al., (2012), who did the same thing for Geman, Greek and Thai. Finally, Susantu et al., (2017) provided a translation into Swedish, Farsi and Thai. GeoQuery LSFB groups together all translations from these multilingual versions of GeoQuery, with the addition of French and French Belgian Sign Language. This repository contains two releases of the dataset: a small version, containing 250 examples, and a larger, synthetically augmented version containing 4519 examples. Both versions are available in xml and json format. Each example contains a natural language expression in English, Spanish, Japanese, Turkish, German, Greek, Thai, Chinese, Farsi, Indonesian, Swedish, French, and LSFB, as well as a procedural semantic representation in Geo-FunQL format (Kate et al., 2005), Geo-Prolog format (Zelle & Mooney, 1996) and SQL-format (Iyer et al., 2017).  In addition to the two releases, we also provide the recordings of signed expressions and the raw annotation files that were used to create GeoQuery-LSFB. Because of its multilingual nature and inclusion of procedural semantic representations, GeoQuery-LSFB forms the perfect resource for the comparative study of spoken and signed languages, both from a theoretical and an application-oriented perspective. References Iyer, S., Konstas, I., Cheung, A., Krishnamurthy, J., & Zettlemoyer, L. (2017, July). Learning a Neural Semantic Parser from User Feedback. In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) (pp. 963-973). Jones, B., Johnson, M., & Goldwater, S. (2012, July). Semantic parsing with bayesian tree transducers. In Proceedings of the 50th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) (pp. 488-496). Kate, Rohit & Wong, Yuk & Mooney, Raymond. (2005). Learning to Transform Natural to Formal Languages. Proceedings of the National Conference on Artificial Intelligence.- 3. 1062-1068. Lu, W., & Ng, H. T. (2011, July). A probabilistic forest-to-string model for language generation from typed lambda calculus expressions. In Proceedings of the 2011 Conference on Empirical Methods in Natural Language Processing (pp. 1611-1622). Susanto, R. H., & Lu, W. (2017, February). Semantic parsing with neural hybrid trees. In Proceedings of the AAAI Conference on Artificial Intelligence (Vol. 31, No. 1). Wong, Y. W., & Mooney, R. (2006, June). Learning for semantic parsing with statistical machine translation. In Proceedings of the Human Language Technology Conference of the NAACL, Main Conference (pp. 439-446). Wong, Y. W., & Mooney, R. (2007, June). Learning synchronous grammars for semantic parsing with lambda calculus. In Proceedings of the 45th Annual Meeting of the Association of Computational Linguistics (pp. 960-967). Zelle, J. M., & Mooney, R. J. (1996, August). Learning to parse database queries using inductive logic programming. In Proceedings of the national conference on artificial intelligence (pp. 1050-1055).