A Corpus and Semantic Parser for Multilingual Natural Language Querying of OpenStreetMap

Abstract

We present a corpus of 2,380 natural language queries paired with machine readable formulae that can be executed against world wide geographic data of the OpenStreetMap (OSM) database. We use the corpus to learn an accurate semantic parser that builds the basis of a natural language interface to OSM. Furthermore, we use response-based learning on parser feedback to adapt a statistical machine translation system for multilingual database access to OSM. Our framework allows to map fuzzy natural language expressions such as "nearby", "north of", or "in walking distance" to spatial polygons on an interactive map. Furthermore, it combines syntactic complexity and compositionality with a reasonable lexical variability of queries, making it an interesting new publicly available dataset for research on semantic parsing.

Proceedings of the 21st Conference on Computational Natural Language Learning (CoNLL 2017)

“…When training on combined English and German data (All model), we observe a slight decrease in performance for both English and German. Even when training on the full target language dataset, using crosslingual word embeddings improves the perEnglish German Haas and Riezler (2016) 68.3 -Kočiský et al (2016) 78 formance by about 2% in both English and German which highlights the effectiveness of crosslingual word embeddings. As shown in Figure 4, adding a machine translation system helps immensely for small datasets.…”

“…Semantic parsing is the task of mapping a natural language query to a logical form (LF) such as Prolog or lambda calculus, which can be executed directly through database query Collins, 2005, 2007;Haas and Riezler, 2016;Kwiatkowksi et al, 2010). Semantic parsing needs application or domain specific training data, so the most straightforward approach is to manufacture training data for each combination of language and application domain.…”

Improving a Neural Semantic Parser by Counterfactual Learning from Human Bandit Feedback

“…Each tag consists of a key and an associated value, for example "tourism : hotel". The NLMAPS corpus was introduced by Haas and Riezler (2016) as a basis to create a natural language interface to the OSM database. It pairs English questions with machine readable parses, i.e.…”

Semantic Parsing with Semi-Supervised Sequential Autoencoders

“…We demonstrate the effectiveness of our proposed model on three semantic parsing tasks: the GEO-QUERY benchmark (Zelle and Mooney, 1996;Wong and Mooney, 2006), the SAIL maze navigation task (MacMahon et al, 2006) and the Natural Language Querying corpus (Haas and Riezler, 2016) on Open-StreetMap. As part of our evaluation, we introduce simple mechanisms for generating large amounts of unsupervised training data for two of these tasks.…”