A Python Library for Deep Linguistic Resources

This paper describes PyDelphin: an open-source software library, coded in Python, for working with the resources and results of Deep Linguistic Processing with HPSG (DELPH-IN). These resources and processing results offer rich syntactic and semantic information from linguistically-informed grammars but the original software stack for working with them presents significant technical hurdles for installation and use, particularly for new users and for single experiments. PyDelphin is quick to install and its thorough documentation makes it easy for newcomers to get started. It contains packages that implement a number of DELPH-IN technologies for working with semantic representations, grammar descriptions, tokenization, and corpus databases. It also has client interfaces for related external processors which make it straightforward to integrate these tools into a single workflow. The library has been successfully used in several research projects in topics such as abstractive summarization, machine translation, and natural language generation.