Text2ALM is an advanced information extraction tool turning implicit information in text into set of facts capturing key properties of entities mentioned in text. It relies on multiple linguistic and knowledge representation resources: Verbnet lexicon, Text2DRS tool, CoreNLP Stanford tool, LTH semantic role labeler, CoreALMLib knowledge library, CALM solver for knowledge representation language ALM.
To examplify Text2ALM at work consider a narrative:
John traveled to the garage. John picked up a football. John went to the kitchen.
The output of Text2ALM will be the following:
location(John, garage, 1), location(John, garage, 2), location(John, kitchen, 3), held_by(football, John, 2), held_by(football, John, 3), location(football, garage, 2), location(football, kitchen, 3)
The system is hosted on GitHub:
The repository includes detailed documentation on how to setup this tool (and all relevant systems) on a LINUX system and includes a manual on how to use Text2ALM.
The system is documented in the following publications:
Craig Olson. "Processing narratives by means of action languages". Master Thesis, University of Nebraska Omaha, USA, Spring 2019
The project has been inspired by the ideas discussed in:
Yuliya Lierler, Daniela Inclezan and Michael Gelfond. "Action Languages and Question Answering", Proceeding of the 12th International Conference on Computational Semantics, 2017
Comments, questions, and/or bugs can be reported to Craig Olson and Yuliya Lierler