This repository contains the Circumstantial Event Ontology (CEO). This is version 1.0 (stable).
The Circumstantial Event Ontology (CEO.owl), is a manually constructed resource which formalizes the pre-, during-, and post-situations of events and the roles of the entities affected by an event. Further, it allows to infer circumstantial relations between events in natural language text. A circumstantial relation is defined here as an implicit and weakly causal relation. For instance, events pertaining to class A and event B can happen independently, but if they appear in context, a circumstantial relation is inferred based on shared properties in the class assertions.
Example: "Three months ago she arsoned her ex-husband stables. She was arrested soon after; the historical buildings are currently being restored by an expert builder." For a human reader, it is clear that 'arson', 'arrest' and 'restoration' are causally related in this context. However, there are no clues in these sentences that explicitly connect them. Our system, based on the CEO ontology, infers exactly these implicit relations. It does not define explicitly that e.g. 'arson' causes 'restoration' as the relation is plausible but certainly not necessary; therefore the ontology is designed to infer it. In the case of ceo:Arson we defined e.g. that after some arson there is damage. And before some restoration takes place, there must be some damage. Hence, we connect the events in this sentence with a so called circumstantial relation. We have created large sets of class assertions in order to connect events based on shared and implicit semantics. As such, we can not only connect the events, but also explain why something happened. In the case of the arson and restoration example, we know that the stables were restored because they were damaged due to some arson.
CEO is fully mapped to SUMO on class level and to FrameNet on class and role level. As input format, CEO can work on unannotated and unprocessed data. In this case, the model will infer circumstantial relations only based on surface forms of the event mentions. To get you started, we provided a vocabulary with event surface forms mapped to CEO classes. However, in order to employ the full inference power of the model, we advise to use processed documents with FrameNet based Semantic Role Labeling. CEO will run on top of this output.
- 223 event classes
- 121 properties
- 79 roles
- 456 situation rule assertions
- skos mappings to SUMO (195) and FrameNet Frames (265) and Frame Elements (263)
CEO-1.owlThis is ESO version 1 (stable)CEO-HumanReadable.pdfThis file contains a transcription of all CEO classes and provides examples of the instantiation of the assertions.CEO-VocabularyThis folder contains the manual mappings between event mentions extracted from the ECB+/CEO corpus and the CEO ontology. It includes a readme. Note that a new file is in the making with proper lemmatization and manual mappings to Princeton WordNet.EvaluationCorpusThis folder contains the ECB+/CEO corpus, annotated with circumstantial relations between event mentions and instances. It includes the CAT annotation task and statistics on the corpus annotation.CEO-Poster.pdfThis file presents a schematic overview of the ontology metamodel.CEO-LatestPaperOnCEO.pdfThis file contains the latest publication on the CEO ontology. (published in Proceedings of LREC 2018.
If you use this resource, please cite: Segers, R., T.Caselli, P. Vossen (2018). “The Circumstantial Event Ontology (CEO) and ECB+/CEO; an Ontology and Corpus for Implicit Causal Relations between Events”. In: Proceedings of the 11th edition of the Language Resources and Evaluation Conference LREC, Miyazaki, Japan, May 7-12, 2018.
roxane.segers (the email symbol) gmail.com
This work was co-funded by:
- Netherlands Organization for Scientific Research (NWO) via the Spinoza grant, awarded to Piek Vossen in the project Understanding Language by Machines
This repository is an adapted clone of this original repository. Cloned an adapted in April 2018.
The Circumstantial Event Ontology(CEO) by Roxane Segers is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
Based on a work at https://github.com/RoxaneSegers/CEO-Ontology.
