narrative schemas

Automatic Extraction of Narrative Schemas from Text

This page contains sets of narrative schemas available for download and their associated documentation.

You may also want to compare these with Nate Chambers' schemas.

Sets of Schemas and Associated Documentation

Formats

Generated schemas are available in two formats currently: python pickles and string dumps.

Python pickles are serializations of python objects. This is how the schemas are piped around from generation to other scripts. If you're coding in Python, it's extremely easy to load the list of schemas in:

schemas = pickle.load(open("schemas.pkl","rb"))

Alternatively, you can get string dumps that have most of the original content of the schemas. These are printed in a format that was easy to hack togehter and generate and is marginally legible:

2
A	pummel	C	_
_	shelter	C	B
A	burn	C	B
A	heat	_	
A	beguile	C	
A	disparage	_	_
A=['PERSON', 'PERSON', 'PERSON', 'PERSON']
C=['PEOPLE', 'PEOPLE', u'people']
B=['LOCATION']
E=[]
D=[]
G=[]
F=[]
H=[]

The following rows–typically the maximum six, but fewer are possible if no verb were found to fit–indicate events in each schema. Each row contains, separated by tabs:

Each of the slot indicators is filled by either a capital letter–referring to a chain expressed underneath the events–an underscore–referring to a slot attested in the data, but not linked with any other slots during schema induction–or nothing, indicating a slot that was not attested in the data. The chain types, expressed below the schema, are lists of tuples of (type, score). These are not ordered.

ACL 2015 CnewS Workshop

Associated Papers

Simonson, D. and Davis, A. (2015, July). Narrative Schemas and Document Categories. In the ACL 2015 CnewS Workshop, Beijing. [paper]

Download (by format)

Properties

Other Notes

EMNLP 2016 CnewS Workshop

Associated Papers

Simonson, D. and Davis, A. (2016, November). NASTEA: Investigating Narrative Schemas through Annotated Entities. In the Second CnewS Workshop, EMNLP 2016, Austin, TX. [paper]

Download (by format)

Properties

Other Notes

Dan Simonson