DEFT English Light and Rich ERE Annotation
| Item Name: | DEFT English Light and Rich ERE Annotation |
| Author(s): | Song Chen, Ann Bies, Kira Griffitt, Joe Ellis, Stephanie Strassel |
| LDC Catalog No.: | LDC2023T04 |
| ISLRN: | 712-226-273-489-1 |
| DOI: | https://doi.org/10.35111/ctq4-mn59 |
| Release Date: | April 17, 2023 |
| Member Year(s): | 2023 |
| DCMI Type(s): | Text |
| Data Source(s): | discussion forum, newswire |
| Project(s): | DEFT |
| Application(s): | information extraction |
| Language(s): | English |
| Language ID(s): | eng |
| License(s): |
LDC User Agreement for Non-Members |
| Online Documentation: | LDC2023T04 Documents |
| Licensing Instructions: | Subscription & Standard Members, and Non-Members |
| Citation: | Chen, Song, et al. DEFT English Light and Rich ERE Annotation LDC2023T04. Web Download. Philadelphia: Linguistic Data Consortium, 2023. |
| Related Works: | View |
Introduction
DEFT English Light and Rich ERE Annotation was developed by the Linguistic Data Consortium (LDC) and consists of 1190 English discussion forum, newswire and proxy documents annotated for entities, relations and events (ERE).
DARPA's Deep Exploration and Filtering of Text (DEFT) program aimed to address remaining capability gaps in state-of-the-art natural language processing technologies related to inference, causal relationships and anomaly detection. LDC supported the DEFT program by collecting, creating and annotating a variety of data sources.
Light ERE annotation labels entity mentions for the target set of entity, relation and event types between and among those entities including coreference. Rich ERE annotation expands types and tagging in the entities, relations, and events annotation tasks and replaces strict event coreference with a more loosely defined event hopper annotation. Further information about the annotation methodology is contained in the documentation accompanying this release.
Data
The source data consists of English discussion forum web text collected by LDC for the DARPA BOLT program and contained in BOLT English Discussion Forums (LDC2017T11); newswire documents published in various data sets released in the TAC KBP project (Text Analysis Conference Knowledge Base Population; and proxy documents intended to mimic government analysis reports of newswire content published in DEFT Narrative Text (LDC2016T07).
902 documents were annotated following Light ERE annotation guidelines. 288 documents were labeled with Rich ERE annotation in a second pass after being annotated for Light ERE.
Below is a data summary:
| Light ERE | Rich ERE | |
| Files | 902 | 288 |
| Words | 505,837 | 180,040 |
| Entity Mentions | 76,528 | 34,606 |
| Entities | 23,313 | 11,546 |
| Fillers | n/a | 2,324 |
| Relations | 9,313 | 4,193 |
| Event Mentions | 4,066 | 5,763 |
| Event Hoppers | 2,898 | 4,143 |
Source documents are in plain text format, annotation is in XML format, and both are UTF-8 encoded.
Samples
Please view the following samples:
Updates
None at this time.
Acknowledgement
This material is based on research sponsored by Air Force Research Laboratory and Defense Advance Research Projects Agency under agreement number FA8750-13-2-0045. The U.S. Government is authorized to reproduce and distribute reprints for Governmental purposes notwithstanding any copyright notation thereon. The views and conclusions contained herein are those of the authors and should not be interpreted as necessarily representing the official policies or endorsements, either expressed or implied, of Air Force Research Laboratory and Defense Advanced Research Projects Agency or the U.S. Government.