Update README.md

README.md

@@ -66,6 +66,16 @@ Please use this reference to cite our work
 > and the entity-related events with an F-score of 0.859. Finally, the model was used for performing an analysis of biases about women and non-Western people in Wikipedia biographies.",\
 > }
 
+# Structure of the dataset
+Each row in the corpus has information about the target entity of the biography (entity), the annotated token, the label, the label id, and the sentence number
+
+Labels are the following:
+* ENTITY (a mention of the entity)
+* EVENT (all generic event like *win*, *study*)
+* EVENT-GR (semantically void predicates like *be* and *have*. Take a look at the work of Bonial and Palmer on light verbs for more information)
+* EVENT-ASP (aspectual events like *start*, *finish*)
+* EVENT-MOD (events or modifier that modify the actual occurrence of the event. E.g., *not*, *try*)
+
 # Data Statement for WikiBio corpus
 
 Data set name: WikiBio