Improving Event Coreference Resolution by Modeling Correlations between Event Coreference Chains and Document Topic Structures
- Run https://github.com/prafulla77/TAC-KBP-2017-Participation to extract event mentions and obtain pairwise score for coreference relations.
- Run https://github.com/PrincetonML/SIF to generate pairwise sentence similarity scores.
- Extract subevents using the patterns described in paper (a sequence of two or more verb event mentions that modify their parent verb with 'conj:and' or 'ccomp' dependency relations in a sentence are extracted as subevents).
- Run new_test_ILP.py.
-
Coreference Scores:
-
Get Subevents using get_subevents.py
-
For sentence similarity: Use:
def _get_sim(emb1, emb2):
inn = (emb1 * emb2).sum(axis=1)
emb1norm = np.sqrt((emb1 * emb1).sum(axis=1))
emb2norm = np.sqrt((emb2 * emb2).sum(axis=1))
scores = inn / emb1norm / emb2norm
return scoresFormat: Dict{Filename:{previoussentence__currentsentence: score}}
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent