Project information
Semantic Role Labeling is a fundamental NLPtask, which aims to find semantic roles for each predicate in a
sentence. The goal of the SRL is to extract the predicate-argument structure of a sentence, identifying"
who did what to "whom" , "when", "where", etc. For example, consider this sentence: The cat eats a fish.
Eats is the verb, The cat is the subject, and a fish is the object complement. We are not interested in
the meaning of" cat" or" fish," but we want to identify and classify them, i.e., associate each argument
with its corresponding role. LSTM-based models in different configurations were used in this paper to
solve this task, including pre-trained word embeddings and contextualized word embedding from BERT and
Graph Convolutional Network. Furthermore, the subtask of the disambiguation of predicates is also
considered because often, the datasets provided have information on the predicates present in the
sentences but not the clarification of the meaning.
