SemEval 2017 - International Workshop on Semantic Evaluation 2017

SemEval 2017 introduces a new task named ScienceIE, which aims to advance machine reading techniques for the greater scientific community. Scientific research and practice is founded upon gaining, maintaining and understanding the body of existing scientific work.
It focuses on the following fundamental objects: PROCESSes, TASKs and MATERIALs. The task is to automatically read scientific documents, determine the boundaries of those objects, label them with their types and extract hyponymy and synonymy relations between them.
Resources:
A corpus for the task consisting of ScienceDirect open access publications will be made available to participants. It will consist of 500 articles (350 for training, 50 for development and 100 for testing) evenly distributed among the domains Computer Science, Material Sciences and Physics. The training data part of the corpus will consist of 350 documents.
Participates can take part in all or any three subtasks: (A) identification of keyphrases, (B) classification of keyphrases, (C) extracting relationship between keyphrases.
Participants may use resources in addition to the ones provided by the organizers, e.g. other existing data annotated with keyphrases, as long as they declare this at submission time. However, they may not manually annotate the test data. For each of the subtasks, the team with the highest micro F1 score is declared the winner.