In this competition, we will use F1-score as the evaluation metric for information extraction. The F1-score is the harmonic mean of precision and recall.
Precision = # of data fields extracted correctly / # of data fields extracted
Recall = # of data fields extracted correctly / # of data fields
F1-score = (2 * Precision * Recall) / (Precision + Recall)
To extract the correct data fields, you need accurately find out the words with the data fields' information, and categorize the data field correctly.
The test has been divided into two parts. Only the score of the public test set will be revealed. However, the final leaderboard is decided by the score of the private test set.
2019 Datagrand Cup: Text Information Extraction Challenge