Towards Personal Relation Extraction Based on Sentence Pattern Tree

Extracting personal relation triple (S, P, O) from large number of unstructured text is crucial to the construction of knowledge graph, knowledge representation and reasoning of personal relation. Aiming at low accuracy in extracting triples from unstructured text, we present a supervised approach to judge whether extracted triples are correct. The approach need to build a knowledge base which contain peoples attributes first, then a sentence pattern tree is learnt according the people attribute knowledge base and the training data. When training, triples are extracted from the text automatically and labelled whether correct or not manually. Then patterns are constructed layer-by-layer according the position of “triple”, “pronoun” and “word” in sentence. At the same time, the correct and error number of triples are recorded on each pattern. When testing, the correctness of triples can be judged by the number recorded in matched patterns. According the test result, our approach does better in the training time, the testing time and the F1-value (76.6%) than the ordinary approach based on feature engineering (75.7%). At last, we make the judgement result of sentence pattern tree as a feature to improve the feature engineering approach (77.5%). In addition, this approach has a better expansibility than the traditional one and has guiding significance to the construction of the training set.