Temporal relation extraction using Apriori algorithm

Existing systems for temporal relation extraction is spread over many domains from statistical approaches to hand crafted inference rules. Systems using hand crafted rules show good results in spite of the fact that they require human expertise that can be costly sometimes. Therefore, there is a need for fully automated systems independent of human interactions. Most of the existing temporal relation extraction systems suffer from sparseness of the available dataset. We aim to minimize this problem using an Apriori based algorithm which is proved to efficiently deal market basket datasets. Market basket datasets are sparse by nature; it is therefore hoped that Apriori will perform better in temporal relation extraction. Our proposed method first extract features from the dataset, then vectorize the features so that Apriori algorithm can be applied on the data. Then frequent itemsets are generated by the Apriori algorithm from which the temporal relations are extracted.