Using Time-Series for Temporal Intent Disambiguation in NTCIR-12 Temporalia

Our group DUT-NLP-EN participated in the TID subtask (English) of NTCIR-12 Temporalia and submitted three runs. The temporal intent probability distribution of four categories (past, recency, future and atemporal) for the 300 test queries are predicted through logistic regression model in all the three runs. In RUN1, four groups of features are used including trigger word, word POS, explicit time gap, temporal probability of words. Implicit time gap is added in the form of rule-based time gap in RUN2 and in the form of timeseries statistics in RUN3. RUN2 performs slightly better than the rest two runs with AvgCosine of 0.732 and AvgAbsLoss of 0.210.