Verified the effect on embedding temporal features into words vectors by increasing the temporal expressions in the sentences. Furthermore, using LSTM to capture the words sequence and considered focus time calculating as document classification tasks which means classify documents to certain time period.
Temporal Information Retrieval
Taking document focus time as a new feature during ranking as well as an application of focus time calculation. The focus time will be compared to query issue time and calculate the weight scores based on query type and time distance. Index, retrieval and rank rely on Apache Solr.
Recent work is using knowledge graph to constract historical sequence of events. This temporal knowledge graph structures can deal with queries like what happens before/after certain events.
