By Craveiro, O.; Macedo, J.; Madeira, H.
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
There is a lot of recent work aimed at improving the effectiveness in Information Retrieval results based on temporal information extracted from texts. Some works use all dates but others use only document creation or modification timestamps. However, no previous work explicitly focuses on the use of dates within in the document content to establish temporal relationships between words in the document. This work estimates these relationships through a temporal segmentation of the texts, exploring them to expand queries. It was achieved very promising results (13% improvement in Precision@15), especially for temporal aware queries. To the best of our knowledge, this is the first work using temporal text segmentation to improve retrieval results.