Abstract | ||
---|---|---|
Temporal information extracted from texts and normalized to some standard format has been exploited in a variety of tasks such as information retrieval and question answering. Classifying documents into categories using temporal features has not yet been tried. Such a method might be particularly valuable when classifying sensitive texts such as patient records, i.e., whenever the pure content of the documents should not be used for the classification. In this paper, we describe, as a proof-of-concept, our work on classifying news articles exploiting only features defined over extracted and normalized temporal expressions. Our evaluation of two classification models on large German and English news archives shows promising results and demonstrates the discriminative power of temporal features for topically classifying text documents.
|
Year | DOI | Venue |
---|---|---|
2019 | 10.1145/3308560.3315000 | Companion Proceedings of The 2019 World Wide Web Conference |
Keywords | Field | DocType |
news classification, temporal information, temporal n-grams | Data mining,Normalization (statistics),Question answering,Computer science,Temporal expressions,Artificial intelligence,Natural language processing,Discriminative model,German | Conference |
ISBN | Citations | PageRank |
978-1-4503-6675-5 | 0 | 0.34 |
References | Authors | |
0 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Lukas Lange | 1 | 0 | 2.03 |
Omar Alonso | 2 | 855 | 65.44 |
Jannik Strötgen | 3 | 492 | 38.20 |