Konferenzbeitrag
Comparing Relevance Feedback Techniques on German News Articles
Lade...
Volltext URI
Dokumententyp
Text/Conference Paper
Dateien
Zusatzinformation
Datum
2017
Autor:innen
Zeitschriftentitel
ISSN der Zeitschrift
Bandtitel
Verlag
Gesellschaft für Informatik e.V.
Zusammenfassung
We draw a comparison on the behavior of several relevance feedback techniques on a corpus of German news articles. In contrast to the standard application of relevance feedback, no explicit user query is given and the main goal is to recognize a user’s preferences and interests in the examined data collection. The compared techniques are based on vector space models and probabilistic models. The results show that the performance is category-dependent on our data and that overall the vector space approach Ide performs best.