Development of a Swedish Corpus for Evaluating Summarizers and other IR-tools
2001 (English)Report (Other academic)
We are presenting the construction of a Swedish corpus aimed at research1on Information Retrieval, Information Extraction, Named Entity Recognitionand Multi Text Summarization, we will also present the results on evaluatingour Swedish text summarizer SweSum with this corpus. The corpus has beenconstructed by using Internet agents downloading Swedish newspaper textfrom various sources. A small part of this corpus has then been manuallyannotated. To evaluate our text summarizer SweSum we let ten studentsexecute our text summarizer with increasing compression rates on the 100manually annotated texts to find answers to predefined questions. The resultsshowed that at 40 percent summarization/compression rate the correct answerrate was 84 percent.
Place, publisher, year, edition, pages
Stockholm: KTH , 2001. , 7 p.
IdentifiersURN: urn:nbn:se:kth:diva-14077OAI: oai:DiVA.org:kth-14077DiVA: diva2:329590
QC 201007122010-07-122010-07-122010-07-16Bibliographically approved