Single-Document Keyphrase Extraction for Multi-Document Keyphrase Extraction

Abstract

Here, we address the task of assigning
relevant terms to thematically and semantically related
sub-corpora and achieve superior results compared to
the baseline performance. Our results suggest that
more reliable sets of keyphrases can be assigned to
the semantically and thematically related subsets of
some corpora if the automatically determined sets of
keyphrases for the individual documents of an entire
corpus are identified first. The sets of keyphrases
assigned by our proposed method for the workshops
present in the ACL Anthology Corpus over a 6-year
period were considered better in more than 60% of
the test cases compared to our baseline system when
evaluated against an aggregation of different human
judgements.