The automatic recognition of the rhetori-
cal function of citations in scientic text
has many applications, from improvement
of impact factor calculations to text sum-
marisation and more informative citation
indexers. Citation function is dened as
the author's reason for citing a given pa-
per (e.g. acknowledgement of the use
of the cited method). We show that
our annotation scheme for citation func-
tion is reliable, and present a supervised
machine learning framework to automati-
cally classify citation function, which uses
several shallow and linguistically-inspired
features. We nd, amongst other things, a
strong relationship between citation func-
tion and sentiment classication.
We study the interplay of the discourse struc-
ture of a scientic argument with formal ci-
tations. One subproblem of this is to clas-
sify academic citations in scientic articles ac-
cording to their rhetorical function, e.g., as a
rival approach, as a part of the solution, or
as a awed approach that justies the cur-
rent research. Here, we introduce our anno-
tation scheme with 12 categories, and present
an agreement study
"Concordle has one point common with Wordle: it makes word clouds. But these are only text, and in a browser in general the choice of fonts is limited, so the clouds are not so very pretty. But it is much more clever: All the words in the cloud are clickable, i.e. they have links to concordancer function. "
Min -- can't remember if I shared this with you before. A qurkiy little tool that might be useful in terms of training. See 'wikisheet' at http://kristinaweb20.pbworks.com/activity+-+introduction+to+ddl for an 'introduction to data-driven learning' with a Concordle task.