Exploring structural features for position analysis in political discussions
Cäcilia Zirn, Michael Schäfer, Michael Strube,
Simone Ponzetto and Heiner Stuckenschmidt
Data and Web Science Group, University of Mannheim, Germany
Heidelberg Institute for Theoretical Studies, Heidelberg, Germany
caecilia@informatik.uni-mannheim.de
In the context of the
NLP Unshared Task
in PoliInformatics 2014
, we analyze the
structure of FOMC discussions as potential
features for position analysis.
We access the length of discussion statements
and show that the distinction between long
opinionated statements and
short spontaneous discussion elements improves
the analysis of similarity among
speakers. Furthermore, we explore the
structure within dialogs by dividing them
into subdialogs and representing the subsequence
of speakers as graphs.
In this
web demo, we present visualizations of
our analysis including the subgraphs and
the similarity among speakers.
Download the full publication
Jump directly to the list of detailed meeting anaylses.
As we do not have a gold standard evaluating the correctness of
the identified speaker similarities, we investigate how stable the similarities of the speaker pairs are across all meetings.
Two speakers that are close in one meeting should
be close in the other meetings, too, as they are not
likely to change their position while being on the
committee.
The following boxplot depicts the stability of
the speaker similarity over all meetings for a given pair
of speakers. (Currently four speaker pairs are shown in this demo).
The following heatmap visualization depicts the average speaker similarity values over all analyzed meetings for a selected set of speakers. The cosine similarities for each speaker pair are ranging from dark red ( ≥ 0.1) to dark green ( ≥ 0.64).
We analyzed the transcripts of 27 FOMC meetings between 2005 and 2007. The detailed results of each meetings analysis can be accessed through the following links.
Detailed analysis of meeting FOMC20050202meeting
Detailed analysis of meeting FOMC20050322meeting
Detailed analysis of meeting FOMC20050503meeting
Detailed analysis of meeting FOMC20050630meeting
Detailed analysis of meeting FOMC20050809meeting
Detailed analysis of meeting FOMC20050920meeting
Detailed analysis of meeting FOMC20051101meeting
Detailed analysis of meeting FOMC20051213meeting
Detailed analysis of meeting FOMC20060131meeting
Detailed analysis of meeting FOMC20060328meeting
Detailed analysis of meeting FOMC20060510meeting
Detailed analysis of meeting FOMC20060629meeting
Detailed analysis of meeting FOMC20060808meeting
Detailed analysis of meeting FOMC20060920meeting
Detailed analysis of meeting FOMC20061025meeting
Detailed analysis of meeting FOMC20061212meeting
Detailed analysis of meeting FOMC20070131meeting
Detailed analysis of meeting FOMC20070321meeting
Detailed analysis of meeting FOMC20070509meeting
Detailed analysis of meeting FOMC20070628meeting
Detailed analysis of meeting FOMC20070807meeting
Detailed analysis of meeting FOMC20070810confcall
Detailed analysis of meeting FOMC20070816confcall
Detailed analysis of meeting FOMC20070918meeting
Detailed analysis of meeting FOMC20071031meeting
Detailed analysis of meeting FOMC20071206confcall
Detailed analysis of meeting FOMC20071211meeting