\subsection{Alternative Features}\r
\label{altfeatures}\r
\r
-In PAN 2012, we have used word 5-grams and stop-word 8-grams.\r
-This year we have experimented with different word $n$-grams, and also\r
+In PAN 2012, we used word 5-grams and stop-word 8-grams.\r
+This year we experimented with different word $n$-grams, and also\r
with contextual $n$-grams as described in \cite{torrejondetailed}.\r
Modifying the algorithm to use contextual $n$-grams created as word\r
5-grams with the middle word removed (i.e. two words before and two words\r
\r
\plagdet{0.7421}{0.6721}{0.8282}{1.0000}\r
\r
-We have then made tests with plain word 4-grams, and to our surprise,\r
+We then made tests with plain word 4-grams, and to our surprise,\r
it gave even better score than contextual $n$-grams:\r
\r
\plagdet{0.7447}{0.7556}{0.7340}{1.0000}\r
of the corpus (in terms of plagdet score), except the 02-no-obfuscation\r
part.\r
\r
-In our final submission, we have used word 4-grams and stop-word 8-grams.\r
+In our final submission, we used word 4-grams and stop-word 8-grams.\r
\r
\subsection{Global Postprocessing}\r
\r
%for development, where it has provided a significant performance boost.\r
%The official performance numbers are from single-threaded run, though.\r
\r
-For PAN 2010, we have used the following postprocessing heuristics:\r
+For PAN 2010, we used the following postprocessing heuristics:\r
If there are overlapping detections inside a suspicious document,\r
keep the longer one, provided that it is long enough. For overlapping\r
-detections up to 600 characters, drop them both. We have implemented\r
-this heuristics, but have found that it led to a lower score than\r
+detections up to 600 characters, drop them both. We implemented\r
+this heuristics, but found that it led to a lower score than\r
without this modification. Further experiments with global postprocessing\r
of overlaps led to a new heuristics: we unconditionally drop overlapping\r
detections with up to 250 characters both, but if at least one of them\r
is longer, we keep both detections. This is probably a result of\r
plagdet being skewed too much towards recall (because the percentage of\r
-plagiarized cases in the corpus is way too high compared to real world),\r
+plagiarized cases in the corpus is way too high compared to real-world),\r
so it is favourable to keep the detection even though the evidence\r
for it is rather low.\r
\r
\r
\subsection{Evaluation Results and Future Work}\r
\r
- The evaulation on the competition corpus had the following results:\r
+ The evaluation on the competition corpus had the following results:\r
\r
\plagdet{0.7448}{0.7659}{0.7251}{1.0003}\r
\r
especially well for human-created plagiarism (the 05-summary-obfuscation\r
sub-corpus), which is where we want to focus for our production\r
systems\footnote{Our production systems include the Czech National Archive\r
-of Graduate Theses, \url{http://theses.cz}}.\r
+of Graduate Theses,\\ \url{http://theses.cz}}.\r
\r
% After the final evaluation, we did further experiments\r
%with feature types, and discovered that using stop-word 8-grams,\r
\r
We plan to experiment further with combining more than two types\r
of features, be it continuous $n$-grams or contextual features.\r
-This should allow us to tune down the aggresive heuristics for joining\r
+This should allow us to tune down the aggressive heuristics for joining\r
neighbouring detections, which should lead to higher precision,\r
-hopefully without sacrifying recall.\r
+hopefully without sacrificing recall.\r
\r
As for the computational performance, it should be noted that\r
our software is prototyped in a scripting language (Perl), so it is not\r