Recommendation: keeping a logbook

By now, the reader will have realised that multiple detection tasks are needed per work-pair in order to tease out as many text reuse styles from the texts as possible. To help keep track of these experiments, we recommend the creation of a logbook to record each and every task with the respective result-set. We find this helps to avoid duplicating analyses and more easily identify the optimal settings for a particular detection task. An example logbook might look something like this:

WORD-BASED FEATURING

Task N Preprocessing Selection Scoring Result
1 syn-repl.; lemmatiz. local-max-feat-freq.; 0.6 feat-dens. 0.5 sim-thresh. 5 reuses
2 ... ... ... ...
3 ... ... ... ...

BIGRAM FEATURING

Task N Preprocessing Selection Scoring Result
1 ... ... ... ...
2 ... ... ... ...

TRIGRAM FEATURING

Task N Preprocessing Selection Scoring Result
1 ... ... ... ...
2 ... ... ... ...

results matching ""

    No results matching ""