Submission guidelines for CCCS at Multiling 2015 ------------------------------------------------ The CCCS task consists in generating short summaries, called synopses, from call-centre conversation transcripts. Each participant can choose to submit system outputs for the French data (DECODA corpus), the Italian data (LUNA corpus) or the English data (translations from both corpora). For each language, a participant can submit up to 3 runs using the submission page at http://pageperso.lif.univ-mrs.fr/~benoit.favre/cccs/submissions/. For each language, one of the run should be primary. If no primary run is indicated, an arbitrary run will be chosen by the organisers. One run corresponds to a zip file containing two files: - cccs--.tsv is the output of the system for that run - cccs--.README containing a description of the run (system configuration, etc.) The format for the system output file is one synopsis per line with the following tab-separated format (without spaces added for clarity): \t : \t \n Spaces can be used in . They should not be used in , and . A script is provided for checking the conformance of a submission to that format. You can run it as follows (if your team id is "teamid" and the run id is "runid"): python2 cccs_check_submission.py teamid runid submission.zip There is no naming convention for the zip file itself. The script does not enforce that each conversation of the test set has a synopsis, but it checks for duplicates.