#TTG tweet timeline generation

#Usage python cluster.py ../data/query.txt ../data/candidate/ 100 corpus.info.txt 1.77 0.23 ../data/result/ 0.9978 0.022

#Format of files ../data/query.txt qid \t qcontent ../data/candidate/ The directory includes 55 candidate files: qid \t Qid \t wid \t rank \t score \t runTag \t wcontent \t qcontent corpus.info.txt First line contains the total number of words in corpus. And the format of the following each line is: word \t tfValue

#Notice In the usage command, 100 stands for mu in Dirichlet smoothing. And 1.77, 0.23, 0.9978 are the relevance, novelty and coverage threshold respectively. Although 0.022 the threshold for Jaccard, it is not applied in the best performance, which means the file "jaccard.py" is useless here. Likewise, the usage of embedding doesn't work for our TTG system, so I leave out the file "w2v.py".