anomaly_detection_in_PTT_forum

script description:

step_1. Because the orginal result of crawler is json, we split to article and reply with csv format.

step_2. Merge data and check date range.

step_3. Use self defined dict of political keywords to check article's political spectrum.

step_4. Use neo4j to get community with interaction between PTT ID by modularity algorithms.

step_5. Parse neo4j result of community to formal csv.

step_6. Use package CKIP's word cut funciton with article's title and content, then use NTUSD's sentiment lable to calculate pos/neg words, and then merge id's commnuty by its ID with step_4's result.

step_7. Use step5's result check personal political spectrum by her/his most posted type of political spectrum.