This is the source code of SAG, which is used to extract the video tags from time-sync comments by a semantic associate graph (SAG). This program can be run by mingw32-g++ compiler (Code::Blocks IDE 16.01) under Windows. We provide sample data with 239 time-sync comments for debugging.

sentence_vector: The folder contains the sentence vector of time-sync comments. The suffix of the file in sentence_vector should be ".txt". In the sample data, the "sample.txt" contains 2*239 lines. Each time-sync comment in sample data contains two lines. The first line contains the timestamp and a time-sync comment. And the second line contains the 200-dimensional sentence vector. In this paper,

word_segmentation: The folder contains the participle results of time-sync comments with part of speech. The suffix of the file in sentence_vector should be ".txt". The file name in “word_segmentation” must be same in "sentence_vector". In the sample data, the "sample.txt" contains 239 lines. Each time-sync comment in sample data contains one line: Chinese word segmentation results (including part of speech).

idf2.txt: The word IDF scores.

result: The folder contains the results of video tag extraction.

dialogue.cpp: The model with dialogue-based cluster algorithm.

topic_center.cpp: The model with topic center-based cluster algorithm.

A detailed explanation of the programs can be found in the program annotation.

This program only supports running under the Windows operating system. The full data set for the experiment could not be published due to copyright reasons. You can email [email protected] to request part of the data.