論文: A Model of Text for Experimentation in the Social Sciences
python stm_main.py -f <filename> -k <number of topic> -i <iteration count>
optional arguments:
-h, --help show this help message and exit
-f FILENAME Set corpus filepath. Fileformat is csv
-d DOCUMENT Set document field name
-c CORPUS Using range of Brown corpus' files(start:end)
--alpha ALPHA Parameter alpha for LDA(default=1.0)
--beta BETA Parameter beta for LDA(default=0.1)
-k TOPICS Number of topics(default=20)
-i ITERATION Iteration count(default=100)
-x X Set prevalences column name
-y Y Set covariates column name
--parser PARSER Select parser eng_nltk or mecab(default=mecab)
--sigma SIGMA Initial value of sigma diagonals(default=0.1)
--stopwords Exclude stop words by using corpus from nltk
--seed SEED Random seed
--df DF Threshold of document freaquency to cut words
--interact Consider interaction between covariates adn topics
--sinit Smart initialize of parameters for LDA
注意
-yをつけた場合、計算速度が落ちます。また、perplexityがあまり下がらない傾向があります。