Requirements:
- Adaptor Grammar software (py-cfg) compiled in PY_CFG (Makefile) (http://web.science.mq.edu.au/~mjohnson/Software.htm)
- Python
- gensim
- (optional) pattern
To launch it on default settings (see in the Makefile, Providence corpus, children Naima, start age 11 months, end age 22 months), you need to put a .txt file in ProvidenceFinal/ToSegment/my_corpus.txt with document boundaries @ (known) or @? (possible boundary but unknown). Then run:
- make prepare_topics
- make all
You can play with the CHILD / EAGE / SAGE variables in the Makefile. E.g. :
make just_basic_and_single CHILD=naima SAGE=11 EAGE=22 NITER=500
make test_wo_prefix_topic CHILD=naima SAGE=11 EAGE=22 NITER=500
for eage in range 12 22; do make just_basic_and_single CHILD=naima SAGE=11 EAGE=$eage NITER=500; done
To get all the data points and plot them, use e.g.:
for eage in range 12 22; do make just_basic_and_single CHILD=naima SAGE=11 EAGE=$eage NITER=500; done
python src/plot_AGs_results.py
the 500-520 iterations and 1000-1010 iterations are ran with a Gamma(100,0.001) PYP b parameter prior the 600-620 iterations are ran with a Gamma(100,0.01) PYP b parameter prior