5

I am looking for a good tool in R or Python or any other implementation that can help to me generate sampling from hierarchical Pitman-Yor processes (HPY) (one of the recent and popular nonparametric Bayesian methods) to do language modeling.

Any help or pointer is appreciated.

user10899
  • 61
  • 1
    Does this help? http://homepages.inf.ed.ac.uk/s0562315/progs/index.html#hpylm – jthetzel Apr 26 '12 at 18:40
  • http://www.cs.colorado.edu/~mozer/Teaching/syllabi/ProbabilisticModels/lectures/Teh2006.pdf – Tristan Apr 27 '12 at 00:11
  • Thanks so much. I find that both these sources may help. I will have a try. But We can not see Songfang Huang's implementation in detail. – user10899 Apr 27 '12 at 01:58
  • I think Teh's presentation is very instructive and lead a good implementation approach. Is there any other sources? I think many people in NLP are facing the same problem. – user10899 Apr 27 '12 at 02:21
  • 1
    It is often appropriate to politely and concisely contact the authors of presentations and manuscripts to ask if they are willing to share the code used in their analyses. For example, the first author of http://www.ncbi.nlm.nih.gov/pubmed?term=hierarchical%20Pitman-Yor used HPY in a proteomic study and presumably implemented it in R, given the R-ish graphics of the paper and that he was at R-core developer Martin Mächler's ETH Zurich at the time. – jthetzel Apr 27 '12 at 19:13

1 Answers1

2

This question is a bit old, but the implementations at https://github.com/vchahun/vpyp and https://github.com/fgnt/nhpylm might be helpful for anybody looking.

There are also implementations in C++ and blog posts explaining them: https://github.com/musyoku/hpylm. The comments and posts are in Japanese however.

xji
  • 273