Abstract
LiStr - a toolkit for induction of linguistic structures without annotated corpora. This is the second version of the toolkit, released at December 2015.
It comprises the tools and supporting scripts for induction of dependency trees from texts with already assigned part-of-speech tags. In this version, the delexicalized part-of-speech tag guesser is comprised.
The UDPC was improved, now it may work with unsupervised word-classes and with universal-dependencies prior probabilities.