To compute the Maximum Likelihood Estimator (MLE) of Collins Parser, Collins Model 1 first generates the head of the rule according to its probability, then generates the head’s dependents. Below is an example from part of the parse tree of the sentence “Workers dumped sacks into a bin.”

University of Notre Dame

Collins parser is a parser that divides constituent into a non-terminal head, non-terminals that are on the left of the head, and non-terminals that are on the right of the head. Abstractly speaking, $LHS=L_n L_{n-1}...L_1 H R_1 ...R_{n-1}R_n$, where $L_i$, $H$ and $R_i$ are symbols that representing its category, head and head tag, such as VP(dumped, VP) in the “Workers dumped sacks into a bin.” example.  

Collins Parser

An on-going but a helpful book resource about NLP
https://web.stanford.edu/~jurafsky/slp3/

Learn Before

Related