安装

不需要安装，直接从github上download下来就行
https://github.com/slavpetrov/berkeleyparser

使用方法

java -jar BerkeleyParser-1.7.jar -gr <grammar>

然后就启动了，你输入一句话，它会返回这句话的parse结果。

查看更多参数

java -jar BerkeleyParser-1.7.jar

-render                       Write rendered tree to image file. (Default: false)
-maxLength                    Maximum sentence length (Default = 200).
-binarize                     Output binarized trees. (Default: false)
-variational                  Use variational rule score approximation instead of max-rule (Default: false)
-useGoldPOS                   Read data in CoNLL format, including gold part of speech tags.
-dumpPosteriors               Dump max-rule posteriors to disk.
-ec_format                    Use Eugene Charniak's input and output format.
-nThreads                     Parse in parallel using n threads (Default: 1).
-modelScore                   Output effective model score (max rule score for max rule parser) (Default: false)
-keepFunctionLabels           Retain predicted function labels. Model must have been trained with function labels. (Default: false)
-nGrammars                    Use a product model based on that many grammars
-chinese                      Enable some Chinese specific features in the lexicon.
-scores                       Output inside scores (only for binarized viterbi trees). (Default: false)
-tokenize                     Tokenize input first. (Default: false=text is already tokenized)
-substates                    Output subcategories (only for binarized viterbi trees). (Default: false)
-outputFile                   Store output in this file instead of printing it to STDOUT.
-confidence                   Output confidence measure, i.e. likelihood of tree given words: P(T|w) (Default: false)
-gr                           Grammarfile (Required)
 [required]
-accurate                     Set thresholds for accuracy. (Default: set thresholds for efficiency)
-inputFile                    Read input from this file instead of reading it from STDIN.
-kbest                        Output the k best parse max-rule trees (Default: 1).
-tree_likelihood              Output joint likelihood of tree and words: P(t,w) (Default: false)
-viterbi                      Compute viterbi derivation instead of max-rule tree (Default: max-rule)
-sentence_likelihood          Output sentence likelihood, i.e. summing out all parse trees: P(w) (Default: false)
-h                            help message

我的使用方法

import os

def berkeley_parse(input_file, output_file, berkeley_jar='./berkeleyparser/BerkeleyParser-1.7.jar', gr='./berkeleyparser/eng_sm6.gr'):
    os.system('java -jar ' + berkeley_jar + ' -gr ' + gr + ' -inputFile ' + input_file + ' -outputFile ' + output_file)

berkeley_parse(input_file, output_file)