wd5 / manalyzer

MAnalyzer 0.1 {#mainpage}

MAnalyzer is fast morph analyzer library, written in C++. It is based on pymorphy algorithm [(read in russian)] (http://packages.python.org/pymorphy/algo.html) and dictionaries.

Features:

• Return grammar info of the word and normal form with grammar info for the word.
• Prediction for unknown words.

MAnalyzer lib released under MIT License (look for analyzer/LICENSE file). create_dics/ script and examples in ./main.cpp and ./cache.cpp files have nothing in common with MAnalyzer library, they are released under MIT License too.

Compilation

Just:

cd analyzer
make

To see debug information while MAnalyzer using use make debug=true instead of make.

Also you can copy libmanalyzer.so file to /usr/lib/ directory (or just use -L flag of g++ compiler).

Dictionary

First of all you need dictionary. In create_dics/ directory there is script, that will make dictionary from sqlite-json-dictionaries for pymorphy. You can download them here.

Anyway you always can download complete dictionaries from here.

Usage

Usage is simple. Include two header files:

#include "analyzer/word_infos.hpp"
#include "analyzer/analyzer.hpp"

And somewhere add these lines:

// Creating analyzer. (dics is directory with dictionary).
Analyzer * analyzer = analyzer_new("dics");
// Array with results. (1024 is array size).
WordInfos * wi = infos_new(1024);
// Analyzing word.
analyzer_get_word_info(analyzer, word, word_size, wi);

To get information about word use these functions:

// Gets amount of results.
int infos_get_size(WordInfos * wi);
// Gets normal form for result.
char * infos_get_normal_form(WordInfos * wi, unsigned int id);
// Gets normal form's id.
unsigned short int infos_get_normal_form_id(WordInfos * wi, unsigned int id);
// Gets form's id for analyzed word.
unsigned short int infos_get_form_id(WordInfos * wi, unsigned int id);

To compile program use g++ main.cpp -lmanalyzer.

More information you can find out in documentation.

ToDo

• Add dawgdic source code and compile library using it.
• Finish documentation (LemmasRules', EndingsRules' descriptions and algorithm's description).
• Refactoring of create_dics/* code is needed!
• Prevent program from crashing when there is no dictionaries.

ChangeLog

0.1

• Initial release.