Welcome to CrfCom's Home!
CRF and CrfCom
CRF ( Conditional Random Field ) is a probabilistic model used in labeling or parsing sequential data. It has been proved to be effective in fulfilling such tasks as Named Entity Recognition, Information Extraction and Text Chunking.
CRF++(CrfPP) is a well-versed C++ implementation of CRF by Taku Kudo. It provides a simple, yet powerful and portable tool for researchers and developers who wish to try and test their research scheme based on CRF.
There are other open-source CRF implementations in C++ available (eg. hCRF, CRFsuite etc.), but most of them are provided as standalone tools and are not intended to be readily used as a component for GUI applications.
CrfCom is a COM library based on CrfPP. It is a not only a wrapper of CrfPP, but also provides a number of extra funtional features that CrfPP and other similar tools have not been equipped with.
Main Features
- CrfCom is a dual-interface in-proccess COM server that uses Free memory model and is thread-safe.
- CrfCom supports events and enables clients to be notified of the operation status during lengthy operations and when the tasks are done.
- CrfCom provides increment building method so that additional training data can be added to an existing model file.
- CrfCom implements ISupportErrorInfo and IProvideClassInfo2 and enables clients to receive error messages generated in the server and obtain type information on its object's outgoing interfaces.
How to Use
CrfCom is a very easy-to-use component for clients implemented in different languages like C#,VB,C++ and other script languages.
To start using
CrfCom, please refer to the Getting Started section of the
online documentation.
References
- J. Lafferty, A. McCallum, and F. Pereira. Conditional random fields: Probabilistic models for segmenting and labeling sequence data, In Proc. of ICML, pp.282-289, 2001
- Taku Kudo, CRF++: Yet Another CRF toolkit V0.51, Jul 11 2008
© HBCHEN. All Rights Reserved.