Conditional random field

from Wikipedia, the free encyclopedia

A Conditional Random Field (CRF) is a type of undirected probabilistic model that is used in machine learning (a branch of artificial intelligence ). The term is often used to refer to a special form with a linear structure, the linear-chain conditional random field . This is typically used to segment sequences. This means that the CRF receives a sequence as input and outputs a sequence of the same length . In contrast to hidden Markov models (HMMs; a different, but directed model for sequential data), a CRF can access the complete information of the input sequence at any point, whereas an HMM only sees the current input. This means that complex sets of features can be used.

training

Like all machine learning models , CRFs must be trained ; that is, their parameters must be estimated from data . Various learning methods exist for this, such as the gradient method or the quasi-Newton method . A number of sequences are specified, of which both the input and the desired output are known (this is monitored learning ). The learning process then tries to adapt the parameters in the CRF so that the correct output sequence is predicted for as many sequences as possible in the training data.

Applications

CRFs have been successfully applied to various problems such as:

See also

credentials

  • J. Lafferty, A. McCallum, F. Pereira: Conditional random fields: Probabilistic models for segmenting and labeling sequence data . In: Proc. 18th International Conf. on machine learning . Morgan Kaufmann, San Francisco, CA 2001, p. 282-289 .
  • A. McCallum: Efficiently inducing features of conditional random fields . In: Proc. 19th Conference on Uncertainty in Artificial Intelligence . 2003.
  • F. Sha, F. Pereira: Shallow parsing with conditional random fields . University of Pennsylvania, 2003 (Technical Report MS-CIS-02-35).
  • HM Wallach: Conditional random fields: An introduction . University of Pennsylvania, 2004 (Technical Report MS-CIS-04-21).
  • C. Sutton, A. McCallum: An Introduction to Conditional Random Fields for Relational Learning . In: Lise Getoor, Ben Taskar (Ed.): Introduction to Statistical Relational Learning . MIT Press, 2006.
  • R. Klinger, K. Tomanek: Classical Probabilistic Models and Conditional Random Fields . Dortmund University of Technology, December 2007, ISSN  1864-4503 ( Online PDF - Algorithm Engineering Report TR07-2-013).
  • T. Kudo, K. Yamamoto, Y. Matsumoto: Applying Conditional Random Fields to Japanese Morphological Analysis . 2004 ( PDF online ).