Conditional Random Fields : Probabilistic Models for Segmenting and Labeling Sequence Data

主要介紹CRF的概念，參考一下兩篇相關文獻 :
[1] J. Lafferty, A. McCallum, and F. Pereira. Conditional random fields: probabilistic models for segmenting and labeling sequence data. In International Conference on Machine Learning, 2001.
[2] Hanna M. Wallach. Conditional Random Fields: An Introduction. University of Pennsylvania CIS Technical Report MS-CIS-04-21.

過去針對sequential data進行segmenting或是labeling的動作，有HMM、MEMMs等技術，以Generative和Discriminative Model區分之。CRFs則是近年來提出的新穎做法，其模型為無向圖，在給定一個觀察序列下，求算整體狀態序列的條件機率。參數的估計方式原先提出iterative scaling algorithm找出Log-likelihood objective function中的最大值，而後的相關文獻中也針對參數預估提出不同的計算方式，以增快求算的效率。

Conditional Random Fields

View more presentations from lswing.