Anyone recommend a good tutorial on conditional random fields

Question

I am trying to find a good textbook on conditional random fields and not yet find one that has not begun to send my brain into fusion. I am good at HMM, and I get the difference between discriminatory and generative models ... but so far I have not been able to find a resource that can give a good comparison of HMM and CRF, which makes sense to me. Any help would be appreciated.

Answer 1

One of the best resources that I have found is actually a section in Christopher Bishop's book Pattern Recognition and Machine Learning (which I highly recommend, by the way) regarding Markov random fields (CRFs are specialized Markov random fields). This is even an example that, I am sure you have already noticed, is incredibly difficult to find for this subject. Now I must point out that this section will not give you a full understanding of CRF, but it, we hope, at least for me, will help you navigate these treacherous CRF textbooks.

In addition, I did not find anything but stunning academic papers on the subject. Here are a few that I found useful, though:

• An Introduction to Conditional Random Fields for Relational Learning - This is thorough enough.

• Conditional Random Fields: Probabilistic Models for Segmentation and Labeling Sequence Data - This is the document that originally outlined the CRF structure. It’s a little easier for me to read than those that came after him. It also compares CRF with HMM and MEMM (including charts.)

Sorry, that’s all I can contribute. I'm still trying to learn CRF myself.

Answer 2

Classical probabilistic models and conditional random fields

http://www.scai.fraunhofer.de/fileadmin/images/bio/data_mining/paper/crf_klinger_tomanek.pdf

This is by far the best textbook I have so far managed to find. As the name suggests, he develops the idea of ​​CRM, first creating on top of more well-known models, including Naive Bayes, HMM and Maximum Entropy. The use of colors and patterns also enhances clarity.

Answer 3

the most amazing introduction to CRF.

In addition, this handout in the classroom pretty well explains the “designation” for the linear chain CRF.

Answer 4

A very interesting CRF video tutorial presented by Professor Charles Elkan (UCSD): http://videolectures.net/cikm08_elkan_llmacrf

And lecture notes can be downloaded from his homepage: http://cseweb.ucsd.edu/users/elkan/250B/cikmtutorial.pdf

Hurrah! Hung Ngo.

Answer 5

I would also recommend this Ph.D. thesis, it has one chapter on graphic models and one in CRF. He introduces all the concepts necessary for understanding CRF.

Update: replaced the link, in case the link is slowed down again, the title of the dissertation is "Scaling conditional random fields for natural language processing". I must add that it discusses the difference between HMM and CRF.

Answer 6

Also "Shallow Parsing with CRFs" by Sha and Pereira here