Dynamic conditional random fields

摘要: In sequence modeling, we often wish to represent complex interaction between labels, such as when performing multiple, cascaded labeling tasks on the same sequence, or long-range dependencies exist. We present dynamic conditional random fields (DCRFs), a generalization of linear-chain (CRFs) in which each time slice contains set state variables and edges---a distributed representation Bayesian networks (DBNs)---and parameters are tied across slices. Since exact inference can be intractable models, perform approximate using several schedules for belief propagation, including tree-based reparameterization (TRP). On natural-language chunking task, show that DCRF performs better than series CRFs, achieving comparable performance only half training data.