An Introduction to Conditional Random Fields

By Charles Sutton, Andrew McCallum. Found. Trends Mach. Learn., 2010

1,292 citations recorded by the source corpus.

Many tasks involve predicting a large number of variables that depend on each other as well as on other observed variables. Structured prediction methods are essentially a combination of classification and graphical modeling. They combine the ability of graphical models to compactly model multivariate data with the ability of classification methods to perform prediction using large sets of input features. This survey describes conditional random fields, a popular probabilistic method for structured prediction. CRFs have seen wide application in many areas, including natural language processing, computer vision, and bioinformatics. We describe methods for inference and parameter estimation for CRFs, including practical issues for implementing large-scale CRFs. We do not assume previous knowledge of graphical modeling, so this survey is intended to be useful to practitioners in a wide variety of fields.

Subfields: Artificial Intelligence

View Paper: https://www.semanticscholar.org/paper/eb9c924383aca0d137c736a41271d569a12cf518

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

Confidence Estimation and Deletion Prediction Using Bidirectional Recurrent Neural Networks (2018, methodology): “This includes simple piece-wise linear mappings in the form of decision trees [9], generalised linear models [10], highly non-linear feed-forward neural networks [11] and sequence models, such as conditional random fields (CRF) [12, 13] and recurrent neural networks [14].”
RisQ: Recognizing Smoking Gestures with Inertial Sensors on a Wristband (2014, neutral): “regression and log-linear CRF models [19].”
SLiMFast: Guaranteed Results for Data Fusion and Source Reliability (2015, methodology): “For learning, we can either use expectation maximization (EM) [32, 34] or empirical risk minimization (ERM) [24] when ground truth is available.”
Joint Semantic Segmentation and 3D Reconstruction from Monocular Video (2014, methodology): “We utilize a discriminative CRF model on P (M|D) to avoid directly modeling the complex dependencies [21,28] among correlated sources of with-depth and semantic-only measurements.”
The Trajectron: Probabilistic Multi-Agent Trajectory Modeling With Dynamic Spatiotemporal Graphs (2018, background): “Probabilistic Graphical Models (PGMs) are a principled instantiation of graphical models [7, 13, 35, 47, 48].”
Weakly Supervised Complementary Parts Models for Fine-Grained Image Classification From the Bottom Up (2019, methodology): “Then a conditional random field (CRF) [40] is used to extract higher-quality object instances.”
Text Classification Algorithms: A Survey (2019, background): “CRFs are essentially a way of combining the advantages of classification and graphical modeling which combining the ability to compactly model multivariate data, and the ability to leverage a high dimensional features space for prediction [172] (This model is very powerful for text data due to high ”
Ontology-Based Semantic Image Segmentation Using Mixture Models and Multiple CRFs (2016, methodology): “Moreover, blind combinations of heterogeneous features and contextual cues exploitation through limited neighborhood relations in the CRFs tend to degrade the labeling performance.”
A review of natural language processing techniques for opinion mining systems (2017, background): “For more about CRF, Sutton and McCallum [49] presented a comprehensive introduction.”
Toward systematic review automation: a practical guide to using machine learning tools in research synthesis (2019, background): “The conditional random field (CRF) is amongst the most prominent of these [18].”
Globally Convergent Newton Methods for Ill-conditioned Generalized Self-concordant Losses (2019, neutral): “1]), which include conditional random fields [33].”
Face Clustering in Videos with Proportion Prior (2015, background): “It directly describes how to take observations and assign them labels [Sutton and McCallum, 2012; Lafferty et al., 2001].”
Sequence Classification Restricted Boltzmann Machines With Gated Units (2020, background): “M ODELING sequences is an important research topic with a variety of applications, ranging from natural language processing [1], [2] to computer vision [3].”