Generalized Expectation Criteria for Semi-Supervised Learning of Conditional Random Fields

By Gideon S. Mann, Andrew McCallum. 2008

139 citations recorded by the source corpus.

Subfields: Representation & Self-Supervised Learning

How researchers see this paper

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

In Representation & Self-Supervised Learning

Citations in the corpus

Benefits of Intermediate Annotations in Reading Comprehension (2020, background): “Our auxiliary training with intermediate annotations is inspired by extensive related work on training models using side information or domain knowledge beyond labels (Mann and McCallum, 2008; Chang et al., 2007; Ganchev et al., 2010; Rocktaschel et al., 2015).”
Injecting Logical Background Knowledge into Embeddings for Relation Extraction (2015, methodology): “…Supervised Learning Our work is also inspired by weakly supervised approaches (Ganchev et al., 2010) that use structural constraints as a source of indirect supervision, and have been used for several NLP tasks (Chang et al., 2007; Mann and McCallum, 2008; Druck et al., 2009; Singh et al., 2010).”
Generating More Interesting Responses in Neural Conversation Models with Distributional Constraints (2018, methodology): “Our approach, which incorporates distributional constraints into the decoding objective, is related to prior work on posterior regularization (Mann and McCallum, 2008; Ganchev et al., 2010; Zhu et al., 2014).”
An Introduction to Conditional Random Fields (2010, background): “expectation criteria [81], discriminative methods [70], posterior regularization [39, 45], and measurement-based learning [73].”
Machine Learning Paradigms for Speech Recognition: An Overview (2013, background): “Generalized expectation criteria [147] represent prior knowledge as labeled features,”
Generalized Expectation Criteria for Semi-Supervised Learning with Weakly Labeled Data (2010, methodology): “GE criteria can be used as a sole criterion for an objective function (e.g., Mann and McCallum 2008).”
Code Mixing: A Challenge for Language Identification in the Language of Social Media (2014, methodology): “and sequence labelling with Conditional Random Fields trained with Generalized Expectation criteria (Mann and McCallum, 2008; Mann and McCallum, 2010), which achieved the highest scores.”
Active Learning by Labeling Features (2009, methodology): “3.1 Learning with feature-label distributions Mann and McCallum (2008) apply GE to a linear-chain, ﬁrst-order CRF.”
Posterior Regularization for Structured Latent Variable Models (2010, methodology): “The GE objective is where ||·|| b is typically the l 2 norm (Druck et al., 2009 use l 2 2 ) or a distance based on KL divergence (Mann and McCallum, 2008), and the model is a log-linear model such as maximum entropy or a CRF.”
(Almost) No Label No Cry (2014, unknown): “The latter, including AMM, (conv/alter)-∝SVM, iteratively optimize a cost over labelings (always consistent with label proportions for AMM, not always for (conv/alter)-∝SVM).”
Active Learning Literature Survey (2009, background): “” The learning algorithm then tries to find a set of model parameters that match expected label distributions over the unlabeled pool U against these user-specified priors (for details, see Druck et al., 2008; Mann and McCallum, 2008).”
Toward interactive training and evaluation (2011, methodology): “As in previous work [6, 7, 16], in this paper we predominantly use input feature label distribution constraints.”
Sequential Labeling Using Deep-Structured Conditional Random Fields (2010, methodology): “Following the earlier work [4], [7], [16], [18] on this task, the annotated ads are divided into a 102-ads training set, a 100-ads development set, and a 100-ads test set.”
Learning from measurements in exponential families (2009, methodology): “# labeled examples 10 25 100 CCR07 74.7 78.5 81.7 MM08 74.6 77.2 80.5 no labeled predicates 67.7 75.6 81.5 + 33 labeled predicates 71.4 76.5 82.5 sidered two types of measurements: fully-labeled examples and labeled predicates where we provide the frequency of the most common label for a word ty”
Structured learning with constrained conditional models (2012, background): “For example, Mann and McCallum (2008), Bellare et al. (2009) proposes a Generalized Expectation Criteria, which uses a different distance function in the E-step.”
Modeling Relations and Their Mentions without Labeled Text (2010, methodology): “Our work is based on constraint-driven semi-supervised learning [6,16].”
Semi-Supervised Learning via Generalized Maximum Entropy (2010, background): “In (Mann & McCallum , 2008) the same authors extend ER to CRFs for semi-supervised structured prediction.”