Tagging English Text with a Probabilistic Model

By Bernard Merialdo. 1994

How researchers see this paper

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

Citations in the corpus

Scaling to Very Very Large Corpora for Natural Language Disambiguation (2001, background): “…some amount of annotated data is available and a much larger amount of data exists unannotated, e.g. Marialdo's HMM part-of-speech tagger training (1994), Charniak's parser retraining experiment (1996), Yarowsky's seeds for word sense disambiguation (1995) and Nigam et al's (1998) topic…”
Simple, robust, scalable semi-supervised learning via expectation regularization (2007, background): “Merialdo (1994), in a now famous negative result, attempts semi-supervised learning to improve HMM part-of-speech tagging and .nds that EM with unlabeled data reduces accuracy.”
Unsupervised Risk Estimation with only Structural Assumptions (2016, background): “However, a wide body of empirical evidence shows that, when the generative model is misspecified, the unlabeled examples can actually degrade performance (Merialdo, 1994; Cozman & Cohen, 2006; Liang & Klein, 2008; Li & Zhou, 2015).”
WordICA—emergence of linguistic representations for words by independent component analysis (2010, methodology): “Johnson (2007) has taken this line of research further by comparing diﬀerent methods for estimating the HMMs.”
A fully Bayesian approach to unsupervised part-of-speech tagging (2007, methodology): “Perhaps the most well-known is that of Merialdo (1994), who used MLE to train a trigram hidden Markov model (HMM).”
Distributed Asynchronous Online Learning for Natural Language Processing (2010, background): “We can surmise from the log-likelihood plot that the drop in accuracy is not due to the optimization being led astray, but probably rather due to the complex relationship between likelihood and taskspecific evaluation metrics in unsupervised learning (Merialdo, 1994).”
Generalized Expectation Criteria for Semi-Supervised Learning with Weakly Labeled Data (2010, background): “Merialdo (1994) gives a classic example where EM fails to help part-of-speech tagging.”
Contrastive Estimation: Training Log-Linear Models on Unlabeled Data (2005, methodology): “Our experiments are inspired by those in Merialdo (1994); we train a trigram tagger using only unlabeled data, assuming complete knowledge of the tagging dictionary.”
A High-Performance Semi-Supervised Learning Method for Text Chunking (2005, background): “Although useful under some circumstances, when a relatively large amount of labeled data is available, the procedure often degrades performance (e.g. Merialdo (1994)).”
Automating the Construction of Internet Portals with Machine Learning (2000, background): “Merialdo (1994) nds a similar eect on tagging accuracy when training part-of-speech taggers using Baum-Welch training when starting from well-estimated initial parameter estimates.”
Learning Syntactic Categories Using Paradigmatic Representations of Word Context (2012, background): “…disambiguation systems, which use labeled training data (Church, 1988), unsupervised disambiguation systems, which use a dictionary of possible tags for each word (Merialdo, 1994), or prototype driven systems which use a small set of prototypes for each class (Haghighi and Klein, 2006).”