Enriching the Knowledge Sources Used in a Maximum Entropy Part-of-Speech Tagger

By Kristina Toutanvoa, Christopher D. Manning. Conference on Empirical Methods in Natural Language Processing, 2000

1,677 citations recorded by the source corpus.

This paper presents results for a maximum-entropy-based part of speech tagger, which achieves superior performance principally by enriching the information sources used for tagging. In particular, we get improved results by incorporating these features: (i) more extensive treatment of capitalization for unknown words; (ii) features for the disambiguation of the tense forms of verbs; (iii) features for disambiguating particles from prepositions and adverbs. The best resulting accuracy for the tagger on the Penn Treebank is 96.86% overall, and 86.91% on previously unseen words.

Subfields: Artificial Intelligence

View Paper: https://www.semanticscholar.org/paper/1504a9d5829033a8cb4cf37b8bb13dfd4baddc7b

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

Chinese Word Segmentation as Character Tagging (2003, neutral): “We then briefly describe the maximum entropy model, which has been successfully applied to POS tagging as well as parsing [Ratnaparkhi, 1996; Ratnaparkhi, 1998].”
Linguistically Motivated Large-Scale NLP with C&C and Boxer (2007, methodology): “The taggers are based on Maximum Entropy tagging methods (Ratnaparkhi, 1996), and can all be trained on new annotated data, using either GIS or BFGS training code.”
Single Document Keyphrase Extraction Using Neighborhood Knowledge (2008, methodology): “We used the Stanford log-linear POS tagger (Toutanova and Manning, 2000) in this study.”
Wide-Coverage Efficient Statistical Parsing with CCG and Log-Linear Models (2007, methodology): “Log-linear models have been applied to a number of problems in NLP, for example, POS tagging (Ratnaparkhi 1996; Lafferty, McCallum, and Pereira 2001), named entity recognition (Borthwick 1999), chunking (Koeling 2000), and parsing (Johnson et al. 1999).”
Discriminative Training Methods for Hidden Markov Models: Theory and Experiments with Perceptron Algorithms (2002, methodology): “For both methods, we tried the algorithms with feature count cut-o s set at 0 and 5 (i.e., we ran experiments with all features in training data included, or with all features occurring 5 times or more included { (Ratnaparkhi 96) uses a count cut-o of 5).”
Discriminative models for information retrieval (2004, methodology): “Both techniques have been successfully applied in several natural language processing tasks such as tagging [24] and text-classification [20, 12].”
Towards Holistic and Automatic Evaluation of Open-Domain Dialogue Generation (2020, methodology): “(Toutanova and Manning, 2000) and the WordNet by Miller (1998) were utilized to do WordNet substitution.”
Language With Character: A Stratified Corpus Comparison of Individual Differences in E-Mail Communication (2006, methodology): “First, there is the issue of positivity.”
Question classification with log-linear models (2006, background): “Maximum Entropy ( ME ) or log-linear models [5] have been successfully applied to many Natural Language Processing ( NLP ) problems which require complex and overlapping features.”
Generating captions without looking beyond objects (2016, methodology): “the dataset but are automatically extracted from the captions generated by the Karpathy et al. system to increase comparability of the results. For this, the Stanford log-linear part-of-speech tagger [6,5] is used to extract all noun tags. The model for the blind noun translation task is trained on ”
Design and realization of a modular architecture for textual entailment (2013, methodology): “The English LAP includes a pipeline based on the Stanford NLP tools (Toutanova and Manning 2000; Finkel, Grenager and Manning 2005), an easy-first parser (Goldberg and Elhadad 2010), and the Ark coreference resolver (Haghighi and Klein 2009).”
TectoMT: Modular NLP Framework (2010, neutral): “TagMxPost uses Adwait Ratnaparkhi’s tagger [1].”
Keyphrase Extraction in Scientific Publications (2007, background): “The plain text form is ﬁrst processed to delimit sentences, then passed to a modern maximum entropy based part-of-speech (POS) tagger [11].”