The Use of Classifiers in Sequential Inference

By Vasin Punyakanok, Dan Roth. 2001

219 citations recorded by the source corpus.

We study the problem of combining the outcomes of several different classifiers in a way that provides a coherent inference that satisfies some constraints. In particular, we develop two general approaches for an important subproblem - identifying phrase structure. The first is a Markovian approach that extends standard HMMs to allow the use of a rich observation structure and of general classifiers to model state-observation dependencies. The second is an extension of constraint satisfaction formalisms. We develop efficient combination algorithms under both models and study them experimentally in the context of shallow parsing.

Subfields: Artificial Intelligence, Probabilistic & Bayesian Methods

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

1. Dan Roth
2. Percy Liang
3. Andrew McCallum
4. Mohit Bansal
5. Luke S. Zettlemoyer
6. Christopher D. Manning
7. Noah A. Smith
8. Hannaneh Hajishirzi

In Probabilistic & Bayesian Methods

1. Wen-tau Yih
2. Dan Roth
3. Andrew McCallum
4. Hal Daumé Iii
5. Daniel Khashabi
6. Xin Li
7. Lev-Arie Ratinov
8. Haoruo Peng

Citations in the corpus

• Reasoning about Quantities in Natural Language (2015, methodology): “We compare the accuracy and running times of the Semi-CRF model (SC) (Sarawagi and Cohen, 2004) and the bank of classifiers model (C+I) (PR) (Punyakanok and Roth, 2001), using 10-fold cross-validation.”
• Learning Question Classifiers (2002, methodology): “Chunks are extracted using a previously learned classifier (Punyakanok and Roth, 2001; Li and Roth, 2001).”
• Search-based structured prediction (2009, neutral): “In the rst variety, the structure in the problem is ignored, and a single classier is trained to predict each element in the output vector independently [ 43 ] or with dependence created by enforcement of membership inYx constraints [45].”
• Illinois-LH: A Denotational and Distributional Approach to Semantics (2014, neutral): “with the Illinois Chunker (Punyakanok and Roth, 2001).”
• Solving Hard Coreference Problems (2019, methodology): “We start by extracting noun phrases using the Illinois-Chunker (Punyakanok and Roth, 2001).”
• Question Answering as Global Reasoning Over Semantic Abstractions (2018, methodology): “For all of these Table 3: The semantic annotator combinations used in our implementation of S EMANTIC ILP. annotator combinations, we let the representation of the answers be K ( A ) = { Shallow-Parse , Tokens } .”
• Shallow Parsing with Conditional Random Fields (2003, background): “, 2000; Ratnaparkhi, 1996) and a variety of other linear classifiers, including winnow (Punyakanok and Roth, 2001), AdaBoost (Abney et al.”
• Design Challenges and Misconceptions in Named Entity Recognition (2009, methodology): “Systems based on perceptron have been shown to be competitive in NER and text chunking (Kazama and Torisawa, 2007b; Punyakanok and Roth, 2001; Carreras et al., 2003) We specify the model and the features with the LBJ (Rizzolo and Roth, 2007) modeling language.”
• Convolutional Neural Networks vs. Convolution Kernels: Feature Engineering for Answer Sentence Reranking (2016, methodology): “We used the Illinois chunker (Punyakanok and Roth, 2001), question class and focus classifiers trained as in (Severyn and Moschitti, 2013) and the Stanford CoreNLP (Manning et al.”
• Probabilistic Reasoning for Entity & Relation Recognition (2002, methodology): “Recognizing and classifying entities and relations in text data is a key task in many NLP problems such as information extraction (IE) (Califf and Mooney, 1999; Freitag, 2000; Roth and Yih, 2001), question answering (QA) (Voorhees, 2000) and story comprehension (Hirschman et al., 1999).”
• Hidden Markov Support Vector Machines (2003, methodology): “Our work continues a recent line of research that includes Maximum Entropy Markov Models (MEMMs) (McCallum et al., 2000; Punyakanok & Roth, 2001), Conditional Random Fields (CRFs) (Lafierty et al., 2001), perceptron re-ranking (Collins, 2002; Collins & Dufiy, 2002) and label sequence boosting (Altun…”
• The University of Illinois submission to the WMT 2015 Shared Translation Task (2015, methodology): “We used a state-of-the-art shallow parser (Punyakanok and Roth, 2001) in conjunction with a constituent parser (Socher et al.”
• Modeling Latent-Dynamic in Shallow Parsing: A Latent Conditional Model with Improved Inference (2008, methodology): “To accommodate multiple overlapping features on observations, some other approaches view the sequence labeling problem as a sequence of classification problems, including support vector machines (SVMs) (Kudo & Matsumoto 2001) and a variety of other classifiers ( Punyakanok & Roth 2001; Abney et al.”
• Efficiently Inducing Features of Conditional Random Fields (2002, methodology): “999), sequence tagging (Rat naparkhi, 1996; Punyakanok & Roth, 2001; McCallum et a!., 2000; Lafferty et a!., 2001; Sha & Pereira, 2003) however, all these examples have used hand-generated fea tures.”
• Understanding the Value of Features for Coreference Resolution (2008, methodology): “We resolve coreference on unannotated text as follows: First we detect mention heads following a state of the art chunking approach (Punyakanok and Roth, 2001) using standard features.”