The end-to-end performance of natural language processing systems for compound tasks, such as question answering and textual entailment, is often hampered by use of a greedy 1-best pipeline archi tecture, which causes errors to propagate and compound at each stage. We present a novel architecture, which models these pipelines as Bayesian networks, with each low level task corresponding to a variable in the network, and then we perform approximate inference to find the best labeling. Our approach is extremely simple to apply but gains the benefits of sampling the entire distribution over labels at each stage in the pipeline. We apply our method to two tasks – semantic role labeling and recognizing textual entailment – and achieve useful performance gains from the superior pipeline architecture.
%0 Generic
%1 finkel06solving
%A Finkel, Jenny Rose
%A Manning, Christopher D.
%A Ng, Andrew Y.
%B Conference on Empirical Methods in Natural Language Processing
%D 2006
%K 2006 stanford parsetree parser NT2OD nlp
%T Solving the Problem of Cascading Errors: Approximate Bayesian Inference for Linguistic Annotation Pipelines
%U http://www.stanford.edu/~jrfinkel/papers/pipeline.pdf
%X The end-to-end performance of natural language processing systems for compound tasks, such as question answering and textual entailment, is often hampered by use of a greedy 1-best pipeline archi tecture, which causes errors to propagate and compound at each stage. We present a novel architecture, which models these pipelines as Bayesian networks, with each low level task corresponding to a variable in the network, and then we perform approximate inference to find the best labeling. Our approach is extremely simple to apply but gains the benefits of sampling the entire distribution over labels at each stage in the pipeline. We apply our method to two tasks – semantic role labeling and recognizing textual entailment – and achieve useful performance gains from the superior pipeline architecture.
@misc{finkel06solving,
abstract = {The end-to-end performance of natural language processing systems for compound tasks, such as question answering and textual entailment, is often hampered by use of a greedy 1-best pipeline archi tecture, which causes errors to propagate and compound at each stage. We present a novel architecture, which models these pipelines as Bayesian networks, with each low level task corresponding to a variable in the network, and then we perform approximate inference to find the best labeling. Our approach is extremely simple to apply but gains the benefits of sampling the entire distribution over labels at each stage in the pipeline. We apply our method to two tasks – semantic role labeling and recognizing textual entailment – and achieve useful performance gains from the superior pipeline architecture. },
added-at = {2007-02-25T19:34:27.000+0100},
author = {Finkel, Jenny Rose and Manning, Christopher D. and Ng, Andrew Y.},
biburl = {https://www.bibsonomy.org/bibtex/2d194e966b42648a00c1cb4e220fa5d75/butonic},
booktitle = {Conference on Empirical Methods in Natural Language Processing},
interhash = {45a38ebe929871d57ea6720de279882c},
intrahash = {d194e966b42648a00c1cb4e220fa5d75},
keywords = {2006 stanford parsetree parser NT2OD nlp},
organization = {The Stanford Natural Language Processing Group},
school = {Stanford University},
timestamp = {2007-02-25T19:34:27.000+0100},
title = {{S}olving the {P}roblem of {C}ascading {E}rrors: {A}pproximate {B}ayesian {I}nference for {L}inguistic {A}nnotation {P}ipelines},
url = {http://www.stanford.edu/~jrfinkel/papers/pipeline.pdf},
year = 2006
}