We propose a unified neural network architecture and learning algorithm that can be applied to various natural language processing tasks including part-of-speech tagging, chunking, named entity recognition, and semantic role labeling. This versatility is achieved by trying to avoid task-specific engineering and therefore disregarding a lot of prior knowledge. Instead of exploiting man-made input features carefully optimized for each task, our system learns internal representations on the basis of vast amounts of mostly unlabeled training data. This work is then used as a basis for building a freely available tagging system with good performance and minimal computational requirements.
%0 Journal Article
%1 Collobert:2011:NLP:1953048.2078186
%A Collobert, Ronan
%A Weston, Jason
%A Bottou, Léon
%A Karlen, Michael
%A Kavukcuoglu, Koray
%A Kuksa, Pavel
%D 2011
%I JMLR.org
%J J. Mach. Learn. Res.
%K dfg-antrag-steckbriefe ma-zehe neuralnet sentimentanalysis
%P 2493--2537
%T Natural Language Processing (Almost) from Scratch
%U http://dl.acm.org/citation.cfm?id=1953048.2078186
%V 12
%X We propose a unified neural network architecture and learning algorithm that can be applied to various natural language processing tasks including part-of-speech tagging, chunking, named entity recognition, and semantic role labeling. This versatility is achieved by trying to avoid task-specific engineering and therefore disregarding a lot of prior knowledge. Instead of exploiting man-made input features carefully optimized for each task, our system learns internal representations on the basis of vast amounts of mostly unlabeled training data. This work is then used as a basis for building a freely available tagging system with good performance and minimal computational requirements.
@article{Collobert:2011:NLP:1953048.2078186,
abstract = {We propose a unified neural network architecture and learning algorithm that can be applied to various natural language processing tasks including part-of-speech tagging, chunking, named entity recognition, and semantic role labeling. This versatility is achieved by trying to avoid task-specific engineering and therefore disregarding a lot of prior knowledge. Instead of exploiting man-made input features carefully optimized for each task, our system learns internal representations on the basis of vast amounts of mostly unlabeled training data. This work is then used as a basis for building a freely available tagging system with good performance and minimal computational requirements.},
acmid = {2078186},
added-at = {2016-11-12T17:27:52.000+0100},
author = {Collobert, Ronan and Weston, Jason and Bottou, L{\'e}on and Karlen, Michael and Kavukcuoglu, Koray and Kuksa, Pavel},
biburl = {https://www.bibsonomy.org/bibtex/23f6c52d6baafef5ab3d65facb5500ddc/albinzehe},
description = {Natural Language Processing (Almost) from Scratch},
interhash = {c1e968fc1903e842ab3c638cd5ffca61},
intrahash = {3f6c52d6baafef5ab3d65facb5500ddc},
issn = {1532-4435},
issue_date = {2/1/2011},
journal = {J. Mach. Learn. Res.},
keywords = {dfg-antrag-steckbriefe ma-zehe neuralnet sentimentanalysis},
month = nov,
numpages = {45},
pages = {2493--2537},
publisher = {JMLR.org},
timestamp = {2021-01-12T14:09:47.000+0100},
title = {Natural Language Processing (Almost) from Scratch},
url = {http://dl.acm.org/citation.cfm?id=1953048.2078186},
volume = 12,
year = 2011
}