Чанкинг текста

Many machine learning methods have recently been applied to natural lan- guage processing tasks. Among them, the Winnow algorithm has been ar- gued to be particularly suitable for NLP problems, …...

Combination of features contributes to a significant improvement in accuracy on tasks such as part-of-speech (POS) tag- ging and text chunking, compared with us- ing atomic features. However, selecting combination …...

This paper proposes two techniques for fast sequential labeling such as part-of-speech (POS) tagging and text chunking. The first technique is a boosting-based al- gorithm that learns rules represented by …...

This paper proposes a boosting algorithm that uses a semi-Markov perceptron. The training algorithm repeats the training of a semi-Markov model and the update of the weights of training samples. …...

Most of the sequence tagging tasks in nat- ural language processing require to recog- nize segments with certain syntactic role or semantic meaning in a sentence. They are usually tackled …...

AbstractThough feature extraction is a necessary first step in statistical NLP, it is often seen as a mere preprocessing step. Yet, it can dominate computation time, both during training, and …...

AbstractToday, the dominant paradigm for training neural networks involves minimizing task loss on a large dataset. Using world knowledge to inform a model, and yet retain the ability to perform …...