|
|
|
|
|
|
Department of Machine Learning
SENNA is a Neural Network Architecture for Semantic Extraction.
We employ a single convolutional neural network architecture that, given a sentence, outputs a host of language processing predictions: part-of-speech tags, chunks, named entity tags, semantic roles and semantically similar words (learnt via a language model). The entire network is trained jointly on all these tasks using multitask learning. All the tasks use labeled data except the language model which learns from unlabeled text and is thus a novel semi-supervised learning method for the shared tasks. Both multitask learning and semi-supervised learning improve the generalization of the shared tasks, resulting in state- of-the-art performance.
Publications :3Related Projects :semantic_extraction. |
|
