TY - CONF
T1 - A comparative study on generalizability of information extraction models on protest news
AU - Başar, Erkan
AU - Ekiz, Simge
AU - Van Den Bosch, Antal
PY - 2019
Y1 - 2019
N2 - Information Extraction applications can help social scientists to obtain necessary information to understand the reasons behind certain social dynamics. Many recent state-of-the-art information extraction approaches are based on supervised machine learning which can recognize information that has similar patterns with previously shown ones. Recognizing relevant information with never-shown patterns, however, is still a challenging task. In this study, we design a Recurrent Neural Network (RNN) architecture employing ELMo embeddings and Residual Bidirectional Long-Short Term Memory layers to overcome this challenge in the context of CLEF 2019 ProtestNews shared task. Furthermore, we train a classical Conditional Random Fields (CRF) model as our strong baseline to display a contrast between a state-of-the-art classical machine learning approach and a recent neural network method both in performance and in generalizability. We show that RNN model outperforms classical CRF model and shows a better promise on generalizability.
AB - Information Extraction applications can help social scientists to obtain necessary information to understand the reasons behind certain social dynamics. Many recent state-of-the-art information extraction approaches are based on supervised machine learning which can recognize information that has similar patterns with previously shown ones. Recognizing relevant information with never-shown patterns, however, is still a challenging task. In this study, we design a Recurrent Neural Network (RNN) architecture employing ELMo embeddings and Residual Bidirectional Long-Short Term Memory layers to overcome this challenge in the context of CLEF 2019 ProtestNews shared task. Furthermore, we train a classical Conditional Random Fields (CRF) model as our strong baseline to display a contrast between a state-of-the-art classical machine learning approach and a recent neural network method both in performance and in generalizability. We show that RNN model outperforms classical CRF model and shows a better promise on generalizability.
KW - Conditional random fields
KW - Information extraction
KW - Recurrent neural networks
KW - Word embeddings
UR - https://www.mendeley.com/catalogue/0a87a2e2-ed49-3b7d-bbcc-9cc79dc65c06/
UR - https://www.mendeley.com/catalogue/0a87a2e2-ed49-3b7d-bbcc-9cc79dc65c06/
M3 - Paper
ER -