Improving Distant Supervision for Information Extraction Using Label Propagation Through Lists
By Lidong Bing, Sneha Chaudhari, Richard C. Wang, William W. Cohen. 2015
39 citations recorded by the source corpus.
Because of polysemy, distant labeling for information extraction leads to noisy training data. We describe a procedure for reducing this noise by using label propagation on a graph in which the nodes are entity mentions, and mentions are coupled when they occur in coordinate list structures. We show that this labeling approach leads to good performance even when off-the-shelf classifiers are used on the distantly-labeled data.
Subfields: Natural Language Processing
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Natural Language Processing
- Ruslan Salakhutdinov
- Hamed Zamani
- Zhilin Yang
- Mostafa Dehghani
- William W. Cohen
- Mohit Bansal
- Jianfeng Gao
- Jie Tang
Citations in the corpus
- Revisiting Semi-Supervised Learning with Graph Embeddings (2016, methodology): “We next considered the DIEL (Distant Information Extraction using coordinate-term Lists) dataset (Bing et al., 2015).”
- Neural Ranking Models with Weak Supervision (2017, background): “Weak supervision is a natural way to benefit from unsupervised data and it has been applied in NLP for various tasks including relation extraction [3, 15], knowledge-base completion [17], sentiment analysis [34], etc.”
- Graph Partition Neural Networks for Semi-Supervised Classification (2018, methodology): “Note that since Freebase is used as ground truth and some entities are not present in texts, the upper bound of recall given by [6] is 61 .”
- Semi-supervised Learning on Graphs with Generative Adversarial Nets (2018, methodology): “r determining a definite label. Pearson correlation coefficient rp = cov(||д||,pf ) σ ||д||σp f exceeds 0.6, indicating obvious positive correlation. 5.3 Results on Entity Extraction The DIEL dataset [3] is a dataset for information extraction. It contains pre-extracted features for each entity ment”
- A clinical text classification paradigm using weak supervision and deep representation (2019, background): “It is proposed primarily for relation extraction from text, wherein a known relation from a knowledge base (e.g. Freebase) is likely to express that relation in an input corpus [26, 27].”