Keywords (3)

Related Publications (25)

Academic

Active Learning for Natural Language Parsing and Information Extraction

Active Learning for Natural Language Parsing and Information Extraction (Citations: 136)


	BibTex \| RIS \| RefWorks Download

, ,

Conference: International Conference on Machine Learning - ICML , pp. 406-414, 1999

Cumulative Annual

View Publication

- ...Most existing work in active learning has concentrated on two data selection strategies: certainty-based [5][6][7][8] and committee-based selection [9][10][11][12]...
Weiwei Yuan, et al. Initial training data selection for active learning
- ... New approaches based on active learning may help select cases for manual labeling for construction of training sets....
Prakash M Nadkarni, et al. Natural language processing: an introduction
- ...The selection of only one sample per iteration yields the highest information gain with respect to the number of labeled samples [13], but—depending on the application—it might be more convenient for a human domain expert or even less expensive to label a set S of samples (with |S| > 1) in each iteration...
Tobias Reitmaier, et al. Active classifier training with the 3DS strategy
- ...One of the earliest active learning works for complex prediction tasks is [69] which studies active learning for both natural language parsing and information extraction from the perspective of unreliability sampling [10]...
Kevin Small, et al. Margin-based active learning for structured predictions
- ...Regarding an appropriate data selection, there are two basic strategies, namely certainty-based selection [7] and committee-based selection [8]...
- ...corresponding training algorithm) and is based on the certainty-based selection strategy [7]...
Edwin Lughofer, et al. On Dynamic Selection of the Most Informative Samples in Classification...