Joint Models for Extracting Adverse Drug Events from Biomedical Text
Authors: Fei Li, Yue Zhang, Meishan Zhang, Donghong Ji
IJCAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results on a standard ADE corpus show that the discrete joint model outperforms a state-of-the-art baseline pipeline significantly. In addition, when discrete features are replaced by neural features, the recall is further improved. |
| Researcher Affiliation | Academia | Fei Li,1 Yue Zhang,2 Meishan Zhang,3 Donghong Ji1 1. State Key Laboratory of Software Engineering, School of Computer, Wuhan University, China 2. Singapore University of Technology and Design 3. School of Computer Science and Technology, Heilongjiang University, China |
| Pseudocode | Yes | Algorithm 1 shows pseudocode of the greedy decoding algorithm... |
| Open Source Code | Yes | Our code is publicly available under GPL at: https://github.com/foxlf823/ade. |
| Open Datasets | Yes | We use the ADE corpus [Gurulingappa et al., 2012], which consists of 1644 Pub Med abstracts for evaluation. |
| Dataset Splits | Yes | We evaluate all the models using 10-fold cross-validation, where 10% of the data are used as the development set, 10% as the test set and the remainder for training. |
| Hardware Specification | No | The paper does not specify any particular hardware components (e.g., CPU, GPU models, memory) used for running the experiments. |
| Software Dependencies | No | The paper mentions using 'The Stanford Core NLP toolkit' for preprocessing but does not provide specific version numbers for this or any other software dependencies. |
| Experiment Setup | Yes | For all the models, we set the initial Ada Grad learning rate and regularization parameter β to 0.01 and 10−8, respectively. For the neural models, embeddings are randomly initialized in the range (-0.01, 0.01), and we set the dimension D to 200 by default. The dropout rate is 0.5. The window size C of the CNN filter is 2 and the size H1 of the CNN output layer is 200. The hidden layer size H2 is 200. |