Multi-Task Medical Concept Normalization Using Multi-View Convolutional Neural Network
Authors: Yi Luo, Guojie Song, Pengyu Li, Zhongang Qi
AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Comprehensive experimental analysis indicates our model outperforms existing baselines which demonstrates the effectiveness of our model. |
| Researcher Affiliation | Academia | Yi Luo,1 Guojie Song,2 Pengyu Li,2 Zhongang Qi3 1Department of Computer Science and Engineering, University of California, San Diego, USA 2Key Laboratory of Machine Perception, Ministry of Education, Peking University, China 3School of Electrical Engineering and Computer Science, Oregon State University, USA |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide concrete access to source code for the methodology described in this paper. |
| Open Datasets | No | The dataset used in this paper comes from Information Center of a cooperative Chinese hospital including discharge summaries from 153 triple-A hospitals in 31 provinces and regions. It has 3125 types of disease and 3154 of procedures. Each record contains a primary diagnosis and main procedure for a patient. Annotations are performed by medical experts producing initial dataset of 7000 entries... The final dataset with 58031 records for training and 6899 for testing is carefully examined by two medical undergraduates. The paper describes the dataset but does not provide any concrete access information (link, DOI, repository, or formal citation for external access). |
| Dataset Splits | No | The final dataset with 58031 records for training and 6899 for testing is carefully examined by two medical undergraduates. The paper specifies training and testing sets but does not explicitly mention a validation set or provide details for a three-way split (train/validation/test). |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments. |
| Software Dependencies | No | The paper mentions software components and methods like "word2vec", "ReLU", "dropout", and "Adam", but does not provide specific version numbers for any software dependencies, libraries, or frameworks used in the experiments. |
| Experiment Setup | Yes | We trained word2vec model (Mikolov et al. 2013) on over 10 million Chinese clinical narrative corpora with word and character vector dimension of 100. For tensor size l, we set it to be 10, 20 for disease and procedure respectively by experimental statistics. For sentence modeling we conduct a Siamese Bi-LSTM model with 15-dimensional hidden vectors ht and memory cells ct. The kernel sizes for CNN model are summed up in Table 2 and 8, 16 feature maps are produced in two convolutional layers. We choose rectifier linear unit (ReLU) (Nair and Hinton 2010) as activation functions and apply droupout (Srivastava et al. 2014) strategy. For training, we use stochastic gradient descent (SGD) Adam (Kingma and Ba 2014) method with shuffled minibatches of size 128 and adopt early-stopping strategy. The learning rate is 0.001 and all trainable parameters are initialized randomly with truncated normalization. We use coefficient α of 0.5. In joint loss, the regularization term λ and β controlling sparsity are set to be 0.1 and 0.001 respectively. |