Weakly-Supervised Hierarchical Text Classification

Authors: Yu Meng, Jiaming Shen, Chao Zhang, Jiawei Han6826-6833

AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on three datasets from different domains demonstrate the efficacy of our method compared with a comprehensive set of baselines.
Researcher Affiliation Academia Yu Meng, Jiaming Shen, Chao Zhang, Jiawei Han University of Illinois at Urbana-Champaign, Urbana, IL, USA {yumeng5, js2, czhang82, hanj}@illinois.edu
Pseudocode Yes Algorithm 1: Overall Network Training.
Open Source Code No The paper does not provide a direct link to its own open-source code or explicitly state that its code is available.
Open Datasets Yes Yelp Review: We use the Yelp Review Full dataset (Zhang, Zhao, and Le Cun 2015) and take its testing portion as our dataset.
Dataset Splits No The paper mentions pre-training and self-training processes but does not specify a distinct validation set or its split for hyperparameter tuning or model selection.
Hardware Specification No The paper does not provide specific hardware details such as GPU models, CPU types, or memory specifications used for running the experiments.
Software Dependencies No The paper mentions using a 'Skip-Gram model' and 'CNN model' but does not provide specific version numbers for any software dependencies, libraries, or frameworks used (e.g., Python, PyTorch, TensorFlow versions).
Experiment Setup Yes For all datasets, we use Skip-Gram model (Mikolov et al. 2013) to train 100-dimensional word embeddings... we use CNN model with one convolutional layer as local classifiers. Specifically, the filter window sizes are 2, 3, 4, 5 with 20 feature maps each. Both the pre-training and the self-training steps are performed using SGD with batch size 256.