A Neural Model for Joint Event Detection and Summarization

Authors: Zhongqing Wang, Yue Zhang

IJCAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments show that our proposed neural joint model is more effective compared to its pipeline baseline.
Researcher Affiliation Academia Zhongqing Wang , and Yue Zhang Soochow University, Suzhou, China Singapore University of Technology and Design, Singapore
Pseudocode No The paper describes algorithmic steps and models using mathematical formulations and descriptive text (e.g., in Section 3.2), but it does not include explicitly labeled pseudocode or algorithm blocks.
Open Source Code Yes We release the code and data sets of this paper at https://github. com/wangzq870305/joint_event_detection.
Open Datasets Yes We collect and annotate two datasets for evaluating the performance of our proposed system, one from the earthquake domain, and another from DDo S attack domain. ... We release the code and data sets of this paper at https://github. com/wangzq870305/joint_event_detection.
Dataset Splits Yes We randomly choose 10 events as training data for the earthquake domain, 80 events as training data for the DDo S domain, and the remaining events as testing data.
Hardware Specification No The paper describes the software components and training parameters, but does not specify any particular hardware (e.g., CPU, GPU models, memory) used for running the experiments.
Software Dependencies No The paper mentions using Adagrad for optimization, dropout, and training word embeddings using the Skip-gram algorithm, but it does not specify version numbers for these or other software libraries or frameworks (e.g., Python, PyTorch, TensorFlow versions).
Experiment Setup Yes Our training objective is to minimize the cross-entropy loss between the gold labels and predicted labels on those three tasks. We apply online training, adjusting model parameters using Adagrad [Duchi et al., 2011]. In order to avoid overfitting, dropout is used on word embeddings with a ratio of 0.2 [Hinton et al., 2012]. The size of the hidden layers Hd, Hc, and Hs are equally set to 32. We train word embeddings using the Skip-gram algorithm1, and fine-tune them during training. The size of word embeddings is 128.