SegBot: A Generic Neural Text Segmentation Model with Pointer Network

Authors: Jing Li, Aixin Sun, Shafiq Joty

IJCAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In our experiments, SEGBOT outperforms state-of-the-art models on both topic and EDU segmentation tasks. We conduct two sets of experiments to evaluate the effectiveness of SEGBOT: segmenting a document into topically coherent segments, and segmenting a sentence into EDUs.
Researcher Affiliation Academia Jing Li, Aixin Sun and Shafiq Joty School of Computer Science and Engineering, Nanyang Technological University, Singapore jli030@e.ntu.edu.sg, {axsun,srjoty}@ntu.edu.sg
Pseudocode No The paper describes the model with equations but does not include any structured pseudocode or algorithm blocks.
Open Source Code No The paper states: 'An online version of SEGBOT (EDU segmentation) is available at http://138.197.118.157:8000/segbot/.' This refers to a demo or online service, not the open-source code for the methodology itself.
Open Datasets Yes For evaluating topic segmentation models, we use the commonly used Choi dataset [Choi, 2000].
Dataset Splits Yes We use the first 10% data of shuffled training set as development set for both Choi dataset and RST-DT dataset.
Hardware Specification Yes SEGBOT is implemented with Py Torch framework and evaluated on NVIDIA Tesla P100 GPU.
Software Dependencies No The paper states 'SEGBOT is implemented with Py Torch framework' but does not provide specific version numbers for PyTorch or any other software dependencies.
Experiment Setup Yes Table 1 gives the details of other hyper-parameter settings. We use Adam optimizer to update model parameters. In addition, we use gradient clipping by a max norm of 5 and l2 -regularization during training.