Neural Sentence Simplification with Semantic Dependency Information

Authors: Zhe Lin, Xiaojun Wan13371-13379

AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate SDISS on three benchmark datasets and it outperforms a number of strong baseline models on the SARI and FKGL metrics. Human evaluation also shows SDISS can produce simplified sentences with better quality.
Researcher Affiliation Academia Zhe Lin, Xiaojun Wan Wangxuan Institute of Computer Technology, Peking University Center for Data Science, Peking University The MOE Key Laboratory of Computational Linguistics, Peking University {linzhe, wanxiaojun}@pku.edu.cn
Pseudocode No The paper provides model architecture descriptions and mathematical equations but does not include explicit pseudocode or algorithm blocks.
Open Source Code Yes Our code is publicly available at https://github.com/ L-Zhe/SDISS.
Open Datasets Yes We evaluate our SDISS model on three benchmark datasets: Newsela, Wiki Small and Wiki Large. Newsela (Xu, Callison-Burch, and Napoles 2015)... Wiki Large (Zhang and Lapata 2017)... Wiki Small was built by Zhu, Bernhard, and Gurevych (2010)...
Dataset Splits Yes Zhang and Lapata (2017) provide standard splits and the train/dev/test sets contain 94,208/1,129/1,076 sentence pairs, respectively. (for Newsela) [...] Wiki Large... with 296,402/2,000/359 complex-simple sentence pairs for train/dev/test sets. [...] Wiki Small... with 88,837/205/100 pairs provided by Zhang and Lapata (2017) as train/dev/test sets.
Hardware Specification No The paper does not provide specific details about the hardware used for experiments, such as GPU models, CPU models, or cloud computing instances.
Software Dependencies No The paper mentions using the 'Stanford Core NLP tool (Manning et al. 2014)' but does not specify a version number for it or for any other software dependencies.
Experiment Setup Yes We set the dimensions of word embedding and hidden units dmodel to 256. For multi-head attention, we set the number of heads to 4. The number of layers for graph encoder, sentence encoder and decoder are all set to 6. We set λ to 5 for Newsela, and 10 for Wiki Small and Wiki Large. We set tmax to 15 for Newsela and 20 for Wiki Small, and tmin is 0 for both datasets. For Wiki Large, because of its complexity, we set tmin to 5 and tmax to 50.