This is a re-implementation of the following paper:
Improved Semantic Representations From Tree-Structured Long Short-Term Memory Networks Kai Sheng Tai, Richard Socher, and Christopher Manning.
The provided implementation can achieve a test accuracy of 51.72 which is comparable with the result reported in the original paper: 51.0(±0.5).
- MXNet nightly build
- requests
- nltk
pip install mxnet --pre
pip install requests nltk
The script will download the [SST dataset] (http://nlp.stanford.edu/sentiment/index.html) and the GloVe 840B.300d embedding automatically if --use-glove
is specified (note: download may take a while).
DGLBACKEND=mxnet python3 train.py --gpu 0
See https://docs.google.com/spreadsheets/d/1eCQrVn7g0uWriz63EbEDdes2ksMdKdlbWMyT8PSU4rc .
The code can work with MXNet 1.5.1