humanity123/memnet_absa

Aspect Level Sentiment Classification with Deep Memory Network

TensorFlow implementation of Tang et al.'s EMNLP 2016 work.

Problem Statement

Given a sentence and an aspect occurring in the sentence, this task aims at inferring the sentiment polarity (e.g. positive, negative, neutral) of the aspect.

Example

For example, in sentence ''great food but the service was dreadful!'', the sentiment polarity of aspect ''food'' is positive while the polarity of aspect ''service'' is negative.

Quick Start

Download the 300-dimensional pre-trained word vectors from Glove and save it in the 'data' folder as 'data/glove.6B.300d.txt'.

Train a model with 7 hops on the Laptop dataset.

python main.py --show True

Note this code requires TensorFlow, Future and Progress packages to be installed.

memnet_absa's Issues

Batching the data

Hi, I was trying to re-implement the work, clean up the code and sort of work with batch data instead of 1 at a time. I notice your code breaks when batch size is increased, but it also looks like going by the paper - this work is not meant to be batched. Am I right in thinking that?

attention weights

how can we see attention weights in different hops like shown in table 4 and table 5 in the paper https://arxiv.org/pdf/1605.08900.pdf

Thank you