hurshprasad / absa

Aspect Based Sentiment Analysis (ABSA)

Deep Learning using Neural Attention and Multi-Tasking

Applied to the SEMEVAL research task 2016 ABSA

The task was removed from SEMEVAL-2017, but still looking to slowly update/structure codebase from Tensorflow 0.10. This is partial code base of prior implementation.

Focus on multi-class/label tasks and used target phrases for attention.

Refrences

• Neural Attention & RNNs

  • Neural machine translation by jointly learning to align and translate
  • Effective Approaches to Attention-based Neural Machine Translation https://arxiv.org/abs/1508.04025
  • Neural and Syntactic Models of Entity-Attribute Relationship for Aspect-based Sentiment Analysis

• Multi Tasking

  • Multitask Learning Author: Rich Caruana http://link.springer.com/article/10.1023/A:1007379606734
  • https://www.ncbi.nlm.nih.gov/pubmed/21992749

• Convolutions for NLP

  • Convolutional Neural Networks for Sentence Classification https://arxiv.org/abs/1408.5882
  • WildML - UNDERSTANDING CONVOLUTIONAL NEURAL NETWORKS FOR NLP
  • https://github.com/yoonkim/CNN_sentence
  • INSIGHT-1 at SemEval-2016 Task 5: Deep Learning for Multilingual Aspect-based Sentiment Analysis

absa's Issues

Problem with reading xml file

Hi,
I have a reading error when I try to read xml file. the output of the error is:

ParseError: not well-formed (invalid token): line 16, column 75

How can I solve this error?

Thanks, Ali