OhioState at IJCNLP-2017 Task 4: Exploring Neural Architectures for Multilingual Customer Feedback Analysis

TL;DR

This paper explores simple neural network architectures like CNN and Bi-LSTM, combined with Fasttext, for multilingual customer feedback analysis, achieving competitive results in the IJCNLP 2017 shared task.

Contribution

It demonstrates that shallow neural architectures can perform well in multilingual customer feedback classification, with top-tier results on Spanish and French datasets.

Findings

Top 5 performance in Spanish and French tasks

Achieved 85.28% exact accuracy for Spanish

Outperformed other models on comment and meaningless tags in French

Abstract

This paper describes our systems for IJCNLP 2017 Shared Task on Customer Feedback Analysis. We experimented with simple neural architectures that gave competitive performance on certain tasks. This includes shallow CNN and Bi-Directional LSTM architectures with Facebook's Fasttext as a baseline model. Our best performing model was in the Top 5 systems using the Exact-Accuracy and Micro-Average-F1 metrics for the Spanish (85.28% for both) and French (70% and 73.17% respectively) task, and outperformed all the other models on comment (87.28%) and meaningless (51.85%) tags using Micro Average F1 by Tags metric for the French task.