Born for Auto-Tagging: Faster and better with new objective functions

TL;DR

This paper introduces BAT, a new auto-tagging model with innovative objective functions and learning strategies that outperform state-of-the-art models in speed and accuracy for keyword extraction tasks.

Contribution

The paper presents BAT, a novel auto-tagging model with new objective functions and a revamped learning rate strategy that achieve faster convergence and higher F scores.

Findings

BAT converges faster than other SOTA models.

BAT achieves higher F scores at 50 epochs compared to competitors at 100 epochs.

New objective functions improve F1 and F2 scores simultaneously.

Abstract

Keyword extraction is a task of text mining. It is applied to increase search volume in SEO and ads. Implemented in auto-tagging, it makes tagging on a mass scale of online articles and photos efficiently and accurately. BAT is invented for auto-tagging which served as awoo's AI marketing platform (AMP). awoo AMP not only provides service as a customized recommender system but also increases the converting rate in E-commerce. The strength of BAT converges faster and better than other SOTA models, as its 4-layer structure achieves the best F scores at 50 epochs. In other words, it performs better than other models which require deeper layers at 100 epochs. To generate rich and clean tags, awoo creates new objective functions to maintain similar ${\rm F_1}$ scores with cross-entropy while enhancing ${\rm F_2}$ scores simultaneously. To assure the even better performance of F scores awoo revamps the learning rate strategy proposed by Transformer \cite{Transformer} to increase ${\rm F_1}$ and ${\rm F_2}$ scores at the same time.