Word Vector Enrichment of Low Frequency Words in the Bag-of-Words Model for Short Text Multi-class Classification Problems
Authors:
Bradford Heap,
Michael Bain,
Wayne Wobcke,
Alfred Krzywicki,
Susanne Schmeidl
Abstract:
The bag-of-words model is a standard representation of text for many linear classifier learners. In many problem domains, linear classifiers are preferred over more complex models due to their efficiency, robustness and interpretability, and the bag-of-words text representation can capture sufficient information for linear classifiers to make highly accurate predictions. However in settings where…
▽ More
The bag-of-words model is a standard representation of text for many linear classifier learners. In many problem domains, linear classifiers are preferred over more complex models due to their efficiency, robustness and interpretability, and the bag-of-words text representation can capture sufficient information for linear classifiers to make highly accurate predictions. However in settings where there is a large vocabulary, large variance in the frequency of terms in the training corpus, many classes and very short text (e.g., single sentences or document titles) the bag-of-words representation becomes extremely sparse, and this can reduce the accuracy of classifiers. A particular issue in such settings is that short texts tend to contain infrequently occurring or rare terms which lack class-conditional evidence. In this work we introduce a method for enriching the bag-of-words model by complementing such rare term information with related terms from both general and domain-specific Word Vector models. By reducing sparseness in the bag-of-words models, our enrichment approach achieves improved classification over several baseline classifiers in a variety of text classification problems. Our approach is also efficient because it requires no change to the linear classifier before or during training, since bag-of-words enrichment applies only to text being classified.
△ Less
Submitted 18 September, 2017;
originally announced September 2017.
