t-SS3: a text classifier with dynamic n-grams for early risk detection over text streams

TL;DR

t-SS3 enhances the SS3 text classifier by dynamically recognizing important n-grams in text streams, improving early risk detection and explanation quality in mental health applications.

Contribution

It introduces t-SS3, an extension of SS3 that dynamically learns and utilizes n-grams for better classification in text streams.

Findings

Improves early depression and anorexia detection performance.

Enhances the richness of visual explanations.

Outperforms previous SS3 results in eRisk tasks.

Abstract

A recently introduced classifier, called SS3, has shown to be well suited to deal with early risk detection (ERD) problems on text streams. It obtained state-of-the-art performance on early depression and anorexia detection on Reddit in the CLEF's eRisk open tasks. SS3 was created to deal with ERD problems naturally since: it supports incremental training and classification over text streams, and it can visually explain its rationale. However, SS3 processes the input using a bag-of-word model lacking the ability to recognize important word sequences. This aspect could negatively affect the classification performance and also reduces the descriptiveness of visual explanations. In the standard document classification field, it is very common to use word n-grams to try to overcome some of these limitations. Unfortunately, when working with text streams, using n-grams is not trivial since the system must learn and recognize which n-grams are important "on the fly". This paper introduces t-SS3, an extension of SS3 that allows it to recognize useful patterns over text streams dynamically. We evaluated our model in the eRisk 2017 and 2018 tasks on early depression and anorexia detection. Experimental results suggest that t-SS3 is able to improve both current results and the richness of visual explanations.