Overlap** Word Removal is All You Need: Revisiting Data Imbalance in Hope Speech Detection
Authors:
Hariharan RamakrishnaIyer LekshmiAmmal,
Manikandan Ravikiran,
Gayathri Nisha,
Navyasree Balamuralidhar,
Adithya Madhusoodanan,
Anand Kumar Madasamy,
Bharathi Raja Chakravarthi
Abstract:
Hope Speech Detection, a task of recognizing positive expressions, has made significant strides recently. However, much of the current works focus on model development without considering the issue of inherent imbalance in the data. Our work revisits this issue in hope-speech detection by introducing focal loss, data augmentation, and pre-processing strategies. Accordingly, we find that introducin…
▽ More
Hope Speech Detection, a task of recognizing positive expressions, has made significant strides recently. However, much of the current works focus on model development without considering the issue of inherent imbalance in the data. Our work revisits this issue in hope-speech detection by introducing focal loss, data augmentation, and pre-processing strategies. Accordingly, we find that introducing focal loss as part of Multilingual-BERT's (M-BERT) training process mitigates the effect of class imbalance and improves overall F1-Macro by 0.11. At the same time, contextual and back-translation-based word augmentation with M-BERT improves results by 0.10 over baseline despite imbalance. Finally, we show that overlap** word removal based on pre-processing, though simple, improves F1-Macro by 0.28. In due process, we present detailed studies depicting various behaviors of each of these strategies and summarize key findings from our empirical results for those interested in getting the most out of M-BERT for hope speech detection under real-world conditions of data imbalance.
△ Less
Submitted 11 April, 2022;
originally announced April 2022.
On the decrease of sunspot activity and absence of severe space weather conditions during the multiple planetary conjunction periods of 1850-2000 AD
Authors:
N. G. Nisha,
T. E. Girish
Abstract:
We have studied solar activity and space weather conditions during multiple planetary conjunction periods ( MPC) as observed from Earth during the years 1850-2000 AD. Significant decreases in sunspot activity and solar 10.7 cm radio flux and absence of severe space weather conditions is found during most of these MPC . Possible planetary influences in longitudinal organisation of sunspot activity…
▽ More
We have studied solar activity and space weather conditions during multiple planetary conjunction periods ( MPC) as observed from Earth during the years 1850-2000 AD. Significant decreases in sunspot activity and solar 10.7 cm radio flux and absence of severe space weather conditions is found during most of these MPC . Possible planetary influences in longitudinal organisation of sunspot activity is suggested by these results. The success of major space missions ( lunar and human) during MPC is also noticed during 1957-2000 AD. The applications of the above results include: short term space mission planning and inference/prediction of solar-terrestrial conditions during multiple planetary conjunction periods.
△ Less
Submitted 23 August, 2020;
originally announced September 2020.
