FENCE: Financial Exaggerated Numeral ClassifiEr

EasyChair Preprint 10151

Abstract

The presence of exaggerated numbers i.e. inflated figures in financial texts allude investors to making wrong decisions. This justifies the need for an automated system for detecting such numerals. In this paper, we present Financial Exaggerated Numeral ClassifiEr (FENCE) to do so. It consists of a neural network based classifier trained with the context-based embeddings of numerals present in financial market comments and blog titles. FENCE has been deployed in Hugging Face spaces and open sourced under the MIT license.

Keyphrases: Exaggerated Numeral Detection, Financial Natural Language Processing, Natural Language Processing

@booklet{EasyChair:10151,
  author    = {Sohom Ghosh and Sudip Kumar Naskar},
  title     = {FENCE: Financial Exaggerated Numeral ClassifiEr},
  howpublished = {EasyChair Preprint 10151},
  year      = {EasyChair, 2023}}