Carnegie Mellon University
Dikshit_Thomas_MOMQ_2023.pdf (226.49 kB)

Financial Claim Detection using Binary Classification

Download (226.49 kB)
posted on 2023-09-18, 16:09 authored by Erin Susan Thomas, Malika Dikshit

This is a FinTech research project that aims at exploring and detecting financial claims, a typical information retrieval task that analyzes managers' speeches in Earnings Conference Calls (ECCs). The goal of claim detection in argument mining is to sort out the key points from a long narrative. In this project, we focus on distinguishing between numerals in the ECCs that are relevant (in-claim) to the company's financial performance and those that are irrelevant (out-of-claim). Given a sentence and the target numeral, we formulate the problem as a binary classification task to tell if the given numeral is an in-claim numeral or not. We explored and evaluated different approaches exploiting a range of machine learning and deep learning algorithms. Our best model yields an F1 score of 86.5%.




Academic Program

  • Information Systems


Houda Bouamor

Usage metrics