notesum.ai

Published at November 13

cs.CL

cs.AI

Released Date: November 13, 2024

Authors: Junhua Liu¹, Kwan Hui Lim², Roy Ka-Wei Lee²

Aff.: ¹Forth AI; ²Singapore University of Technology and Design

Model	Precision	Recall	F1	Accuracy
Human Evaluation
Shortlisting	0.8464	0.8418	0.8156	0.8155
Interview-Rec	0.8011	0.8193	0.8321	0.8087
Traditional Machine Learning Models
XGBoost	0.7902	0.7859	0.7878	0.7931
TF-IDF	0.6938	0.6527	0.6488	0.6839
Neural Network Models
MLP	0.7967	0.7990	0.7911	0.7989
HAN	0.7716	0.7707	0.7711	0.7759
BiLSTM-Indv	0.7963	0.7612	0.7667	0.7816
BiLSTM-Concat	0.8291	0.8178	0.8176	0.8276
Retrieval-Based Models
FAISS-L2	0.6886	0.6707	0.6897	0.6897
FAISS-CS	0.6659	0.6724	0.6654	0.6713
FAISS-CV	0.7796	0.7511	0.7558	0.7640
Large Language Models
GPT-4o	0.5579	0.5114	0.4111	0.5600
GPT-4o-RA	0.7347	0.7365	0.7352	0.7371
Proposed Models
BGM-HAN	0.8622	0.8405	0.8453	0.8506
BGM-HAN- $\mathcal{W}_{SAR}$	0.8995	0.8918	0.8945	0.8966