Fake News Detection Project

---

Fake News Detection with Language Models(KoBERT).

Personal Project , Jan. 2024 ~ Feb. 2024

---

Software Stacks

---

Project Motivation

• The amount of information has increased due to the development of information production technology.
• This has led to an increase in the production of fake information.
• Fake information can confuse readers and spread misinformation, which can lead to a variety of problems.
• we study the detection of “click-bait”, one of the types of fake news.

---

Implementation

1. Preprocessing Data

• Before tokenize, add "Special Token" between title and context

2. DL Model Design

• Using "Pre-trained Ko-BERT Model"
• Ko-BERT Model(link)

3. Training Result

loss	accuracy
0.3304	98.05 %

4. Crawling and inferring real-world news articles

• We crawled 355 "Naver News Articles", Inference with our model.
• The result is shown below

---

Outputs

• Publication Conference Paper in Korean Society For Internet Information Spring Conference (Apr. 2024)