/team-night-coding-sergeants

Night Coding Sergeants

Primary LanguageJupyter Notebook

Team Night Coding Sergeants

image

Hi ! We are the "Nighht Coding Sergeants" team. It cosnists of 4 students from Military University of Technology in Warsaw (Wojskowa Akademia Techniczna), Poland: Dawid MAŚLANIK, Hubert SIKORSKI, Rafał WASILUK and Kamil WIŚNIEWSKI. We decided to take part in DISInformation Analyser Challenge during NATO Tide Hackathon 2023. We collectively believe that these days we are vulnerable to fake news and other forms of disinformation in both military and civilian affairs. The war in the neighboring country of Ukraine, which has been going on for almost a year, shows how great the role of information is and how much the enemy can gain by manipulating what we know and how the public reacts to specific events. We would like to contribute as much as we can to improve our ability to protect against such threats. The aforementioned goal and the fact that we would like to learn as much as possible during the event are the reasons why we signed up for this year's hackathon.

Here is a diagram depicting our solution: NTC - diagramy (1)

The following datasets, among others, were used during the manufacturing proces: image

source: https://arxiv.org/ftp/arxiv/papers/2112/2112.11185.pdf

Here is a brief diary describing what we did during event:

  • MONDAY (20.02.2023):

    • we introduced our team to the NATO Tide Hackathon community,
    • we downloaded provided datasets and we got acquainted with the problem,
    • we carried out a lot of analysis of the datasets provided,
    • we did a lot of painstaking research on how to deal with the dataset we had recived.

  • TUESDAY (21.02.2023):

    • we continued our research on fake news detection methods, we found that there is a lot of different types of methods to perform this task: image source: https://arxiv.org/ftp/arxiv/papers/2112/2112.11185.pdf
    • we decided to try to implement a way of verifying information on the basis of the knowledge base created from ground truth Guardians and New York Times articles,
    • we decided to create a solution consisting of frontend and backend (API that will provide results from different approaches to the problem on the reliability of the information submitted for checking),
    • we started collecting available datasets from the Internet and teaching models on them.

  • WEDNESDAY (22.02.2023):

    • we continued work on our solutions,
    • we saw that extraction of knowledge from provided articles is a tought task,

  • THURSDAY (23.02.2023):

    • we presented our solutions and discused them with the challenge judges,
    • we have bundled all the solutions developed so far into one system,

  • FRIDAY (24.02.2023): TBA