What tools are required in producing the output from the input

[UP2040499]

[UP2040499]

• At present, it appears that most methods of validation will have to be implemented manually, as during initial research, tools for some of these methods either do not exist or exist only in closed sources (such as the MoD).
• Some methods are presented in dissertation section 2.5.3, however these only detail the method not any tools or implementations.
• For evaluating credibility, both the US army and UK police have scales for determining provenance, credibility, reliability & veracity.
  • UK police use a 5x5x5 matrix to evaluate provenance
  • US army uses a six-point reliability scale and an eight-point credibility scale individually to evaluate a source

[UP2040499]

So nearly all methods will be implemented by me unless I find any tools or libraries that can do these methods.

[UP2040499]

Tools:

• Git/Github for source control and management
• Python 3.8 (same version as google colab, which is where the BERT model is hosted/demoed)
• Google colab TPUs using google cloud
  • Will create as much code locally as possible
  • Avoids costs and complications with google cloud
  • Need to talk to IT help regarding organisational access to google cloud TPUs.
• Internet crawlers, for as many sites as possible (conventional news, social media)
• BERT QA model to help with crawling through websites (in document form) and finding key information
  • See ​​(Upadhayay et al., 2020)​ for their use of the model
• Reply with other libraries & modules that are used (collect references in this discussion)