kenrick95 / aba-web

Final Year Project: Web-based Argumentation

Web-based Argumentation

Test status:

Extra short summary of this project:

• Assumption-based Argumentation solver
• Plus a web interface
• Input: argumentation rules
• Output: which argument is conflict-free, stable, complete, admissible, ideal, grounded.
• FYP report: link.
• How to setup this project for local development: link

Architecture

• Static page served by Python Flask web-framework
• Input parsed by Parser
• Parser attempts to construct the ABA Framework, including
  • ABA Arguments
  • ABA Dispute Trees
• Outputs the visualization of the arguments and dispute trees, including
  • semantics, elaborated in subsequent section

Parser

a, b, c |- d.

Rule

• LHS contains zero or more space-separated symbols
• RHS conrains exactly one symbol

Meaning

• Symbols in LHS supports the RHS symbol, making a potential argument
• If LHS is empty, then the potential argument is a ground truth
• If the symbols contained in LHS are ground truths or assumptions or both, then the potential argument is an actual argument

Notes

• If the sentence is potential argument, it will not be used in the dispute tree derivation

contrary(a, b).

Rule

• a function containing two parameters, both accepting a symbol

Meaning

• first symbol is a contrary of second symbol
• first symbol is an assumption

assumption(x).

Rule

• a function containing one parameter accepting a symbol

Meaning

• explicitly state the parameter is an assumption

Argument semantics implemented

• Conflict-free
• Admissible
• Complete
• Grounded
• Stable
• Ideal

Testing

Using Python unittest framework, tests mainly sourced from examples presented at Craven, Toni "Argument Graphs and Assumption-Based Argumentation" (2016).

Notes

This work is based mainly on the book chapter authored by Dung, Kowalski, Toni titled "Assumption-based Argumentation" (2009).

requirements.txt contains libraries needed for setting up the project.

The following files are related to deployment in Microsoft Azure, please disregard them when evaluating the project:

• ptvs_virtualenv_proxy.py
• runtime.txt
• web.config
• wheelhouse/ujson-1.35-cp34-none-win32

Performance Testing

• perf_memory_limitter.py is for limitting RAM usage, returning MemoryError if Python overshoot a certain amount of RAM usage
• perf_test.py is the main file to test performance of this project.
• perf_test_runonly.txt is the list of test cases used in perf_test.py
• pert_test_proxdd.py is a wrapper to test the performance of proxdd, a Prolog program by Toni (2012).
• perf_test_proxdd_setup.txt is the list of test cases used in pert_test_proxdd.py, essentially the same as perf_test_runonly.txt

aba-web's Issues

Add help page

Having a help page would be useful. The things you put in the help can be part of your report, e.g., definition of the implemented semantics and examples.

AND & OR of ABA_Rule

Fan Xiuyi:

However, I’m still a bit concerned about your program.

It seems you don’t distinguish between:

b |- a
c |- a

and

b, c |- a

As you can see, these two sets of rules mean very different things. First set says, if b OR c holds, then a holds; whereas the second set says if b AND c holds, then a holds.

---

Haven’t read your code, but it more or less means that you need to duplicate your search structure for each of these OR branches. And you need to search through all possible branches and check for loops. (I know, I know, this is very tedious … But there isn’t any other way to do it, as far as I know …)

Resubmitting appends output, not replaces it

Do lazy-rendering of D3 graphs

Current implementation renders all graphs, and then hide it. It might cause performance issue on browser when it gets a lot of symbols.

Still display the argument tree for potential argument; but only show the dispute tree for actual argument

Currently, I do not show the argument tree and dispute tree for potential argument