Propositional Logic Inference Engine, working with defined format and natural language.

• Aleksandar Manev 100586500
• Cameron Watts 7713266

Group Number: COS30019_A02_T034

• Truth Table:
  • Missing: N/A
  • Bugs: None known.
  • Feature: Accepts more complicated KB. (eg. a&b=>b&c)
  • Feature: You can ask it a negation using "~". (eg. ~a)
• Forward Chaining:
  • Missing: N/A
  • Bugs: None known.
  • Feature: No extra features.
• Backwards Chaining:
  • Missing: N/A
  • Bugs: None known.
  • Feature: No extra features.
• Natural Language:
  • Missing: N/A
  • Bugs: None known.
  • Feature: No extra features.

TELL
a=>b; b=>c; b&c=>a;
ASK
a
TELL
p2=> p3; p3 => p1; c => e; b&e => f; f&g => h; p1=>d; p1&p3 => c; a; b; p2;
ASK
d
TELL
if I'm happy then I win. If I win then You Loose. if you play then you have fun. If I have fun and you have fun then It's a good day. If It's a good day and I dance then you dance. If you loose, you are broke. If you loose and I win then you play. You have fun. I'm millionaire. I have fun. I'm happy.
ASK
You Loose.
TELL
p=>q; l&m=>p;b&l=>m;a&p=>l;a&b=>l;a;b;
ASK
l

For our research, we did Natural Language Processing. Our strategy was to detect patterns in a sentence and try to convert it into variables and symbols. Using Regular expression we were able to manipulate with the sentence string and identify the key values. We compare them and construct Horn Clause rules of propositional logic.

We firstly defined all the rules that one sentence must or could have.

Using Regular expressing,

• then([\s\S]*) - select everything after "then"
• ,([\s\S]*) - select everything after ","
• ^(.*?)then - select everything before "then"
• ^(.*?), - select everything before ","

We were able to split the rule sentence into two sides, left and right side of the implementation. Form there we split each side on "and" and each one of the values seperated by "and" vas representing a variable.

If I am happy, I am winning. If I am winning and I am dancing then you are winning. I am happy.

Translation: a=>b;b&c=>d;a;

Having this translated KB I can use it my TT, BC, FC algorithms.

The output will be printed in Natural Language.

We used the content from the Chapter 7 of the book "Artificial Intelligence, A modern Approach III Edition" by Stuart J Russell and Peter Norvig to understand better how the the Forward Chaining and Backward Chaining algorithms work.

• http://stackoverflow.com/questions/18413204/regular-expression-to-select-everything-before-and-up-to-a-particular-text
  • Used Regular expression for the natural language processing.
• https://people.cs.pitt.edu/~milos/courses/cs2740/Lectures/class6.pdf
  • This slides helped us understand backwards and forward chaining.
• http://stackoverflow.com/questions/10723168/generating-truth-tables-in-java
  • Generating truth table using binary representation of the row number.

The UI works same as its specified on the requirements.

• java -jar engine.jar TT input.txt
• java -jar engine.jar FC input.txt
• java -jar engine.jar BC input.txt

Natural Language works on Horn Clause type of kb

If [variable] then [variable].
If [variable] and [variable] then [variable].
If [variable],[variable].
[variable].
every rule must end with .

You can see examples up in the test cases.

We had 5 tasks to complete in total:

1. Truth Tabe - Aleksandar Manev
2. Forward Chaining - Cameron Watts
3. Backwards Chaining - Aleksandar Manev
4. Research - Aleksandar Manev
5. Documentation - Cameron Watts

• Aleksandar developed 60% of the assigment. Due to he's more free scadule he was confident to do the Trurh table, Backwards chaining and the Research part of the assigment. Aleks's development went smoothly being able to finish both of the algorithms for TT and BC in the first two days of the assigment period. This gave him time to work on the Reasearch and finish it by the end of the week. Aleks's found some bugs in the BC implementation and he re-build the algorythm to fix this issue. It took him 1 day to finish finalize the BC algorithm.
• Cameron developed 40% of the assigment. He worked on the forward chaining algorithm and documentation of the code. His development went smooth with being able to finish his algorythm in the first week of the assigment period. There was only one bug that we fixed it later on.