Timeshift

Requirements

This project requires Python 2.2. The module sets.py from Python 2.3 is included.

Usage

1. Create a seperate Python module creating the events and a TimeNet to store the time relations between them:

   example.py:

   #!/usr/bin/env python
   
   from timeshift import *
   import relation
   
   E1 = Event("E1", "Mary went to the airport")
   E2 = Event("E2", "Peter bought a car")
   E3 = Event("E3", "Peter drove Mary to the airport")
   E4 = Event("E4", "Mary flew from Trondheim to Oslo")
   
   net = TimeNet()
   net.addrelation(relation.before, E1, E4) # went before flew
   net.addrelation(relation.before, E2, E3) # bought before drove
   net.addrelation(relation.equals, E1, E3) # drove and went at same time
   
   # Check one particular relation, what about her flight
   # and Peters car purchase?
   print E4, net.net[E4,E1], E1
   
   # Print all 'known' knowledge
   net.print_knowledge()
   

2. Run the module and examine the results:

   : stain@ozelot ~/timeshift;python example.py
   E4 Set([>]) E1
   Mary flew from Trondheim to Oslo is after Mary went to the airport
   Mary flew from Trondheim to Oslo is after Peter bought a car
   Mary flew from Trondheim to Oslo is after Peter drove Mary to the airport
   Mary went to the airport is before Mary flew from Trondheim to Oslo
   Mary went to the airport is after Peter bought a car
   Mary went to the airport equals Peter drove Mary to the airport
   Peter bought a car is before Mary flew from Trondheim to Oslo
   Peter bought a car is before Mary went to the airport
   Peter bought a car is before Peter drove Mary to the airport
   Peter drove Mary to the airport is before Mary flew from Trondheim to Oslo
   Peter drove Mary to the airport equals Mary went to the airport
   Peter drove Mary to the airport is after Peter bought a car
   

   The first outputted line shows that the only possible relation between E4 and E1 is > - meaning that E4 is after E1, Mary's flight was after Peter's car purchase

   The other output is all the knowledge that the system knows 'for sure'. Such knowledge is relations between nodes where there is only one possible relation. Note that this includes the basic facts initially added and their inverse values. (Ie both "E1 before E4" and "E4 after E1").

   Relations between other nodes (events) are not printed, as there is only partly or no knowledge on what relations might exist. In more advanced examples, this could be relations such that 'E1 is before, finishes, during or meeting E2'.

Included files

These are the files included:

Note

There is a (intended?) bug in the Add-algorithm in [Allen]. Allen's first for-loop finds new transitivity information for N(k,j), but then checks R(k,i) for new information instead of R(k,j). The fix is to replace (k,i) with (k,j) in the if-test:

if R(k,j) ispropersubset N(k,j)
    then add <k,j> to ToDo;

References