Deep QA - A Deep Reasoning Question Answering System

This is the homepage of Deep QA, which is a question answering system designed for answering simulation based reasoning questions. We used answering questions about biological pathways as our inspiration, but it can be applied to other domains as well.

Features

The system provides a high-level language for pathway and query specification, which is inspired by high level action and query languages. Features of Pathway Specification Language include:

  • The pathway specification uses Petri Net execution semantics.
  • Natural actions, that execute automatically when their preconditions are met.
  • Actions can have different effect based on conditions.
  • The level of parallelism of actions can be varied in the pathway specification, between serial, arbitrary amount, or maximum parallelism.

Features of the Query Specification Language include:

  • Ability to specify interventions to change pathways, and the ability to filter the evolution trajectories based on observations.
  • Point and interval queries that can be quantitative or qualitative in nature.
  • Aggregate quantitative queries over a set of trajectories.
  • Comparative quantitative queries over two sets of trajectories.

Publications and Further Information

In progress


Examples

We suggest to try some of the examples to get familiar with the system.

Example Pathway Specification:
domain of 'f16bp' is integer, 'dhap' is integer,
    'g3p' is integer, 'bpg13' is integer
't4' may execute causing 'f16bp' change value by -1,
    'dhap' change value by +1, 'g3p' change value by +1
't5a' may execute causing dhap change value by -1,
    'g3p' change value by +1
't5b' may execute causing 'g3p' change value by -1,
    'dhap' change value by +1
't6' may execute causing 'g3p' change value by -1,
    'bpg13' change value by +2
initially 'f16bp' has value 0, 'dhap' has value 0,
    'g3p' has value 0, 'bpg13' has value 0
firing style max


Example Query Specification:
direction of change in average rate of production of 'bpg13' is d
when observed between time step 0 and time step 5;
comparing nominal pathway with modified pathway obtained
due to interventions : remove 'dhap' as soon as produced;
using initial setup : continuously supply 'f16bp' in quantity 1;

Simulation parameters:
Simulation steps = 5
Max tokens = 20

Results:
 The direction "d" will be "<", suggesting that the rate of "bpg13" production will be lowered when a nominal pathway (given by the pathway specification) is modified such that DHAP is removed as soon as it is produced.