Simple knowledge based question answering system for factoid questions with the help of wikipedia.
First clone our repository:
git clone --depth 1 https://github.com/QABot-Team/twitter-bot.git --branch master --single-branch
Switch to the main project directory and install all dependencies:
cd twitter-bot/qa
./install.sh
Then you can run our application with the following command:
python src/main.py [--help]
The application requires a running elastic search instance with a indexed wikipedia dump. Host, port and index name of
the elastic instance can be configured in our config.py.
To write test cases we use the python unittest framework. To run our tests simply execute the following command:
python setup.py test
This uses pytest as test runner. If pytest is not installed
it will be downloaded automatically.
For logging we use the Logger facade class which internal uses the python logger module.
Our facade automatically adds the module name from which the log message was created.
Example:
from utils.logger import Logger
Logger.info('Start answer processing pipeline')
We use different log level for different purposes.
| Level | When it’s used | Method call |
|---|---|---|
| DEBUG | Detailed information, typically of interest only when diagnosing problems. | Logger.debug(message) |
| INFO | Confirmation that things are working as expected. | Logger.info(message) |
| WARNING | An indication that something unexpected happened, or indicative of some problem in the near future (e.g. ‘disk space low’). The software is still working as expected. | Logger.warning(message) |
| ERROR | Due to a more serious problem, the software has not been able to perform some function. | Logger.error(message) |
| CRITICAL | A serious error, indicating that the program itself may be unable to continue running. | Logger.critical(message) |
- To learn more about question answering have a look at 21 - 1 - What is Question Answering-NLP-Dan Jurafsky & Chris Manning or the whole draft about Speech and Language Processing.
- START is another question answering system developed by the MIT InfoLab. How the system works is superficially explained here.
- yodaqa is a further question answering system developed in python which is unfortunately currently not working since the necessary services provided by the auther are not reachable at the moment (see issue #73).