This projects implements a Python Wrapper for the temporal tagger HeidelTime. To achieve that, it makes use of the HeidelTime-standalone application. We only tested this code for Debian-based systems and cannot guarantee functionality for other systems.
For convenience, we provide you with an installation script for HeidelTime standalone that does everything (including changing the appropriate config files) for you. Navigate to the top-level folder of python_heideltime and type the following into your console:
chmod +x install_heideltime_standalone.sh
./install_heideltime_standalone.sh
Once the script has successfully installed the dependencies, you can install the package with
python3 -m pip install .
In general, to use Python_HeidelTime, it is necessary to install HeidelTime-standalone first. If you choose not to go with the provided script, you may download the current standalone version from Heideltime's releases page. At the point of the development of this project, version 2.2.1 was the most current one. Support for newer or older versions is not guaranteed. A detailed description of the installation process can be found in the HeidelTime Standalone Manual.
If your operating system is Mac OSX, you need to download an appropriate treetagger for your OS. Replace line 15 in install_heideltime_standalone.sh with
wget --no-verbose https://www.cis.uni-muenchen.de/~schmid/tools/TreeTagger/data/tree-tagger-MacOSX-3.2.3.tar.gzYou then need to manually set the path to the treetagger in config.props for the HeidelTime-standalone installation.
Do not move the files after the installation; HeidelTime requires absolute paths in its config files.
Then you may similarly install the package with
python3 -m pip install .
In the config.props file of HeidelTime, you can also specify if you want to limit HeidelTime to detect only Timex3 expressions of certain types (data, duration, set, and time).
When using Python_HeidelTime, please also cite HeidelTime itself as stated on their project page.
The main class of Python_HeidelTime is simply called Heideltime and is initialized without any parameters. All paramaters x can be specified by the set_x methods. Typically, they are named as described in the HeidelTime Standalone Manual. The correct use of all parameters is also given in the manual. When certain parameters are renamed, it is stated so in the source code.
After initialization, a text can be parsed by the parse function by passing the argument as a string object.
The following code snippet works a simple example.
from python_heideltime import Heideltime
heideltime_parser = Heideltime()
heideltime_parser.set_document_type('NEWS')
print(heideltime_parser.parse('Yesterday, I bought a cat! It was born earlier this year.'))Which should result in the following output (with regard to your current date).
<?xml version="1.0"?>
<!DOCTYPE TimeML SYSTEM "TimeML.dtd">
<TimeML>
<TIMEX3 tid="t1" type="DATE" value="2019-06-13">Yesterday</TIMEX3>, I bought a cat! It was born <TIMEX3 tid="t3" type="DATE" value="2019" mod="START">earlier this year</TIMEX3>.
</TimeML>