Missing Role Labels When Using Open Information Extraction

I wanted to use the Open Information Extraction tool (https://demo.allennlp.org/open-information-extraction) as it gives me spatial labels when I use it in the demo mode, e.g. “He walks along the river” labels “along the river” as LOCATION.
However, when I use the tool in python these precise labels are missing and everything is just labeled as ARG0, ARG1, etc.

Here the code I use in Google Colab:

!pip install allennlp==0.9
!pip install spacy==2.2.0

from allennlp.predictors.predictor import Predictor

predictor = Predictor.from_path(“https://s3-us-west-2.amazonaws.com/allennlp/models/openie-model.2018-08-20.tar.gz”)

predictor.predict(sentence=“He was walking along the river”)

Also note that I have to use version 0.9 and 2.2.0 of allennlp and spacy as using the versions given in the Usage tab of the demo lead to error messages.

Hi duplo,
I have also noticed the pip install for reading comprehension is incorrect on the usage tab, it only works with 0.9 and i had to guess which model to use with 0.9. Like you, the model was not stored in google it was stored in amazon. How did you know which model to use and where it was stored?

Hi roist,

it was pretty much trial and error until something I found on stackoverflow did the trick. It was not on the list from their official github which includes the pretrained models.