A Commonsense-Infused Language-Agnostic Learning Framework for Enhancing Prediction of Political Polarity in Multilingual News Headlines
Predicting the political polarity of news headlines is a challenging task as they are inherently short, catchy, appealing, context-deficient, and contain only subtle bias clues. It becomes even more challenging in a multilingual setting involving low-resource languages. Our research hypothesis is that the use of additional knowledge, such as commonsense knowledge can compensate for a lack of adequate context. However, in a multilingual setting, it becomes ineffective as the majority of the underlying knowledge sources are available only in high-resource languages, such as English. To overcome this barrier, we propose to utilise the Inferential Commonsense Knowledge (IC_Knwl) via a Translate-Retrieve-Translate strategy to introduce a learning framework for the prediction of political polarity in multilingual news headlines. To evaluate the effectiveness of our framework, we present a dataset of multilingual news headlines.
For More details refer our paper (Coming Soon!!)
We recommend Conda with Python3. Use requirements.yml to create the necessary environment.
The dataset and its generation scripts are stored in the data folder.
Follow https://github.com/allenai/comet-atomic-2020/ to retrieve the Inferential Commonsense Knowledge (IC_Knwl).
Use https://cloud.google.com/translate for translations.
To replicated the reported dataset run:
python3 main.py eventRegistry_apiKey
To generate custom dataset, pass the deseired values in the commandline arguments. For example, to retrieve events in the 'Business' category in the 'Slovene' language reported by 'Delo' run:
python3 main.py eventRegistry_apiKey --lang slv --category news/Business --source delo.si
For headlines only
python3 Headline.py
For IC_Knwl only
python3 IC_Knwl.py
For headlines with IC_Knwl
python3 Headline+IC_Knwl.py
For headlines with attended IC_Knwl
python3 Headline+Attn(IC_Knwl).py
MIT License