Multilingual News Article Framing Classifier
A news article can address a story from a number of different perspectives or "frames". This service is able to identify the main frames used in a particular article; it is a multilingual classifier capable of processing 104 languages, but it will perform the best on the six languages on which it was trained: English, French, German, Italian, Polish and Russian.
It is developed as part of the EDMO Ireland project and is based on our submission to the SemEval 2023 Task 3, where it performed the best on English data.
The classifier recognises 9 principal frames:
- Economy and Resources
- Religious, Ethical and Cultural
- Fairness, Equality and Rights
- Law and Justice System
- Crime and Punishment
- Security, Defense and Well-being
- Health and Safety
- Politics
- International Relations
The maximum input document length is 512 words. This means that any text following the first 512 words will be ignored by the classifier.
In addition to the overall framing classification, the classifier is able to indicate which sentences in the document (within the first 512 words) it considered to be the most important - note this does not identify which sentences are relevant to which frames, necessarily, but it provides a guide as to which parts of the document were most informative to the decision process. To request this information, enable the :Important_Sentence annotation type.
| Default annotations | |
| Each frame annotation has a score that represents the probability assigned to the label by the classifier. | |
| News topic | Definition |
| :Economy_and_Resources | costs, benefits or other financial implications; availability of physical, human or financial resources, and capacity of current systems |
| :Religious_Ethical_and_Cultural | religious or ethical implications; traditions, customs or values in relation to a policy issue |
| :Fairness_Equality_and_Rights | balance or distribution of rights, responsibilities, and resources |
| :Law_and_Justice_System | rights, freedoms, and authority of individuals, corporations, and government |
| :Crime_and_Punishment | effectiveness and implications of laws and their enforcement |
| :Security_Defense_and_Well_being | threats to welfare of the individual, community, or nation; threats and opportunities for the individualâs wealth, happiness, and well-being |
| :Health_and_Safety | health care, sanitation, public safety |
| :Politics | considerations related to politics and politicians, including lobbying, elections, and attempts to sway voters; attitudes and opinions of the general public, including polling and demographics; specific politics aimed at addressing problems |
| :International_Relations | international reputation or foreign policy |
| Additional annotations available if selected | |
| :Important_Sentence | Identifies the sentences in the document that were most influential to the classifier in making its decision. The sentences are not explicitly linked to a particular frame but they give an indication of what led the classifier to make the choice it did. |
Use this pipeline
You can process up to 1,200 documents per day free of charge using the REST API, at an average rate of 2 documents/sec. Higher quotas are available for research users by arrangement, contact us for details.
The API endpoint for this pipeline is:
