Category Classifier

This classifier is a specialized multiclass text classifier that categorizes media posts based on the IPTC 17 Media Topic NewsCodes for the Media extracted from the given text. “other” label will be displayed when it has multiple topics or not enough information. The output would be following 12 labels which includes ITPC combined NewsCodes topics :

• entertainment -> arts, culture and entertainment
• conflicts -> conflicts, war and peace
• crime -> crime, law and justice
• disaster -> disaster, accident and emergency incident
• business -> economy, business and finance
• health -> health
• lifestyle -> lifestyle and leisure
• politics -> politics, religion and labor
• technology -> science and technology, environment
• society -> society and education
• Sport -> sport
• other -> multiple topics, short text or not enough information

The category classifier is designed to filter blogs and news articles based on 12 different topics. For example, marketing and security companies can use a category classifier in combination with hard news classifiers to identify and track recent news about organizations or people. Media companies can aggregate technology news based on categories and published dates.

The Category classifier can add one metadata section as an enrichment with the following labels.

The result is one of the labels along with the predicted confidence value:

“enrichment”:
{
“category”: {
“label”: “health”,
“confidence”: 0.95
},
}