Today’s social media landscape has seen an increasing number of content producers. Social media platforms are competing engagement time away from news outlets as consumers access more information and entertainment online. News is a business that relies on engagement to make a profit. To stay in business, news outlets are working hard to come up with different strategies.
One strategy has been that of publishing factual yet sensational headlines and phrases. These headlines and phrases are often designed to capture and arouse the reader, not to accurately reflect the news in the article. News Shield redacts sensational headlines and sentences from your browser, presenting the reader with only those sentences that are relevant to the story.
News Shield reads an article on a white-listed page and separates the article into individual sentences.
News Shield classifies each sentence as “News” or “Not News.”
News Shield replaces the sentences classified as “Not News” with redacted block text.
Natural Language Processing sanitizes each sentence by stripping invalid Unicode symbols from the author’s text. After the symbols are stripped, the Linguistic Tagger is used to split articles by sentence terminators and return a trimmed sentence that can be used as input to the machine learning classifier.
Machine Learning uses the output from the Natural Language Processing (NLP) string as input into the Text Classifier. The Text Classifier identifies sentences as either News or Not News. The model is built using transfer learning with a dynamically embedded feature extractor. Sentences that are tagged as Not News are replaced with redacted blocks to obscure sensational content from the reader.
Identifies sensational writing in news publications and redacts the sentences in-line for each article
Use Apple’s native Safari Application Extension API to securely read and replace text
Executes natural language processing and machine learning algorithms in real-time
No data is ever sent to any publication or third-party to track what is being redacted
Heuristics for the classifier are driven by community contributions to our open source listLearn More