The amount of news information a person can routinely access these days would have been unimaginable a hundred years ago. But we still have just 24 hours in a day, and only a single pair of eyes to read, and so the question arises: how to get as much valuable news as possible in a limited time? Global media organizations are seeking the best ways to share the latest and most interesting news with their users, and AI is coming up with solutions.
A “news push service” is designed to effectively deliver the most relevant and valuable news content to individual readers. In order to realize such a service, the system needs to analyze and judge a user’s preferences and all current news content, which requires processing a huge amount of data. Fortunately, meeting high data processing demands is something natural language understanding technology is good at.
Natural language understanding technology can assist news media platforms in building high-quality and accurate news information to enhance their content value. The technology can also analyze a particular news item for prejudice and ambiguity to help readers separate fake news bits from the sea of information.
For News Providers
Two different type of data collection are required for news media companies to create an efficient news feed service. First, it’s necessary to get semantic information about text, audio and video content. This includes topic classification, relevant keywords, sentiments, entities, etc. All this data will be used to build a knowledge graph for easy content search, content syndication and recommendations. Secondly, the system needs to collect time series data content consumed by users. This data will be used by the algorithm to recognize user interest along with content metadata and relevant content data.
Case study：Content understanding and classification
Robuzz is a New York-based startup that specializes in smart news feeds. Robuzz allows users to push real-time news alerts about tasks, businesses, and any related content topics in a customized format. Robuzz analyzes massive news data through deep learning and natural language understanding techniques to quickly find news content of interest to users. Robuzz also provides network monitoring and extraction services, collating information from different websites and other unstructured sources, and automatically classifying news based on different themes, keywords, people and businesses. Robuzz’s news pushes can directly interface with third-party applications.
Case study：Preference analysis
Recent Media is a smart news app powered by artificial intelligence. The program analyzes the content of news texts through deep learning and natural language understanding techniques, understands the meaning of news, and pushes news for users. Recent Media also learns user interests through big data analytics technology, enabling it to suggest relevant articles and topics that users may like. Based in San Francisco, Recent Media released iOS and Android versions in 2015 and has developed a beta version for testers.
For News Readers
When human beings read a sentence or a paragraph, we interpret the words within a whole document context for better understanding. NLP technology can teach machines to read using context to help them understand the differences between real news and fake news. One of the most popular methodologies is a TF-IDF (term frequency–inverse document frequency) vectorizer, which is used to determine an article’s word importance from the entire corpus.
Case study：Fake news identification
NewsAnglr is an application providing sports news recommendation services. The Belgian company leverages deep learning algorithms, text analysis, and natural speech analysis techniques to provide users with real and valuable news content. NewsAnglr quickly compares thousands of news sources in different categories and then automatically groups articles to provide different angles of interpretation and analysis on the same news story. NewsAnglr believes artificial intelligence technology gives it an edge over traditional news recommendation services in helping users separate opinions from facts.
Case study：Analysis of news content bias
Knowherenews is a San Francisco-based startup that uses AI to produce a website with “the world’s most unbiased news.” The company’s AI tracks stories on politics, business, technology, etc. on dozens of media sources across a left-to-right political spectrum, weighing their credibility and so on to output “impartial” coverage. The final stories are however human edited “for grammar, style, and bias.” The site also produces fascinating examples of how the language used in coverage of the same story can be tweaked to reflect very different perspectives.
Natural language understanding technology continues to optimize news push services. For news media platforms, AI can improve news information content and accuracy classification structures. NLP technology can also help a platform accurately predict user preferences to effectively improve user satisfaction. NLP technology can also help readers detect and purge subjective or fake news.
Continuing and significant improvements in natural language understanding technology can be expected to further accelerate its deployment across a variety of news services and platforms in the short-term future.
Analyst: Ying Shan| Editor: Michael Sarazen