The media industry lends itself to a wide variety of applications of Artificial Intelligence (AI). In this post we will help you understand the breadth of possible applications of AI in media. AI can offer unexpected solutions to some of your current challenges and can help you identify new interesting business cases.
We will briefly describe several representative use cases dealing with both text or multimedia content, explaining for each what is their business value and how the solution works.
Automated fact- and fairness-checking
Improve the quality and reliability of news articles by performing automated fact-checking that allows the writer to identify inaccurate or unreliable statements in a text. In addition, it’s possible to flag inadvertently discriminative or unfair statements.
Natural language processing (NLP) is an AI technique that allows a machine to read and understand human language. In this example, an AI system can intelligently match the working document with other reference documents and sources discussing similar topics, and comparing the statements in each document. To build such a solution, you will need a dataset of news articles or other documents about the topic of interest.
Trend & innovation detection in news
Identify strategic opportunities early on by automatically identifying emerging trends and innovations based on the analysis of news articles.
By analyzing the stream of news articles being published every day, an NLP system can learn to differentiate usual content from brand new topics and trends (e.g. the coronavirus in December 2019). As soon as new topics emerge, they can be flagged and reported to a human user to take relevant actions on (early reporting, strategic business decisions, …).
Increase reader engagement by producing “snackable content”, i.e. personalized summaries of articles that are tailored to specific individuals.
Automated summarisation is a common application of NLP, where artificial intelligence is used to identify and select the core informational content within a text. By taking into account user data, this application can make more personal summaries. Behavior data of readers, such as the topics and articles they are interested in as well as their engagement with specific articles, allow to create a reader profile. This reader profile can help focus the summarisation on the most relevant aspects of the text.
Video tagging and search
Improve user experience by allowing users to find specific multimedia content by analyzing and automatically tagging images and video with relevant hashtags, as well as allowing users to search for specific scenes or snippets within video content.
Computer vision techniques such as object detection and instance segmentation can be used to identify the content of a video frame by frame. This information can in turn be converted into hashtags to classify videos. In addition, temporal patterns across frames can be extracted from videos. This can help classify sequences into scenes. Specific scenes can then be searched and browsed by end users.
Improve user experience and engagement by providing shortened, summarised versions of your video content to your end-users.
Once the content of a video has been classified into scenes, the video can be manipulated for other applications, such as shortening and summarization. In the context of videos, specific scenes can be selected based on their content and combined into a shorter video, acting as a video summary of the initial content.
Maintain user engagement by providing relevant recommendations for multimedia content based on the user profile.
Multiple approaches are possible for media recommendation. A classical type of recommender system, called “collaborative filtering”, focuses on finding similar readers and recommending to each other the content they have consumed. Another approach, more centred on the content, learns to match specific content aspects (identified through NLP or computer vision depending on the content format) with specific interests of users (learned from analysing their engagement with various content).
The possibilities for AI in the Media industry are numerous and varied. Each application will present its own challenges. To maximize the chances of success of an AI project, we shared a list of best practices in our post Five keys to set your AI project up for success.
In the spirit of the agile project management methodology, we strive to set up a basic yet functional proof of concept of your project as soon as possible. This allows us to gather two main types of insights: on the one hand, whether the data is sufficient (quantitatively and qualitatively) to support our client’s project, and on the other hand, whether their project is delivering value and meeting the expectations of its end users.
Based on these key observations, rapid iterative improvements in line with user feedback will guarantee success and impact for your organization.