Over the past few decades, images and videos have become an important type of media and are used in various sectors as an effective means to increase sales. Based on ReportLinker, the AI market is expected to grow to US$ 312.4bn by 2027.
Apart from the production, consumption, and exploitation of digital image and video material by TV stations (commercial, public, and pay TV), film production companies and streaming providers (video on demand), as well as by advertising agencies and publishers, images and videos also play an essential role in quality control, digital asset management systems, autonomous driving, or in municipal archives.
However, the current market conditions do not reflect one aspect: Due to the rapidly increasing amount of such assets across various industries, most users of visual data lack adequate and innovative tools for automated analysis, tagging, and management of their media content in order to make informed decisions and optimize processes. The steadily growing number of photos and recordings is becoming an escalating problem for companies, which to date can only be solved by investing a lot of time and money. A new business need emerged that must be met by innovative companies using AI and machine learning tools to deeply understand multimedia content.
One huge challenge for content producers is sourcing archival material for high-quality, user-focused content and doing so for a breadth of consumers. The challenge is to create better content in less time and resources. For many companies, this is a potentially impossible task.
How has metadata extraction worked so far?
The main reason is how editors, journalists, and broadcasters manage their media data repository, based on which new content is created. Photos and video clips are typically stored in different file systems or internal folders, either only partially tagged or without any supplementary information.
Presently, this target group needs to tag their visual media manually to gain insights, which is time-consuming, expensive, and error-prone. Analyzing visual data takes an operator about 24 times longer than using an AI-based recognition service as offered by DeepVA. The time saved is precious and can be spent on other tasks and challenges.
If a content producer comes up with an idea for a new piece of content, there is a lot of work to be done before it can be realized. With a bit of luck, some suitable material may be found in the distributed file system based on one or two keywords, provided that the file has been stored correctly labeled. More realistically, however, the content producer must sift through numerous images before finding worthwhile material, and even then, there remains doubt whether a better photo might exist somewhere else. Yet, due to time constraints, further research is omitted, and the resulting quality, and thus user engagement rate, will suffer as a result.
Broadcasters find themselves in an even more complex situation. Audiences are much bigger, and therefore the preferences in terms of media content and end devices differ considerably. Consequently, the diversity and sheer volume of material make it even more challenging to retrieve the required content quickly.
Technology evolution has made live and on-demand coverage more accessible. People no longer consume media and news solely on TV and radio, [TV1] but across all devices and platforms—backed by seamless on-demand access. Plus, they consume mass media in a variety of formats.
Broadcasters are forced to produce better and more creative content every day without wasting additional resources. This causes the data pool to grow continuously to the point where employees can no longer keep up with manually tagging and managing their datasets. Hence, the productivity of the various workflows diminishes rapidly, along with the quality of the published content.
Leveraging digital tools can boost efficiency and quality when dealing with metadata and alleviate workloads. Today, numerous instruments exist, such as digital/media asset management systems that, e.g., facilitate the work of online editorial teams, unburden journalists, or simplify publishing processes. One of the technological developments that is seen as adding tremendous value across industries is artificial intelligence (AI), which substantially enhances the power of such digital tools.
AI to increase efficiency in the media industry
The main advantage of AI, though, is its ability to analyze vast amounts of data within a short period of time and to draw actionable conclusions from it. It acts like a human being, except that its decisions are strictly fact-based, using a wealth of information. In doing so, it can access both external (social media, news aggregators, search engines, etc.) and internal data sources (e.g., archive or user behavior on a website).
Utilizing computer vision is more than promising for all kinds of companies. To pinpoint needed content with suggestions for related material and navigate through a massive pool of data to create fresh content or reuse old material is expected to make many professionals’ jobs easier and more rewarding. Creators can devote more time to the creative part of their work, editors become more productive, and consumers will find interesting content more easily.
Thus, AI is an omnipresent topic that will play a vital role in the future.
Despite the great potential, AI and computer vision are used by only a few media firms in production so far. In addition, many organizations are still unfamiliar with this technology but need to catch up quickly to remain competitive.
The fundamental problem is that many organizations have such a large amount of heterogeneous data that traditional recognition services and dataset management tools can only recognize and properly manage a fraction of it. As a result, much content remains hidden in the database and fails to offer any added value. Solving this issue involves manual keywording of data by human workers—a process that is too costly and time-consuming for businesses due to the sheer amount of data at hand; accordingly, the use of AI is further postponed.
What specifically is AI used for in media companies?
As with many other industries, the power of AI and machine learning lies in the very data they generate and process for media companies. Possible use cases include a more searchable repository and business process optimization for a personalized news stream for readers and users, market research, and user analytics alike.
But using AI in media operations also enables a whole new set of services, such as automatic tagging or inferring plots in videos, as made possible by DeepVA’s software. Moreover, thanks to various AI tools, such as dataset management, automated training data creation—and with it the possibility to build their own AI models—, as well as indexing of unknown content, companies can eliminate manual work and leverage the full potential of their media repositories and content. Moreover, these tools can be applied quite intuitively on DeepVA’s AI platform, enabling a quick and easy introduction to AI without any expert knowledge.
Why should media companies apply AI?
Tools with integrated artificial intelligence have great potential to change how media companies work. AI tools can take over simple editorial tasks. A prominent example is the preparation of standardized reports of matches in amateur soccer.
But more development work is needed until AI can deliver high-quality reports and stories. More mature, however, has become the use of AI to support editorial work—especially in the online environment. Here, artificial intelligence can relieve newsrooms of repetitive tasks and thus create breathing space for original journalistic work. Such tools can help editors cope more easily with the new demands of the digital world. It makes editorial workflows more efficient, and output can be automatically enhanced and post-processed.
Publishers, media outlets, and especially online media editorial offices also hold a key asset for the use of AI: Thanks to their publications, they possess large amounts of (digitized) information that can be used to train AI, i.e., to design their own AI models. The advantage of deploying AI is that separate, subsequent keywording of content is no longer necessary. Instead, the AI analyzes existing texts and assigns them to topics and keywords. This is true for articles as well as for web content.
Learning from experience, the use of AI in the media industry realized significant productivity gains and time savings. JournalismAI Report also confirms this, stating that the primary reason for using artificial intelligence in editorial work is to increase efficiency. That is, media content creation will accelerate. AI can support the entire digital publishing process, from topic identification to distribution and analysis, and takes much of the “technical” article optimization requirements away from editorial teams.
“Artificial intelligence will be responsible for groundbreaking changes in all industries—including the media industry. In the past, companies that have adopted digital transformation have prevailed. That’s why media companies should embrace new technologies today rather than tomorrow, being bold to try things out.”
When it comes to news, viewers want up-to-the-minute coverage and will abandon one media channel in favor of another if reports arrive there more quickly. Concerning entertainment, they want simple-to-use interfaces and ample on-demand content to choose from. Netflix is leading the way because of its convenience combined with its vast pool of unique content, and media providers and news outlets need to move quickly to deliver simple user experiences beating the competition.
AI vs. human
Going forward, the question is: Will AI make the work of journalists, producers, and archivists obsolete? Experts agree by answering no. Algorithms will handle those tasks that machines are able to accomplish faster and better. But algorithms must be developed, controlled, and tested, too. Human-machine collaboration is the next big thing in the media industry.
Experts do not see the danger that artificial intelligence will replace media professionals. Yet, there are already AI systems that compose music, write texts, or paint pictures, but they are rather exceptions. Genuine creativity is a quality that is inherent in us humans. Nevertheless, the machine can beat humans, primarily when data processing and preparation are required. This is where AI development can streamline workflows and create rudimentary content.
Therefore, machine learning and creativity should not be seen as opponents: Ideally, creativity and Data Science work hand in hand right from the start. Successful production of content, articles, and videos is always a combination of creativity and knowledge of the data.
