There are multiple challenges in delivering the best monitoring results from TV, radio and other sources, which eMedia Monitor meets with a variety of advanced technologies. The procedure starts with recording and processing data from all required channels. There is much more to the task of processing audio and video than one would first assume: Audio needs to be segmented and converted into text via speech-to-text technology, while from a video additional and often crucial information is extracted via OCR and logo detection. After collecting all this information, the semantics and sentiments are analyzed.
When relevant content is found, the user is alerted. Then he also needs access to the text, source, and various other information that was extracted during the process, which our company provides in the form of an accessible web platform and tools like the auto-playing waterfall.
eMedia Monitor is a technology company which continually sets new benchmarks in natural language processing. Our technological advantage allows us to introduce technology-based services ahead of our competitors and open new markets. E.g. we pioneered real-time analysis of broadcast content in 2007 — a full two years ahead of anybody else in the content-management industry.
The natural language is easy to learn and understand for humans, but very hard to master for computers. In the past years, the performance of our system in understanding spoken language has improved dramatically, including speech recognition, speech understanding, speaker recognition and language recognition.Read More
Content can be extracted from both textual and visual sources. As video information contains many levels (colour, shape, actors and text e. g. subtitles), video processing is inherently multidisciplinary. It requires high competence in signal processing, acoustics, phonetics, linguistics and computer science.Read More
eMedia Monitor provides real-time media monitoring and instant alerts around the clock. The decisive advantage of real-time data, especially in critical situations, influences marketing, sales and therefore business success.Read More
A continuous stream of data becomes better manageable when it is split up into segments. On these we use a semantic-analysis tool to find boundaries between topics. Emotions play a key role in the industry, and organizations can benefit from sentiment analysis to determine emotional value.Read More
Deep learning refers to a method of machine learning that uses artificial neural networks with numerous intermediate layers between the input layer and the output layer, thereby forming an extensive internal structure.
The training of such is relatively simple in theory, but incredibly resource- and time-intensive in practice. How does it functon?
For developing a neural network to use for speech or image recognition or similar use-cases, it needs to be trained with a set of already evaluated training data. For example, for training it to recognize a certain object in an image, it needs to be shown a multitude of images labeled as containing or as not containing the targeted object, so that the network learns to differentiate between them and recognize the correct ones.
The difficulties lie in providing a good and sizeable training set. The system needs a certain amount of images to be able to grasp the similarities, but they should not be too similar to each other either (a famous urban legend about tanks comes to mind, where the system learned to distinguish between times of day instead of detecting tanks). Yet, with sufficient training data, the process needs time – even with a cluster of servers and powerful GPUs it takes weeks to train just for a single training shot.
So, deep learning is a challenging task.
For proper performance, a certain data infrastructure is required, which eMedia Monitor has at its disposal by permanently recording the highest number of radio and TV channels in the industry. Furthermore, expertise in statistical analysis and data science is needed. To provide this and a fundamental understanding of those knowledge-spaces, in fact, the majority of eMM’s highly-trained workforce are university graduates.
Find more information about the latest technology advances in: