Technology

Common video search engines have one problem: Video and audio files are “invisible” for them, so they are heavily dependent on meta data describing the media content. In contrast to common search engines, spactor.com uses a technology allowing us to “look into” media files and analyze them in-depth.

spactor.com is based on THASEUS, an advanced analyzing system for audio and video content. THASEUS combines elements from artificial intelligence, speech processing, and natural linguistics into one powerful framework. Running on state-of-the art algorithms, it offers an automatic generation of meta data in real time and stunning quality.

Speech recognition engine

The core of THASEUS is our world-class speech-to-text engine: With more than 150.000 trained words, THASEUS works on a massive pool of terms. The system even extracts automatically new words from online sources and learns to recognize them. THASEUS can process speech independent from gender, age and accent. With a multitude of additional features, such as language recognition, speech/non-speech segmentation, or speaker change detection, it ensures that every word of rich-media content is decoded into text in high quality.

Semantic Analysis

In order to set the resulting text into a larger context, THASEUS offers language processing tools based on knowledge ontologies and thesauri. This way, the system can automatically determine the topic of a media clip, or highlight important persons, locations, or brands named within the media content. As a result, the technology provides detailed meta data allowing spactor.com to search on.

Scalable infrastructure

Designed as a modular framework running on a cluster architecture, it allows us to scale our indexing performance highly flexibly. We constantly benchmark and improve THASEUS in order to boost the recognition quality.