326 Personal Content Experience: Managing Digital Life in the Mobile Age translators, but even that does not guarantee interoperability. Deliber- ately faked metadata or forged ontologies will destroy many schemes based on open contributions. Moreover, knowledge acquisition will prove too hard for most real-world problems. If the metadata cannot be captured as a by-product of existing applications, it will remain uncovered. • Fake metadata and privacy are key concerns It is of crucial importance to start attaching metadata to content now. The more we have standardized annotations and associations available, the less likely we will have unorganized personal content fl oating around in the future – that is, content beyond most GEMS actions. The question is: how will this take place? 8.1.1 Metadata Enablers We have stressed the importance of automatically created metadata. Several related existing technologies that are currently under heavy research can be identifi ed in such fi elds as image analysis, language understanding, sensor processing, context awareness, pattern recogni- tion, and data mining in general. • Metadata magic comes from machine cognition, data mining, and context awareness We could consider a brute force approach, where metadata is abstracted from raw data, once again regardless of content type. A typical example might be scanning through all content, at the binary level, and identifying ASCII or Unicode strings that make up human- readable words. These words are potential candidates for describing the fi le in some aspect. However, there will be plenty of noise in the results, so consistency and integrity checking will be needed. Besides, much of the interesting content today comes in non-textual forms. Metadata should be streamlined not only across applications, but also across content types. Metadata-based similarity recognition across different media formats could lead us to automatic cross-media catego- rization. Once metadata is created for a certain content object, other related pieces should benefi t from the same metadata, regardless of the content type. • Metadata will propagate between related content items and applications