How does Language Technology (LT) as a sector stack up against current IT predictions for the next decade or so? The latest, and possibly most famous crystal ball gazing based on existing data is Mary Meeker’s (massive) “State of the Internet” slideshow available here .

To get a handle on some of the key trends identified for the decade ahead, Tom Kravitz at Gigaom neatly summarized  them as “the next ten years will be about wearable computing and personal data”. If so, they’ll involve interesting challenges for language & speech technology. Here’s a brief take on some of these.

Big photo as stories. Meeker points out that people are sharing around 500 million photos a day, doubling the size of the global digital album each year. But can the inherent reference and knowledge in this colossal image repository be harnessed for use? Instead of depending on either the up-loader or the crowd to add encoded context to these visual objects, the semantic enrichment of photos (remember that one picture is worth a thousand words) looks like a suitable case for LT automation as Google  is showing. Building up the knowledge that enables a photo to be fully tagged and hence widely searchable and usable as a source of as yet unimagined data applications will be a long process. But if it proves possible to semanticize any picture using recognition tech and automatic metadata generation, then we can link photos together and mix, match and compare them to tell better stories about the world and also about the photos themselves (one more big linked data source).

Video is data. Meeker claims the web is increasingly awash with video content, and video sharing is booming, especially short-form videos that send personal messages (video chat). Video too becomes data once it’s published and can be enriched and linked in the same way as photos. But with the added value of (often) having a spoken sound track that contains opinion and sentiments. So it will be vital to automatically tag videos to boost their searchability and linkability to other semantic eco-systems. This company offers an example of how you can tag with minimal geo-semantics today. There could also be a market for transcribing and leveraging the linguistic content of some of those video sound tracks for legal or other reasons that could benefit from a secure automated or possibly a crowdsourced solution. And in the not so distant future, the intrinsic “meaning/content” of the images themselves will be available for semantic inference in ways we can only dream about.

Wearable computing: the dress rehearsal: Kravitz suggests that the “next ten years will be about wearable computers with sensors”, following the PC decade and more recently the mobile decade. This means that personal sound files and other data generated by wearable computers will enrich the bottomless vault of potentially trackable, comparable and searchable information.

Meeker reckons fitness and health data will be a huge percentage of that personal data as more and more people “realize how behavioural shifts can improve their health and how new devices can help people track those behaviours.” One plausible accessory to make sense of the hosepipe of figures for citizens without medical training will be better data visualization.

One reasonable approach would not be simply to showcase your personal figures as a fancy diagram, but to embed the data into a story that speaks uniquely to each individual. Surely automated narratives – probably embedded in videos with a speech track - could be a powerful way to get individuals to change their habits where necessary. Whatever happens, we think that the power to transform complex data points into convincing storylines could be a winning fashion statement for wearable computer users, as in other personal data futures.