Image courtesy: Flickr/by Filter Forge

Image courtesy: Flickr/by Filter Forge

Ikuya Yamada, co-founder and CTO of Studio Ousia, the company behind Linkify – the technology to automatically extract certain keywords and add intelligent hyperlinks to them to accelerate mobile search – recently sat down with The Semantic Web Blog to discuss the company’s work, including its vision of Semantic AR (augmented reality).

The Semantic Web Blog: You spoke at last year’s SEEDS Conference on the subject of linking things and information and the vision of Semantic AR, which includes the idea of delivering additional information to users before they even launch a search for it. Explain your technology’s relation to that vision of finding and delivering the information users need while they are consuming content – even just looking at a word.

Yamada: The main focus of our technology is extracting accurately only a small amount of interesting keywords from text [around people, places, or things]. …We also develop a content matching system that matches those keywords with other content on the web – like a singer [keyword] with a song or a location [keyword] with a map. By combining keyword extraction and the content matching engine, we can augment text using information on the web.

So, basically we augment real-world things with information on the web. We would like to realize this vision, called Semantic AR. Devices like Google Glass [and other wearable technologies] are very important for our company [when it comes to real-world applicability of this vision].”

The Semantic Web Blog: Is this only about semantically enhancing experiences around text?

yamadaYamada: We also want to use OCR technology so we can extract text from images on the web. That way we can augment real-world images with information from the web.

The Semantic Web Blog: What are some of the specific components behind the capabilities you deliver today as part of Linkify that wrap into the bigger Semantic AR vision?

Yamada: Phroni is our backend engine that is basically our [proprietary] machine-learning technology to extract interesting keywords from text automatically and turn them into links.

We use Linked Open Data sources such as DBpedia and Freebase, and also we extract information from a Wikipedia dump, so these three data sources are very important. We obtain a large amount of text from the web using our web crawler that is open source. We conduct some NLP on this [to help] provide contexts around keywords; we use terms that are nearby the keywords to detect some kind of meaning or similarity. We combine all this to give accurate results for links [so that users can gain more information about highlighted keywords]. Converting keywords into links basically improves the user interface on smart phones and tablets.

Linkify uses the engine but it also is available in the SDK for mobile application developers to easily integrate Linkify and Phroni functionality into the app.