Grapeshot Inside IBM

In 2005 Grapeshot started working with IBM, integrating Grapeshot's automatic content extraction web service with IBM's WebSphere Portal product.

First showcased in 2006 at NEXPO in Chicago USA, the Grapeshot/IBM "Audience Portal" demonstrated a portal interface where each user's content experience was personalised. The Grapeshot webservice extracts the essence of the page and creates a personal user profile of important keywords to then fetch personal headlines and advertisements for each user.

IBM has already spent vast resources on building tools for the "Semantic Web" - a grand project to categorise the growing internet into a universal set of categories, with an intelligent semantic layer that can link "penguins" to the "Antarctica", or "penguins" to "chocolate biscuits", based on the context of words.

IBM's Head of Global Publishing back in 2005 found the Grapeshot web service was more immediate and useful than trying to develop a rule based set of taxonomies and semantic techniques to improve users' navigation of media content. Whilst IBM's Media & Entertainment customers may have already invested millions of dollars in content creation - there had yet to be better ways to improve user navigation, and revenue yield of portal websites. It seemed that portals were still serving the same content to every visitor with no regard for their personal interests or skew of enquiry. What IBM was looking for was a simple, non-rules-based method for linking content, without the cost of implementing a detailed ontological semantic layer.

Fast track to the International Broadcasting Convention in Amsterdam 2009 where IBM showcased the News Discovery Portal - the latest integration of Grapeshot with IBM WebSphere. As journalists write new scripts and stories, the Grapeshot service discovers archive and related content (scripts, audio, video) that the journalist can combine into his or her fresh script.

IBM demonstrates a 25% cost saving through improved productivity as broadcasters leverage their existing assets intelligently into the editorial workflow.

The journalists concentrate on writing their scripts without having to manually tag or connect content elements together, as IBM's LanguageWare extracts the entities, and Grapeshot discovers the relatedness of different assets, automatically.

Grapeshot and IBM have signed several teaming agreements since 2005, to develop and distribute Grapeshot technology within IBM applications. IBM and Grapeshot have led a joint Research & Development project together since 2007.

Partly funded by the UK Technology Strategy Board, the project models the "velocity and trajectory of a news story" where the context of each news story is measured against "the grapeshot" of other relates stories. This work discovers how new stories break on the internet and explores which stories have more consequences than others.