Grapeshot USPs

Algorithms

Grapeshot's algorithms include the latest algorithms developed by Cambridge academics in the BM25 suite of probabilistic information retrieval algorithms (2000-); built upon the founders' rich pedigree of producing advanced search algorithms and business solutions in Muscat since 1984.

Grapeshot's WordRank algorithms apply weights to words that change, based on user behaviour:

• 1. Multiple-word querying: whole documents or excerpts from emails or browser pagers can be queried in milliseconds. WordRank establishes the best words to consider (where word weights can be different between two users)
• 2. Dynamic word suggestion: like a folksonomy or taxonomy, Grapeshot suggests useful related terms with no upfront taxonomy required. WordRank uses probability relationships to find other related words.
• 3. Variable word-weights: the core of WordRank which offers personalisation and auto-categorization opportunities. As the user looks at particular documents, the clickstream can update the user profile, and Grapeshot can constantly monitor and adapt the word weights for each user.

Size & Scale

Grapeshot code is incredibly small - just 300k for a full XML indexing and search system that scales to provide millisecond searches across millions of documents. It means the code is small to insert inside your own applications, is not memory hungry and opens up new opportunities for intelligent mobile devices, where memory and space may be at a premium.

Grapeshot's small memory footprint underpins why the search can be so fast - "market leading" was the verdict of a recent technology benchmark.

Multiple language Capability

Mining unstructured text is important for conceptual advertising or reading the page that a user is looking at. It assists "more like this" searches and links documents by shared concepts. However, rather than develop specialist taxonomies or word synonym lists per language, the Grapeshot technology can be applied to any text in any language. This cuts set-up and taxonomy maintenance costs and provides a fast reliable way to scale into Chinese, Arabic and other global markets. Grapeshot can suggest related terms using just probability math and the interrelationships between words in the existing corpus of document; although existing taxonomies can always be plugged in too if they already exist.

Pedigree

Few search vendors have been doing commercial search systems since 1984 whilst also positioned at the forefront of academic endeavour. Grapeshot's founder Martin Porter is not only well known for his Porter Stemmer used by small and large software companies worldwide (IBM, Microsoft and Google included); but Porter has been internationally acclaimed for his "outstanding contribution to information retrieval".

Grapeshot co-founder John Snyder has worked on commercial intranet and content solutions since 1992, solving business problems for corporations worldwide.

Grapeshot code reflects the fusion of pioneering academic work and the commercial needs of businesses that need to deploy solutions to data-mine and leverage user patterns of content consumption into information intelligence and advertising performance.