With the Christmas period drawing to a close, it seems a good time to reflect on our progress in December towards the grammar tool.

While a holiday instead of work may have been a good idea, we chose instead to see how much further we could push the technology, keeping the core engine unchanged, but adding new patterns to the existing ones in the form of the Wordnet dictionary.

While the dictionary, from the Princeton source, needed a bit of massaging to fit, it did so nicely adding approximately 150,000 words and phrases, and 300,000 new connections. It also came with a number of ambiguities that had not previously been an issue. Who uses the word "are" to describe an area, I ask you? Fortunately, it is not a big task to add another pattern to the existing mix to remove the impact of the redundancies.

The major benefit in having a world-class dictionary within the Project Turing code comes in the form of its associations. By providing multiple word-senses to the program, it is able to use the word and phrase hierarchies to determine appropriate meanings of the words. In other words, it makes the program an intelligent dictionary by firstly identifying valid word senses within the sentence's grammar. And further, and much better still, where additional patterns are stored, the program shows only the word senses that are valid within the semantic patterns available (the program has no context pattern store at this stage).

The new features will be displayed on the web site as samples in the next week.

We will re-assess the impact to our product schedule by the end of the month. As the benefits in the grammar tool providing an intelligent dictionary appear to be more compelling than just a grammar tool, it is likely that we will rework our product plan to incorporate the additional capabilities.

Watch this space, and yes, Happy New Year to you all!

John Ball