Concept Searching

Due to the multiple meanings accorded to many words and the complexity of its structure, a concept is generally expressed using short sequences of words. It is these patterns that convey most of the meaning in any language. The key patterns are mostly proper nouns, noun phrases and verb phrases. Any attempt to analyse text based on single words in isolation is likely to introduce significant ambiguity since it is the context within which word patterns appear that identifies the real meaning. For example, the phrase 'foul play' has a very precise meaning to most readers - on their own, however, the words would be unlikely to immediately bring to mind the original concept.

conceptSearching automatically identifies the most significant patterns in any text and uses these compound terms to rank results based on an understanding of meaning - rather than simply based on finding the required words. This is significantly more adaptive and flexible than exact phrase or proximity searching. Also, queries can be expressed in natural language, with no need for complex query syntax associated with traditional Boolean techniques. It is also able to recognise and accept the 'stemmed' version of each word - from the original 'deal', the words 'deals', 'dealing', 'dealings' and so on would be valid.

conceptSearching is the first true concept search engine in that it combines Bayesian Inference with Shannon's Information Theory to correctly weight not only single words but also compound terms (i.e. multi-word phrases). Since most concepts are expressed in short phrases, rather than in single words used in isolation, its unique technology is being used to build ground-breaking applications.

Some examples of the disparity in meaning between a single word and its inclusion as part of a phrase are listed below.