Reading, Analysing and Activating Complex Text
The Orion Semantic AI system has a vocabulary of about 45,000 words, and about 10,500 phrases. The first step in analysing a new piece of text is for the machine to read the text, looking for words or phrases it doesn’t know, and for any words it doesn’t know in the dictionary definitions of the unknown words or phrases. The resulting word files are checked by hand if time allows (a dictionary definition can be very narrow, or missing an obvious definition, including literal use), and then the word and its meaning structure is created in the network. A word can have several parts of speech (up to 5), if a verb then several types of verb, then possibly many senses. When reading the text, direct connections are immediately made so that the word is identified within the grammatical structure . Often, that is not possible – the POS or senses are reduced, but not reduced to one – the word is left to be revisited when more information is available – sometimes the necessary context is a hundr...
