Summary
The invention relates to searching through an electronic text document for passages relating to a desired concept by using a conditional random field algorithm (a class of statistical modelling). A set of training texts are deconstructed by a computer processor to extract features including the text of complete sentences, tokens used in those sentences, the sequencing of those sentences, and the layout and typography of the text. The conditional random field algorithm applies one of two labels to each sentence: relevant to the concept being searched for (also known as -State A-), or background information (-State B-). A search algorithm then returns all those sentences which have been labelled with -State A-, i.e. those sentences relevant to the concept being searched for.
Applying the Aerotel approach, the Hearing Officer found the contribution fell within the excluded matter as a program for a computer as such. The application was refused.
Full decisionO/057/21 310Kb