TOPIC SPECIFIC LANGUAGE MODEL AND TEXT SEGMENT DIVISION AND LABEL APPLICATION USING USER DIALOGUE BY TOPIC SPECIFIC LABELLING STATISTIC

摘要

PROBLEM TO BE SOLVED: To provide a method for executing segment division and labelling of unstructured text in response to judgment by a user.SOLUTION: This invention is pertaining to a system for segment division and a user interface for structuring unstructured text by using a trained statistic model based on training data with notes. The segment division and application to be executed are presented to the user for an overall review and an alternative segment division and label application are presented to the user. The user can either select the alternative segment division and alternative labelling or input a segment division defined by the user or labelling defined by the user and a plurality of different actions are started in reactions to correction by the user. In addition, this method includes learning function, log recording and analysis for a correction a user has introduced in order to suit this method to the user's preference and to further train the statistic model.