Example-Based Sentence Reduction Using Hidden Markov Model

Sentence reduction is the problem of removing redundant words or phrases from an input sentence by creating a new sentence, in which the gist of the meaning of the original sentence is unchanged. All most previous methods required a syntax parser before reducing sentence. However, these methods were difficult to apply to a language in which there was not a reliable parser. In this paper, we propose two new sentence reduction algorithms without using syntactic parsing for the input sentence. In the first algorithm, we present an novel application of using one of ExampleBased Machine Translation method, the template translation learning algorithm. This algorithm works well in reduction, but the problem of using it is the computational calculation problem. To solve this problem, we extend the template translation algorithm by making an innovative use of Hidden Markov Model based on a set of template rules that obtained by learning from the examples. Experiments on applying the proposed algorithms shows a promising result without complex processing.