Context Based Automated Image Selector for Articles using Text Illustrator

Abstract

Mainly journalists are facing the problem of finding suitable images for their articles. The concept of automated text illustrator has been around for a while. These systems take in an article, identify the keywords of the article and output images from an image database based on the keywords. Such systems may not be suitable as it is difficult to identify keywords in an article to identify images that best suit their texts. It is important to consider the meaning of an article and then use keywords based on those to search for matching images. Most of the available systems have given more emphasis on the image retrieval component and less emphasis on the text processing component. Our attempt is to concentrate on the text processing component, so that it can be used by existing text illustrators. We try to identify the correct meaning of the keywords based on the context of the verbs with which they appear. It also identifies the best possible word that represents the identified meaning.