Extractive and Abstractive Caption Generation Model for News Images

-This paper provides a model for automatically generating captions for news images, which is used to support development of news media management and many multimedia applications. In the existing method, the captions for the news images are given manually by reading the text content. Thus the caption generation task requires human involvement and hence a time consuming process. The proposed system uses a two-stage framework for automatically generating captions for news images: Content Selection and Surface realization. Content Selection identifies what the image and accompanying article are about, whereas surface realization determines how to verbalize the chosen content. The images are analyzed using the image annotation technique. It uses a multimodal vocabulary consisting of textual words and visual terms. The textual words obtained from annotation are clustered with the words in the news document. Here the extractive and abstractive models for generating short, meaningful and precise captions for the news image are used. The advantages of this model are: (1) It does not require manual annotation of images (2) It reduces the need for human supervision. Index Terms --Caption generation, Image annotation, Stemming, Summarization. ________________________________________________________________________________________________________________