Section-wise indexing and retrieval of research articles

摘要

Relevant information extraction is a dire need of the scholarly community. There are a number of systems available to find relevant information from scientific literature such as search engines, citation indexes, digital libraries etc. For a search query, a long list of irrelevant documents is presented to the users mainly due to the huge number of availability of the full-text document, and furthermore due to the unstructured nature of indexed scientific resources. The contemporary systems have formally defined the structure of scientific documents. However, populating the already available enriched scientific structure from unstructured/semi-structured scientific documents has not been addressed previously. In this research paper, we have designed, implemented, and evaluated an automated technique that is able to tag each paper's content with logical sections appearing in the scientific document. The proposed system has been evaluated against the benchmark, subsequently, the proposed system have been also compared with machine learning techniques that may be used for the same task. It has been empirically shown that the overall correctness and completeness of our proposed technique is 0.78 and 0.79 respectively and thus the overall accuracy of about 0.78 was achieved. The achieved results are good as compared to machine learning based classification. The developed system may help future information retrieval systems, digital libraries, and citation indexes to index, retrieve, rank and visualize most relevant scientific documents for the scientific community.