BibPro: A Citation Parser Based on Sequence Alignment Techniques

Abstract

The dramatic increase in the number of academic publications has led to a growing demand for efficient organization of the resources to meet researchers' specific needs.

BibPro: A Citation Parser Based on Sequence Alignment

“…They include HTML structural analysis, natural language processing, machine learning, data modeling, and ontology. For citation domain, numerous works reported in the literature, e.g., [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23] use similar concepts to extract metadata from citations. The approaches can be roughly classified into two categories: learning-based and knowledge-based approaches.…”

“…In this paper, we propose a new citation parser called BibPro, which retains the advantages of our previous work [22], [23] (e.g., it uses protein sequences to represent citations and employs BLAST to find similar templates), and integrate the concept of knowledge-based approach and learning-based approach. Instead of relying on a knowledge database and heuristic rules, BibPro developed a canonicalization algorithm to systematically capture the structural features of a citation string and store these features in a sequence template.…”

“…In the third experiment, we compare BibPro with its previous version [23]. The main difference between the two versions is that the previous version does not have the blocking mechanism and the score matrix is constructed Fig.…”

Parsing Publication Lists on the Web

“…Bibro [13] is a template-based citation parser, and the key idea of BibPro is using the order of punctuation marks and reserved words in a citation string to represent its citation style. For a given citation string, BibPro encodes it as a protein sequence, which preserves citation style information.…”

“…By using these two properties, our approach first analyzes the DOM tree and find out a tree level where nodes are most likely to represent citation records. To estimate whether a node is represented as a citation record, our previous work "BibPro" [13] is applied to calculate the probability, which was designed for parsing a citation record into several fields (e.g., author, title, venue, etc.). When a string of a node is given, BibPro can output the probability that the given string is a citation string, hence we can find out one tree level in the DOM tree where citation records exist.…”