The Band Assignment Parser: A Tool to Identify Band Assignments in Research Publications

摘要

The process of extracting spectroscopic band assignments from the literature is a necessary but laborious part of the interpretation of spectra, particularly for those new to the field. As part of a wider project we have developed a tool called the Band Assignment Parser to aid the process of finding band assignments. The method is implemented on the web at http://umber.sbs.man.ac.uk/BAP/ and http://www.S-A-S.org/BAP. It accepts HTML and Adobe PDF versions of research publications and extracts sentences containing band assignments, discarding the rest of the information. Based on our sample data, this enables the user to recover >90percent of the band assignment sentences (which, due to redundancy in the sentences, is almost 100percent of band assignments) while having to read only 12percent of total sentences. The method follows a computational approach to processing literature known as 'text mining'. Text mining commonly applies methods from the fields of Natural Language Processing and Artificial Intelligence to capture important components of the text. However, for the Band Assignment Parser such sophistications are unnecessary as the spectroscopic community is consistent and reliable when referring to specific bands in a spectrum, as spectroscopic units are almost always reported. Therefore, the task of identifying sentences containing band assignments can be reduced to the simpler problem of isolating sentences containing units. The Band Assignment Parser simply looks for unit notation, such as cm~(-1), nm, or eV.