With a growing importance for assigning subject categories to research publications, we have developed a technique based on text analysis to assist with choosing a suitable category for a publication. The analysis is done based on a publication’s abstract and/or ISSN.
To find a subject area for a publication, copy the abstract into the appropriate box. If the publication is published in a journal, give the ISSN (either of print or electronic, in format 4 digits a "-" and 4 more digits). The ISSN helps to focus the choice of subject areas to those most relevant to the publication.
You can also specify whether you would like level 2 (3-digit) or level 3 (5-digit) categories.
Language options
When you paste an abstract into the box below, an attempt will be made to automatically identify the language; you can change the suggestion if you wish. Note that while there are some 250 000 data points to create the subject area definitions in English, there are "only" 25 000 for Swedish.
Presentation of results
In result you will get two sections. The first (based on supplied ISSN) results from a comparison of the given abstract against subject areas relevant for a given journal; the second results from comparing the given abstract against all 255 subject areas. The former tends to give more reliable results, but in cases where an ISSN is missing or where we have no information about it, the latter is the only choice.
Up to five subject areas are listed for both of the scenarios above. The "score" is a measure how well the subject area matches with the submitted abstract. The higher the number the better.
If you leave the abstract field empty and enter an ISSN then you will see all the subject areas that are connected with the given journal.
Results as XML
For those interested in a more automated approach, you can HTTP POST "abstract", "issn", "language" (the latter "en" or "sv") and "subjectlevel" (the latter a "2" or "3") to https://ep.liu.se/subject_categories/Default.aspx. You will get back some simpler XML with suggested subject areas as per above.
If anyone is interested in a brief overview of the method, then it is described here.
Let us know what you think
Any feedback is welcome: ep@ep.liu.se. We are particularly interested in cases where the match against all subject areas gave a better fit than against a journal's limited list. If you send me the subject area that seemed best and ISSN, we can add it to a list of subject areas associated with the journal.
