Author Generated JATS XML Markup Andy Gajetzki CIO, ispub.com Olivier Wenker, MD, MBA Founder and CEO, ispub.com.

Author Generated JATS XML Markup

Andy GajetzkiCIO, ispub.com

Olivier Wenker, MD, MBAFounder and CEO, ispub.com

How We Started• Co-Founded Worldwide Cars Online in 1990– Sent images of cars and car parts via Compuserve

emails (modem speed 7kb/sec)– No official Internet – Closed the company in 1994

• Created online content while at Baylor in 1994• Netscape goes public in 1995• Officially launched 1st online journal in 1995

How We Continued• Started with The Internet Journal of

Anesthesiology• Added more journal over time• All were open access from the beginning no

registration required as reader)• Some of the first articles were submitted in

print via mail and I retyped them with Word• Articles were then submitted to me via email

(attached as Word document)

How We Continued• Initially used a Mosaic Browser tool and then a

Netscape Browser tool to create HTML for the web pages

• Then used 1st version of FrontPage to create a more complex web site

• We decided in 1997 to convert Word documents into SGML data sets and then to use XML in 1998

What We Are Today• We currently publish 82 titles (online medical

journals) at www.ispub.com • We use our own article submission system

(home-grown) at www.quickmedpub.com • We just implemented a new backend for

article submissions and article flow• We decided to have authors generate much of

the markup

And Now Lets Get Technical

Author Generated JATS XML Markup

by Andy Gajetzki

What is our JATS editor?• Represents a move to author generated

markup for our XML• Based on a customizable and reusable PHP

component – Symfony2 – popular PHP framework

• Easy to use– Form based, WYSYWIG and linear workflow

Our old workflow• How we used to do things:• Three separate workflows for each article:

1. Header generation2. Body markup3. Conversion from proprietary XML to JATS as the

last step

Word Macros

Problems with our current method

• Time consuming– Delays in publishing

• Error prone– Data entry is performed by programmers

• Authors don’t like the delay to publish and the delay to correct errors

Design Rational• We can’t support the whole spec.– How did we determine what to support?• Statistical analysis of most markup in our current article

corpus

How can we offset as much markup to the author as possible but still have a clean and intelligible end product?

What is supported• NLM Blue 3.0• Two separate support levels– Inline-level– Block-level

• Our level of JATS support is determined by each level.

Inline Level• Italics, bold, and all other presentation layer

markup supported

Block level• Single level sections only as WYSIWYG editor is

based on the HTML DOM– Other tools providing a more XML approach are

expensive, and more difficult for the author to use• General structure is

<sec> <title>

<xyz>

• <Sec> – > Boxed-text, fig, graphic, preformat, table-wrap, p, list

Titles• Support of presentational elements with, for

the most part, a non-mixed content-type

Contributors• Flexible• Single / collaborative

authors• Most JATS

<contrib-group>markup supported

• Inline-level formatting in block elements

Keywords• Keywords should be based

on MeSH entries• Validation constraints can

be applied based on that

Other article-meta• Article ID’s• Author notes• Supplemental content• Funding/grants• Article history• Permissions

Abstract / Body / Appendices

• Currently a moving target• MathML is not currently supported• Current subset of JATS covers 99% of our

cases, but we will always try to expand coverage

• WYSIWYG HTML Editor• Utilize a specific subset of HTML that we can

unambiguously map to JATS via data transformations– XSLT– regexp

• If no mapping is possible, another method must be devised

Images / Table Capture / Media• Images / Figures are handled via out-of-band

file upload on a separate page• Authors are requested to upload highest

quality format that they can• Tables can either be captured as an image, or

inserted via a Word style table creation tool • Other media types have not been

implemented yet

Endnote Handling – Document references

• JavaScript annotation tool• Endnote number / reference is highlighted in

the text and a resolution is made to a back-matter citation entry

Supported Back Matter• Acknowledgments• Appendices• Biography• Glossary’s• Citations• Notes – Content-type attribute of note element supported

Citation Handling – Back matter• One citation per line• Regular expression search for meta-data

service identifiers at PMC and Crossref– If a match is found, correct metadata is pulled

from the service• Simple JavaScript annotation tool to tokenize

citation string• Before submission, author must resolve all

endnote problems

Citation Tokenization Example

From browser to JATS XML• The block level components operate on the

HTML DOM• CSS classes are added to elements to

distinguish content types• Through various transformations, we interpret

the resultant DOM and produce the JATS XML

HTML mapping JATS XML

Validation• When things go wrong

1) XSD Validation- Intervention required by staff

2) Style/presentation problems- Intervention required by author/staff

3) Copy editing4) Peer review

Amazon Mechanical Turk• For predictable failures, Amazon Mechanical

Turk, a platform for “human intelligence tasks”, can be used

• For a small price, work units are created and human workers get paid to perform the task– 24x7 availability

Summary

Contact For Questions

Technical questions:Andy Gajetzki

[email protected]

General questions:Olivier Wenker, MD, MBA

[email protected]