�ݺ�ߣ

Editor's Notes

• #3: OSA is a scholarly publisher with 18 current and legacy journals, 300+ conference proceedings, a member magazine, and a growing number of additional products. Legacy journals go back to 1917 and have heavy physics content. Also content many surprises, e.g., foreign language articles, society news, yearbook-like pix, cartoons, doodles.
• #6: Citation data modeled in various ways to help researcher size up the article quickly.
• #7: All equations captured as MathML; display solution with MathJax running on OSA server. Works on all major browsers (including on smart phone) with no user download. MathJax performance will improve with better MathML native support from browsers.
• #8: Go to Optics ImageBank to browse all figure images across all content!
• #10: Legacy journals served as the membership magazine but also at times like a yearbook. Particularly challenging were 100s of photos of people—unrelated to the article—who where tacked to the ends of articles as filler.
• #14: In developing specs, we had to achieve balance between two extremes. The method we settled on was highly collaborative and required both parties to be agile. The balance of the presentation will describe the process that OSA and DCL used to work collaboratively to convert 750,000 pages of legacy journal content to well-polished NLM XML.
• #17: Phased by source material, journal type, newer-to-oldest -> enabled full collection in shortest amount of time.
• #18: >>Source types we dealt with in the different phases of this project. OSA DTD is XML is created as part of the production process and is used for publishing…not done as an afterthought. It’s XML first. NLM 2.3 and SGML material was xml/sgml last.
• #19: Because of the wide range of inputs and many years of varied material, we wanted to build flexibility into the process – and these are three ways. Weekly meetings with OSA – scheduled with an agenda… in addition to close communication anytime something new came up OSA had to make business decisions about the repercussions regarding going back to make changes…
• #21: No superset of the JATS DTD was created This is a list of some of the conversion details that had to be dealt with. Fitting structures in the NLM – make sure that all tags and information converts and is tagged properly in NLM tagging. Tables – the tagging structure of CALS vs. HTML is different and so empty cells, spanned cells etc. all have to be properly accounted for otherwise the HTML table will be skewed. MathML – equations often appear on several lines in the pdf. Retain these line breaks where possible. In this SGM part of the conversion we are converting the Math as is, so if there are no line breaks in the SGML, then the XML does not have either. Cross reference range information – these are the textual callouts to structures (references, figures, tables etc.). Should we retain all of the xrefs within the range or just the end values?
• #22: These are examples of cross-reference ranges in the source. In the case of citations – we include only xrefs of the upper and lower limits of the range. For figures, tables and equations – we include <xref> tagging for the items found in the range as well.
• #23: Since the CSS does not support character table alignment, we have to align columns using only right, left and center. Note how the middle column is aligned center since left/right would not work, so center is the default value for numeric data.
• #24: As we converted unexpected scenarios came up where text was missing in the source. In cases like these we would turn to OSA for guidance in how to handle the missing text. If another page was available, it would be replaced. Using a placeholder to mark missing text was a typical solution – like [[missing text]].
• #25: The requirements for jumping pages is to include all necessary page numbers in the metadata in <fpage>, <lpage> and <page-range>, and then combine the different pages into one PDF – so that one XML and one PDF (that contains all necessary pages) is delivered for an article containing jumping pages.
• #26: Here’s an example of a chemical double bond using Unicode 2550 (renders in red).  The point this example brings out, is that since there is no specific Unicode entity for a chemical double bond, it has been mapped to an alternate entity (from a box drawing set of entities), which in this case fortunately looks rather close (or the same) to what was needed.  
• #27: A non-standard structure was found – OSA was consulted, and they opted to have it tagged as boxed text.
• #28: OSA had to determine whether or not they actually wanted to retain this information since this is filler info – not really part of the article. In the end they decided to retain it in the <back> - in AO as <bio> and JOSA title as <sec> with a attribute value of “end-matter”.
• #30: A visual review is done to make sure the text is visually correct. These are 4 specific things to look out for.
• #31: Mention Sasha Essential collaboration between DCL and OSA. OSA with subject matter expertise and DCL with technical conversion expertise. This warning was issued because (to be published) was found in a citation, indicating that it has not yet been published, and for unpublished material the citation type attribute should be “other”. This warning was issued because the title of the section designated for Tables/Figures has only Figures yet the title is “Figures and Tables”. This error was issued because the citation contains multiple <article-title> tags, and only one is allowed.
• #33: The sw finds different variations of the same word and outputs this report to help find extra hyphenation. It also checks hyphenation of all words – to see if both
• #34: This report displays the metadata components in a table format, which is visually easy to review. Discrepancies are easily seen and can be highlighted.