Automating the production of bibliographic records for MEDLINE

Links to Feeds:
Publications	RSS
CEB News	RSS

CEB Projects

page

Automating the production of bibliographic records for MEDLINE

10 Production supervision and control tools

In a production system the management and control of the workflow is exercised by the facility supervisors who need to transmit necessary information to the workstation operators. Observing that some of this information flow could be automated, tools were developed to assist in this process. The objective is to meet the supervisors' administrative and management requirements while realizing productivity gains. CheckIn, Admin and CrashPatrol are such tools.

10.1 CheckIn

The first stage in the MARS workflow is CheckIn,⁴⁴ immediately followed by Scan. CheckIn's role is to collect and organize information necessary to track the physical journal issues from their arrival at the MARS production facility until processing is completed. Prior to the development of this module, the facility supervisor manually researched key information related to the journal issue to be processed, hand wrote the information on a paper form, and attached the form to the issue before sending it on its way to the various operator workstations. CheckIn exploits the fact that most of this information is already in the DCMS (Data Creation Maintenance System) and the MARS databases, and may be automatically extracted from these.

The process is as follows: as a journal issue arrives at the production facility, the supervisor wands the NLM-applied barcode to read the "MRI" number that uniquely identifies the issue. CheckIn sends an on-demand query via network to the DCMS database at NLM, which returns data about the journal title (such as ISSN and journal name) and the particular issue (such as volume number, issue number, and cover date of publication). A concurrent query to the MARS database at the facility retrieves some other items of related information, namely, the journal's priority and notes about how it should be processed.

The retrieved information is used for two purposes:

It is printed on a cover page (Figure 10.1.1) that the supervisor paperclips to the journal issue. This information helps keep track of the issue as it moves through the facility and provides guidelines and special instructions for the operators. The cover page is subsequently retained in a manager's notebook as supplemental information.
The information, after any adjustment by the supervisor (e.g., change the priority for handling the journal, or to add special instructions for the operators), is stored in the MARS database, initializing the workflow for this issue. The data then flows to the Scan and Edit stations, where the operators do not have to type or pick it manually.

Design and Architecture

The design of CheckIn evolved from the rapid prototyping of several dialog-box-based applications built in VC++/MFC. CheckIn's main dialog window looks like the printed cover page of Figure 10.1.1, minus the grid. CheckIn is a manned application, rather than a daemon, because the "Notes" field, for instance, may need to be edited to pass on supplemental, transient information to the MARS operators.

Figure 10.1.1 Cover Sheet. Shown reformatted for conciseness from its actual full-page size. The grid at the bottom is penciled in by each task's operator as the attached physical journal is relinquished. Citational information above the grid, prior to the development of CheckIn, was entered by hand.

Activated by the supervisor, CheckIn passes a journal's MRI as a URL parameter to the DCMS website. Live data is fetched from there in XML format, then parsed using Microsoft's MSXML implementation of the standard XML DOM object^45,47 model. DOM is a good choice for small XML files (as here) since the entire DOM parse tree resides in memory. The data format is defined by the "NLM MEDLINE" DTD⁴⁶. While neither this DTD nor the DCMS web service (i.e., that delivers XML-over-HTTP) were specifically designed for MARS, they offer vital data that MARS needs: ISSN, journal title (in full and short forms), cover date, volume and issue number. CheckIn also retrieves the remaining useful data that is not available with this DTD (e.g., initial/default values for priority, or for in-house "notes" to operators) from the MARS database.

In addition to reading from the MARS database, CheckIn also writes data to it, as purpose (2) suggests. Such database access is through client-side middleware. The original MARS implementation deployed a C++ class library for this: RogueWave's DBTools++. But this has the disadvantage of on-going per-seat costs. The design of CheckIn pioneered our use of Microsoft's ActiveX Data Objects (ADO) instead, a more economical option in the long run.

Besides the addition of role (2) functionality discussed above, CheckIn includes a "Lookup Wizard", to enable the operator to look up a journal from the MARS database by either journal title or ISSN, in case of the following difficulties with retrieving information from DCMS:

No response, or an erroneous response, from the DCMS web service;
The requested MRI is unknown to DCMS;
The returned information is missing an ISSN;
The returned ISSN is unknown to MARS.

Case 3 can occur because, in the DCMS database, unlike in MARS, ISSN is neither a key nor a required field. (If a journal really has no ISSN, the facility supervisor enters a placeholder number.) For case 3, CheckIn automatically tries to match the fetched journal title against those in the MARS database. For case 4, the user is asked if such a match attempt by title is desirable.

For all cases, the fallback is Lookup Wizard invocation. The wizard first explains the problem and allows the user to ask for a search by ISSN or title. The second step presents an ordered scrollable list of all journals known to MARS, and an entry field pre-initialized if possible. As the user edits the entry field, automatic prefix matching occurs against the list, so that the field may be completed with minimal typing.

Performance in production

CheckIn's operation has been successful. Each day, the supervisor uses it to check in about 40 journals with at least 600 articles. Prior to CheckIn, the information in the cover pages was hand researched and written, a process that took over an hour. It now takes under 15 minutes. We expect to see additional labor savings upon completion of the rollout of data deposition by CheckIn and its pick up by Scan and Edit.

Next steps

Planned improvements to CheckIn include:

Enabling the supervisor to add a new journal title and its ISSN to the MARS database at CheckIn, without requiring intervention by the MARS database administrator as at present;
"Instrumenting" the module so that its performance can be quantitatively assessed;
Contributing to a projected MARS resource management system, that will replace the cover sheets with a paperless system.

page

Return to top