Editing NZ Online: The Review Process
For the reviewer, the edit process begins by logging in. This takes the editor to a page with account details and information about the block of records being reviewed. If the "Resume Editing" link is clicked the edit application loads and brings up the next unreviewed record in the block. The reviewer can review as many records as desired in a session and log out at any time.
The review process entails comparing the digital record with the print record. The edit tool page provides a screen shot and outlines the process.
Submit Review Decision
Review of a record results in one of three possible outcomes.
Important to reviewers: Please ensure you understand the distinction between a correction and a comment. The purpose of a correction is to match the digital record with the printed record. To suggest a change to the original printed record use the "Comment" option. Changes to the original printed record will be reviewed by a review process to be determined. Any addenda will be attached as supplementary records cross-referenced to the original.
Repeat for each record
At the end of a review session the reviewer logs out. Any corrections made to NZ records may not appear in the database immediately. The correction submission process creates a new temporary record that is appended to the database. An admin tool allows a subsequent review of groups of new corrections by an NZ editor. These corrections are reviewed prior to committing the changes to the database permanently.
Notes and Observations
Reviewing falls into two categories: typographical accuracy and parsing accuracy.
Typographical accuracy: We are primarily interested in typographical accuracy. The double-keying process appears to have maintained a high degree of accuracy in the conversion. The only exception is minor and easily addressed with an automated pass. This is a terminal period (.) missing at the end of many instances of the taxonomic category. ("Dipt" instead of "Dipt.", "Mamm" instead of "Mamm.")
Missing periods within the Category: One exception is a relatively consistent missing period on the abbreviated taxonomic categories at the end of the records. This we can fix automatically in post-processing so that alone should not warrant a correction. Do not bother correcting a missing period in the taxonomic groups.
Asterisks within the name record: New name records within the Addenda and Corrigenda records are indicated with an asterisk preceding the name string in the print record. This character affects querying of the database record. We addressed this by removing the asterisk and setting the following attributes: newRecordFlag=1 CorrigendaFlag=1 addendaFlag=0. This allows this class of records to be uniquely identified.
Parsing accuracy: The process of parsing most records is a relatively straightforward affair but some records are more difficult to interpret how the text might be broken into constituent fields. A record may be typographically correct but not correctly parsed into appropriate fields. Some judgement may be required to assess how a record should be parsed.
For example the genus Acanthohoplites is parsed as follows:
The publication citation contains additional nomenclatural comments similar to that which is often parsed into the notes field. Many cases like this exist where there is some question as to how best to parse the record.
Parsing issues will ultimately fall to the judgement of the reviewer and final edit process when the corrected record is committed to the database.
Note: One of the most common (but still rare) causes for an erroneous record come from problems with the automated parsing process. The parsing of the record used automated methods that made assumptions about the data record. For example, a dash always precedes a taxonomic category that in turn is always at the end of a record. In some cases, a dash appears within a publication and the parser apparently took this to indicate the end of the record and truncated the record prematurely.
This was done at the contracter and a note to future conversion efforts (ourselves and others) is to request a copy of the unparsed file as well as the resultant parsed file. A pass comparing the string lengths of each record would detect these unexpected truncations.