There are a number of aspects to using the Match Review application in OEDQ. Click on the links below to jump to the relevant Help sections:
The following terms are central to an understanding of the Match Review application:
Issues are the units of review work used in the Match Review application. The assigning user creates an issue and assigns it to either a specific user, or to a group of users. The issue includes a URL that can be used to open Match Review, so that reviewing users do not require access to OEDQ itself.
Issues are also used as a more general tracking mechanism in OEDQ. For more information, see Issues.
A review group is a network of records that are related to one another according to the match rules defined in the match processor. For more information, see What is a review group?
A match group is a network of records which have been positively confirmed as related, either by definite match decisions made by the match processor, or by resolving possible matches with a Match decision.
The Decision Key is a set of input attributes that is used when re-applying manual match decisions to a new set of results. If any of the data in these attributes changes between one run of the matching process and the next, manual match decisions will not be re-applied, triggering a re-review of the relevant relationships. A decision key should include sufficient attributes to ensure that significant changes to the data are not missed, but should not contain attributes which do not affect the match decision.
When the development of a match processor is complete - that is, the rules for automatic matching have been finalized, and the match processor has been run - the matching results can be assigned to multiple users for review. There are two parts to the review process - relationships review, where possible matches between records are resolved to be either matches or no matches, and merged review, where, if merging matching records, the user can check and change how the final output records are formed.
Note that these two stages of the review process may be conducted at the same time, or separately. However, for a given review group, the relationships formed using the automatic matching rules should always be resolved as either matches or no matches, in order to confirm the final structure of the match group (or groups) formed from the review group. The merged output can then be reviewed for each match group with confidence that the structure of the group will not change.
In some cases, the designer of the match process may decide to formalize this separation by assigning the relationships in a match process for review, but not assigning the merged output for review until the relationships review process is complete. In other cases, it may make sense for the reviewers to decide on the best output for a set of matching records at the same time as confirming whether or not records should match.
To assign the relationships formed by a match processor for review:
To assign the merged output from a Deduplicate, Enhance, Consolidate or Advanced Match processor for review:
Note that no record merging occurs in a Link match processor, so this option is not available.
Both of these options bring up a dialog to create a new issue. The assignment of relationships or merged output for a match processor is created as an issue in order to utilize the workflow management of OEDQ, so that the issue can be worked on and closed off like any other. The users that are assigned the issue will be able to link directly through to the results of the match processor.
For example, the following issue may be created when assigning the relationships in a match processor for review:
A web link is automatically inserted on the top line of the Issue Description. This link may be sent to other members of staff to enable them to review it (see Reviewing results, below).
Note that you can assign the issue to a single user, or to a group of users. If you want many users to review your results, but those users are not in the same user group, you can assign multiple issues - one to each user - by repeating the assignment process for each user you want to review the results.
To access the review application, the user must either already be logged on to OEDQ or must have been granted login permissions for the Issue Application. There are three ways of accessing the results requiring review from a given match processor:
The first two methods will normally be used by reviewers working on production systems. These users may not have the ability to view or change the match process.
The third method will normally be used by the developer of the matching process, to check results before the development of the process has been completed.
The review screen displays matching records organized into review groups, as shown below. See What is a review group? for more information on review groups.
All the review groups (of both definite and possible matches) are available for review. To filter the groups, click on the Filter icon, and select your filter criteria. For example to filter the groups to see only those groups that contain relationships with a Decision of Possible Match:
Or, to filter the groups to display only those with relationships formed by a specific match rule:
You can also use a number of filters in combination. To use multiple
filters, click on the Plus button 
and add another filter.
By default, only groups that meet both filter criteria will be displayed
(that is, AND logic is used). However, you can choose to display groups that
meet any of the filter criteria by selecting the Use
OR logic checkbox. For example, the below filters will find all
groups that contain a relationship with a Decision of Match, and all groups
that contain a relationship created by the 'Forename stand, surname exact,
address exact' match rule:
Note that by default, the options specified to filter groups will only filter the view down to the set of groups that meet your criteria. For large review groups that contain a lot of relationships, it may be useful to filter the group itself down so that you are only comparing records for one or two relationships. You can do this either by manually selecting relationships within each group in the lower pane, or by ticking the option to Auto-select relationships based on search, which will then automatically select only those relationships that meet your stated filter criteria (for example, those created by a specific match rule).
If you do use these options, you should be aware that there may be other records in the same review group that are currently not being displayed, though you will still be able to see the relationships that join these records to the review group in the lower pane. In the below screenshot, for example, the review group contains three records, but only the two records matched using the Name and Postcode rule are being displayed because the Auto-select option is ticked:
The Matching Records Review screen allows you to make a final decision for all possible matches by making a manual decision of Match or No Match for each pair of records that matched using a match rule with an outcome of Awaiting Review. It is also possible to set the decision to Pending - this enables an extra step in the manual matching workflow, to indicate that a relationship has been reviewed, but no final decision has been made yet.
|
Note: It is possible to customize the match decisions and decision transitions to support specific workflow requirements. This includes the ability to restrict certain decision transitions so that they can only be made by groups of users. If match decisions have been customized, you may see different values in the Decision dropdown. Customizing match decisions can be done by the administrator of the OEDQ server. |
All manual decisions are remembered in subsequent executions of the process containing the match processor, provided the records being matched are not changed. Therefore, for each possible match, you will only have to decide once whether it should, or should not, be considered as a positive match.
To resolve a possible match as either Match, No Match or Pending, change the Decision on the relationship:
It is also possible to change previous definite decisions (either Match or No Match decisions) back to Possible Match (or Pending), if they were found to be made in error during the review process.
Note that the Review Status of the relationship changes to User Reviewed, and the structure of the match groups may change, as records are definitively related, or not related, to one another. This is reflected in the colors used in displaying the review group and the Match Group Ids of the records in the review group. The user and date/time the decision was made is added automatically.
It is also possible to add comments on relationships, for example to store details of how a manual match decision was reached. To add a comment, use the green plus icon underneath the speech bubble icon:
When a comment has been added to a relationship, the green plus icon changes to a speech bubble. Click on the speech bubble to display all the comments on the relationship, and to add or delete comments.
All manual decisions that have been made can be viewed by changing the filter to see groups where the Review Status is 'User Reviewed'.
You can change a previously made match decision on a relationship provided the review group that the relationship is not locked by another user.
OEDQ stores a full history of all decisions made on a relationship. To see this history, click on the calendar icon on the relationship.
The Review Merged Output tab of the review screen allows you manually to resolve any errors that were encountered during the automatic selection of a merged output record for each match group, by selecting or specifying the required output. It also allows you to over-rule the automatically generated output, if you think that the automatic output rules have not selected the best output data.
Note that this process is not relevant for Link processors, or other match processes where there is no need to merge records together in the final output.
Merged output is automatically generated using the rules specified in the Merge sub-processor. However, those rules may cause errors in output selection (for example because there is no 'Most Common' value for an attribute). If a match group contains one or more errors, it is failed. The output values for these groups must be specified manually.
To specify the output for failed output groups:
You can the review the merged output record created for the match group in the lower half of the screen. The attribute or attributes with selection errors are highlighted so you can quickly see the decisions that you need to make.
To specify the desired output value manually, simply type it into the grid. Alternatively, you can copy and paste the value from the matching records in the group displayed above.
You can also add a comment when making manual output decisions, for example to capture how the decision was reached. The latest comment is displayed in the window to the right of the merged output record. To add a comment, click on the ellipsis button above this window.
As well as resolving output where the automatic output rules have failed to generate it, you can review and over-rule the automatically generated output for a match group. To do this, simple navigate to any match group using the filters available, and overwrite the automatically generated output by manually clicking in the cell.
You can change a previously made output decision at any time.
On occasions, it is useful for the administrator of a review to make a large number of manual match decisions in a single automated operation.
For example, in order to speed up the process of making manual match decisions, it may be decided only to make positive Match decisions. All remaining Review decisions might effectively be considered as No Matches, once a review has been completed. If the review process for a set of match results has been completed, therefore, it may be useful for the administrator to set all remaining Review decisions to No Match. This then ensures that when the match process is rerun on a larger set of data, possible matches that have already been reviewed will not need to be reviewed again.
Bulk review decisions are configured by selecting the Configure Bulk Review Rules link with a match processor open:
It is possible to filter the decisions to be updated to a new Match Status by Rule, Review Status and Match Status. It is also possible to apply multiple bulk review rules, for example to reset all decisions for one rule to Review, and to set all Pending decisions for another rule to No Match.
The following screenshot shows an example where all manual decisions with a current Match Status of Pending will be set to No Match:
The configured bulk review rules will be applied the next time the process containing the match processor is run. They will not reapply on subsequent runs of the match processor, however, as it is assumed that bulk review decisions should be applied as a one-off operation.
Oracle ® Enterprise Data Quality Help version 9.0
Copyright ©
2006,2011 Oracle and/or its affiliates. All rights reserved.