WorldServer - What are the rules for Ranking Translation Memory Matches?

What are the WorldServer Rules for Ranking Translation Memory Matches and how do they work?

Please refer to this article to learn where to find documentation related to how Translation Memory works in WorldServer:

Where can I find detailed infomation about how Translation Memory works in WorldServer?

Please also review the attached Translation Memory User Guide to find more information about WorldServer Rules for Ranking Translation Memory Matches on page 32. This is the relevant section:

During the leverage process, there may be more than one appropriate match candidate. For instance, there may be multiple ICE matches or multiple exact matches. This section describes the rules with which TM matches will be ranked.

These rankings apply for both exact and ICE matches. They do not, however, apply to SPICE matches, because there can never be more than one SPICE match in the system for a given segment. They also do not apply to fuzzy matches, because fuzzy matches are never automatically populated into segments during the leverage process.

The rankings are listed in order of precedence. The subsequent ranking criteria are leveraged only in the event that the higher-ranking criteria are unable to lead to a single choice.

1. ICE Context: ICE matches are ranked above non-ICE matches, and full context ICE matches are ranked above partial ICE matches. Exact matches are ranked below all ICE matches.
2. Metadata-based Matching: TM match entries that contain the same AIS-to-TM mapped attribute values as the asset being leveraged will be ranked above the other matches. Attribute matching is all-or-nothing; there is no support for matching just some of the attributes. All mapped TM attributes must match the values from the asset or else there is no elevation in ranking. For instance, if there are two mapped attributes, and a match from the TM contains attribute values that map to only one of the values from the asset, then this match is not treated as an attribute match.
3. AIS Context: Matches from the same asset are preferred over matches from other documents. If the asset in question is being re-translated, then the system prefers matches that were generated from a previous translation of the current asset.
4. Position Rank: Only used for ICE matches, and when the ICE match is from the same asset being leveraged. This is currently, this is used for re-translated assets only. This ranking allows retranslation of a given asset to have multiple translations for a repeated segment, provided that the repeated segments were originally translated differently.
5. Most Recent: This is the final rank condition. Effectively, if all other things are equal, then take the most recent entry.

This article is relevant:

WorldServer - how to control the behavior of ICE Matches and set Penalties where required