POS Attribute
- Last Updated: July 24, 2026
- 2 minute read
- Semaphore
- Documentation
Specifies that Part of Speech tags are matched (when available).
Attribute added in semaphore 4.0
When using language packs POS tagging is available for most supported languages - setting this attribute will check the POS tags for the data for this rule and will only match tokens in a document which have the same part-of-speech.
When using language packs pos=“1” is the default if stem="1" is asked for (ie stem matches are always POS restricted) since this appears to be the most wanted behaviour.
If a particular term is not matching due to an error in guessing the part of speech when processing the rulebase switching off pos matching for that term may help.
Example
<sentence pos="1" data="a green pub"/>
This will not fire for “Mr Green was seen entering a pub” since the part of speech for green will be determined as an adjective for the sentence rule - this will not match “Green” as a proper-noun
You could override by setting the pos attribute directly on the rule and/or adjust the default value which stem=“1” would set by editing the Classification Server configuration file, and adding / adjusting this property:
<property name=“ForcePOSMatch” value=“[X]” />
to the ruleloader bean.
Where X is true(1) or false (0) which sets the pos attribute to 1 or 0 respectively when stem=“1” applies
As an example, the bean may look like this:
<property name="allocation" value="3000000"/> <property name="ForcePOSMatch" value="false" />Also note that when the config file change is made, not only does the Classification Server need to be re-started (as usual for any config file changes), but also the rules need to be re-compiled (e.g. by re-Publishing, or forcing Classification Server to re-compile from a .pak file (e.g. by deleting the \compiled directory)).
For Language Packs version 4.1 this behaviour has been changed and stem and pos are orthoganal attributes and inherited as normal