SelectType Attribute

Specifies the type of SELECT rule behaviour.

Added Semaphore 10.1

Normal use is to use the TYPE attribute and let CS work out that it is applying to a SELECT rule and so means SELECTTYPE

Applies to

• SELECT

Values

• “overlapping” - The default behaviour - finds any intersection between children
• “contains” - Will only select evidence if it fully contains other evidence
• “contained” - Will only select evidence if it is fully contained in other evidence

Description

By default the select rule will “select” both sides of an intersection between children’s evidence.

The word intersection here can be confusing since it is normally used for sets - and sets of atomic values ie A is a member of set1 and set2 therefore A is a member of the intersection between set1 and set2. This is more complex here since the sets being considered are sets of phrase ranges and each phrase range is actually itself a set of tokens (with the constraint that they are contiguous in the document). So you have to be careful which type of set you are talking about an intersection for. To try to avoid confusion we have used the word overlap for phrase range “intersection” itself whilst intersection applies to sets of phraseranges.

If there is any overlap between phrase ranges from different children the select rule will “select” both phrase ranges from the respective children rather than creating a phrase range to describe the overlap and adding that to the intersection set.

You can modify this behaviour so that the select will pick only contained (the overlap fully covers the phrase range) or containing (the overlap fully covers the other phrase range) phrase ranges.

Example

    <select type="contained" >
         <text data="A" />
         <text data="A B" />
    </select>

On

 This has A B C in it.

Here the SELECT rule will fire with only a single phrase range “A” since that is the only contained phrase range

    <select type="contains" >
         <text data="A" />
         <text data="A B" />
    </select>

Here the SELECT rule will fire with “A B” as the only phrase range (since that fully contains “A” )

However

    <select type="contains" >
         <text data="A B" />
         <text data="B C" />
    </select>

Will not fire the SELECT rule (there is no overlap which is fully contained)

This behaviour is available through the use of other rules but you have to use a “double negative” to express contained which people found very difficult to understand.

   <select>
      <text data="A" />
      <intersection not="1" >
         <text data="A" />
         <text not="1" data="A B" />
      </intersection>
    </select>

It is not clear at first glance that the above will find any “A” that is contained in “A B” (but not other “A”’s in the document)