Collocated Verbs

The current collocated verb structure has a connection directly from the base verb to one of the maps for the preposition. This has no node to allow connection to synonyms or multiple meanings. It also fails where two prepositions are used in series.

 Collocated verbs can be classified into four forms within the system: 

Mainly synonym of a single relation – “kick out”   ->    “reject”
Additional relation – “fight on” -> “continue to fight”
Direct build instruction - "agree with thing" - Relation2Object
Better index into preposition maps – “cut with a knife” -> WithUse 

Synonym of Single Relation

“Kicked out” can be seen as an invocation of ToReject and, as a possible meaning of ToKickOut, we could use the relation structures of ToReject to verify that this meaning was meant – ToReject

            “The cow kicked out”

 would be rejected, and a prepositional map would need to be found.

The structure, with change of meaning from parent to collocated verb:

Collocated Verb.JPG (115747 bytes)

We find ToKick, then find its collocations, and find the common element with the known preposition, in this case ToKickOut. ToKickOut itself can have relation structures, or it can acquire them from its different meanings, if it has them.

 This structure aids searching, as the search structure can use ToReject, ToKickOut, or any other synonym or false antonym - a false ToAccept will do as a synonym for ToReject– all will be resolved to the same ToReject relation.

 The use of simpler forms – ToReject for ToKickOut or ToDiscard for ToThrowOut – serves to formalise the built structure – making search faster.

 Sometimes the appropriate relation is not a direct one to one mapping – a 2-place relation should turn into a 3-place, or an active use turn into a passive use, as

     John came up against an obstacle -> An obstacle blocked John

 A simple mapping operator can be interposed, to allow for remapping of the parameters. See Synonym Map.

Matching 

We will often need to match the symbols we are given, as

The accident came about as a result of carelessness.
The package came about 3 o’clock.
I have come about the rug.
The ship came about in the gale.

 There are two levels of matching - multiple independent tests and the use of an Active Map.

1) Operators connected to the collocation provide tests - is the relation expressed as a noun, is it in a passive grammatical context, is the object a person. An example is the collocations for ToAgree.

allagreecolls.png (13111 bytes)

Note the inherited collocation - InTime - which applies to all actions. A specific collocation has tests.

agreewith.png (22375 bytes)

The collocation AgreeWithFred requires the object of the preposition to be a person - the checked test in the bottom left hand corner of the window. The action to be taken on matching is to create an ObjectGroup on the subject connection of the relation, as

AgreeWith.JPG (65003 bytes)

If the statement had been "I agree with the decision", then a different structure would have been built -

agreeWithApproach.png (51825 bytes)

The collocation structure in the network looks like

agreenodes.png (62790 bytes)

The collocation is found as a common object between ToAgree and With, a check is run (MustBePerson) and then the network construction command is executed.

2) A PREPOSITION1 operator attached to the Collocated Verb is an indication that a test using an Active Map must be run to determine if the synonym is valid - the map allows much more precise testing than can separate independent tests on the relation and the object.

 Some collocated verbs have no exact single word relation synonym, and for those we would need relation structures hung off the collocated verb node. We will also have the case where several collocated verb forms have the same meaning, with no single word relation as a synonym. In these cases, the other forms will point to a more formal collocated form, such as

 to ease out” -> “to remove from”

 The method needs to work successfully for different forms of the relation

They kicked out the application.
They kicked the application out.
The application was kicked out.
John suggested kicking out the application.
The kicking out of the application came as a surprise.

Collocations Requiring Additional Relation

 John worked on through the night.
The rope was cut with a knife.
He voted for many reasons.

Multiple Collocations

We also need to handle cases where there are two prepositions together:

Why do I have to put up with that?
John doesn’t get on with Fred.
Fred and John don’t get on.
He should stop messing about and get on with it. (two different meanings, depending whether the object is a person or not)
Keeping up with the Joneses.
I want to find out about whales.
He came up against a problem.

 We can handle these using an extension of the method for a single preposition - ToFindOut is connected through another MEANING2 operator to ToFindOutAbout.

Direct Build Instruction

For simple collocations (the bulk of them), a build instruction is sufficient. It might be "take the left hand object, make it a relation, make the right hand object the subject of that relation. The "by" in "a man was hit by a truck" would use such an instruction, after checking that the left hand object was in passive form and the right hand object was not a time value.

Indexing into Prepositional Maps

Some prepositions have many potential meanings, and it is more efficient to index into a specific meaning characteristic of the relation.

He voted for the candidate.

He voted for many different reasons.

He voted for the first time.

 The other uses of ‘for’ can be determined by searching through different maps to find a match – the ‘vote for a candidate’ is sufficiently distinctive to provide an index to it.

Collocated Verb Index.JPG (109601 bytes)

 The indexed map can be on a different preposition to the one which locates the collocated verb.

The conditions for the map still have to be satisfied – if not, and all maps indexed from the collocated verb fail, the search will restart using the preposition only.

 See

Prepositional Maps

Synonym Map

Communication Relations