Linguistic Commonsense

Let me introduce 2 kinds of verbs – ‘Free’ and ‘Attached’. Attached verbs are those verbs which have a variable in their definition. They are dependent upon an external variable entity for the completion of their Semantics. Free verbs are those which don’t have a variable in their definitions; they are “free” and self-sufficient.

For example, consider ‘preparing’. Preparing =  all actions that go into the creation / making of {something}. Here, the ‘something’ is the variable. So, ‘preparing’ is an ‘attached’ verb.  What is the consequence of this? The occurrence of ‘preparing’ has to be accompanied, more often than not, with a partner – this ‘something’ – somewhere in the data, where ‘preparing’ occurs.
But now consider a verb I define – say, ‘debbing’. ‘Debbing’ means the collective act of walking through the corridor of the house to the hall, then walking to the sofa and finally sitting on the sofa. There is no apparent variable here; hence, ‘debbing’ is a ‘free’ verb. Consequence? A free verb like ‘debbing’ needn’t be accompanied by some partner (filling the slot of any variable in its definition), in the piece of data/text it appears. It is free to be tagged along with other linguistic fragments like – debbing so late in the night, debbing so fast etc. but doesnt have any imperative companion.

This concept has Cognitive bearings. “I am preparing” is more “incomplete” than “I am debbing”. that is, if someone says – I am debbing – it seems fairly stable and doesnt lead to further imperative questions. But if someone just says – I am preparing – it immediately sparks a query in the mind like – preparing what? Hence this “Linguistic-Cognitive-inspiration” is present in attached verbs. 

Now, this slot-filling of variables will lead to commonsense-connections to the original word, and they come from the definitions of these words. Hence, this is what I call ‘Linguistic Commonsense’.

One application : 
Creating a mass of commonsense reasoning over a text – 

Extending this to words in general – There will be some ‘free’ words like say ‘calculator’ wherein there isn’t any appreciable/noticeable variable, and there will be some ‘attached’ words like ‘preparing’ where there is clearly a variable. The latter is what we have to exploit for mining out commonsense reasoning links and commonsense activity in general over any given text!

You can derive a lot of commonsense using this so called Linguistic-Cognitive-inspiration inherent in ‘attached’ words. The variable part in the definition of some word is the commonsense attached with the occurrence of that word in any data. So, for example, ‘dancing’ has the definition – moving body and body parts to some sound. Here the variable is ‘some’ (sound). So the commonsense attached with dancing is that – there should be some sound accompanying dancing, if and when there is dancing going on. Or, ‘preparing something‘ (omelette, breakfast, a trap etc.) is the commonsense attached with the word ‘preparing’ in any data. 

(Note :  We are here only talking about Linguistic-Commonsensically inspired questions and not general conventional commonsense. In the latter sense, there are other questions accompanying ‘dancing’ like dancing on what? dancing where? dancing with whom (if anyone)? etc.)

So, given a text, take each word and its definition. If there is a variable in there, question for the entity present in that text, filling that slot. That will lead the machine to think (question) commonsensically over that text, in cases where there isn’t an answer to the question present in that text!

One thought on “Linguistic Commonsense

  1. Suggested reading to add to annotated bibliography:
    Kiryu K. Types of verbs and functions of the causative suffix-k in Newar. Kobe Papers in Linguistics. Vol. 3. 1. 2001;9.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: