Collocates and Mutual Information

 

Top  Previous  Next

Here are the collocates of AGO computed using the written section of the BNC, ordered by frequency.

 

collocates_of_ago

 

There are nearly 17,000 instances of AGO, and YEARS is the top collocate, found 9,000 times near AGO. The "Relation" column is blank and at this point it is only possible to sort each column by its frequency or to sort the words alphabetically. What's needed is a way of knowing how closely each of these collocates of AGO is related to it. Are A, THE, WAS etc. really closely linked to AGO

 

If we now choose Compute | Mutual Information in the menu,

 

compute_MI_choice

 

and select a suitable wordlist to use for the comparison:

 

choose_BNC_wordlist

 

then we get the following list when sorted by clicking the Relation column:

 

collocates_of_ago_sorted_by_MI

 

The top items in the list now reflect much better the tendency of AGO to accompany periods of time and numbers. [The top collocate (HENSLEY) only occurs 5 times with AGO (out of 17 altogether in the whole BNC Written).]