You are on page 1of 4

Project 2: Word Blends

Thomas Bulick and William Clark


Ling 401
Due 12/1/15
Work Breakdown:
WilliamWrote the containment, consonant blend, polysyllable blend, general blend function, combined code,
edited paper.
Thomas Wrote the paper and wrote code for the vowel blend function

I. Goal
The purpose of this project was to create all the possible forms that could possibly occur
as a result of blending between two given words. To this end, we designed a code that produces
forms that account for the three main ways two words can be blended: at shared consonants, at
similar vowels, and by various combinations of the syllables composing the two words.
II. Methods
We used four basic functions and then one more complex function to accomplish our
task. Our first function, containment(wd1,wd2), takes two arguments and checks to ensure that
the two words provided are actually blend-able. Our native speaker intuition tells us that two
words like bicycle and cycle cannot be blended together because one word is contained
within the other and a blend would just result in a recreation of the longer word. Specifically, the
code checks three scenarios, if the length of word 1 is greater than that of word 2 it ensures that
word 2 is not contained in 1, if word 2 is longer than word 1 it ensures that word 1 is not
contained in word 2, and if word 1 and word 2 are the same length, it ensures they are not
actually the same word.

Next we had our consonant blending function, constBlend(wd1,wd2). This function also
takes two words as arguments and works by analyzing all the consonants within each word and
then identifying the consonants shared by them. The function then creates combined words at
each shared consonant. An important note that applies to all of our combining functions is that
we have decided that the order in which words are presented seems to matter. In our own native
speaker intuition, if asked to blend dog and cat we definitely would produce dat but if
asked to blend cat and dog it would produce cog. Therefore when blending two words the
order is preserved in that the first part of the blend comes from the first word and the second part
of the blend comes from the second word. However, in the case that some speakers argue that
order is irrelevant it would be a relatively simple task to adapt the function to account for the
extra possibilities.
Then we had a function that blends based on vowels, vowelBlend(wd1,wd2). This
function works in a very similar way to the consonant blending function, also taking two word
arguments. It identifies the vowels in each word and then accounting for order as in the
consonant blend function, it creates a list of possible blends at those vowels.
Our fourth function, polyBlend(wd1,wd2), then accounted for possible blends based on
combinations of syllables from two separate words, the process by which blends that do not
share segments are generally formed. Firstly, the code does its best to break down each word into
syllables, consisting primarily of a CVC structure. Although this isnt perfect and has room for
improvement, without a very detailed analysis of possible English syllable structures it is the best
our code can account for. The code then assembles a combination of the first syllable of the first
word and everything but the first syllable of the second word, and then returns that as the most
likely syllable based blend.

Finally, our last function, blend(wd1,wd2), was more complex and tied everything else
together. This function takes two words as arguments and then applies all the other functions
previously mentioned to those two words. First it uses the containment function to ensure that the
two words can in fact be blended in the first place. Then it assembles all the possible blends that
could be formed from the two words, shared consonant blends, similar vowel blends, and
syllable combination blends. The function then produces lists of each type of blend, labeled as
consonant blends, vowel blends, or polysyllabic blends, giving us our list of possible ways that
any two given words could be blended. Additionally, if any of our individual blending functions
failed to return anything, then the function will output that failure with a statement like No X
blends found.
III. Conclusions and Improvements
Overall our code did produce the desired results. Given an input like crocadile and
alligator we are given a list of all possible blends that could result: ['cr', 'crocadilligator',
'crocadiligator', 'calligator', 'cigator', 'cator', 'cor', 'crocligator']. Also from analyzing the data
from our function and comparing it to native speaker intuitions we can identify which of these
are most likely to be what would actually be produced as a blend.
One improvement for our code would be in our vowel and consonant blending functions.
As in the example given, currently the functions will return words like cr and cor which our
intuition call invalid blends for these two words. To remedy this I believe the simplest method
would be to write in a check that ensures the blends that are returned contain at least as many
syllables as the smaller of the two inputs. In general, no blend made by either of us as native
speakers ever reduces the syllable count to be less than both of the provided input words..

One improvement that could be made is to do a more comprehensive analysis of syllable


boundaries to use within our polysyllabic blend function. English has very well defined
acceptable clusters and ideally we would want to adapt our function to better account for the
exact acceptable clusters and syllable structure. Id imagine the most efficient way to do this is to
actually just use a non-alphanumeric character to encode syllable breaks into the input itself,
although one could also write a code with a set of acceptable onset clusters and coda clusters and
identify those clusters accordingly.
Another improvement would be, ideally, to convert this entire procedure into phonetic
transcriptions. Especially with the vowel blending function, sometimes the orthographic
characters correspond to different phonetic sounds and blending by vowels generally occurs at
similar phonetic occurrence, not orthographic ones. Therefore in an ideal world, we would have
each word written in IPA to account for more accurate vowel, and also consonant, blending.

You might also like