Linguistics 431/631: Connectionist language modeling

Ben Bergen

 

Meeting 6: Problems in morphology

September 26, 2006

 

Pinker and Prince

 

Pinker and Prince (1988) made two main critiques of the RM model

      They argue that it does not in fact behave like a child learning language

      Its failings are due to properties of connectionist systems, which means that any similar connectionist system will fail for the same reasons

 

Today well talk about basic issues PP take with RMs model and their analysis of its performance.

 

RMs assumptions

 

They first consider the implicit assumptions about language made by RM.

      That the Wickelphone/Wickelfeature provides an adequate basis for phonological generalization, circumventing the need to deal with strings.

      That the past tense is formed by direct modification of the phonetics of the root, so that there is no need to recognize a more abstract level of morphological structure.

      That the formation of strong (irregular) pasts is determined by purely phonetic considerations, so that there is no need to recognize the notion 'lexical item' to serve as a locus of idiosyncrasy.

      That the regular system is qualitatively the same as the irregular, differing only in the number and uniformity of their populations of exemplars, so that it is appropriate to handle the whole stem/past relation in a single, indissoluble facility.

 

1.   Problems with Wickelphones

      No way to represent certain distinctions

o      Albal

o      Albalbal

      Dont capture similarities

o      Slit and silt have no Wickelphones in common

      Fails to exclude the impossible

o      ABC -> CBA is a easy as ABC -> ABC

 

2.   The variation of the past tense suffix is explainable as a general phonological principle

      Consider other English suffixes, like the 3rd person present indicative s, or the possessive s

      These show the same type of allomorphy, as well as a regular-irregular distinction

      Coda clusters must be matched for voicing assimilation deals with this

      You cant have two identical sounds in a coda schwa insertion takes care of this

      So PP critique RMs model for only treating the past tense, where more general phonological principles seem to apply

 

3.   Because the relation between present and past tense is truly arbitrary, RM cannot explain the prevalence across languages of inflectional alternations that preserve of stem and affix identities.

 

4.   Because the RM model has no representation for a lexical item, but rather is based entirely on the sounds of the present tense form, it cannot deal with homonymic forms with different past tenses

       He braked the car suddenly.  *broke

       He flied out to center field.  *flew

       He ringed the city with artillery.  *rang

       He sleighed down the hill.  *slew

       He spitted the pig.  *spat

       He righted the boat.  *rote

 

5. The RM model does not distinguish between strong and weak verb patterns. While the weak pattern is applied to any verb, the strong patterns have certain characteristics

       Family resemblances

       Prototypicality

       Failures of predictability

 

PP argue that the differences arise from the strong classes being memorized and the weak classes being the product of a rule (note dual-route models).

 

Problems with the results of the model

 

When RM tested the model with 72 new regular verbs, 24 produced incorrect outputs.

 


Some were simply perverse:

a. squat - squakt

b. mail - membled

c. tour - toureder

d. mate - maded

 

Along with some strange vowel changes:

a. shape - shipt

b. sip - sept

c. slip - slept

d. brown - brawned

e. mail membled

Others overgeneralized the cut pattern, which should be restricted to verbs that end in t or d (as has been shown in children).

a. hug - hug

b. smoke - smoke

c. brown - brawned

 

Finally, some showed a strong or exclusive tendency to double marking with the regular past tense morpheme:

a. type - typeded

b. step - steppeded

c. snap - snappeded

d. map - mappeded


 

PP argue that this is not indicative of adult-like morphological knowledge.