Deeper into Linguistic Variation

The Atkins-Levin paper begins by giving a thorough account of the uses of "shake" verbs in a large corpora.  They found unusual transitive uses of shake and its near synonyms exclusively in fiction and attempted to isolate the reason for these differences.  The main observation they drew is the fact that the various shake near-synonyms differ in their uses based on the semantic types of their subject nouns.  They were able to isolate a possible reason for the distinction between these words, conjecturing that "the semantic concept of internal versus external causation can account for the apparent idiosyncratic behavior of certain verbs, explaining why near-synonyms may differ syntactically despite their semantic closeness" (107).  It is fascinating to me that the lexical meaning actually influences the syntactic role a word is able to play in a sentence.  

Last week, we read about syntactic trees--ones that can generate syntactically valid sentences.  After reading Atkins and Levin, I was left wondering whether we could create a sense-parser that would be able to generate semantically valid sentences.  It would be interesting to write an algorithm that extracts the various properties of the words that surround other words, and tries to generate sentences using this kind of information learned from large corpora.  However, I am not sure this kind of parser would be able to draw high-level abstractions about near-synonyms like the distinction between externally caused and internally caused verbs.  This would require perhaps an ability to recognize patterns in meaning that I am not sure computers are able to do yet.

In Haspel's two chapters, we are introduced to many new linguistic concepts.  He discusses how the words we use in language can be grouped into categories based on their syntactic relationships.  Many “word-forms” we use are grouped together into a “paradigm” under one “lexeme”, which is a dictionary entry that encapsulates these many forms.  These words can be further broken down into morphemes and the study of how morphemes come together is called morphology.  An interesting point in this piece was the idea that morphology is more than just morpheme concatenation.  The example put forth by Haspel was in German.  In German, to pluralize some words, an umlaut is added to a vowel early in the word.  (Mutter -> Mütter).  In contract, in Turkish, almost all morphology is done through concatenation of prefixes and suffixes to create nouns, adjectives and verbs of similar semantic meaning from one stem.  

Slobin brings the distinction between languages beyond mere syntactical and morphological distinction.  He introduces the idea that we need to understand the “rhetorical style” of languages, in addition to the semantics and syntax, to really be able to understand the language.  A deep understanding of narrative style is essential in being able to successfully translate narrative stories between languages, especially the S and V languages he focuses on in the piece.