QP Fest 2019: The Sixteenth Annual QP Fest

Please join us to hear presentations of recently completed qualifying papers!

Abstracts

 Scott Borgeson
 Estonian is famous for its three-way length contrast in consonants and vowels. In this talk, I will examine the origin and structure of the longest of these, the so-called "Q3". I will examine several existing accounts--in particular, the monosyllabic-foot analysis of Prince (1980) and the trimoraic-syllable analysis of Hayes (1989) and Bye (1997)--and will show that they are not able to adequately explain the properties and distribution of Q3.

 I will therefore propose a hybrid approach: the definition of Q3 is a syllable that is both trimoraic and exhausts its foot. I will show that their trimoraicity results from vowel deletion and shortening followed by long-distance compensatory lengthening, and will argue that their syllable-as-foot structure results from their trimoraicity: because trimoraic syllables are ternary-branching, they must be reparsed so that their third mora is attached directly to the foot node, forcing the second syllable out.

 Michael Hahn
Cross-linguistic universals of word order, and the search for explanations for these, have been a topic of intense study in linguistics. A prominent line of research has argued that these universals arise because languages are optimized for human communication and language processing, with a variety of proposals for specific criteria that natural language might be optimized for. We test these explanations by formalizing and computationally implementing three such optimization criteria: dependency locality, predictability, and parseability. We implement each criterion as an objective function to be applied to a word order grammar: a probabilistic grammar that generates word order given a fixed dependency tree topology. Using tree structures from 51 diverse languages in the Universal Dependencies corpora, we find that the optimized word order grammars recover most of the Greenberg word order correlation universals. Our results provide direct computational evidence that universals of word order can be explained as optimizing ease of human language processing.

 Dora Demszky
We provide an NLP framework to uncover four linguistic dimensions of political polarization in social media: topic choice, framing, affect and illocutionary force. We quantify these aspects with existing lexical methods, and propose clustering of tweet embeddings as a means to identify salient topics for analysis across events; human evaluations show that our approach generates more cohesive topics than traditional LDA-based models. We apply our methods to study 4.4M tweets on 21 mass shootings. We provide evidence that the discussion of these events is highly polarized politically and that this polarization is primarily driven by partisan differences in framing rather than topic choice. We identify framing devices, such as grounding and the contrasting use of the terms “terrorist” and “crazy”, that contribute to polarization. Results pertaining to topic choice, affect and illocutionary force suggest that Republicans focus more on the shooter and event-specific facts (news), while Democrats focus more on the victims and call for policy changes. Our work contributes to a deeper understanding of the way group divisions manifest in language and to computational methods for studying them.

 Erika Petersen
 In this talk, I propose an intervention-based account of parasitic gap (pg) behavior inside temporal adverbial clauses in Spanish. I advance the hypothesis that there is a relation between  the left edge of the adverbial clause and the parasitic gap (pg), the latter a null pronoun which  requires wh-licensing by matrix C. This relation is Person-feature-based, driven by a Person  feature with the EPP property in the head of the adverbial clause (Adv). The proposal unifies  two heretofore separate observations about the syntax of parasitic gaps in Spanish: first, the  nominal identity of the pg; second, the fact that pg is licensed in non-finite, but not in finite,  adverbial clauses. Previous approaches to PGC in Spanish offer a subjacency-based account of the finiteness contrast, while stipulating that the categorial restriction of the pg-antecedent  is due to the pronominal nature of the pg (García Mayo 1993, García Mayo and Kempchinsky 1994). The analysis proposed here derives these characteristics from a single analytical commitment, namely the Person probe on Adv. In finite temporal clauses, the Person feature of  finite TP (TP+FIN) or the DP in Spec,TP+FIN intervenes in the probing process, precluding the licensing of the pg.

Yiwei Luo
 Focusing on the case of English deadjectival adverbs like awfully, insanely and remarkably, this paper argues that the semantic bleaching of a literal adverb (awfully behaved) into an intensifying adverb (awfully nice) is a process that lexicalizes a scale structural component of the adjective root as the amount the bleached adverb intensi- fies. We argue that, with the exception of adverbs of magnitude (greatly, abundantly), semantic dissimilation is the mechanism behind bleaching. To test these proposals, we develop robust methods based on relationships between historical word embeddings and n-gram collocations to model the bleaching of a set of 250 English deadjectival intensifiers. We show that our theory of bleaching accounts for systematic variation across adverbs in terms of the kinds of adverbs that bleach, the semantic facets of their literal meaning that bleach, how much they bleach, and how rapidly they bleach.

Chantal Gratton

TBA