2016 Annual Meeting
Till Poppels & Roger Levy (UCSD) Structure-sensitive noise inference: undoing exchange errors / 12
The father bought his son for a bicycle. literal non-literal Was something bought for the son? 100% 0% “No” “Yes” Till Poppels & Roger Levy (UCSD) Structure-sensitive noise inference: undoing exchange errors / 12
The father bought his son for a bicycle. literal non-literal Was something bought for the son? 33% 66% “No” “Yes” Till Poppels & Roger Levy (UCSD) Structure-sensitive noise inference: undoing exchange errors / 12
The father bought The cook baked his son for a bicycle. Lucy for a cake. Was something Was something The apprentice bought for the son? baked for Lucy? fetched a hammer the carpenter. 47% 53% Was something 33% 66% fetched for the carpenter? literal non-literal literal non-literal 33% 67% The bartender poured the customer The man ordered his for a drink. literal non-literal girlfriend for some Was something champagne. The charity built a poured for the house the hurricane Was something customer? victim. ordered for the champagne? Was something 79% 67% 21% built for the hurricane victim? 33% literal non-literal 75% literal non-literal 25% literal non-literal Till Poppels & Roger Levy (UCSD) Structure-sensitive noise inference: undoing exchange errors / 12
𝑸 𝑵 𝑱 Listener error Speaker error Environmental noise Anderson (1990); Levy (2008) Till Poppels & Roger Levy (UCSD) Structure-sensitive noise inference: undoing exchange errors / 12
𝑸 𝑵 𝑱 ∝ 𝑸 𝑱 𝑵 𝑸(𝑵) Plausibility Listener error Speaker error Environmental noise Till Poppels & Roger Levy (UCSD) Structure-sensitive noise inference: undoing exchange errors / 12
𝑸 𝑵 𝑱 ∝ 𝑸 𝑱 𝑵 𝑸(𝑵) Was something bought for the son? DO/PO benefactives Yes No 1 insertion 𝑄 𝐽 𝑁 = 1 𝑄 𝐽 𝑁 = 1 𝑄(𝐽|𝑁) = 𝑄(𝑗𝑜𝑡: 𝑔𝑝𝑠) The father bought his son for a bicycle. The father bought his son a bicycle. Did the dryer shrink something? Transitive/Intransitive The father bought a bicycle for his son. The father bought a bicycle his son. Yes No 1 deletion 𝑄 𝐽 𝑁 = 1 𝑄 𝐽 𝑁 = 1 1 deletion 𝑄(𝐽|𝑁) = 𝑄(𝑒𝑓𝑚: 𝑗𝑜𝑡𝑗𝑒𝑓) 𝑄 𝐽 𝑁 = 1 𝑄 𝐽 𝑁 = 1 𝑄(𝐽|𝑁) = 𝑄(𝑒𝑓𝑚: 𝑔𝑝𝑠) The t-shirt shrank the dryer. The t-shirt shrank inside the dryer. Did the girl kick something? Active/Passive The dryer shrank the t-shirt. The dryer shrank inside the t-shirt. Yes No 2 insertions 𝑄 𝐽 𝑁 = 1 𝑄 𝐽 𝑁 = 1 1 insertion 𝑄(𝐽|𝑁) = 𝑄 𝑗𝑜𝑡: 𝑥𝑏𝑡 𝑄(𝑗𝑜𝑡: 𝑐𝑧) 𝑄 𝐽 𝑁 = 1 𝑄 𝐽 𝑁 = 1 𝑄(𝐽|𝑁) = 𝑄(𝑗𝑜𝑡: 𝑗𝑜𝑡𝑗𝑒𝑓) The girl was kicked by the ball. The girl kicked the ball The ball was kicked by the girl. The ball kicked the girl. 2 deletions 𝑄 𝐽 𝑁 = 1 𝑄 𝐽 𝑁 = 1 𝑄 𝐽 𝑁 = 𝑄 𝑒𝑓𝑚: 𝑥𝑏𝑡 𝑄(𝑒𝑓𝑚: 𝑐𝑧) Till Poppels & Roger Levy (UCSD) Structure-sensitive noise inference: undoing exchange errors / 12
100% % literal responses implausible implausible implausible plausible plausible plausible 0% Active/Passive Transitive/Intransitive DO/PO Till Poppels & Roger Levy (UCSD) Structure-sensitive noise inference: undoing exchange errors / 12
100% % literal responses implausible plausible plausible plausible implausible implausible Insert/delete: Insert/delete: Insert/delete: “by” and “was” “for” 1 preposition 0% Active/Passive Transitive/Intransitive DO/PO from to The package fell to the table from the floor. The package fell to the table from the floor. Till Poppels & Roger Levy (UCSD) Structure-sensitive noise inference: undoing exchange errors Structure-sensitive noise inference: undoing exchange errors / 12
The package [ VP fell [ PP to the table] [ PP from the floor]]. The package fell to the table from the floor. cf. “spoonerisms” (e.g. MacKay, 1970) W aste the t erm T aste the w erm F ighting a l iar L ighting a f ire B attle ships and c ruisers C attle ships and b ruisers B usy D ean D izzy b ean Till Poppels & Roger Levy (UCSD) Structure-sensitive noise inference: undoing exchange errors Structure-sensitive noise inference: undoing exchange errors / 12
The package [ VP fell …] 3% 97% 5% implausible plausible non-canonical [ PP to the table] [ PP from the floor] [ PP to the floor] [ PP from the table] canonical [ PP from the floor] [ PP to the table] [ PP from the table] [ PP to the floor] 95% Plausibility Norming Canonicality Norming [ PP from …] [ PP to …] 95% - 5% [ PP to …] [ PP from …] [ PP with …] [ PP about …] 80% - 20% [ PP about …] [ PP with …] [ PP to …] [ PP about …] 81% - 19% [ PP about …] [ PP to …] [ PP from …] [ PP about …] 67% - 33% [ PP about …] [ PP from …] [ PP for …] [ PP in …] 51% - 49% [ PP in …] [ PP for …] [ PP in …] [ PP at …] 58% - 42% [ PP at …] [ PP in …] [ PP to …] [ PP for …] 97% - 3% [ PP for …] [ PP to …] response ~ plausibility + canonicality + (1 + plausibility + canonicality || item) + (1 + plausibility + canonicality || subject) Till Poppels & Roger Levy (UCSD) Structure-sensitive noise inference: undoing exchange errors / 12
𝑸 𝑵 𝑱 ∝ 𝑸 𝑱 𝑵 𝑸(𝑵) Predictions 1. Noise inference whenever prior probabilities permit 2. Additive effects of Plausibility plausibility and canonicality Canonicality Listener error Speaker error Environmental noise Till Poppels & Roger Levy (UCSD) Structure-sensitive noise inference: undoing exchange errors / 12
𝑸 𝑵 𝑱 ∝ 𝑸 𝑱 𝑵 𝑸(𝑵) Predictions 1. Noise inference whenever prior probabilities permit 2. Additive effects of Plausibility plausibility and canonicality Canonicality Listener error Speaker error Environmental noise Till Poppels & Roger Levy (UCSD) Structure-sensitive noise inference: undoing exchange errors / 12
𝑸 𝑵 𝑱 ∝ 𝑸 𝑱 𝑵 𝑸(𝑵) * Plausibility * Canonicality Listener error Speaker error 100% Environmental noise * % literal responses non-canonical canonical plausible plausible Predictions non-canonical implausible canonical implausible 1. Noise inference whenever prior probabilities permit 2. Additive effects of plausibility and canonicality 0% Till Poppels & Roger Levy (UCSD) Structure-sensitive noise inference: undoing exchange errors / 12
[ PP to the table] [ PP from the floor] [ PP to the floor] [ PP from the table] [ PP from the floor] [ PP to the table] [ PP from the table] [ PP to the floor] 100% % literal responses implausible plausible plausible plausible implausible implausible 0% Active/Passive Transitive/Intransitive DO/PO Till Poppels & Roger Levy (UCSD) Structure-sensitive noise inference: undoing exchange errors / 12
Swapping nouns in active/passive? [ PP to the table] [ PP from the floor] [ PP to the floor] [ PP from the table] The ball was kicked by the girl. [ PP from the floor] [ PP to the table] [ PP from the table] [ PP to the floor] What’s the difference? Function vs. content words? * Opposite pattern in spoonerisms. 100% (MacKay, 1987) Interim Summary Adjuncts vs. Complements? * Possible, but speculative. We know that prepositions can be % literal responses exchanged. non-canonical We don’t know that nouns can’t implausible plausible plausible plausible canonical plausible plausible be exchanged. non-canonical implausible canonical implausible implausible Why exchanges don’t occur in active/passive sentences is an open question. implausible 0% Active/Passive Exchanges Transitive/Intransitive DO/PO Till Poppels & Roger Levy (UCSD) Structure-sensitive noise inference: undoing exchange errors / 12
1. Do people REALLY consider all conceivable interpretations during language comprehension? “That’s not the right kind of process , intuitively.” “That’s not a computationally feasible mechanism.” Till Poppels & Roger Levy (UCSD) Structure-sensitive noise inference: undoing exchange errors Structure-sensitive noise inference: undoing exchange errors / 12
1. Do people REALLY consider all conceivable interpretations during language comprehension? “That’s not the right kind of process, intuitively .” “That’s not a computationally feasible mechanism.” Till Poppels & Roger Levy (UCSD) Structure-sensitive noise inference: undoing exchange errors Structure-sensitive noise inference: undoing exchange errors / 12
Marr (1982) “In order to understand bird flight , we have to understand aerodynamics ; only then do 1. Do people REALLY consider the structure of feathers and all conceivable interpretations the different shapes of birds’ during language comprehension? wings make sense. “That’s not the right kind of process, intuitively .” “That’s not a computationally feasible mechanism .” 2. If we open the door to non-literal interpretations, does that mean that anything goes? What about: “The cat is on the mat.” Till Poppels & Roger Levy (UCSD) Structure-sensitive noise inference: undoing exchange errors Structure-sensitive noise inference: undoing exchange errors / 12
Recommend
More recommend
