![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
Recent invited speakers
Feb. 20th, 2012 07:29 pmThere's been some interesting talks lately, but today was the first one in a while that made me think "I should blog about that". But since I also would like records of the other talks, I'm going to start trying to summarise the ones I found interesting.
Julie Van Dyke - on language processing using cue retrieval
* Language processing is really heavily dependent on working memory.
* But we don't actually know much about working memory (eg. how much of it we have), so to be safe let's assume that a hypothetical person can only remember the last item they heard/read. This isn't as insane as it sounds - computer models have indicated that processing can do pretty well even with such an impoverished working memory. Everything that isn't in active working memory is absorbed passively and can be called upon (albeit not as easily)
* So let's consider a few hypothetical sentences: 1) the book ripped 2) the book recommended by the editor ripped 3) the book from Michigan by Anne Rice that was recommended by the editor ripped. How does a listener tell if 'ripped' forms a grammatical sentence with 'the book'? There are a few ways: they could search forwards or backwards through the sentence, in which case you would expect processing times to reflect the amount of material between "the book" and "ripped". Or you could do cue-based retrieval, where you filter the sentence for words that have the features you're looking for, in which case you wouldn't expect there to be significant time difference in retrieval. As the name of the talk might suggest, people use cue-based retrieval.
* So now we have a model where we store words as bundles of semantic/phonological/etc features and then retrieve them by using those features. But what if the sentence has several possible words that have the features you're looking for? In that case, retrieval might get blocked due to interference from the other items. This, according to Julie Van Dyke, is why people forget. (I don't know whether she meant in general or when processing sentences. Hopefully the latter)
* And the main difference between people who are good at processing (eg. fast readers) vs those who aren't, is almost entirely based on how detailed your representations are. Because if your word representations are super detailed with lots of features, it's easier to zero in on them. And, good news, the main factor in how good your representations are (after controlling for IQ and a bunch of other bothersome details) is practice. So if you suck at reading, all you need to do to fix it is read more.
Julie Van Dyke - on language processing using cue retrieval
* Language processing is really heavily dependent on working memory.
* But we don't actually know much about working memory (eg. how much of it we have), so to be safe let's assume that a hypothetical person can only remember the last item they heard/read. This isn't as insane as it sounds - computer models have indicated that processing can do pretty well even with such an impoverished working memory. Everything that isn't in active working memory is absorbed passively and can be called upon (albeit not as easily)
* So let's consider a few hypothetical sentences: 1) the book ripped 2) the book recommended by the editor ripped 3) the book from Michigan by Anne Rice that was recommended by the editor ripped. How does a listener tell if 'ripped' forms a grammatical sentence with 'the book'? There are a few ways: they could search forwards or backwards through the sentence, in which case you would expect processing times to reflect the amount of material between "the book" and "ripped". Or you could do cue-based retrieval, where you filter the sentence for words that have the features you're looking for, in which case you wouldn't expect there to be significant time difference in retrieval. As the name of the talk might suggest, people use cue-based retrieval.
* So now we have a model where we store words as bundles of semantic/phonological/etc features and then retrieve them by using those features. But what if the sentence has several possible words that have the features you're looking for? In that case, retrieval might get blocked due to interference from the other items. This, according to Julie Van Dyke, is why people forget. (I don't know whether she meant in general or when processing sentences. Hopefully the latter)
* And the main difference between people who are good at processing (eg. fast readers) vs those who aren't, is almost entirely based on how detailed your representations are. Because if your word representations are super detailed with lots of features, it's easier to zero in on them. And, good news, the main factor in how good your representations are (after controlling for IQ and a bunch of other bothersome details) is practice. So if you suck at reading, all you need to do to fix it is read more.
