Mental Lexicon Flashcards
In Elman’s view of the mental lexicon:
A. The meaning of a word is accessed by retrieving it from the lexicon entry.
B. A word is represented as a list of attributes associated with the word in a kind of dictionary.
C. The lexicon primarily stores syntactic and semantic associations of words.
D. The meaning of a word arises from the mental states that are produced by the word stimulus.
D. The meaning of a word arises from the mental states that are produced by the word stimulus.
Which of the following IS NOT consistent with Elman’s model of the mental lexicon?
A. Word access is seen as more dynamic than passive.
B. The lexicon consists of a process that is evoked by words, and unfolds over time.
C. The lexicon contains sets of attributes linked to each word in long term memory that are retrieved when words are recognized.
D. Words are more like “operators” than “operands” during word processing.
C. The lexicon contains sets of attributes linked to each word in long term memory that are retrieved when words are recognized.
Which of the following IS NOT A CORRECT description of Elman’s Simple Recurrent Network (SRN)?
A. The context units represent the hidden unit activations at the time of input (t).
B. The SRN is an example of a connectionist model.
C. The SRN was designed to model the dynamic changes that unfold over time in the state of a system.
D. The SRN is capable of “learning” from previous activation states.
A. The context units represent the hidden unit activations at the time of input (t).
Elman describes the use of SRNs to learn to predict words in sentences. Which of the following is one reason he gave for why prediction might be relevant to language?
A. Because it accelerates instrumental learning.
B. Because prediction in the form of expectancy generation plays a role in language comprehension.
C. Because the reward system responds to violations of expectations.
D. Because word order follows a highly variable pattern in sentences.
B. Because prediction in the form of expectancy generation plays a role in language comprehension.
After training on a corpus of sentences, the pattern of word activations present in the hidden units of an SRN reflected:
A. The distinction between nouns and verbs, but not between animate and inanimate nouns
B. The distinction between animate and inanimate nouns, but not the distinction between nouns and verbs.
C. Only word order distinctions, since no semantic information was provided to the SRN during training.
D. The distinction between nouns and verbs as well as between animate and inanimate nouns.
D. The distinction between nouns and verbs as well as between animate and inanimate nouns.
In symbolic linguistic theories, a lexicon entry is a “type” rather than a “token”, which means:
A. an abstract representation of the meaning of a word, rather than the actual word stimulus
B. a word root rather than the other forms and tenses of a word
C. a group of words that are semantically related, rather than a specific word
D. a word category, such as animate nouns or inanimate nouns, rather than an individual word
A. an abstract representation of the meaning of a word, rather than the actual word stimulus
Elman suggests that new categories, or concepts, within the mental lexicon, might be discoverable from:
A. Changing the weights assigned to the context units of the SRN.
B. Using different kinds of sentences to train the SRN.
C. Changing the viewing perspective on the high-dimensional state space generated by the SRN.
D. Adding more category information to the input provided to the SRN for training.
C. Changing the viewing perspective on the high-dimensional state space generated by the SRN.
When Elman says that SRNs exhibit generalization, he means:
A. That predictions made by SRNs are only to very general categories of word “types”.
B. That SRNs exhibit an increasing rate of learning during training, as is seen in the “vocabulary burst” of human toddlers.
C. That late in training only a few instances of a new word are needed for SRNs to represent them in the correct category space and to predict them in the correct contexts.
D. That learning in SRNs is initially slow, because there is no category structure yet to draw on.
C. That late in training only a few instances of a new word are needed for SRNs to represent them in the correct category space and to predict them in the correct contexts.
What Elman is talking about in the section on semantic combination is how SRNs:
A. distinguish “tokens” from “types”
B. distinguish the meanings of a word in different phrase and sentence contexts
C. distinguish the different probabilities of particular words following other ones
D. distinguish the meaning of one word from a related one in the same category
B. distinguish the meanings of a word in different phrase and sentence contexts
