University of Stuttgart > Institute for Natural Language Processing > StatNLP group > Lukas Michelbacher > The AAM-experiment info page
Last changed: Tue 01 Nov 2011 06:24:48 PM CET

Asymmetry in Corpus-Derived and Human Word Associations (Web Appendix)

This pages contains additional information about to the elicitation experiment that was carried out for the publication.

Contents:

Instructions

Dear Participant,

Please read the instructions carefully. Since this is an experiment about English it is vital that you only continue if you are a native speaker of English. Thank you.

What you have to do for this experiment is to type in words. You'll be presented a word pair where either the first or the second word has been blanked out. Your task is to fill in the blanks with as many words as you can think of.

It is important that you only give words that would appear right after or right before the displayed word in normal speech (depending on the position of the blank line). The experiment is not a Free Association experiment. In a Free Association experiment you are presented a cue word and the goal is to give words that come to your mind after seeing the cue word. For example 'boy -> girl' or 'food -> drink'. This is not what this experiment is about. There may be overlaps between the two kinds of experiments but here the goal is to give answers so that the blank line and the word that is already there make up a fixed expression. For example 'boy -> scout' or 'food -> court'.

Try to think of as many words as possible for each answer. It often helps to imagine the words in several different contexts or to say it out loud. Let your mind wander - as long as you are sure that the answers you give are actually being used by English native speakers. Give all the words that you can think of but try not to spend too much time on a single question. if you can't think of anything, press the 'next' button. But don't overuse it ;-).

Please put commas between the words you type in. That will make it easier to automatiocally process your results.

For convenience, there's also a 'back' button to go back to previous questions.

Have fun!

top

Subject responses

stimulus stimulus id # subjects subject id comma separated list of responses
_ silos 3127 94 119 feed, wheat, grain, pig, storage
alarming _ 3091 94 33 rate, noise, statistics, fact, increase, decrease

The file containing the response data has five colums. See the table above for a brief overview. The first column contains the stimulus as it was presented to the subjects. The underscore marks the blank the subjects had to fill out. The second column contains a unique stimulus id. Column three indicates how many subjects responded to the stimulus. Column four contains a unique subject id for inter-subject comparisons. The last column contains all the responses a subject gave to the stimulus. None of the responses have been removed except for cases where a subjects commented on the stimulus or the like. Some minor spelling corrections have been made and the spelling was changed to British English to conform with the corpus data used in our study. All responses have been converted to lower case. There are two versions available for download:

In the first version, every response was run through a lemmatizer.

top

Word pair list

top