How a Word of a Text Selects the Related Words in a Human Association Network

According to tradition, experimentally obtained human associations are analyzed in themselves, without relation to other linguistic data. In rare cases, human associations are used as the norm to evaluate the performance of algorithms, which generate associations on the basis of text corpora. This chapter will describe a mechanical procedure to investigate how a word embedded in a text context may select associations in an experimentally built human association network. Each association produced in the experiment has a direction from stimulus to response. On the other hand, each association is based on the semantic relation between the two meanings, which has its own direction which is independent from the direction of associations. Therefore, we may treat the network as a directed or an undirected graph. The procedure described in this chapter uses both graph structures to produce a semantically consistent sub-graph. A comparison of the results shows that the procedure operates equally well on both graph structures. This procedure is able to distinguish those words in a text which enter into a direct semantic relationship with the stimulus word used in the experiment employed to create a network, and is able to separate those words of the text which enter into an indirect semantic relationship with the stimulus word.

 
Source
< Prev   CONTENTS   Source   Next >