50-something years of work on collocations

Abstract

This paper explores ways in which research into collocation should be improved. After a discussion of the parameters underlying the notion of 'collocation' , the paper has three main parts. First, I argue that corpus linguistics would benefit from taking more seriously the understudied fact that collocations are not necessarily symmetric, as most association measures imply. Also, I introduce an association measure from the associative learning literature that can identify asymmetric collocations and show that it can also distinguish collocations with high and low association strengths well. Second, I summarize some advantages of this measure and brainstorm about ways in which it can help re-examine previous studies as well as support further applications. Finally, I adopt a broader perspective and discuss a variety of ways in which all association measures -directional or not -in corpus linguistics should be improved in order for us to obtain better and more reliable results.

Collocations in context

“…Gries 2013). Unlike traditional metrics such as MI, MI2, MI3, loglikelihood, T-score, etc., which in a symmetrical collocation window produce the same association value for node-collocate, collocate-node combinations (i.e.…”

“…On the other hand, love is much more strongly and exclusively connected with the noun affair; when the word affair appears in text, there is a large probability that the preceding word is love. In addition to the three criteria discussed above, Gries (2013) points out three other criteria that should be considered: (iv) directionality, (v) dispersion and (vi) type-token distribution among collocates.…”

“…Baker et al 2008, Xiao & McEnery 2006, Syanova & Schmitt 2008. However, more than fifty years into the research on collocations, many of the lessons learned from this research have yet to be systematically evaluated and fully implemented in the tools that corpus linguists use (see Gries 2013 for an important discussion about how research into collocations can be improved). Traditionally, three criteria for identifying collocations have been proposed.…”

“…Yet the traditional association measures do not capture this difference because the majority of those commonly used in corpus linguistics are symmetrical measures. 1 Gries (2013) therefore suggests using Delta P as a measure that takes directionality into account by producing two different values of collocational strength for any pair of words. Dispersion is the distribution of the node and the collocates in the corpus (cf.…”

“…This is a relatively even distribution compared to another potential collocate agape (a Greek term for nonromantic love), which occurs 9 times but only in 2 texts. Finally, Gries (2013) raises type-token distribution as a desirable criterion which has been partly operationalized through the lexical gravity G measure in Daudaravičius & Marcinkevičienė (2004). This criterion takes into account not only the strength of a given collocational relationship (say between love and affair), but also the level of competition for the slot(s) around the node word from other collocate types.…”