cuatro.cuatro Results
The contingency tables of the clustering results with three clusters are depicted in Table 5. Part A of the table depicts the solution obtained with theoretical features, while Part B represents the solution obtained with POS features. Rows are gold standard classes and columns are clusters, labeled with the cluster number provided by the algorithm. The ordering of the cluster numbers corresponds to the quality of the cluster, measured in terms of the clustering criterion (see Equation (2)), 0 representing the cluster with the highest quality. In each cell Cij of Table 5, the number of adjectives of class i that are assigned to cluster j by the algorithm is given. The largest value for each class is highlighted (see gray cells).
First model: Three-way solution contingency tables for theoretical and POS features. Rows are gold standard classes, columns are clusters. Row TotalGS shows the number of Gold Standard lemmata and row Totalcl the total number of lemmata contained in each cluster. Note that the column labeled Total represents the row sum for each part (as the number of items per class is identical).
There is certainly that team (class 0 in both solutions) that features most relational adjectives regarding the standard. Here is the really compact people according to the clustering requirement.
This new discussion concentrates on this new cluster analyses having three and you may five clusters as our foundation is about three classes (intensional, qualitative, and relational) so we thought a total of five groups (basic categories together with polysemous kinds: intensional-qualitative and you will qualitative-relational)
Several other cluster (dos inside the service A, one in service B) has got the greater part of qualitative adjectives on standard, together with all of the intensional and you may IQ adjectives.
Adjectives which can be polysemous ranging from a beneficial qualitative and you will an effective relational studying (QR) are thrown as a consequence of most of the groups, even though they inform you a tendency to getting ascribed on the relational group in the services B (class 0).
The 5-method email address details are represented for the Desk six. On one-hand, the brand new desk means that the five-ways build found because of the clustering algorithm is very similar to the three-way build for the Desk 5. As a result the three clusters in Good and you may B possess basically already been duplicated from the three very first groups inside C and you will D, correspondingly. Simultaneously, the differences between the formations gotten using theoretical in place of POS has actually become more visible regarding the four-ways options. Regarding the lay-upwards of one’s check out, we transgenderdate had questioned one to class for each and every classification, and additionally QR and you will IQ adjectives isolated in the a cluster of the own. This can be demonstrably perhaps not borne call at Table six. Everything we get a hold of as an alternative would be the fact (a) the latest combined groups persist and you can rating filled up with the new clustering traditional (find groups 0 for the solution C and you may 0–1 in solution D, with a mixture of Q, QR, and Roentgen adjectives), and (b) several additional short clusters are formulated (clusters 3 and 4 in options) with no obvious translation, suggesting your three-ways set-upwards matches top the dwelling bare of the clustering algorithm.
About conversation away from Tables 5 and you will six i conclude one to the three-ways clustering matches the target category much better than the five-ways clustering, which polysemous adjectives aren’t identified as an alternative classification. Such performance advise that modeling polysemous adjectives with respect to most, cutting-edge groups is not a sufficient method (we go back to this aspect after that).
Bear in mind that people outlined theoretic and you may POS enjoys examine the newest formations gotten using technically advised and theory-independent enjoys. After that function data, not said here getting space reasons, suggests a leading correlation between your really detailed attributes of alternatives A beneficial and you can B. step three So it shows the brand new interaction between the two element representations having respect on clustering show: The newest POS have elicited as most discriminative by clustering formula is truthfully people who correspond to the new theoretical features. So it communications explains the similarity between the choices gotten towards two types of symbol as well as the same time will bring support to your present concept of the fresh new theoretic possess.