4.cuatro Efficiency
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 one to group (cluster 0 in solutions) containing the majority of relational adjectives on the gold standard. This is actually the really compact people depending on the clustering standards.
The fresh new dialogue is targeted on the latest team analyses that have three and four groups because the all of our foundation try about three groups (intensional, qualitative, and relational) and in addition we believe a total of five classes (very first kinds as well as polysemous categories: intensional-qualitative and you will qualitative-relational)
Some other group (dos inside service A, 1 in services B) provides the most qualitative adjectives regarding gold standard, and most of the intensional and you can IQ adjectives.
Adjectives that will be polysemous anywhere between a qualitative and you may good relational understanding (QR) is actually strewn because of most of the clusters, while they inform you a propensity to be ascribed towards relational party in the solution B (people 0).
The 5-method results are illustrated in the Table 6. To your one hand, the fresh desk implies that the five-method build found because of the clustering algorithm is quite similar to the 3-way design in Dining table 5. This means that the 3 groups in A beneficial and you can B has fundamentally already been replicated by the three earliest groups in C and you will D, correspondingly. At the same time, the distinctions between the formations gotten having fun with theoretic instead of POS provides become more noticeable on the four-method choices. Regarding lay-up of one’s test, we’d asked you to definitely people for every single category, including QR and IQ adjectives isolated from inside the a group of the own. This might be obviously maybe not borne call at Desk 6. That which we look for rather would be the fact (a) the new blended clusters persist and you will get stuffed with the fresh new clustering criterion (see clusters 0 inside the solution C and you may 0–one in provider D, which have a mixture of Q, QR, and you can Roentgen adjectives), and you will (b) a few extra quick groups manufactured (clusters 3 and you may 4 in alternatives) and no clear translation, suggesting the around three-means set-right up fits finest the dwelling uncovered by clustering algorithm.
On the conversation out-of Tables 5 and you will six we stop one the 3-ways clustering meets the mark class much better than the 5-means clustering, and this polysemous adjectives commonly identified as a special classification. These efficiency advise that acting polysemous adjectives in terms of additional, cutting-edge kinds isn’t an adequate means (we go back to this aspect after that).
Recall that people discussed theoretical and you may POS have to compare new formations acquired having fun with technically informed and you can theory-separate enjoys. Then function analysis, maybe not said right here to own space grounds, reveals a leading correlation within most descriptive attributes of options An excellent and B. 3 This features the brand new communications among them function representations having value to the clustering efficiency: Brand new POS has actually elicited as most discriminative from the clustering algorithm was accurately those people that match the latest theoretical keeps. That it correspondence teaches you the latest resemblance amongst the alternatives gotten with the 2 kinds of signal and also at once brings help into the introduce concept of the brand new theoretic enjoys.