Search

The tasks of grouping, classifying, and clustering data are widely encountered nowadays in different areas of activity. These tasks need to be solved in librarianship and even in social networks. One of the complex and important tasks is a grouping of normative and reference information. Solving this problem will allow us to perform complex operations with the normativereference information: building a hierarchical structure from scratch, adding new entities or groups to the existing, and combining several lists of normative-reference information independently of the original hierarchical structure presence. This article includesdescription of using DBSCAN as an agglomerative-iterative clusterization algorithm. The iterative part of this algorithm is necessary forfull entity list clusterization on every hierarchical level. Clustering metrics such as adjusted Rand index, Jaccard index, Foulkes-Mallows index are considered. A new metric based on the previously mentioned metrics has been proposed. A combination of Word2Vec and TF-IDF algorithms is used to convert the textual names of objects to numerical form.

Improving the interfaces in molecular graphics applications, making them more natural and easy to use, is an important task, given the current complexity of the displayed objects and of modeling operations. Clicking near an atom center is the usual method of atom selection. However, this method has certain disadvantages when working with images composed of different atomic representations such as sticks, CPK, or dotted surfaces. We propose another technique allowing the user to obtain the correct answer when he or she clicks on any element of the atom image.