The networks of KONECT are classified into categories, for instance social networks, interaction networks and rating networks.

• Affiliation networks are bipartite networks of people's inclusion in groups. In most cases, these are from social networks where people can join groups, but can also denote other types of affiliation, such as athletes being parts in teams. Group networks are bipartite, unweighted, and do not allow multiple links.
• Authorship networks are unweighted bipartite networks consisting of links between authors and their works. In some authorship networks such as that of scientific literature, works have typically a few authors, whereas works in other authorship networks may have many authors, as in Wikipedia articles. In an authorship network, vertices corresponding to works are added one by one with all their incident edges, and the link prediction problem is not meaningful. Therefore, we also consider author-author networks of co-authorships. In these networks, each edge represents a joint authorship, with the timestamp representing the publication date. Example of authorship networks are the Wikipedia user–article edit network, and the DBLP network of scientists and their publications.
• Co-occurrence networks represent the simultaneous appearance of items. Co-occurrence networks are unipartite, undirected and unweighted. Examples are the co-purchase network from Amazon, and the is-similar relationship of DBpedia.
• Communication networks contain edges that represent individual messages sent from one person to another. Communication networks are directed and may have multiple edges. Timestamps denote the date of a message. Communication networks may or may not allow loops. Examples of communication networks include email networks, Facebook wall posts and Twitter posts addressed using the “@name” notation.
• Feature networks are bipartite, and denote any kind of feature assigned to entities. Examples are the genre of songs and the occupation of people. Feature networks are unweighted and have edges that are not annotated with edge creation times.
• Folksonomies consist of tag assignments connecting a user, an item and a tag. For folksonomies, we follow the 3-bipartite approach and consider the three possible bipartite networks, i.e. the user–item, user–tag and item–tag networks. This allows us to apply methods for bipartite graphs to hypergraphs, which is not possible otherwise. Examples of folksonomies are Delicious for bookmarks, CiteULike for scientific publications and MovieLens for movies.
• Interaction networks represent collections of single events between entities. Most interaction networks have edges with timestamps. Examples are the interaction of proteins and other molecules in biological organisms and the Last.fm user–song listening network.
• Lexical networks are networks of words connected by various word–word relationships. For instance, links can denote semantic similarity, but also phonological similarity or other relations. Lexical networks are unipartite.
• Physical networks represent physically existing network structures. This includes physical computer networks and road networks. These networks result from an underlying two-dimensional geographical layout. Examples are road networks and the autonomous systems of the Internet.
• Rating networks consist of assessments given to entities by users, weighted by a rating value. Most rating networks are bipartite, when users rate items. A few rating networks are unipartite and directed, when users rate other users. Most rating networks are weighted; others are signed, when there is an explicit neutral rating value of zero. Examples are the Netflix movie ratings and Jester joke ratings.
• Reference networks consist of citations or hyperlinks between publications, patents or web pages. Reference networks are directed. In most reference networks, edges are created along with nodes. Therefore, the link prediction problem cannot be applied to them. An exception are hyperlink networks, where links can be added at any time. Examples are hyperlinks between pages on the World Wide Web and citations between scientific publications.
• Semantic networks are generic networks of entities connected by relationships. Our dataset collection contains a single semantic network, DBpedia, containing data extracted from the English Wikipedia, in which entities are individual lemmas and relationships are inferred from infoboxes.
• Social networks represent relations of friendship between persons. Some social networks have negative edges, which denote enmity. All social networks have simple or signed edges, since it is not possible to add another user multiple times to one's friend (or foe) list. In social networks, timestamps denote when a link was established. Examples are Facebook friendships, the Twitter follower relationship, and friends and foes on Slashdot.
• Text networks are a set of text documents connected to the words they contain. Text networks are bipartite, and have multiple edges, reflecting the fact that words can appear multiple times in a single document. Documents can range from individual tweets to full encyclopedia pages.