In this paper we propose a Twitter recommender based on a semantic description of users' interests. To express interests we use friendship information, which is readily available in users' profiles, not only in Twitter but in the majority of Social Networks, thus presenting substantial advantage in terms of computational complexity with respect to methods based on content mining. To obtain a synthetic representation of interests, we first identify, for each user, his/her topical friends, i.e. Those friends for which there is at least one associated page in Wikipedia. Then, category vectors at different levels of generality are created for each user, exploiting WiBi, the Wikipedia Bitaxonomy. The recommendation task is accomplished in three steps: First, we classify users as belonging to a target community, which is again inferred from highly popular topical friends, such as Democrat Party, Starbucks, etc. Then, for each community, we identify the relevant item sets (efficiently computed using a minimum, rather than maximum, support count). Our items are either topical friends or higher level Wikipedia categories, and the generality level is a tunable parameter. Finally, we fine-tune recommendations for each community, using semantic association rules.

In this paper we propose a Twitter recommender based on a semantic description of users' interests. To express interests we use friendship information, which is readily available in users' profiles, not only in Twitter but in the majority of Social Networks, thus presenting substantial advantage in terms of computational complexity with respect to methods based on content mining. To obtain a synthetic representation of interests, we first identify, for each user, his/her topical friends, i.e. Those friends for which there is at least one associated page in Wikipedia. Then, category vectors at different levels of generality are created for each user, exploiting WiBi, the Wikipedia Bitaxonomy. The recommendation task is accomplished in three steps: First, we classify users as belonging to a target community, which is again inferred from highly popular topical friends, such as Democrat Party, Starbucks, etc. Then, for each community, we identify the relevant item sets (efficiently computed using a minimum, rather than maximum, support count). Our items are either topical friends or higher level Wikipedia categories, and the generality level is a tunable parameter. Finally, we fine-tune recommendations for each community, using semantic association rules.

A semantic recommender for micro-blog users

STILO, GIOVANNI
2015-01-01

Abstract

In this paper we propose a Twitter recommender based on a semantic description of users' interests. To express interests we use friendship information, which is readily available in users' profiles, not only in Twitter but in the majority of Social Networks, thus presenting substantial advantage in terms of computational complexity with respect to methods based on content mining. To obtain a synthetic representation of interests, we first identify, for each user, his/her topical friends, i.e. Those friends for which there is at least one associated page in Wikipedia. Then, category vectors at different levels of generality are created for each user, exploiting WiBi, the Wikipedia Bitaxonomy. The recommendation task is accomplished in three steps: First, we classify users as belonging to a target community, which is again inferred from highly popular topical friends, such as Democrat Party, Starbucks, etc. Then, for each community, we identify the relevant item sets (efficiently computed using a minimum, rather than maximum, support count). Our items are either topical friends or higher level Wikipedia categories, and the generality level is a tunable parameter. Finally, we fine-tune recommendations for each community, using semantic association rules.
2015
9781467372787
In this paper we propose a Twitter recommender based on a semantic description of users' interests. To express interests we use friendship information, which is readily available in users' profiles, not only in Twitter but in the majority of Social Networks, thus presenting substantial advantage in terms of computational complexity with respect to methods based on content mining. To obtain a synthetic representation of interests, we first identify, for each user, his/her topical friends, i.e. Those friends for which there is at least one associated page in Wikipedia. Then, category vectors at different levels of generality are created for each user, exploiting WiBi, the Wikipedia Bitaxonomy. The recommendation task is accomplished in three steps: First, we classify users as belonging to a target community, which is again inferred from highly popular topical friends, such as Democrat Party, Starbucks, etc. Then, for each community, we identify the relevant item sets (efficiently computed using a minimum, rather than maximum, support count). Our items are either topical friends or higher level Wikipedia categories, and the generality level is a tunable parameter. Finally, we fine-tune recommendations for each community, using semantic association rules.
File in questo prodotto:
Non ci sono file associati a questo prodotto.
Pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11697/133252
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 4
  • ???jsp.display-item.citation.isi??? 4
social impact