Felice Ferrara

The role of tags for recommendation: A survey

Social tagging is an innovative and powerful mechanism introduced with Web 2.0: it shifts the task of classifying resources from a reduced set of knowledge engineers to the wide set of Web users. Users of social tagging systems define personal classifications which can be used by other peers for browsing available resources. However, due to the absence of rules for managing the tagging process, and to the lack of predefined schemas or structures for inserting metadata and relationships among tags, current user generated classifications dop not produce sound taxonomies. This is a strong limitation which prevents an effective and informed resource sharing. For this reason researchers are modeling innovative recommender systems capable to better support tagging, browsing, and searching for new resources. This paper is a survey which discusses the role of tags in recommender systems: starting from social tagging systems, we analyze various techniques for suggesting content and we introduce the approaches exploited for proposing tags for classifying resources, considering both personalized and not-personalized recommendation.

A Keyphrase-Based Paper Recommender System

Current digital libraries suffer from the information overload problem which prevents an effective access to knowledge. This is particularly true for scientific digital libraries where a growing amount of scientific articles can be explored by users with different needs, backgrounds, and interests. Recommender systems can tackle this limitation by filtering resources according to specific user needs. This paper introduces a content-based recommendation approach for enhancing the access to scientific digital libraries where a keyphrase extraction module is used to produce a rich description of both content of papers and user interests.

Supporting Personalized User Concept Spaces and Recommendations for a Publication Sharing System

Current publication sharing systems weakly support creation and personalization of customized user concept spaces. Focusing the attention on the user, SharingPapers, the adaptive publication sharing system proposed in this paper, allows users to organize documents in flexible and dynamic concept spaces; to merge their concept map with a social network connecting people involved in the domain of interest; to support knowledge expansion generating adaptive recommendations. SharingPapers presents a multi-agent architecture and proposes a new way of representing user profiles, their evolution and views of them.

On Social Semantic Relations for Recommending Tags and Resources Using Folksonomies

Social tagging is an innovative and powerful mechanism introduced by social Web: it shifts the task of classifying resources from a reduced set of knowledge engineers to the wide set of Web users. However, due to the lack of rules for managing the tagging process and of predefined schemas or structures for inserting metadata and relationships among tags, current user generated classifications do not produce sound taxonomies. This is a strong limitation which prevents an effective and informed resource sharing; for this reason the most recent research in this area is dedicated to empower the social perspective applying semantic approaches in order to support tagging, browsing, searching, and adaptive personalization in innovative recommender systems. This paper proposes a survey on existing recommender systems, discussing how they extract social semantic relations (i.e. relations among users, resources and tags of a folksonomy), and how they utilize this knowledge for recommending tags and resources.

Extracting Keyphrases from Web Pages

Social tagging systems allow people to classify Web resources by using a set of freely chosen terms commonly called tags. However, by shifting the classification task from a set of experts to a larger and untrained set of people, the results of the classification are not accurate. The lack of control and guidelines generates noisy tags (i.e. tags without a clear semantic) which lower the precision of the user generated classifications. In order to face this limitation several tools have been proposed in the literature for suggesting to the users tags which properly describe a given resource. On the other hand we propose to suggest n-grams (named keyphrases) by following the idea that sequences of two/three terms can better face potential ambiguities. More specifically, in this work, we identify a set of features which characterize n-grams adequate for describing meaningful aspects reported in the Web pages. By means of these features, we developed a mechanism which can support people when classifying Web pages by automatically suggesting meaningful keyphrases.

Personalized Access to Scientific Publications: from Recommendation to Explanation

Several recommender systems have been proposed in the literature for adaptively suggesting useful references to researchers with different interests. However, in order to access the knowledge contained in the recommended papers, the users need to read the publications for identifying the potentially interesting concepts. In this work we propose to overcome this limitation by utilizing a more semantic approach where concepts are extracted from the papers for generating and explaining the recommendations. By showing the concepts used to find the recommended articles, users can have a preliminary idea about the filtered publications, can understand the reasons why the papers were suggested and they can also provide new feedback about the relevance of the concepts utilized for generating the recommendations.

Evaluating the results of methods for computing semantic relatedness

The semantic relatedness between two concepts is a measure that quantifies the extent to which two concepts are semantically related. Due to the growing interest of researchers in areas such as Semantic Web, Information Retrieval and NLP, various approaches have been proposed in the literature for automatically computing the semantic relatedness. However, despite the growing number of proposed approaches, there are still significant criticalities in evaluating the results returned by different semantic relatedness methods. The limitations of the state of the art evaluation mechanisms prevent an effective evaluation and several works in the literature emphasize that the exploited approaches are rather inconsistent. In this paper we describe the limitations of the mechanisms used for evaluating the results of semantic relatedness methods. By taking into account these limitations, we propose a new methodology and new resources for comparing in an effective way different semantic relatedness approaches.

Extracting and exploiting topics of interests from social tagging systems

Users of social tagging systems spontaneously annotate resources providing, in this way, useful information about their interests. A collaborative filtering recommender system can use this feedback in order to identify people and resources more strictly related to a specific topic of interest. Such a collaborative filtering approach can compute similarities among tags in order to select resources associated to tags relevant for a specific interest of the user. Several research works try to infer these similarities by evaluating co-occurrences of tags over the entire set of annotated resources discarding, in this way, information about the personal classification provided by users. This paper, on the other hand, proposes an approach aimed at observing only the set of annotations of a single user in order to identify his topic of interests and to produce personalized recommendations. More specifically, following the idea that each user may have several distinct interests and people may share just some of these interests, our approach adaptively filters and combines the feedback of users according to a specific topic of interest of a user.

Improving collaborative filtering in social tagging systems

User-based Collaborative Filtering (CF) systems generate recommendations for a specific user by combining feedback (i.e. information about what is relevant for a user) provided by a set of people similar to that user. In these system the similarity among people is computed by taking into account the set of shared resources. However, there are several application domains, such as social tagging systems, where each user may have several different Topic of Interests (ToIs). In these cases, two users could share only some interests and, therefore, only a part of the feedback should be considered for producing recommendations. Focusing on social tagging systems, we propose here a novel approach to detect ToIs in the collection of the bookmarks of a user. Given a specific ToI, we adaptively identify similar people (i.e., sharing the same ToI) and select only the resources relevant to the specific ToI.

RES: A Personalized Filtering Tool for CiteSeerX Queries Based on Keyphrase Extraction

Finding satisfactory scientific literature is still a very time-consuming task. In the last decade several tools have been proposed to approach this task, however only few of them actually analyse the whole document in order to select and present it to the user and even less tools offer any kind of explanation of why a given item was retrieved/recommended. The main goal of this demonstration is to present the RES system, a tool intended to overcome the limitations of traditional recommender and personalized information retrieval systems by exploiting a more semantic approach where concepts are extracted from the papers in order to generate and then explain the recommendation. RES acts like a personalized interface for the well-known CiteSeerX system, filtering and presenting query results accordingly to individual user’s interests.