Giovanni Semeraro | Università degli Studi di Bari (original) (raw)

Papers by Giovanni Semeraro

Research paper thumbnail of Reproducibility Analysis of Recommender Systems relying on Visual Features: traps, pitfalls, and countermeasures

Proceedings of the 17th ACM Conference on Recommender Systems

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Extracting Relations from Italian Wikipedia Using Self-Training

Proceedings of the Eighth Italian Conference on Computational Linguistics CliC-it 2021

Bookmarks Related papers MentionsView impact

Research paper thumbnail of A Framework for Holistic User Modeling Merging Heterogeneous Digital Footprints

Adjunct Publication of the 26th Conference on User Modeling, Adaptation and Personalization, 2018

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Word Embedding Techniques for Content-based Recommender Systems: An Empirical Evaluation

This work presents an empirical comparison among three widespread word embedding techniques as La... more This work presents an empirical comparison among three widespread word embedding techniques as Latent Semantic Indexing, Random Indexing and the more recent Word2Vec. Specifically, we employed these techniques to learn a lowdimensional vector space word representation and we exploited it to represent both items and user profiles in a content-based recommendation scenario. The performance of the techniques has been evaluated against two state-ofthe-art datasets, and experimental results provided good insights which pave the way to several future directions. 1. MOTIVATIONS AND METHODOLOGY Word Embedding techniques learn in a totally unsupervised way a low-dimensional vector space representation of words by analyzing their usage in (very) large corpora of textual documents. These approaches are recently gaining more and more attention, since they showed very good performance in a broad range of natural language processingrelated scenarios, ranging from sentiment analysis and machine tr...

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Myrror

Adjunct Publication of the 26th Conference on User Modeling, Adaptation and Personalization, 2018

In this paper we present Myrror, a platform that supports the creation of a unique representation... more In this paper we present Myrror, a platform that supports the creation of a unique representation of the user that encodes several facets such as her interests, activities, habits, mood, social connections and so on. Such a representation, that we called holistic user model, is based on the footprints the user spread on social networks and through personal devices. Specifically, our platform acquires personal data coming from several sources, such as Twitter, Facebook, Instagram, Android smartphones and FitBit wristbands, and merges all these information to infer high-level features and populate the facets of the model. Such holistic user models are made available to both the user itself and to third-party services. In the former case, data are shown through a visual interface to improve her self-awareness and her consciousness. In the latter, data are exposed to developers and new personalized services based on these richer user profiles can be created. In both cases, the user has ...

Bookmarks Related papers MentionsView impact

Research paper thumbnail of A Hybrid Recommendation Framework Exploiting Linked Open Data and Graph-based Features

Proceedings of the 25th Conference on User Modeling, Adaptation and Personalization, 2017

In this article we propose a hybrid recommendation framework based on classification algorithms a... more In this article we propose a hybrid recommendation framework based on classification algorithms as Random Forests and Naive Bayes. We fed the framework with several heterogeneous groups of features, and we investigate to what extent features gathered from the Linked Open Data (LOD) cloud (as the genre of a movie or the writer of a book)) as well as graph-based features calculated on the ground of the tripartite representation connecting users, items and properties in the LOD cloud impact on the overall accuracy of the recommendations. In the experimental session we evaluate the effectiveness of our framework on varying of different groups of features, an results show that both LOD-based and graph-based features positively affect the overall performance of the algorithm, especially in highly sparse recommendation scenarios.

Bookmarks Related papers MentionsView impact

Research paper thumbnail of DeCAT 2015 - Workshop on Deep Content Analytics Techniques for Personalized and Intelligent Services

Personal Information Management (PIM) research is challenging primarily due to the inherent natur... more Personal Information Management (PIM) research is challenging primarily due to the inherent nature of PIM. Studies have shown that people often adopt their own schemes when organising their personal collections, possibly because PIM tool-support is still lacking. In this paper we investigate the problem of automatic organisation of personal information into task-clusters by transparently exploiting the user’s behaviour while performing some tasks. We conduct a controlled experiment, with 22 participants, using three different task-execution strategies to gather clean data for our evaluation. We use our PiMx (PIM analytix) framework to analyse this data and understand better the issues associated with this problem. Based on this analysis, we then present the incremental density-based clustering algorithm, iDeTaCt, that is able to transparently generate task-clusters by exploiting document switching and revisitation. We evaluate the algorithm’s performance using the collected datasets...

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Recommender Systems in the Internet of Talking Things (IoTT)

In the Internet of Things, smart devices are connected to collect and to exchange data. In our vi... more In the Internet of Things, smart devices are connected to collect and to exchange data. In our vision, in the Internet of Talking Things, objects such as intelligent fridges will be able to communicate with humans to set up preferences and profiling options which allow a personalized usage of the object. In this paper, we present a recommender system implemented as a Telegram Bot, that can fit with the previous scenario. The system is a movie recommender which exploits the information available in the Linked Open Data (LOD) cloud for generating the recommendations and leading the conversation with the user. It can be easily seen as an intelligent component of a connected TV. ACM Reference format: Fedelucio Narducci, Marco de Gemmis, Pasquale Lops, Giovanni Semeraro. Copyrights held by the authors. Recommender Systems in the Internet of Talking Things (IoTT) . RecSys 2017 Posters, Como, Italy, August 27-31 (RecSys 2017 Poster Proceedings), 2 pages. 1 BACKGROUND AND MOTIVATIONS The ma...

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Exploiting Regression Trees as User Models for Intent-Aware Multi-attribute Diversity

Diversity in a recommendation list has been recognized as one of the key factors to increase user... more Diversity in a recommendation list has been recognized as one of the key factors to increase user’s satisfaction when interacting with a recommender system. Analogously to the modelling and exploitation of query intent in Information Retrieval adopted to improve diversity in search results, in this paper we focus on eliciting and using the prole of a user which is in turn exploited to represent her intents. The model is based on regression trees and is used to improve personalized diversication of the recommendation list in a multi-attribute setting. We tested the proposed approach and showed its eectiveness in two dierent domains, i.e. books and movies.

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Adaptive and Personalized Systems Based on Semantics

Semantics in Adaptive and Personalised Systems, 2019

In the introduction of this book, we have thoroughly discussed the importance of adaptive and per... more In the introduction of this book, we have thoroughly discussed the importance of adaptive and personalized systems in a broad range of applications. In particular, we have motivated the use of content-based information and textual data, and we have analyzed all the possible limitations of approaches based on keyword-based representation. In this chapter, we will focus on the application of semantics-aware representation techniques in recommender systems, user modeling, and social media analysis, and we will show how the exploitation of enhanced representation leads to an improvement of the results.

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Automatic Selection of Linked Open Data Features in Graph-based Recommender Systems

In this paper we compare several techniques to automatically feed a graph-based recommender syste... more In this paper we compare several techniques to automatically feed a graph-based recommender system with features extracted from the Linked Open Data (LOD) cloud. Specifically, we investigated whether the integration of LOD-based features can improve the e↵ectiveness of a graph-based recommender system and to what extent the choice of the features selection technique can influence the behavior of the algorithm by endogenously inducing a higher accuracy or a higher diversity. The experimental evaluation showed a clear correlation between the choice of the feature selection technique and the ability of the algorithm to maximize a specific evaluation metric. Moreover, our algorithm fed with LODbased features was able to overcome several state-of-the-art baselines: this confirmed the e↵ectiveness of our approach and suggested to further investigate this research line.

Bookmarks Related papers MentionsView impact

Research paper thumbnail of CBRecSys 2016. New Trends on Content-Based Recommender Systems: Proceedings of the 3rd Workshop on New Trends on Content-Based Recommender Systems co-located with 10th ACM Conference on Recommender Systems (RecSys 2016)

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Basics of Content Representation

Semantics in Adaptive and Personalised Systems, 2019

The importance of content-based features in intelligent information access systems as search engi... more The importance of content-based features in intelligent information access systems as search engines, information filtering tools, and recommender systems has been thoroughly discussed in the Introduction of this book. All the examples we have provided showed that textual data can be really useful to: (i) tackle some of the issues that affect data representation in intelligent systems and (ii) take the most out of the information that are today available on social networks and in personal devices, by leading to a more effective filtering of the information flow and to more satisfying recommendations.

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Improving the User Experience with a Conversational Recommender System

AI*IA 2018 – Advances in Artificial Intelligence, 2018

Chatbots are becoming more and more popular for several applications like customer care, health c... more Chatbots are becoming more and more popular for several applications like customer care, health care, medical diagnoses. Generally, they have an interaction with users based on natural language, buttons, or both. In this paper we study the user interaction with a content-based recommender system implemented as a Telegram chatbot. More specifically, we investigate on one hand what are the best strategies for reducing the cost of interaction for the users and, on the other hand how to improve their experience. Our chatbot is able to provide personalized recommendations in the movie domain and implements critiquing strategies for improving the recommendation accuracy as well. In a preliminary experimental evaluation, carried out through a user study, interesting results emerged.

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Modeling Community Behavior through Semantic Analysis of Social Data

Proceedings of the 2016 Conference on User Modeling Adaptation and Personalization, 2016

This paper presents the results of The Italian Hate Map, a research project aiming to monitor the... more This paper presents the results of The Italian Hate Map, a research project aiming to monitor the level of intolerance of the Italian country by mining the content posted on social networks. Within the project, a pipeline of algorithms for data extraction, semantic processing, sentiment analysis and content classification has been defined to process huge amounts of Tweets and to build a map of the most at-risk areas, thus identifying the Italian communities tending to have a more intolerant behavior. The outcomes resulting from the analysis of the maps confirmed the insight that the adoption of semantic content analysis techniques can be very useful to create value from the rough content available on the Web, and to go one step further in understanding very complex phenomena by modeling offline behavior of the communities on the ground of their online behavior on social networks.

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Context-aware graph-based recommendations exploiting Personalized PageRank

Knowledge-Based Systems, 2021

Bookmarks Related papers MentionsView impact

Research paper thumbnail of A Multi-criteria Recommender System Exploiting Aspect-based Sentiment Analysis of Users' Reviews

Proceedings of the Eleventh ACM Conference on Recommender Systems, 2017

In this paper we propose a multi-criteria recommender system based on collaborative filtering (CF... more In this paper we propose a multi-criteria recommender system based on collaborative filtering (CF) techniques, which exploits the information conveyed by users' reviews to provide a multi-faceted representation of users' interests. To this end, we exploited a framework for opinion mining and sentiment analysis, which automatically extracts relevant aspects and sentiment scores from users' reviews. As an example, in a restaurant recommendation scenario, the aspects may regard food quality, service, position, athmosphere of the place and so on. Such a multi-faceted representation of the user is used to feed a multi-criteria CF algorithm which predicts user interest in a particular item and provides her with recommendations. In the experimental session we evaluated the performance of the algorithm against several state-of-the-art baselines; Results confirmed the insight behind this work, since our approach was able to overcome both single-criteria recommendation algorithms as well as more sophisticated techniques based on matrix factorization.

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Combining text summarization and aspect-based sentiment analysis of users' reviews to justify recommendations

Proceedings of the 13th ACM Conference on Recommender Systems, 2019

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Tuning Personalized PageRank for Semantics-Aware Recommendations Based on Linked Open Data

The Semantic Web, 2017

Bookmarks Related papers MentionsView impact

Research paper thumbnail of An investigation on the user interaction modes of conversational recommender systems for the music domain

User Modeling and User-Adapted Interaction, 2019

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Reproducibility Analysis of Recommender Systems relying on Visual Features: traps, pitfalls, and countermeasures

Proceedings of the 17th ACM Conference on Recommender Systems

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Extracting Relations from Italian Wikipedia Using Self-Training

Proceedings of the Eighth Italian Conference on Computational Linguistics CliC-it 2021

Bookmarks Related papers MentionsView impact

Research paper thumbnail of A Framework for Holistic User Modeling Merging Heterogeneous Digital Footprints

Adjunct Publication of the 26th Conference on User Modeling, Adaptation and Personalization, 2018

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Word Embedding Techniques for Content-based Recommender Systems: An Empirical Evaluation

This work presents an empirical comparison among three widespread word embedding techniques as La... more This work presents an empirical comparison among three widespread word embedding techniques as Latent Semantic Indexing, Random Indexing and the more recent Word2Vec. Specifically, we employed these techniques to learn a lowdimensional vector space word representation and we exploited it to represent both items and user profiles in a content-based recommendation scenario. The performance of the techniques has been evaluated against two state-ofthe-art datasets, and experimental results provided good insights which pave the way to several future directions. 1. MOTIVATIONS AND METHODOLOGY Word Embedding techniques learn in a totally unsupervised way a low-dimensional vector space representation of words by analyzing their usage in (very) large corpora of textual documents. These approaches are recently gaining more and more attention, since they showed very good performance in a broad range of natural language processingrelated scenarios, ranging from sentiment analysis and machine tr...

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Myrror

Adjunct Publication of the 26th Conference on User Modeling, Adaptation and Personalization, 2018

In this paper we present Myrror, a platform that supports the creation of a unique representation... more In this paper we present Myrror, a platform that supports the creation of a unique representation of the user that encodes several facets such as her interests, activities, habits, mood, social connections and so on. Such a representation, that we called holistic user model, is based on the footprints the user spread on social networks and through personal devices. Specifically, our platform acquires personal data coming from several sources, such as Twitter, Facebook, Instagram, Android smartphones and FitBit wristbands, and merges all these information to infer high-level features and populate the facets of the model. Such holistic user models are made available to both the user itself and to third-party services. In the former case, data are shown through a visual interface to improve her self-awareness and her consciousness. In the latter, data are exposed to developers and new personalized services based on these richer user profiles can be created. In both cases, the user has ...

Bookmarks Related papers MentionsView impact

Research paper thumbnail of A Hybrid Recommendation Framework Exploiting Linked Open Data and Graph-based Features

Proceedings of the 25th Conference on User Modeling, Adaptation and Personalization, 2017

In this article we propose a hybrid recommendation framework based on classification algorithms a... more In this article we propose a hybrid recommendation framework based on classification algorithms as Random Forests and Naive Bayes. We fed the framework with several heterogeneous groups of features, and we investigate to what extent features gathered from the Linked Open Data (LOD) cloud (as the genre of a movie or the writer of a book)) as well as graph-based features calculated on the ground of the tripartite representation connecting users, items and properties in the LOD cloud impact on the overall accuracy of the recommendations. In the experimental session we evaluate the effectiveness of our framework on varying of different groups of features, an results show that both LOD-based and graph-based features positively affect the overall performance of the algorithm, especially in highly sparse recommendation scenarios.

Bookmarks Related papers MentionsView impact

Research paper thumbnail of DeCAT 2015 - Workshop on Deep Content Analytics Techniques for Personalized and Intelligent Services

Personal Information Management (PIM) research is challenging primarily due to the inherent natur... more Personal Information Management (PIM) research is challenging primarily due to the inherent nature of PIM. Studies have shown that people often adopt their own schemes when organising their personal collections, possibly because PIM tool-support is still lacking. In this paper we investigate the problem of automatic organisation of personal information into task-clusters by transparently exploiting the user’s behaviour while performing some tasks. We conduct a controlled experiment, with 22 participants, using three different task-execution strategies to gather clean data for our evaluation. We use our PiMx (PIM analytix) framework to analyse this data and understand better the issues associated with this problem. Based on this analysis, we then present the incremental density-based clustering algorithm, iDeTaCt, that is able to transparently generate task-clusters by exploiting document switching and revisitation. We evaluate the algorithm’s performance using the collected datasets...

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Recommender Systems in the Internet of Talking Things (IoTT)

In the Internet of Things, smart devices are connected to collect and to exchange data. In our vi... more In the Internet of Things, smart devices are connected to collect and to exchange data. In our vision, in the Internet of Talking Things, objects such as intelligent fridges will be able to communicate with humans to set up preferences and profiling options which allow a personalized usage of the object. In this paper, we present a recommender system implemented as a Telegram Bot, that can fit with the previous scenario. The system is a movie recommender which exploits the information available in the Linked Open Data (LOD) cloud for generating the recommendations and leading the conversation with the user. It can be easily seen as an intelligent component of a connected TV. ACM Reference format: Fedelucio Narducci, Marco de Gemmis, Pasquale Lops, Giovanni Semeraro. Copyrights held by the authors. Recommender Systems in the Internet of Talking Things (IoTT) . RecSys 2017 Posters, Como, Italy, August 27-31 (RecSys 2017 Poster Proceedings), 2 pages. 1 BACKGROUND AND MOTIVATIONS The ma...

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Exploiting Regression Trees as User Models for Intent-Aware Multi-attribute Diversity

Diversity in a recommendation list has been recognized as one of the key factors to increase user... more Diversity in a recommendation list has been recognized as one of the key factors to increase user’s satisfaction when interacting with a recommender system. Analogously to the modelling and exploitation of query intent in Information Retrieval adopted to improve diversity in search results, in this paper we focus on eliciting and using the prole of a user which is in turn exploited to represent her intents. The model is based on regression trees and is used to improve personalized diversication of the recommendation list in a multi-attribute setting. We tested the proposed approach and showed its eectiveness in two dierent domains, i.e. books and movies.

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Adaptive and Personalized Systems Based on Semantics

Semantics in Adaptive and Personalised Systems, 2019

In the introduction of this book, we have thoroughly discussed the importance of adaptive and per... more In the introduction of this book, we have thoroughly discussed the importance of adaptive and personalized systems in a broad range of applications. In particular, we have motivated the use of content-based information and textual data, and we have analyzed all the possible limitations of approaches based on keyword-based representation. In this chapter, we will focus on the application of semantics-aware representation techniques in recommender systems, user modeling, and social media analysis, and we will show how the exploitation of enhanced representation leads to an improvement of the results.

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Automatic Selection of Linked Open Data Features in Graph-based Recommender Systems

In this paper we compare several techniques to automatically feed a graph-based recommender syste... more In this paper we compare several techniques to automatically feed a graph-based recommender system with features extracted from the Linked Open Data (LOD) cloud. Specifically, we investigated whether the integration of LOD-based features can improve the e↵ectiveness of a graph-based recommender system and to what extent the choice of the features selection technique can influence the behavior of the algorithm by endogenously inducing a higher accuracy or a higher diversity. The experimental evaluation showed a clear correlation between the choice of the feature selection technique and the ability of the algorithm to maximize a specific evaluation metric. Moreover, our algorithm fed with LODbased features was able to overcome several state-of-the-art baselines: this confirmed the e↵ectiveness of our approach and suggested to further investigate this research line.

Bookmarks Related papers MentionsView impact

Research paper thumbnail of CBRecSys 2016. New Trends on Content-Based Recommender Systems: Proceedings of the 3rd Workshop on New Trends on Content-Based Recommender Systems co-located with 10th ACM Conference on Recommender Systems (RecSys 2016)

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Basics of Content Representation

Semantics in Adaptive and Personalised Systems, 2019

The importance of content-based features in intelligent information access systems as search engi... more The importance of content-based features in intelligent information access systems as search engines, information filtering tools, and recommender systems has been thoroughly discussed in the Introduction of this book. All the examples we have provided showed that textual data can be really useful to: (i) tackle some of the issues that affect data representation in intelligent systems and (ii) take the most out of the information that are today available on social networks and in personal devices, by leading to a more effective filtering of the information flow and to more satisfying recommendations.

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Improving the User Experience with a Conversational Recommender System

AI*IA 2018 – Advances in Artificial Intelligence, 2018

Chatbots are becoming more and more popular for several applications like customer care, health c... more Chatbots are becoming more and more popular for several applications like customer care, health care, medical diagnoses. Generally, they have an interaction with users based on natural language, buttons, or both. In this paper we study the user interaction with a content-based recommender system implemented as a Telegram chatbot. More specifically, we investigate on one hand what are the best strategies for reducing the cost of interaction for the users and, on the other hand how to improve their experience. Our chatbot is able to provide personalized recommendations in the movie domain and implements critiquing strategies for improving the recommendation accuracy as well. In a preliminary experimental evaluation, carried out through a user study, interesting results emerged.

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Modeling Community Behavior through Semantic Analysis of Social Data

Proceedings of the 2016 Conference on User Modeling Adaptation and Personalization, 2016

This paper presents the results of The Italian Hate Map, a research project aiming to monitor the... more This paper presents the results of The Italian Hate Map, a research project aiming to monitor the level of intolerance of the Italian country by mining the content posted on social networks. Within the project, a pipeline of algorithms for data extraction, semantic processing, sentiment analysis and content classification has been defined to process huge amounts of Tweets and to build a map of the most at-risk areas, thus identifying the Italian communities tending to have a more intolerant behavior. The outcomes resulting from the analysis of the maps confirmed the insight that the adoption of semantic content analysis techniques can be very useful to create value from the rough content available on the Web, and to go one step further in understanding very complex phenomena by modeling offline behavior of the communities on the ground of their online behavior on social networks.

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Context-aware graph-based recommendations exploiting Personalized PageRank

Knowledge-Based Systems, 2021

Bookmarks Related papers MentionsView impact

Research paper thumbnail of A Multi-criteria Recommender System Exploiting Aspect-based Sentiment Analysis of Users' Reviews

Proceedings of the Eleventh ACM Conference on Recommender Systems, 2017

In this paper we propose a multi-criteria recommender system based on collaborative filtering (CF... more In this paper we propose a multi-criteria recommender system based on collaborative filtering (CF) techniques, which exploits the information conveyed by users' reviews to provide a multi-faceted representation of users' interests. To this end, we exploited a framework for opinion mining and sentiment analysis, which automatically extracts relevant aspects and sentiment scores from users' reviews. As an example, in a restaurant recommendation scenario, the aspects may regard food quality, service, position, athmosphere of the place and so on. Such a multi-faceted representation of the user is used to feed a multi-criteria CF algorithm which predicts user interest in a particular item and provides her with recommendations. In the experimental session we evaluated the performance of the algorithm against several state-of-the-art baselines; Results confirmed the insight behind this work, since our approach was able to overcome both single-criteria recommendation algorithms as well as more sophisticated techniques based on matrix factorization.

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Combining text summarization and aspect-based sentiment analysis of users' reviews to justify recommendations

Proceedings of the 13th ACM Conference on Recommender Systems, 2019

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Tuning Personalized PageRank for Semantics-Aware Recommendations Based on Linked Open Data

The Semantic Web, 2017

Bookmarks Related papers MentionsView impact

Research paper thumbnail of An investigation on the user interaction modes of conversational recommender systems for the music domain

User Modeling and User-Adapted Interaction, 2019

Bookmarks Related papers MentionsView impact