Mauro Dragoni

Multidimensional relevance: Prioritized aggregation in a personalized Information Retrieval setting

A new model for aggregating multiple criteria evaluations for relevance assessment is proposed. An Information Retrieval context is considered, where relevance is modeled as a multidimensional property of documents. The usefulness and effectiveness of such a model are demonstrated by means of a case study on personalized Information Retrieval with multi-criteria relevance. The following criteria are considered to estimate document relevance: aboutness, coverage, appropriateness, and reliability. The originality of this approach lies in the aggregation of the considered criteria in a prioritized way, by considering the existence of a prioritization relationship over the criteria. Such a prioritization is modeled by making the weights associated to a criterion dependent upon the satisfaction of the higher-priority criteria. This way, it is possible to take into account the fact that the weight of a less important criterion should be proportional to the satisfaction degree of the more important criterion. Experimental evaluations are also reported.

Propagating and Aggregating Fuzzy Polarities for Concept-Level Sentiment Analysis

AbstractAn emerging field within sentiment analysis concerns the investigation about how sentiment polarities associated with concepts have to be adapted with respect to the different domains in which they are used. In this paper, we explore the use of fuzzy logic for modeling concept polarities, and the uncertainty associated with them, with respect to different domains. The approach is based on the use of a knowledge graph built by combining two linguistic resources, namely WordNet and SenticNet. Such a knowledge graph is then exploited by a graph-propagation algorithm that propagates sentiment information learned from labeled datasets. The system implementing the proposed approach has been evaluated on the Blitzer dataset. The results demonstrate its viability in real-world cases.

Multidimensional Relevance: A New Aggregation Criterion

In this paper, a new model for aggregating multiple criteria evaluations for relevance assessment is proposed. An information retrieval context is considered, where relevance is modelled as a multidimensional property of documents. In the paper, the proposed aggregation operator is applied to define a model for personalized Information Retrieval (IR), in which four criteria are considered in order to assess document relevance: aboutness , coverage , appropriateness and reliability . The originality of this approach lies in the aggregation of the considered criteria in a prioritized way, by considering the existence of a prioritization relationship over the criteria. Such a prioritization is modeled by making the weights associated with a criterion dependent upon the satisfaction of the higher-priority criteria. This way, it is possible to take into account the fact that the weight of a less important criterion should be proportional to the satisfaction degree of the more important criterion. In the paper, some preliminary experimental results are also reported.

A conceptual representation of documents and queries for information retrieval systems by using light ontologies

This article presents a vector space model approach to representing documents and queries, based on concepts instead of terms and using WordNet as a light ontology. Such representation reduces information overlap with respect to classic semantic expansion techniques. Experiments carried out on the MuchMore benchmark and on the TREC-7 and TREC-8 Ad-Hoc collections demonstrate the effectiveness of the proposed approach.

A Fuzzy System for Concept-Level Sentiment Analysis

An emerging field within Sentiment Analysis concerns the investigation about how sentiment concepts have to be adapted with respect to the different domains in which they are used. In the context of the Concept-Level Sentiment Analysis Challenge, we presented a system whose aims are twofold: (i) the implementation of a learning approach able to model fuzzy functions used for building the relationships graph representing the appropriateness between sentiment concepts and different domains (Task 1); and (ii) the development of a semantic resource based on the connection between an extended version of WordNet, SenticNet, and ConceptNet, that has been used both for extracting concepts (Task 2) and for classifying sentences within specific domains (Task 3).

Supervised Opinion Frames Detection with RAID

Most systems for opinion analysis focus on the classification of opinion polarities and rarely consider the task of identifying the different elements and relations forming an opinion frame. In this paper, we present RAID, a tool featuring a processing pipeline for the extraction of opinion frames from text with their opinion expressions, holders, targets and polarities. RAID leverages a lexical, syntactic and semantic analysis of text, using several NLP tools such as dependency parsing, semantic role labelling, named entity recognition and word sense disambiguation. In addition, linguistic resources such as SenticNet and the MPQA Subjectivity Lexicon are used both to locate opinions in the text and to classify their polarities according to a fuzzy model that combines the sentiment values of different opinion words. RAID was evaluated on three different datasets and is released as open source software under the GPLv3 license.

An ontological representation of documents and queries for information retrieval systems

This paper presents a vector space model approach, for representing documents and queries, using concepts instead of terms and WordNet as a light ontology. This way, information overlap is reduced with respect to the classic semantic expansion techniques. Experiments carried out on the MuchMore benchmark showed the effectiveness of the approach.

A novel similarity-based crossover for artificial neural network evolution

This work presents an evolutionary approach for the optimization of neural networks design, based on the joint evolution of the topology and the connection weights, providing a novel similarity-based crossover that aims to overcome one of the major problems of this operator, known as the permutation problem. The approach has been implemented and applied to two benchmark classification problems in machine learning, and the experimental results, compared to those obtained by other works in the literature, show how it can produce compact neural networks with a satisfactory generalization capability.

A Knowledge-Based Approach for Aspect-Based Opinion Mining

In the last decade, the focus of the Opinion Mining field moved to detection of the pairs “aspect-polarity” instead of limiting approaches in the computation of the general polarity of a text. In this work, we propose an aspect-based opinion mining system based on the use of semantic resources for the extraction of the aspects from a text and for the computation of their polarities. The proposed system participated at the third edition of the Semantic Sentiment Analysis (SSA) challenge took place during ESWC 2016 achieving the runner-up place in the Task #2 concerning the aspect-based sentiment analysis. Moreover, a further evaluation performed on the SemEval 2015 benchmarks demonstrated the feasibility of the proposed approach.

An Information Retrieval-Based System for Multi-domain Sentiment Analysis

This paper describes the SHELLFBK system that participated in ESWC 2015 Sentiment Analysis challenge. Our system takes a supervised approach that builds on techniques from information retrieval. The algorithm populates an inverted index with pseudo-documents that encode dependency parse relationships extracted from the sentences in the training set. Each record stored in the index is annotated with the polarity and domain of the sentence it represents; this way, it is possible to have a more fine-grained representation of the learnt sentiment information. When the polarity of a new sentence has to be computed, the new sentence is converted to a query and a two-steps computation is performed: firstly, a domain is assigned to the sentence by comparing the sentence content with domain contextual information learnt during the training phase, and, secondly, once the domain is assigned to the sentence, the polarity is computed and assigned to the new sentence. Preliminary results on an in-vitro test case demonstrated promising results.