Nuova ricerca


Ricercatore t.d. art. 24 c. 3 lett. B presso: Dipartimento di Economia "Marco Biagi"

Home | Didattica |


2020 - BLAST2: An Efficient Technique for Loose Schema Information Extraction from Heterogeneous Big Data Sources [Articolo su rivista]

We present BLAST2 a novel technique to efficiently extract loose schema information, i.e., metadata that can serve as a surrogate of the schema alignment task within the Entity Resolution (ER) process — to identify records that refer to the same real-world entity — when integrating multiple, heterogeneous and voluminous data sources. The loose schema information is exploited for reducing the overall complexity of ER, whose naïve solution would imply O(n^2) comparisons, where is the number of entity representations involved in the process and can be extracted by both structured and unstructured data sources. BLAST2 is completely unsupervised yet able to achieve almost the same precision and recall of supervised state-of-the-art schema alignment techniques when employed for Entity Resolution tasks, as shown in our experimental evaluation performed on two real-world data sets (composed of 7 and 10 data sources, respectively).

2020 - Dagger: A Data (not code) Debugger [Relazione in Atti di Convegno]
Kindi Rezig, El; Cao, Lei; Simonini, Giovanni; Schoemans, Maxime; Madden, Samuel; Tang, Nan; Ouzzani, Mourad; Stonebraker:, Michael

With the democratization of data science libraries and frame- works, most data scientists manage and generate their data analytics pipelines using a collection of scripts (e.g., Python, R). This marks a shift from traditional applications that communicate back and forth with a DBMS that stores and manages the application data. While code debuggers have reached impressive maturity over the past decades, they fall short in assisting users to explore data-driven what-if sce- narios (e.g., split the training set into two and build two ML models). Those scenarios, while doable programmati- cally, are a substantial burden for users to manage them- selves. Dagger (Data Debugger) is an end-to-end data de- bugger that abstracts key data-centric primitives to enable users to quickly identify and mitigate data-related problems in a given pipeline. Dagger was motivated by a series of interviews we conducted with data scientists across several organizations. A preliminary version of Dagger has been in- corporated into Data Civilizer 2.0 to help physicians at the Massachusetts General Hospital process complex pipelines.

2020 - Entity resolution on camera records without machine learning [Relazione in Atti di Convegno]
Zecchini, L.; Simonini, G.; Bergamaschi, S.

This paper reports the runner-up solution to the ACM SIGMOD 2020 programming contest, whose target was to identify the specifications (i.e., records) collected across 24 e-commerce data sources that refer to the same real-world entities. First, we investigate the machine learning (ML) approach, but surprisingly find that existing state-of-the-art ML-based methods fall short in such a context-not reaching 0.49 F-score. Then, we propose an efficient solution that exploits annotated lists and regular expressions generated by humans that reaches a 0.99 F-score. In our experience, our approach was not more expensive than the dataset labeling of match/non-match pairs required by ML-based methods, in terms of human efforts.

2020 - JedAI^3 : beyond batch, blocking-based Entity Resolution [Relazione in Atti di Convegno]
Papadakis, George; Tsekouras, Leonidas; Thanos, Emmanouil; Pittaras, Nikiforos; Simonini, Giovanni; Skoutas, Dimitrios; Isaris, Paul; Giannakopoulos, George; Palpanas, Themistoklis; Koubarakis, Manolis

2020 - RulER: Scaling Up Record-level Matching Rules [Relazione in Atti di Convegno]
Gagliardelli, Luca; Simonini, Giovanni; Bergamaschi, Sonia

2020 - Scaling up Record-level Matching Rules [Relazione in Atti di Convegno]
Gagliardelli, L.; Simonini, G.; Bergamaschi, S.

Record-level matching rules are chains of similarity join pred-icates on multiple attributes employed to join records that refer to the same real-world object when an explicit foreign key is not available on the data sets at hand. They are widely employed by data scientists and practitioners that work with data lakes, open data, and data in the wild. In this work we present a novel technique that allows to efficiently exe-cute record-level matching rules on parallel and distributed systems and demonstrate its efficiency on a real-wold data set.

2020 - Three-dimensional Entity Resolution with JedAI [Articolo su rivista]
Papadakis, G.; Mandilaras, G.; Gagliardelli, L.; Simonini, G.; Thanos, E.; Giannakopoulos, G.; Bergamaschi, S.; Palpanas, T.; Koubarakis, M.

Entity Resolution (ER) is the task of detecting different entity profiles that describe the same real-world objects. To facilitate its execution, we have developed JedAI, an open-source system that puts together a series of state-of-the-art ER techniques that have been proposed and examined independently, targeting parts of the ER end-to-end pipeline. This is a unique approach, as no other ER tool brings together so many established techniques. Instead, most ER tools merely convey a few techniques, those primarily developed by their creators. In addition to democratizing ER techniques, JedAI goes beyond the other ER tools by offering a series of unique characteristics: (i) It allows for building and benchmarking millions of ER pipelines. (ii) It is the only ER system that applies seamlessly to any combination of structured and/or semi-structured data. (iii) It constitutes the only ER system that runs seamlessly both on stand-alone computers and clusters of computers — through the parallel implementation of all algorithms in Apache Spark. (iv) It supports two different end-to-end workflows for carrying out batch ER (i.e., budget-agnostic), a schema-agnostic one based on blocks, and a schema-based one relying on similarity joins. (v) It adapts both end-to-end workflows to budget-aware (i.e., progressive) ER. We present in detail all features of JedAI, stressing the core characteristics that enhance its usability, and boost its versatility and effectiveness. We also compare it to the state-of-the-art in the field, qualitatively and quantitatively, demonstrating its state-of-the-art performance over a variety of large-scale datasets from different domains. The central repository of the JedAI's code base is here: A video demonstrating the JedAI's Web application is available here:

2019 - Computing inter-document similarity with Context Semantic Analysis [Articolo su rivista]
Beneventano, Domenico; Benedetti, Fabio; Bergamaschi, Sonia; Simonini, Giovanni

We propose a novel knowledge-based technique for inter-document similarity computation, called Context Semantic Analysis (CSA). Several specialized approaches built on top of specific knowledge base (e.g. Wikipedia) exist in literature, but CSA differs from them because it is designed to be portable to any RDF knowledge base. Our technique relies on a generic RDF knowledge base (e.g. DBpedia and Wikidata) to extract from it a contextual graph and a semantic contextual vector able to represent the context of a document. We show how CSA exploits such Semantic Context Vector to compute inter-document similarity effectively. Moreover, we show how CSA can be effectively applied in the Information Retrieval domain. Experimental results show that our general technique outperforms baselines built on top of traditional methods, and achieves a performance similar to the ones built on top of specific knowledge bases.

2019 - Data civilizer 2.0: A holistic framework for data preparation and analytics [Relazione in Atti di Convegno]
Rezig, E. K.; Cao, L.; Stonebraker, M.; Simonini, G.; Tao, W.; Madden, S.; Ouzzani, M.; Tang, N.; Elmagarmid, A. K.

Data scientists spend over 80% of their time (1) parameter-tuning machine learning models and (2) iterating between data cleaning and machine learning model execution. While there are existing efforts to support the first requirement, there is currently no integrated workflow system that couples data cleaning and machine learning development. The previous version of Data Civilizer was geared towards data cleaning and discovery using a set of pre-defined tools. In this paper, we introduce Data Civilizer 2.0, an end-to-end workflow system satisfying both requirements. In addition, this system also supports a sophisticated data debugger and a workflow visualization system. In this demo, we will show how we used Data Civilizer 2.0 to help scientists at the Massachusetts General Hospital build their cleaning and machine learning pipeline on their 30TB brain activity dataset.

2019 - Entity Resolution and Data Fusion: An Integrated Approach [Relazione in Atti di Convegno]

2019 - Scaling entity resolution: A loosely schema-aware approach [Articolo su rivista]
Simonini, Giovanni; Gagliardelli, Luca; Bergamaschi, Sonia; Jagadish, H. V.

In big data sources, real-world entities are typically represented with a variety of schemata and formats (e.g., relational records, JSON objects, etc.). Different profiles (i.e., representations) of an entity often contain redundant and/or inconsistent information. Thus identifying which profiles refer to the same entity is a fundamental task (called Entity Resolution) to unleash the value of big data. The naïve all-pairs comparison solution is impractical on large data, hence blocking methods are employed to partition a profile collection into (possibly overlapping) blocks and limit the comparisons to profiles that appear in the same block together. Meta-blocking is the task of restructuring a block collection, removing superfluous comparisons. Existing meta-blocking approaches rely exclusively on schema-agnostic features, under the assumption that handling the schema variety of big data does not pay-off for such a task. In this paper, we demonstrate how “loose” schema information (i.e., statistics collected directly from the data) can be exploited to enhance the quality of the blocks in a holistic loosely schema-aware (meta-)blocking approach that can be used to speed up your favorite Entity Resolution algorithm. We call it Blast (Blocking with Loosely-Aware Schema Techniques). We show how Blast can automatically extract the loose schema information by adopting an LSH-based step for efficiently handling volume and schema heterogeneity of the data. Furthermore, we introduce a novel meta-blocking algorithm that can be employed to efficiently execute Blast on MapReduce-like systems (such as Apache Spark). Finally, we experimentally demonstrate, on real-world datasets, how Blast outperforms the state-of-the-art (meta-)blocking approaches.

2019 - SparkER: Scaling Entity Resolution in Spark [Relazione in Atti di Convegno]
Gagliardelli, Luca; Simonini, Giovanni; Beneventano, Domenico; Bergamaschi, Sonia

We present SparkER, an ER tool that can scale practitioners’ favorite ER algorithms. SparkER has been devised to take full ad- vantage of parallel and distributed computation as well (running on top of Apache Spark). The first SparkER version was focused on the blocking step and implements both schema-agnostic and Blast meta-blocking approaches (i.e. the state-of-the-art ones); a GUI for SparkER, to let non-expert users to use it in an unsupervised mode, was developed. The new version of SparkER to be shown in this demo, extends significantly the tool. Entity matching and Entity Clustering modules have been added. Moreover, in addition to the completely unsupervised mode of the first version, a supervised mode has been added. The user can be assisted in supervising the entire process and in injecting his knowledge in order to achieve the best result. During the demonstration, attendees will be shown how SparkER can significantly help in devising and debugging ER algorithms.

2018 - BigDedup: a Big Data Integration toolkit for Duplicate Detection in Industrial Scenarios [Relazione in Atti di Convegno]
Gagliardelli, Luca; Zhu, Song; Simonini, Giovanni; Bergamaschi, Sonia

Duplicate detection aims to identify different records in data sources that refers to the same real-world entity. It is a fundamental task for: item catalogs fusion, customer databases integration, fraud detection, and more. In this work we present BigDedup, a toolkit able to detect duplicate records on Big Data sources in an efficient manner. BigDedup makes available the state-of-the-art duplicate detection techniques on Apache Spark, a modern framework for distributed computing in Big Data scenarios. It can be used in two different ways: (i) through a simple graphic interface that permit the user to process structured and unstructured data in a fast and effective way; (ii) as a library that provides different components that can be easily extended and customized. In the paper we show how to use BigDedup and its usefulness through some industrial examples.

2018 - Enhancing big data exploration with faceted browsing [Relazione in Atti di Convegno]
Bergamaschi, Sonia; Zhu, Song; Simonini, Giovanni

With the modern information technologies, data availability is increasing at formidable speed giving raise to the Big Data challenge (Bergamaschi, 2014). As a matter of fact, Big Data analysis now drives every aspect of modern so- ciety, such as: manufacturing, retail, financial services, etc., (Labrinidis & Jagadish, 2012). In this scenario, we need to rethink advanced and efficient human-computer-interaction to be able to handling huge amount of data. In fact, one of the most valuable means to make sense of Big Data, to most peo- ple, is data visualization. As a matter of fact, data visualization may guide decision-making and become a powerful tool to convey information in all data analysis tasks. However, to be actually actionable, data visualization tools should allow the right amount of interactivity and to be easy to use, under- standable, meaningful, and approachable. In this article, we present a new approach to visualize and explore a huge amount of data. In particular, the novelty of our approach is to enhance the faceted browsing search in Apache Solr∗ (a widely used enterprise search platform) by exploiting Bayesian networks, supporting the user in the explo- ration of the data. We show how the proposed Bayesian suggestion algorithm (Cooper & Herskovits, 1991) be a key ingredient in a Big Data scenario, where a query can generate too many results that the user cannot handle. Our pro- posed solution aim to select best results, which together with the result-path, chosen by the user by means of multi-faceted querying and faceted navigation, can be a valuable support for both Big Data exploration and visualization. In the fellowing, we introduce the faceted browsing technique, then we de- scribe how it can be enhanced exploiting Bayesian networks.

2018 - Enhancing Loosely Schema-aware Entity Resolution with User Interaction [Relazione in Atti di Convegno]
Simonini, Giovanni; Gagliardelli, Luca; Zhu, Song; Bergamaschi, Sonia

Entity Resolution (ER) is a fundamental task of data integration: it identifies different representations (i.e., profiles) of the same real-world entity in databases. To compare all possible profile pairs through an ER algorithm has a quadratic complexity. Blocking is commonly employed to avoid that: profiles are grouped into blocks according to some features, and ER is performed only for entities of the same block. Yet, devising blocking criteria and ER algorithms for data with highly schema heterogeneity is a difficult and error-prone task calling for automatic methods and debugging tools. In our previous work, we presented Blast, an ER system that can scale practitioners’ favorite Entity Resolution algorithms. In current version, Blast has been devised to take full advantage of parallel and distributed computation as well (running on top of Apache Spark). It implements the state-of-the-art unsuper- vised blocking method based on automatically extracted loose schema information. It comes with a GUI, which allows: (i) to visualize, understand, and (optionally) manually modify the loose schema information automatically extracted (i.e., injecting user’s knowledge in the system); (ii) to retrieve resolved entities through a free-text search box, and to visualize the process that lead to that result (i.e., the provenance). Experimental results on real-world datasets show that these two functionalities can significantly enhance Entity Resolution results.

2018 - How improve Set Similarity Join based on prefix approach in distributed environment [Relazione in Atti di Convegno]
Zhu, Song; Gagliardelli, Luca; Simonini, Giovanni; Beneventano, Domenico

Set similarity join is an essential operation in data integration and big data analytics, that finds similar pairs of records where the records contain string or set-based data. To cope with the increasing scale of the data, several techniques have been proposed to perform set similarity joins using distributed frameworks, such as the MapReduce framework. In particular, Vernica et al. [3] proposed a MapReduce implementation of the so-called PPJoin algorithm [2], which in a recent study, was experimentally demonstrated as one of the best set similarity join algorithm [4]. These techniques, however, usually produce huge amounts of duplicates in order to perform parallel processing successfully. The large number of duplicates incurs on both large shuffle cost and unnecessary computation cost, which significantly decrease the performance. Moreover, these approaches do not provide a load balancing guarantee, which results in a skewness problem and negatively affects the scalability properties of these techniques. To address these problems, in this paper, we propose a duplicate-free framework, called TTJoin, to perform set simi- larity joins efficiently by utilizing an innovative filter based on prefix tokens and we implement it with one of most popular distributed framework, i.e., Apache Spark. Experiments on real world datasets demonstrate the effectiveness of proposed solution with respect to either traditional PPJoin and the MapReduce implementation proposed in [3].

2018 - MOMIS Dashboard: a powerful data analytics tool for Industry 4.0 [Relazione in Atti di Convegno]
Magnotta, Luca; Gagliardelli, Luca; Simonini, Giovanni; Orsini, Mirko; Bergamaschi, Sonia

In this work we present the MOMIS Dashboard, an interactive data analytics tool to explore and visualize data sources content through several kind of dynamic views (e.g. maps, bar, line, pie, etc.). The software tool is very versatile, and supports the connection to the main relational DBMS and Big Data sources. Moreover, it can be connected to MOMIS, a powerful Open Source Data Integration system, able to integrate heterogeneous data sources as enterprise information systems as well as sensors data. MOMIS Dashboard provides a secure permission management to limit data access on the basis of a user role, and a Designer to create and share personalized insights on the company KPIs, facilitating the enterprise collaboration. We illustrate the MOMIS Dashboard efficacy in a real enterprise scenario: a production monitoring platform to analyze real-time and historical data collected through sensors located on production machines that optimize production, energy consumption, and enable preventive maintenance.

2018 - Schema-agnostic Progressive Entity Resolution [Relazione in Atti di Convegno]
SIMONINI, GIOVANNI; Papadakis, George; Palpanas, Themis; BERGAMASCHI, Sonia

Entity Resolution (ER) is the task of finding entity profiles that correspond to the same real-world entity. Progressive ER aims to efficiently resolve large datasets when limited time and/or computational resources are available. In practice, its goal is to provide the best possible partial solution by approximating the optimal comparison order of the entity profiles. So far, Progressive ER has only been examined in the context of structured (relational) data sources, as the existing methods rely on schema knowledge to save unnecessary comparisons: they restrict their search space to similar entities with the help of schema-based blocking keys (i.e., signatures that represent the entity profiles). As a result, these solutions are not applicable in Big Data integration applications, which involve large and heterogeneous datasets, such as relational and RDF databases, JSON files, Web corpus etc. To cover this gap, we propose a family of schema-agnostic Progressive ER methods, which do not require schema infor- mation, thus applying to heterogeneous data sources of any schema variety. First, we introduce a na ̈ıve schema-agnostic method, showing that the straightforward solution exhibits a poor performance that does not scale well to large volumes of data. Then, we propose three different advanced methods. Through an extensive experimental evaluation over 7 real-world, established datasets, we show that all the advanced methods outperform to a significant extent both the na ̈ıve and the state-of-the-art schema- based ones. We also investigate the relative performance of the advanced methods, providing guidelines on the method selection.

2018 - Towards Progressive Search-driven Entity Resolution [Relazione in Atti di Convegno]
Pietrangelo, A.; Simonini, G.; Bergamaschi, S.; Koumarelas, I.; Naumann, F.

Keyword-search systems for databases aim to answer a user query composed of a few terms with a ranked list of records. They are powerful and easy-to-use data exploration tools for a wide range of contexts. For instance, given a product database gathered scraping e-commerce websites, these systems enable even non-technical users to explore the item set (e.g., to check whether it contains certain products or not, or to discover the price of an item). However, if the database contains dirty records (i.e., incomplete and duplicated records), a pre-processing step to clean the data is required. One fundamental data cleaning step is Entity Resolution, i.e., the task of identifying and fusing together all the records that refer to the same real-word entity. This task is typically executed on the whole data, independently of: (i) the portion of the entities that a user may indicate through keywords, and (ii) the order priority that a user might express through an order by clause. This paper describes a first step to solve the problem of progressive search-driven Entity Resolution: resolving all the entities described by a user through a handful of keywords, progressively (according to an order by clause). We discuss the features of our method, named SearchER and showcase some examples of keyword queries on two real-world datasets obtained with a demonstrative prototype that we have built.

2017 - BigBench workload executed by using Apache Flink [Relazione in Atti di Convegno]
Bergamaschi, Sonia; Gagliardelli, Luca; Simonini, Giovanni; Zhu, Song

Many of the challenges that have to be faced in Industry 4.0 involve the management and analysis of huge amount of data (e.g. sensor data management and machine-fault prediction in industrial manufacturing, web-logs analysis in e-commerce). To handle the so-called Big Data management and analysis, a plethora of frameworks has been proposed in the last decade. Many of them are focusing on the parallel processing paradigm, such as MapReduce, Apache Hive, Apache Flink. However, in this jungle of frameworks, the performance evaluation of these technologies is not a trivial task, and strictly depends on the application requirements. The scope of this paper is to compare two of the most employed and promising frameworks to manage big data: Apache Flink and Apache Hive, which are general purpose distributed platforms under the umbrella of the Apache Software Foundation. To evaluate these two frameworks we use the benchmark BigBench, developed for Apache Hive. We re-implemented the most significant queries of Apache Hive BigBench to make them work on Apache Flink, in order to be able to compare the results of the same queries executed on both frameworks. Our results show that Apache Flink, if it is configured well, is able to outperform Apache Hive.

2017 - From Data Integration to Big Data Integration [Capitolo/Saggio]
Bergamaschi, Sonia; Beneventano, Domenico; Mandreoli, Federica; Martoglia, Riccardo; Guerra, Francesco; Orsini, Mirko; Po, Laura; Vincini, Maurizio; Simonini, Giovanni; Zhu, Song; Gagliardelli, Luca; Magnotta, Luca

Abstract. The Database Group (DBGroup, and Information System Group (ISGroup, re- search activities have been mainly devoted to the Data Integration Research Area. The DBGroup designed and developed the MOMIS data integration system, giving raise to a successful innovative enterprise DataRiver (, distributing MOMIS as open source. MOMIS provides an integrated access to structured and semistructured data sources and allows a user to pose a single query and to receive a single unified answer. Description Logics, Automatic Annotation of schemata plus clustering techniques constitute the theoretical framework. In the context of data integration, the ISGroup addressed problems related to the management and querying of heterogeneous data sources in large-scale and dynamic scenarios. The reference architectures are the Peer Data Management Systems and its evolutions toward dataspaces. In these contexts, the ISGroup proposed and evaluated effective and efficient mechanisms for network creation with limited information loss and solutions for mapping management query reformulation and processing and query routing. The main issues of data integration have been faced: automatic annotation, mapping discovery, global query processing, provenance, multi- dimensional Information integration, keyword search, within European and national projects. With the incoming new requirements of integrating open linked data, textual and multimedia data in a big data scenario, the research has been devoted to the Big Data Integration Research Area. In particular, the most relevant achieved research results are: a scalable entity resolution method, a scalable join operator and a tool, LODEX, for automatically extracting metadata from Linked Open Data (LOD) resources and for visual querying formulation on LOD resources. Moreover, in collaboration with DATARIVER, Data Integration was successfully applied to smart e-health.

2017 - Sopj: A scalable online provenance join for data integration [Relazione in Atti di Convegno]
Zhu, Song; S., Email Author; Fiameni, Giuseppe; G., Email Author; Simonini, Giovanni; G., Email Author; Bergamaschi, S.

Data integration is a technique used to combine different sources of data together to provide an unified view among them. MOMIS[1] is an open-source data integration framework developed by the DBGroup1. The goal of our work is to make MOMIS be able to scale-out as the input data sources increase without introducing noticeable performance penalty. In particular, we present a full outer join method capable to efficiently integrate multiple sources at the same time by using data streams and provenance information. To evaluate the scalability of this innovative approach, we developed a join engine employing a distributed data processing framework. Our solution is able to process input data sources in the form of continuous stream, execute the join operation on-the-fly and produce outputs as soon as they are generated. In this way, the join can return partial results before the input streams have been completely received or processed optimizing the entire execution.

2017 - SparkER: an Entity Resolution framework for Apache Spark [Software]
Gagliardelli, Luca; Simonini, Giovanni; Zhu, Song; Bergamaschi, Sonia

Entity Resolution is a crucial task for many applications, but its nave solution has a low efficiency due to its quadratic complexity. Usually, to reduce this complexity, blocking is employed to cluster similar entities in order to reduce the global number of comparisons. Meta-Blocking (MB) approach aims to restructure the block collection in order to reduce the number of comparisons, obtaining better results in term of execution time. However, these techniques alone are not sufficient to work in the context of Big Data, where typically the records to be compared are in the order of hundreds of million. Parallel implementations of MB have been proposed in the literature, but all of them are built on Hadoop MapReduce, which is known to have a low efficiency on modern cluster architecture. We implement a Meta-Blocking technique for Apache Spark. Unlike Hadoop, Apache Spark uses a different paradigm to manage the tasks: it does not need to save the partial results on disk, keeping them in memory, which guarantees a shorter execution time. We reimplemented the state-of-the-art MB techniques, creating a new algorithm in order to exploit the Spark architecture. We tested our algorithm over several established datasets, showing that ours Spark implementation outperforms other existing ones based on Hadoop.

2016 - BLAST: a Loosely Schema-aware Meta-blocking Approach for Entity Resolution [Articolo su rivista]
Simonini, Giovanni; Bergamaschi, Sonia; Jagadish, H. V.

Identifying records that refer to the same entity is a fundamental step for data integration. Since it is prohibitively expensive to compare every pair of records, blocking techniques are typically employed to reduce the complexity of this task. These techniques partition records into blocks and limit the comparison to records co-occurring in a block. Generally, to deal with highly heterogeneous and noisy data (e.g. semi-structured data of the Web), these techniques rely on redundancy to reduce the chance of missing matches. Meta-blocking is the task of restructuring blocks generated by redundancy-based blocking techniques, removing superfluous comparisons. Existing meta-blocking approaches rely exclusively on schema-agnostic features. In this paper, we demonstrate how “loose” schema information (i.e., statistics collected directly from the data) can be exploited to enhance the quality of the blocks in a holistic loosely schema-aware (meta-)blocking approach that can be used to speed up your favorite Entity Resolution algorithm. We call it Blast (Blocking with Loosely-Aware Schema Techniques). We show how Blast can automatically extract this loose information by adopting a LSH-based step for e ciently scaling to large datasets. We experimentally demonstrate, on real-world datasets, how Blast outperforms the state-of-the-art unsupervised meta-blocking approaches, and, in many cases, also the supervised one.

2016 - Enhancing entity resolution efficiency with loosely schema-aware techniques - Discussion paper [Relazione in Atti di Convegno]
Simonini, G.; Bergamaschi, S.

Entity Resolution, the task of identifying records that refer to the same real-world entity, is a fundamental step in data integration. Blocking is a widely employed technique to avoid the comparison of all possible record pairs in a dataset (an inefficient approach). Renouncing to exploit schema information for blocking has been proved to limit the chance of missing matches (i.e., it guarantees high recall), at the cost of a low precision. Meta-blocking alleviates this issue by restructuring a block collection, removing redundant and superfluous comparisons. Yet, existing meta-blocking techniques exclusively rely on schema-agnostic features. In this paper, we investigate how loose schema information, induced directly from the data, can be exploited in an holistic loosely schema-aware (meta-)blocking approach that outperforms the state-of-the-art meta-blocking in terms of precision, without renouncing high level of recall. We implemented our idea in a system called Blast, and experimentally evaluated it on real-world datasets.

2016 - Providing Insight into Data Source Topics [Articolo su rivista]
Bergamaschi, Sonia; Ferrari, Davide; Guerra, Francesco; Simonini, Giovanni; Velegrakis, Yannis

A fundamental service for the exploitation of the modern large data sources that are available online is the ability to identify the topics of the data that they contain. Unfortunately, the heterogeneity and lack of centralized control makes it difficult to identify the topics directly from the actual values used in the sources. We present an approach that generates signatures of sources that are matched against a reference vocabulary of concepts through the respective signature to generate a description of the topics of the source in terms of this reference vocabulary. The reference vocabulary may be provided ready, may be created manually, or may be created by applying our signature-generated algorithm over a well-curated data source with a clear identification of topics. In our particular case, we have used DBpedia for the creation of the vocabulary, since it is one of the largest known collections of entities and concepts. The signatures are generated by exploiting the entropy and the mutual information of the attributes of the sources to generate semantic identifiers of the various attributes, which combined together form a unique signature of the concepts (i.e. the topics) of the source. The generation of the identifiers is based on the entropy of the values of the attributes; thus, they are independent of naming heterogeneity of attributes or tables. Although the use of traditional information-theoretical quantities such as entropy and mutual information is not new, they may become untrustworthy due to their sensitivity to overfitting, and require an equal number of samples used to construct the reference vocabulary. To overcome these limitations, we normalize and use pseudo-additive entropy measures, which automatically downweight the role of vocabulary items and property values with very low frequencies, resulting in a more stable solution than the traditional counterparts. We have materialized our theory in a system called WHATSIT and we experimentally demonstrate its effectiveness.

2015 - Big data exploration with faceted browsing [Relazione in Atti di Convegno]
Zhu, Song; Simonini, Giovanni

Big data analysis now drives nearly every aspect of modern society, from manufacturing and retail, through mobile and financial services, through the life sciences and physical sciences. The ability to continue to use big data to make new connections and discoveries will help to drive the breakthroughs of tomorrow. One of the most valuable means through which to make sense of big data, and thus make it more approachable to most people, is data visualization. Data visualization can guide decision-making and become a tool to convey information critical in all data analysis. However, to be actually actionable, data visualizations should contain the right amount of interactivity. They have to be well designed, easy to use, understandable, meaningful, and approachable. In this article we present a new approach to visualize huge amount of data, based on a Bayesian suggestion algorithm and the widely used enterprise search platform Solr. We demonstrate how the proposed Bayesian suggestion algorithm became a key ingredient in a big data scenario, where generally a query can generate so many results that the user can be confused. Thus, the selection of the best results, together with the result path chosen by the user by means of multi-faceted querying and faceted navigation, can be very useful.

2015 - Supporting Image Search with Tag Clouds: A Preliminary Approach [Articolo su rivista]
Guerra, Francesco; Simonini, Giovanni; Vincini, Maurizio

Algorithms and techniques for searching in collections of data address a challenging task, since they have to bridge the gap between the ways in which users express their interests, through natural language expressions or keywords, and the ways in which data is represented and indexed.When the collections of data include images, the task becomes harder, mainly for two reasons. From one side the user expresses his needs through one medium (text) and he will obtain results via another medium (some images). From the other side, it can be difficult for a user to understand the results retrieved; that is why a particular image is part of the result set. In this case, some techniques for analyzing the query results and giving to the users some insight into the content retrieved are needed. In this paper, we propose to address this problem by coupling the image result set with a tag cloud of words describing it. Some techniques for building the tag cloud are introduced and two application scenarios are discussed.

2014 - Discovering the topics of a data source: A statistical approach? [Relazione in Atti di Convegno]
Bergamaschi, Sonia; Ferrari, Davide; Guerra, Francesco; Simonini, Giovanni

In this paper, we present a preliminary approach for automatically discovering the topics of a structured data source with respect to a reference ontology. Our technique relies on a signature, i.e., a weighted graph that summarizes the content of a source. Graph-based approaches have been already used in the literature for similar purposes. In these proposals, the weights are typically assigned using traditional information-theoretical quantities such as entropy and mutual information. Here, we propose a novel data-driven technique based on composite likelihood to estimate the weights and other main features of the graphs, making the resulting approach less sensitive to overfitting. By means of a comparison of signatures, we can easily discover the topic of a target data source with respect to a reference ontology. This task is provided by a matching algorithm that retrieves the elements common to both the graphs. To illustrate our approach, we discuss a preliminary evaluation in the form of running example.

2014 - Keyword Search over Relational Databases: Issues, Approaches and Open Challenges [Relazione in Atti di Convegno]
Bergamaschi, Sonia; Guerra, Francesco; Simonini, Giovanni

In this paper, we overview the main research approaches developed in the area of Keyword Search over Relational Databases. In particular, we model the process for solving keyword queries in three phases: the management of the user’s input, the search algorithms, the results returned to the user. For each phase we analyze the main problems, the solutions adopted by the most important system developed by researchers and the open challenges. Finally, we introduce two open issues related to multi-source scenarios and database sources handling instance not fully accessible.

2014 - Towards declarative imperative data-parallel systems ? [Relazione in Atti di Convegno]
Interlandi, M.; Simonini, G.; Bergamaschi, S.

Pushed by recent evolvements in the field of declarative networking and data-parallel computation, we propose a first investigation over a declarative imperative parallel programming model which tries to combine the two worlds. We identify a set of requirements that the model should possess and introduce a conceptual sketch of the system implementing the foresaw model.

2014 - Using Big Data to Support Automatic Word Sense Disambiguation [Relazione in Atti di Convegno]
Guerra, Francesco; Simonini, Giovanni

Word Sense Induction (WSI) usually relies on data structures built upon the words to be disambiguated. This is a time-consuming process that requires a huge computational effort. In this paper, we propose an approach to automatically build a generic sense inventory (called iSC) to be used as a reference for disambiguation. The sense inventory is built extracting insight from Big Data exploiting a community detection algorithm. Since generate taking into account large corpora of data, the iSCis independent of the domain of application and of predefined target words.