Search

The most popular e-commerce search engines allow the user to run a keyword search, to find relevant results and to narrow down the results by mean of filters. The engines can also keep track of data and activities of the users, to provide personalized content, thus filtering automatically out a part of the results. Issues occur when personalization is not transparent and interferes with the user choices. Indeed, it has been noticed that, in some cases, a different ordering of search results is shown to different users. This becomes particularly critical when search results are associated with prices. Changing the order of search results according to prices is known as price steering. This study investigates if and how price steering exists, considering queries on Google Shopping by users searching from different geographic locations, distinguishable by their values of Gross Domestic Product. The results confirm that products belonging to specific categories (e.g., electronic devices and apparel) are shown to users according to different prices orderings, and the prices in the results list differ, on average, in a way that depends on users' location. All results are validated through statistical tests.

This work presents the solution adopted by the sisinflab team to solve the task NEEL-IT (Named Entity rEcognition and Linking in Italian Tweets) at the Evalita 2016 challenge. The task consists in the annotation of each named entity mention in a Twitter message written in Italian, among characters, events, people, locations, organizations, products and things and the eventual linking when a corresponding entity is found in a knowledge base (e.g. DBpedia). We faced the challenge through an approach that combines unsupervised methods, such as DBpedia Spotlight and word embeddings, and supervised techniques such as a CRF classifier and a Deep learning classifier. Questo lavoro presenta la soluzione del team sisinflab al task NEEL-IT (Named Entity rEcognition and Linking in Italian Tweets) di Evalita 2016. Il task richiede il riconoscimento e l’annotazione del testo di un messaggio di Twitter in Italiano con entità nominate quali personaggi, eventi, persone, luoghi, organizzazioni, prodotti e cose e eventualmente l’associazione di queste entità con la corrispondente risorsa in una base di conoscenza quale, DBpedia. L’approccio proposto combina metodi non supervisionati quali DBpedia Spotlight e i word embeddings, e tecniche supervisionate basate su due classificatori di tipo CRF e Deep learning.

We address the problem of automatically detecting the scope of negations and speculations in clinical notes, by proposing a machine-learning algorithm that analyzes the dependency tree of a sentence. Given a negative/speculative cue, the algorithm tries to extend the boundary of the scope towards the left and the right, by navigating through the parse tree. We report on experiments with the algorithm using the Bioscope corpus. Il lavoro affronta il problema di identificare l’ambito a cui si applica una negazione o un’espressione dubitativa nel testo di un referto medico. Si propone un algoritmo di apprendimento automatico, che analizza l’albero di parsing di ogni frase. Dato un indizio di negazione/ipotesi, l’algoritmo cerca di estendere il confine dell’ambito sia a destra che a sinistra, attraversando l’albero di parsing. Riportiamo infine i risultati di esperimenti con l’algoritmo effettuati usando il corpus Bioscope.

This work presents our team solution for task 4a (Message Polarity Classification) at the SemEval 2016 challenge. Our experiments have been carried out over the Twitter dataset provided by the challenge. We follow a supervised approach, exploiting a SVM polynomial kernel classifier trained with the challenge data. The classifier takes as input advanced NLP features. This paper details the features and discusses the achieved results.

Automatic quality evaluation of Web information is a task with many fields of applications and of great relevance, especially in critical domains like the medical one. We move from the intuition that the quality of content of medical Web documents is affected by features related with the specific domain. First, the usage of a specific vocabulary (Domain Informativeness); then, the adoption of specific codes (like those used in the infoboxes of Wikipedia articles) and the type of document (e.g., historical and technical ones). In this paper, we propose to leverage specific domain features to improve the results of the evaluation of Wikipedia medical articles. In particular, we evaluate the articles adopting an "actionable" model, whose features are related to the content of the articles, so that the model can also directly suggest strategies for improving a given article quality. We rely on Natural Language Processing (NLP) and dictionaries-based techniques in order to extract the bio-medical concepts in a text. We prove the effectiveness of our approach by classifying the medical articles of the Wikipedia Medicine Portal, which have been previously manually labeled by the Wiki Project team. The results of our experiments confirm that, by considering domain-oriented features, it is possible to obtain sensible improvements with respect to existing solutions, mainly for those articles that other approaches have less correctly classified. Other than being interesting by their own, the results call for further research in the area of domain specific features suitable for Web data quality assessment.

Search engines and social media keep trace of profile- and behavioral-based distinct signals of their users, to provide them personalized and recommended content. Here, we focus on the level of web search personalization, to estimate the risk of trapping the user into so called Filter Bubbles. Our experimentation has been carried out on news, specifically investigating the Google News platform. Our results are in line with existing literature and call for further analyses on which kind of users are the target of specific recommendations by Google.

The paper describes our experiments addressing the SemEval 2014 task on the Analysis of Clinical text. Our approach consists in extending the techniques of NE recognition, based on sequence labelling, to address the special issues of this task, i.e. the presence of overlapping and discontiguous mentions and the requirement to map the mentions to unique identifiers. We explored using supervised methods in combination with word embeddings generated from unannotated data.

Due to the large amount of social network data produced at an ever growing speed and their complex nature, recent works have addressed the problem of efficiently querying such data according to social, temporal or spatial dimensions. In this work we propose a data model that keeps into account all these dimensions and we compare different approaches for efficient query execution on a large real dataset using standard relational technologies. Due to the large amount of social network data produced at an ever growing speed and their complex nature, recent works have addressed the problem of efficiently querying such data according to social, temporal or spatial dimensions. In this work we propose a data model that keeps into account all these dimensions and we compare different approaches for efficient query execution on a large real dataset using standard relational technologies.

In embedded Linux systems low power consumption and memory usage are strict constraint. The problem to manage a large quantity of file metadata, that usually represents the most expensive task of such systems, is a task delegated to user space programs, generally, database based. In this work we propose how to deal with metadata at file system level with Hixosfs. Hixosfs is an ext2 based Linux filesystem, designed to easily cataloging and retrieving e-mails, musical files and logs files according final user selected metadata. The core idea is to move the problem complexity from userspace to kernel level to speed up the overall process. The implementation mainly required ext2 file system structure to be extended to store file tags, then they can be accessed and modified directly by system calls. Analysis and comparison works state this approach to be suitable for a large amount of data.

This work describes a clinical process management system aimed to support a process-centred vision of health care practices. The system is founded on knowledge representation and semantic information extraction approaches allowing medical knowledge modelling and acquisition. At the heart of the system there are formalisms and languages well suited for representing, clinical processes as workflows, medical and domain knowledge as ontologies and rules enabling the recognition of semantic patterns representing ontology concepts. The system acquires and stores clinical process instances into a medical knowledge base which parameters are obtained exploiting a semantic information extraction approach enabling automatic medical knowledge acquisition from unstructured clinical documents. The main goal of the system is to assists health care professional in executing and monitoring clinical processes by providing functionalities for automatic knowledge acquisition. Acquired information can be analyzed for identifying main causes of medical errors, high costs and, potentially, to suggest clinical processes restructuring or improvement able to enhance cost control and patient safety.

This work presents our team solution for task 4a (Message Polarity Classification) at the SemEval 2016 challenge. Our experiments have been carried out over the Twitter dataset provided by the challenge. We follow a supervised approach, exploiting a SVM polynomial kernel classifier trained with the challenge data. The classifier takes as input advanced NLP features. This paper details the features and discusses the achieved results.