In many recommendation applications, items may have limited availability thereby causing conflict among users interested in the same items. Over time, this results in unequal user treatment: few users are recommended the limited items and receive preferential treatment, while the rest is left with sub-optimal recommendations, ultimately leading them to leave. In this paper, we formalize the novel problem of compensating users in multi-session recommendations under limited item availability. Our aim is to generate recommendations that not only optimize accuracy, but also compensate users over time for the loss of accuracy incurred in previous iterations. We design compensation strategies that serve users and items in different orders and accommodate various recommendation adoption models. Our algorithms are integrated into SoCRATe (System for Compensating Recommendations with Availability and Time), a framework that enables us to study loss compensation over time. Our experiments on real data demonstrate that to best compensate users for the incurred loss, traditional recommenders need to be revisited to account for item availability. Our experiments on synthetic data explore different parameters of our solution and show that it is much faster than an optimal (brute-force) compensation strategy, while achieving comparable results.

Abstract The ever-increasing number of systems based on semantic text analysis is making natural language understanding a fundamental task: embedding-based language models are used for a variety of applications, such as resume parsing or improving web search results. At the same time, despite their popularity and widespread use, concern is rapidly growing due to their display of social bias and lack of transparency. In particular, they exhibit a large amount of gender bias, favouring the consolidation of social stereotypes. Recently, sentence embeddings have been introduced as a novel and powerful technique to represent entire sentences as vectors. We propose a new metric to estimate gender bias in sentence embeddings, named bias score. Our solution leverages semantic importance of words and previous research on bias in word embeddings, and it is able to discern between neutral and biased gender information at sentence level. Experiments on a real-world dataset demonstrate that our novel metric can identify gender stereotyped sentences. Furthermore, we employ bias score to detect and then remove or compensate for the more stereotyped entries in text corpora used to train sentence encoders, improving their degree of fairness. Finally, we prove that models retrained on fairer corpora are less prone to make stereotypical associations compared to their original counterpart, while preserving accuracy in natural language understanding tasks. Additionally, we compare our experiments with traditional methods for reducing bias in embedding-based language models.

We demonstrate SoCRATe, an online system dedicated to providing adaptive recommendations to users when items have limited availability. SoCRATe is relevant to several real-world applications, among which movie and task recommendations. SoCRATe has several appealing features: (i) watching users as they consume recommendations and accounting for user feedback in refining recommendations in the next round; (ii) implementing loss compensation strategies to make up for sub-optimal recommendations, in terms of accuracy, when items have limited availability; (iii) deciding when to re-generate recommendations on a need-based fashion. SoCRATe accommodates real users as well as simulated users to enable testing multiple recommendation choice models. To frame evaluation, SoCRATe introduces a new set of measures that capture recommendation accuracy, user satisfaction and item consumption over time. All these features make SoCRATe unique and able to adapt recommendations to user preferences in a resource-limited setting. A video of SoCRATe is available at https://youtu.be/4wlaScc_rUo.

In our everyday lives, technologies based on data play an increasingly important role. With the widespread adoption of decision making systems also in very sensitive environments, fairness has become a very important topic of discussion within the data science community. In this context, it is crucial to ensure that the data on which we base these decisions, are fair, and do not reflect historical biases. In this demo, we propose FAIR-DB (FunctionAl dependencIes to discoveR Data Bias), a system that exploiting the notion of Functional Dependency, a particular type of constraint on the data, can discover unethical behaviours in a dataset. The proposed solution is implemented as a web-based application, that, given an input dataset, generates such dependencies, walks the user trough their analysis, and finally provides many insights about bias present in the data. Our tool uses a novel metric to evaluate the unfairness present in datasets, identifies the attributes that encompass discrimination (e.g. ethnicity, sex or religion), and provides very precise information about the groups treated unequally. We also provide a detailed description of the system architecture and present a demonstration scenario, based on a real-world dataset frequently used in the field of computer ethics.

Abstract Nowadays, data integration must often manage noisy data, also containing attribute values written in natural language such as product descriptions or book reviews. In the data integration process, Entity Linkage has the role of identifying records that contain information referring to the same object. Modern Entity Linkage methods, in order to reduce the dimension of the problem, partition the initial search space into “blocks” of records that can be considered similar according to some metrics, comparing then only the records belonging to the same block and thus greatly reducing the overall complexity of the algorithm. In this paper, we propose two automatic blocking strategies that, differently from the traditional methods, aim at capturing the semantic properties of data by means of recent deep learning frameworks. Both methods, in a first phase, exploit recent research on tuple and sentence embeddings to transform the database records into real-valued vectors; in a second phase, to arrange the tuples inside the blocks, one of them adopts approximate nearest neighbourhood algorithms, while the other one uses dimensionality reduction techniques combined with clustering algorithms. We train our blocking models on an external, independent corpus, and then, we directly apply them to new datasets in an unsupervised fashion. Our choice is motivated by the fact that, in most data integration scenarios, no training data are actually available. We tested our systems on six popular datasets and compared their performances against five traditional blocking algorithms. The test results demonstrated that our deep-learning-based blocking solutions outperform standard blocking algorithms, especially on textual and noisy data.