Keyword queries offer a convenient alternative to traditional SQL in querying relational databases with large, often unknown, schemas and instances. The challenge in answering such queries is to discover their intended semantics, construct the SQL queries that describe them and used them to retrieve the respective tuples. Existing approaches typically rely on indices built a-priori on the database content. This seriously limits their applicability if a-priori access to the database content is not possible. Examples include the on-line databases accessed through web interface, or the sources in information integration systems that operate behind wrappers with specific query capabilities. Furthermore, existing literature has not studied to its full extend the inter-dependencies across the ways the different keywords are mapped into the database values and schema elements. In this work, we describe a novel technique for translating keyword queries into SQL based on the Munkres (a.k.a. Hungarian) algorithm. Our approach not only tackles the above two limitations, but it offers significant improvements in the identification of the semantically meaningful SQL queries that describe the intended keyword query semantics. We provide details of the technique implementation and an extensive experimental evaluation.
Information overload occurs when the information available exceeds the user's ability to process it. To manage information overload, a user is required to discriminate among useful, redundant, incorrect, and meaningless information. From a computer science perspective, this means we must provide users with a combination of techniques and tools for collecting, grouping, classifying, selecting, indexing, ranking, and filtering useful information. The articles in this special issue show four different facets of the information overload problem, by providing the readers with a big picture of the main research outcomes in this topic.
Predictive Maintenance (PdM) is the newest strategy for maintenance management in industrial contexts. It aims to predict the occurrence of a failure to minimize unexpected downtimes and maximize the useful life of components. In data-driven approaches, PdM makes use of Machine Learning (ML) algorithms to extract relevant features from signals, identify and classify possible faults (diagnostics), and predict the components’ remaining useful life (prognostics). The major challenge lies in the high complexity of industrial plants, where both operational conditions change over time and a large number of unknown modes occur. A solution to this problem is offered by novelty detection, where a representation of the machinery normal operating state is learned and compared with online measurements to identify new operating conditions. In this paper, a systematic study of autoencoder-based methods for novelty detection is conducted. We introduce an architecture template, which includes a classification layer to detect and separate the operative conditions, and a localizer for identifying the most influencing signals. Four implementations, with different deep learning models, are described and used to evaluate the approach on data collected from a test rig. The evaluation shows the effectiveness of the architecture and that the autoencoders outperform the current baselines.
24 April 2023 in IEEE Transactions on Knowledge and Data Engineering
In the past decade, many approaches have been suggested to execute ML workloads on a DBMS. However, most of them have looked at in-DBMS ML from a training perspective, whereas ML inference has been largely overlooked. We think that this is an important gap to fill for two main reasons: (1) in the near future, every application will be infused with some sort of ML capability; (2) behind every web page, application, and enterprise there is a DBMS, whereby in-DBMS inference is an appealing solution both for efficiency (e.g., less data movement), performance (e.g., cross-optimizations between relational operators and ML) and governance. In this paper, we study whether DBMSs are a good fit for prediction serving. We introduce a technique for translating trained ML pipelines containing both featurizers (e.g., one-hot encoding) and models (e.g., linear and tree-based models) into SQL queries, and we compare in-DBMS performance against popular ML frameworks such as Sklearn and ml.net . Our experiments show that, when pushed inside a DBMS, trained ML pipelines can have performance comparable to ML frameworks in several scenarios, while they perform quite poorly on text featurization and over (even simple) neural networks.