No abstract

We present Ditto, a novel entity matching system based on pre-trained Transformer-based language models. We fine-tune and cast EM as a sequence-pair classification problem to leverage such models with a simple architecture. Our experiments show that a straight-forward application of language models such as BERT, DistilBERT, or RoBERTa pre-trained on large text corpora already significantly improves the matching quality and outperforms previous state-of-the-art (SOTA), by up to 29% of F1 score on benchmark datasets. We also developed three optimization techniques to further improve Ditto's matching capability. Ditto allows domain knowledge to be injected by highlighting important pieces of input information that may be of interest when making matching decisions. Ditto also summarizes strings that are too long so that only the essential information is retained and used for EM. Finally, Ditto adapts a SOTA technique on data augmentation for text to EM to augment the training data with (difficult) examples. This way, Ditto is forced to learn "harder" to improve the model's matching capability. The optimizations we developed further boost the performance of Ditto by up to 9.8%. Perhaps more surprisingly, we establish that Ditto can achieve the previous SOTA results with at most half the number of labeled data. Finally, we demonstrate Ditto's effectiveness on a real-world large-scale EM task. On matching two company datasets consisting of 789K and 412K records, Ditto achieves a high F1 score of 96.5%.

Schema mappings are high-level specifications that describe the relationship between two database schemas. Two operators on schema mappings, namely the composition operator and the inverse operator, are regarded as especially important. Progress on the study of the inverse operator was not made until very recently, as even finding the exact semantics of this operator turned out to be a fairly delicate task. Furthermore, this notion is rather restrictive, since it is rare that a schema mapping possesses an inverse.In this article, we introduce and study the notion of a quasi-inverse of a schema mapping. This notion is a principled relaxation of the notion of an inverse of a schema mapping; intuitively, it is obtained from the notion of an inverse by not differentiating between instances that are equivalent for data-exchange purposes. For schema mappings specified by source-to-target tuple-generating dependencies (s-t tgds), we give a necessary and sufficient combinatorial condition for the existence of a quasi-inverse, and then use this condition to obtain both positive and negative results about the existence of quasi-inverses. In particular, we show that every LAV (local-as-view) schema mapping has a quasi-inverse, but that there are schema mappings specified by full s-t tgds that have no quasiinverse. After this, we study the language needed to express quasi-inverses of schema mappings specified by s-t tgds, and we obtain a complete characterization. We also characterize the language needed to express inverses of schema mappings, and thereby solve a problem left open in the earlier study of the inverse operator. Finally, we show that quasi-inverses can be used in many cases to recover the data that was exported by the original schema mapping when performing data exchange.

We present an archiving technique for hierarchical data with key structure. Our approach is based on the notion of timestamps whereby an element appearing in multiple versions of the database is stored only once along with a compact description of versions in which it appears. The basic idea of timestamping was discovered by Driscoll et al. in the context of persistent data structures where one wishes to track the sequences of changes made to a data structure. We extend this idea to develop an archiving tool for XML data that is capable of providing meaningful change descriptions and can also efficiently support a variety of basic functions concerning the evolution of data such as retrieval of any specific version from the archive and querying the temporal history of any element. This is in contrast to diff-based approaches where such operations may require undoing a large number of changes or significant reasoning with the deltas. Surprisingly, our archiving technique does not incur any significant space overhead when contrasted with other approaches. Our experimental results support this and also show that the compacted archive file interacts well with other compression techniques. Finally, another useful property of our approach is that the resulting archive is also in XML and hence can directly leverage existing XML tools.

We study two classes of view update problems in relational databases. We are given a source database S, a monotone query Q, and the view Q(S) generated by the query. The first problem that we consider is the classical view deletion problem where we wish to identify a minimal set T of tuples in S whose deletion will eliminate a given tuple t from the view. We study the complexity of optimizing two natural objectives in this setting, namely, find T to minimize the side-effects on the view, and the source, respectively. For both objective functions, we show a dichotomy in the complexity. Interestingly, the problem is either in P or is NPhard, for queries in the same class in either objective function. The second problem in our study is the annotation placement problem. Suppose we annotate an attribute of a tuple in S. The rules for carrying the annotation forward through a query are easily stated. On the other hand, suppose we annotate an attribute of a tuple in the view Q(S), what annotation(s) in S will cause this annotation to appear in the view, minimizing the propagation to other attributes in Q(S)? View annotation is becoming an increasingly useful method of communicating meta-data among users of shared scientific data sets, and to our knowledge, there has been no formal study of this problem. Our study of these problems gives us important insights into computational issues involved in data provenance or lineage-the process by which data moves through databases. We show that the two problems correspond to two fundamentally distinct notions of provenance, why and where-provenance.