Automata learning has many applications in artificial intelligence and software engineering. Central to these applications is the L* algorithm, introduced by Angluin (1987). The L* algorithm learns deterministic finite-state automata (DFAs) in polynomial time when provided with a minimally adequate teacher. Unfortunately, the L* algorithm can only learn DFAs over finite alphabets, which limits its applicability. In this paper, we extend L* to learn symbolic automata whose transitions use predicates over rational numbers, i.e., over infinite and dense alphabets. Our result makes the L* algorithm applicable to new settings like (real) RGX, and time series. Furthermore, our proposed algorithm for learning each predicate is optimal in the sense that it asks a number of queries to the teacher that is at most linear with respect to the size of their representation.

We study the task of disambiguation of finite state automata, namely, converting an automaton into an equivalent, unambiguous one. We do this by developing a novel and simple algorithmic framework that generalizes the subset construction for determinization, and that satisfies some desirable properties: (1) it preserves the original automaton if it was already unambiguous, (2) it computes the successor states on-the-fly and (3) computes each new state in polynomial time - this last point is crucial as it guarantees that the running time is polynomial in the size of the output automaton. Then, we show how to apply this framework for partial disambiguation: by changing the criterion that builds the new states, we develop algorithms for different levels of ambiguity, namely, finitely ambiguous, and polynomially ambiguous automata. These algorithms also satisfy condition (1) for their respective levels, and also (2) and (3). Finally, we show that the disambiguation framework can easily be extended to other models of automata like weighted automata.

Complex Event Recognition (CER) systems are used to identify complex patterns in event streams, such as those found in stock markets, sensor networks, and other similar applications. An important task in such patterns is aggregation, which involves summarizing a set of values into a single value using an algebraic function, such as the maximum, sum, or average, among others. Despite the relevance of this task, query languages in CER typically support aggregation in a restricted syntactic form, and their semantics are generally undefined. In this work, we present a first step toward formalizing a query language with aggregation for CER. We propose to extend Complex Event Logic (CEL), a formal query language for CER, with aggregation operations. This task requires revisiting the semantics of CEL, using a new semantics based on bags of tuples instead of sets of positions. Then, we present an extension of CEL, called Aggregation CEL (ACEL), which introduces an aggregation operator for any commutative monoid operation. The operator can be freely composed with previous CEL operators, allowing users to define complex queries and patterns. We showcase several queries in practice where ACEL proves to be natural for specifying them. From the computational side, we present a novel automata model, called Aggregation Complex Event Automata (ACEA), that extends the previous proposal of Complex Event Automata (CEA) with aggregation and filtering features. Moreover, we demonstrate that every query in ACEL can be expressed in ACEA, illustrating the effectiveness of our computational model. Finally, we study the expressiveness of ACEA through the lens of ACEL, showing that the automata model is more expressive than ACEL.

Complex Event Recognition (CER) is a group of data management technologies that model data streams as sequences of events, where users are interested in recognizing complex events, namely, groups of events that represent critical situations in real life. Examples of complex events could include a fire detected by a sensor network in a nature reserve, an accident recognized by cameras in a smart city, or a critical social event in a social network. In these scenarios, event streams are generated continuously at high speed, and the importance of each event decays rapidly over time. To process them, a complex event recognition engine is a data management software that must efficiently process such data and alert on the presence of complex events in real time. In this talk, we will present the dissection of CORE [Kyle Bossonney et al., 2025; Marco Bucchi et al., 2022], a novel complex event recognition engine. The dissection will cover all its internal components: starting with its architecture, we will examine its query language, stream and memory management, query optimization, query evaluation, and complex event outputting. We will focus on the technical solutions and challenges of a CER engine, from both theoretical [Alejandro Grez et al., 2019; Alejandro Grez et al., 2021] and practical [Kyle Bossonney et al., 2025; Marco Bucchi et al., 2022] perspectives. In particular, based on our understanding of its components, we will review several open research problems and possible directions for future work in complex event recognition.

Due to the importance of linear algebra and matrix operations in data analytics, there is significant interest in using relational query optimization and processing techniques for evaluating (sparse) linear algebra programs. In particular, in recent years close connections have been established between linear algebra programs and relational algebra that allow transferring optimization techniques of the latter to the former. In this paper, we ask ourselves which linear algebra programs in MATLANG correspond to the free-connex and q-hierarchical fragments of conjunctive first-order logic. Both fragments have desirable query processing properties: free-connex conjunctive queries support constant-delay enumeration after a linear-time preprocessing phase, and q-hierarchical conjunctive queries further allow constant-time updates. By characterizing the corresponding fragments of MATLANG, we hence identify the fragments of linear algebra programs that one can evaluate with constant-delay enumeration after linear-time preprocessing and with constant-time updates. To derive our results, we improve and generalize previous correspondences between MATLANG and relational algebra evaluated over semiring-annotated relations. In addition, we identify properties on semirings that allow to generalize the complexity bounds for free-connex and q-hierarchical conjunctive queries from Boolean annotations to general semirings.

We study the problem of evaluating a Monadic Second Order (MSO) query over strings under updates in the setting of direct access. We present an algorithm that, given an MSO query with first-order free variables represented by an unambiguous variable-set automaton $\mathcal{A}$ with state set $Q$ and variables $X$ and a string $s$, computes a data structure in time $\mathcal{O}(|Q|^ω\cdot |X|^2 \cdot |s|)$ and, then, given an index $i$ retrieves, using the data structure, the $i$-th output of the evaluation of $\mathcal{A}$ over $s$ in time $\mathcal{O}(|Q|^ω\cdot |X|^3 \cdot \log(|s|)^2)$ where $ω$ is the exponent for matrix multiplication. Ours is the first efficient direct access algorithm for MSO query evaluation over strings; such algorithms so far had only been studied for first-order queries and conjunctive queries over relational data. Our algorithm gives the answers in lexicographic order where, in contrast to the setting of conjunctive queries, the order between variables can be freely chosen by the user without degrading the runtime. Moreover, our data structure can be updated efficiently after changes to the input string, allowing more powerful updates than in the enumeration literature, e.g.~efficient deletion of substrings, concatenation and splitting of strings, and cut-and-paste operations. Our approach combines a matrix representation of MSO queries and a novel data structure for dynamic word problems over semi-groups which yields an overall algorithm that is elegant and easy to formulate.

We present a theoretical framework for the extraction and transformation of text documents as a two-phase process: The first phase uses document spanners to extract information from the input document. The second phase transforms the extracted information into a suitable output. To support several reasonable extract-transform scenarios, we propose for the first phase an extension of document spanners from span-tuples to so-called multispan-tuples, where variables are mapped to sets of spans instead of only single spans. We focus on multispanners described by regex formulas, and we prove that these have the same desirable properties as standard regular spanners. To formalize the second phase, we consider transformations that map every pair document-tuple, where each tuple comes from the (multi)span-relation extracted in the first phase, into a new output document. The specification of the two phases is what we call an extract-transform (ET) program, which covers practically relevant extract-transform tasks. In this paper, our main technical goal is to identify a broad class of ET programs that can be evaluated efficiently. We specifically focus on the scenario of regular ET programs: the extraction phase is given by a regex multispanner and the transformation phase is given by a regular string-to-string function. We show that for any regular ET program, given an input document, we can enumerate all final output documents with output-linear delay after linear preprocessing. As a side effect, we characterize the expressive power of regular ET programs and also show that they have desirable properties, like being closed under composition.

We study the problem of enumerating results from a query over a compressed document. The model we use for compression are straight-line programs (SLPs), which are defined by a context-free grammar that produces a single string. For our queries, we use a model called Annotated Automata, an extension of regular automata that allows annotations on letters. This model extends the notion of Regular Spanners as it allows arbitrarily long outputs. Our main result is an algorithm that evaluates such a query by enumerating all results with output-linear delay after a preprocessing phase which takes linear time on the size of the SLP, and cubic time over the size of the automaton. This is an improvement over Schmid and Schweikardt's result, which, with the same preprocessing time, enumerates with a delay that is logarithmic on the size of the uncompressed document. We achieve this through a persistent data structure named Enumerable Compact Sets with Shifts which guarantees output-linear delay under certain restrictions. These results imply constant-delay enumeration algorithms in the context of regular spanners. Further, we use an extension of annotated automata which utilizes succinctly encoded annotations to save an exponential factor from previous results that dealt with constant-delay enumeration over vset automata. Lastly, we extend our results in the same fashion Schmid and Schweikardt did to allow complex document editing while maintaining the constant delay guarantee.

When query evaluation produces too many tuples, a new approach in query answering is to retrieve a diverse subset of them. The standard approach for measuring the diversity of a set of tuples is to use a distance function between tuples, which measures the dissimilarity between them, to then aggregate the pairwise distances of the set into a score (e.g., by using sum or min aggregation). However, as we will point out in this work, the resulting diversity measures may display some unintuitive behavior. Moreover, even in very simple settings, finding a maximally diverse subset of the answers of fixed size is, in general, intractable and little is known about approximations apart from some hand-picked distance-aggregator pairs. In this work, we introduce a novel approach for computing the diversity of tuples based on volume instead of distance. We present a framework for defining volume-based diversity functions and provide several examples of these measures applied to relational data. Although query answering of conjunctive queries (CQ) under this setting is intractable in general, we show that one can always compute a (1-1/e)-approximation for any volume-based diversity function. Furthermore, in terms of combined complexity, we connect the evaluation of CQs under volume-based diversity functions with the ranked enumeration of solutions, finding general conditions under which a (1-1/e)-approximation can be computed in polynomial time.

Complex Event Recognition (CER) establishes a relevant solution for processing streams of events, giving users timely information. CER systems detect patterns in real-time, producing complex events and generating responses to them. In these tasks, time is a first-class citizen. Indeed, the time-sequence model distinguishes CER from other solutions, like data stream management systems. Surprisingly, until now, time constraints are usually included in CER query languages and models in a restricted way, and we still lack an understanding of the expressiveness and of efficient algorithms concerning this crucial feature: time. This work studies CER under time constraints regarding its query language, computational models, and streaming evaluation algorithms. We start by introducing an extension of Complex Event Logic (CEL), called timed CEL, with simple time operators. We show that timed CEL aids in modeling CER query languages in practice, serving as a proxy to study the expressive power of such languages under time constraints. For this purpose, we introduce an automata model for studying timed CEL, called timed Complex Event Automata (timed CEA). This model extends the existing CEA model with clocks, combining CEA and timed automata in a single model. We show that timed CEL and timed CEA are equally expressive, giving the first characterization of CER query languages under time constraints. Then, we move towards understanding the efficient evaluation of timed CEA over streams concerning its determinization and efficient algorithms. We present a class of timed CEA that are closed under determinization; furthermore, we show that this class contains swg-queries, an expressive class of CER queries recently introduced by Kleest-Meißner et al. Finally, we present a streaming evaluation algorithm with constant update time and output-linear delay for evaluating deterministic monotonic timed CEA with a single clock, which have only less equal or greater equal comparisons.

Due to the importance of linear algebra and matrix operations in data analytics, there is significant interest in using relational query optimization and processing techniques for evaluating (sparse) linear algebra programs. In particular, in recent years close connections have been established between linear algebra programs and relational algebra that allow transferring optimization techniques of the latter to the former. In this paper, we ask ourselves which linear algebra programs in MATLANG correspond to the free-connex and q-hierarchical fragments of conjunctive first-order logic. Both fragments have desirable query processing properties: free-connex conjunctive queries support constant-delay enumeration after a linear-time preprocessing phase, and q-hierarchical conjunctive queries further allow constant-time updates. By characterizing the corresponding fragments of MATLANG, we hence identify the fragments of linear algebra programs that one can evaluate with constant-delay enumeration after linear-time preprocessing and with constant-time updates. To derive our results, we improve and generalize previous correspondences between MATLANG and relational algebra evaluated over semiring-annotated relations. In addition, we identify properties on semirings that allow to generalize the complexity bounds for free-connex and q-hierarchical conjunctive queries from Boolean annotations to general semirings.

The set of answers to a query may be very large, potentially overwhelming users when presented with the entire set. In such cases, presenting only a small subset of the answers to the user may be preferable. A natural requirement for this subset is that it should be as diverse as possible to reflect the variety of the entire population. To achieve this, the diversity of a subset is measured using a metric that determines how different two solutions are and a diversity function that extends this metric from pairs to sets. In the past, several studies have shown that finding a diverse subset from an explicitly given set is intractable even for simple metrics (like Hamming distance) and simple diversity functions (like summing all pairwise distances). This complexity barrier becomes even more challenging when trying to output a diverse subset from a set that is only implicitly given (such as the query answers for a given query and a database). Until now, tractable cases have been found only for restricted problems and particular diversity functions. To overcome these limitations, we focus in this work on the notion of ultrametrics, which have been widely studied and used in many applications. Starting from any ultrametric d and a diversity function δ extending d , we provide sufficient conditions over δ for having polynomial-time algorithms to construct diverse answers. To the best of our knowledge, these conditions are satisfied by all the diversity functions considered in the literature. Moreover, we complement these results with lower bounds that show specific cases when these conditions are not satisfied and finding diverse subsets becomes intractable. We conclude by applying these results to the evaluation of conjunctive queries, demonstrating efficient algorithms for finding a diverse subset of solutions for acyclic conjunctive queries when the attribute order is used to measure diversity.

Hierarchical conjunctive queries (HCQ) are a subclass of conjunctive queries (CQ) with robust algorithmic properties. Among others, Berkholz, Keppeler, and Schweikardt have shown that HCQ is the subclass of CQ (without projection) that admits dynamic query evaluation with constant update time and constant delay enumeration. On a different but related setting stands Complex Event Recognition (CER), a prominent technology for evaluating sequence patterns over streams. Since one can interpret a data stream as an unbounded sequence of inserts in dynamic query evaluation, it is natural to ask to which extent CER can take advantage of HCQ to find a robust class of queries that can be evaluated efficiently. In this paper, we search to combine HCQ with sequence patterns to find a class of CER queries that can get the best of both worlds. To reach this goal, we propose a class of complex event automata model called Parallelized Complex Event Automata (PCEA) for evaluating CER queries with correlation (i.e., joins) over streams. This model allows us to express sequence patterns and compare values among tuples, but it also allows us to express conjunctions by incorporating a novel form of non-determinism that we call parallelization. We show that for every HCQ (under bag semantics), we can construct an equivalent PCEA. Further, we show that HCQ is the biggest class of full CQ that this automata model can define. Then, PCEA stands as a sweet spot that precisely expresses HCQ (i.e., among full CQ) and extends them with sequence patterns. Finally, we show that PCEA also inherits the good algorithmic properties of HCQ by presenting a streaming evaluation algorithm under sliding windows with logarithmic update time and output-linear delay for the class of PCEA with equality predicates.

We present the theoretical foundations and first experimental study of a new approach in centrality measures for graph data. The main principle is straightforward: the more relevant subgraphs around a vertex, the more central it is in the network. We formalize the notion of “relevant subgraphs” by choosing a family of subgraphs that, given a graph G and a vertex v , assigns a subset of connected subgraphs of G that contains v . Any of such families defines a measure of centrality by counting the number of subgraphs assigned to the vertex, i.e., a vertex will be more important for the network if it belongs to more subgraphs in the family. We show several examples of this approach. In particular, we propose the All-Subgraphs (All-Trees) centrality, a centrality measure that considers every subgraph (tree). We study fundamental properties over families of subgraphs that guarantee desirable properties over the centrality measure. Interestingly, All-Subgraphs and All-Trees satisfy all these properties, showing their robustness as centrality notions. To conclude the theoretical analysis, we study the computational complexity of counting certain families of subgraphs and show a linear time algorithm to compute the All-Subgraphs and All-Trees centrality for graphs with bounded treewidth. Finally, we implemented these algorithms and computed these measures over more than one hundred real-world networks. With this data, we present an empirical comparison between well-known centrality measures and those proposed in this work.

Some of the most relevant document schemas used online, such as XML and JSON, have a nested format. In the past decade, the task of extracting data from nested documents over streams has become especially relevant. We focus on the streaming evaluation of queries with outputs of varied sizes over nested documents. We model queries of this kind as Visibly Pushdown Annotators (VPAnn), a computational model that extends visibly pushdown automata with outputs and has the same expressive power as monadic second-order logic over nested documents. Since processing a document through a VPAnn can generate a massive number of results, we are interested in reading the input in a streaming fashion and enumerating the outputs one after another as efficiently as possible, namely, with constant delay. This article presents an algorithm that enumerates these elements with constant delay after processing the document stream in a single pass. Furthermore, we show that this algorithm is worst-case optimal in terms of update-time per symbol and memory usage.

Centrality measures are widely used to assign importance to graph-structured data. Recently, understanding the principles of such measures has attracted a lot of attention. Given that measures are diverse, this research has usually focused on classes of centrality measures. In this work, we provide a different approach by focusing on classes of graphs instead of classes of measures to understand the underlying principles among various measures. More precisely, we study the class of trees. We observe that even in \fix{the} case of trees, there is no consensus on which node should be selected as the most central. To analyze the behavior of centrality measures on trees, we introduce a property of \emph{tree rooting} that states a measure selects one or two adjacent nodes as the most important, and the importance decreases from them in all directions. This property is satisfied by closeness centrality but violated by PageRank. We show that, for several centrality measures that root trees, the comparison of adjacent nodes can be inferred by \emph{potential functions} that assess the quality of trees. We use these functions to give fundamental insights on rooting and derive a characterization explaining why some measure root trees. Moreover, we provide an almost liner-time algorithm to compute the root of a graph by using potential functions. Finally, using a family of potential functions, we show that many ways of tree rooting exist with desirable properties.

In this paper, we present the REmatch system for information extraction. REmatch is based on a recently proposed enumeration algorithm for evaluating regular expressions with capture variables supporting the all-match semantics. It tells a story of what it takes to make a theoretically optimal algorithm work in practice. As we show here, a naive implementation of the original algorithm would have a hard time dealing with realistic workloads. We thus develop a new algorithm and a series of optimizations that make REmatch as fast or faster than many popular RegEx engines while at the same time being able to return all the outputs: a task that most other engines tend to struggle with.

Some of the most relevant document schemas used online, such as XML and JSON, have a nested format. In the last decade, the task of extracting data from nested documents over streams has become especially relevant. We focus on the streaming evaluation of queries with outputs of varied sizes over nested documents. We model queries of this kind as Visibly Pushdown Transducers (VPT), a computational model that extends visibly pushdown automata with outputs and has the same expressive power as MSO over nested documents. Since processing a document through a VPT can generate a massive number of results, we are interested in reading the input in a streaming fashion and enumerating the outputs one after another as efficiently as possible, namely, with constant-delay. This paper presents an algorithm that enumerates these elements with constant-delay after processing the document stream in a single pass. Furthermore, we show that this algorithm is worst-case optimal in terms of update-time per symbol and memory usage.

We introduce annotated grammars, an extension of context-free grammars which allows annotations on terminals. Our model extends the standard notion of regular spanners, and is more expressive than the extraction grammars recently introduced by Peterfreund. We study the enumeration problem for annotated grammars: fixing a grammar, and given a string as input, enumerate all annotations of the string that form a word derivable from the grammar. Our first result is an algorithm for unambiguous annotated grammars, which preprocesses the input string in cubic time and enumerates all annotations with output-linear delay. This improves over Peterfreund's result, which needs quintic time preprocessing to achieve this delay bound. We then study how we can reduce the preprocessing time while keeping the same delay bound, by making additional assumptions on the grammar. Specifically, we present a class of grammars which only have one derivation shape for all outputs, for which we can enumerate with quadratic time preprocessing. We also give classes that generalize regular spanners for which linear time preprocessing suffices.

Abstract We study a variant of the classical membership problem in automata theory, which consists of deciding whether a given input word is accepted by a given automaton. We do so through the lenses of parameterized dynamic data structures: we assume that the automaton is fixed and its size is the parameter, while the input word is revealed as in a stream, one symbol at a time following the natural order on positions. The goal is to design a dynamic data structure that can be efficiently updated upon revealing the next symbol, while maintaining the answer to the query on whether the word consisting of symbols revealed so far is accepted by the automaton. We provide complexity bounds for this dynamic acceptance problem for timed automata that process symbols interleaved with time spans. The main contribution is a dynamic data structure that maintains acceptance of a fixed one-clock timed automaton $${\mathcal {A}}$$ A with amortized update time $$2^{{\mathcal {O}}(|{\mathcal {A}}|)}$$ 2 O ( | A | ) per input symbol.

Complex Event Recognition (CER) systems are a prominent technology for finding user-defined query patterns over large data streams in real time. CER query evaluation is known to be computationally challenging, since it requires maintaining a set of partial matches, and this set quickly grows super-linearly in the number of processed events. We present CORE, a novel COmplex event Recognition Engine that focuses on the efficient evaluation of a large class of complex event queries, including time windows as well as the partition-by event correlation operator. This engine uses a novel automaton-based evaluation algorithm that circumvents the super-linear partial match problem: under data complexity, it takes constant time per input event to maintain a data structure that compactly represents the set of partial matches and, once a match is found, the query results may be enumerated from the data structure with output-linear delay. We experimentally compare CORE against state-of-the-art CER systems on real-world data. We show that (1) CORE's performance is stable with respect to both query and time window size, and (2) CORE outperforms the other systems by up to five orders of magnitude on different workloads.

In the last years, enumeration algorithms with bounded delay have attracted a lot of attention for several data management tasks. Given a query and the data, the task is to preprocess the data and then enumerate all the answers to the query one by one and without repetitions. This enumeration scheme is typically useful when the solutions are treated on the fly or when we want to stop the enumeration once the pertinent solutions have been found. However, with the current schemes, there is no restriction on the order how the solutions are given and this order usually depends on the techniques used and not on the relevance for the user. In this paper we study the enumeration of monadic second order logic (MSO) over words when the solutions are ranked. We present a framework based on MSO cost functions that allows to express MSO formulae on words with a cost associated with each solution. We then demonstrate the generality of our framework which subsumes, for instance, document spanners and adds ranking to them. The main technical result of the paper is an algorithm for enumerating all the solutions of formulae in increasing order of cost efficiently, namely, with a linear preprocessing phase and logarithmic delay between solutions. The novelty of this algorithm is based on using functional data structures, in particular, by extending functional Brodal queues to suit with the ranked enumeration of MSO on words.

Ensuring the correctness of distributed cyber-physical systems can be done at runtime by monitoring properties over their behaviour. In a decentralised setting, such behaviour consists of multiple local traces, each offering an incomplete view of the system events to the local monitors, as opposed to the standard centralised setting with a unique global trace. We introduce the first monitoring framework for timed properties described by timed regular expressions over a distributed network of monitors. First, we define functions to rewrite expressions according to partial knowledge for both the centralised and decentralised cases. Then, we define decentralised algorithms for monitors to evaluate properties using these functions, as well as proofs of soundness and eventual completeness of said algorithms. Finally, we implement and evaluate our framework on synthetic timed regular expressions, giving insights on the cost of the centralised and decentralised settings and when to best use each of them.

Linear algebra algorithms often require some sort of iteration or recursion as is illustrated by standard algorithms for Gaussian elimination, matrix inversion, and transitive closure. A key characteristic shared by these algorithms is that they allow looping for a number of steps that is bounded by the matrix dimension. In this paper we extend the matrix query language MATLANG with this type of recursion, and show that this suffices to express classical linear algebra algorithms. We study the expressive power of this language and show that it naturally corresponds to arithmetic circuit families, which are often said to capture linear algebra. Furthermore, we analyze several sub-fragments of our language, and show that their expressive power is closely tied to logical formalisms on semiring-annotated relations.

Counting the number of words of a certain length accepted by a non-deterministic finite automaton (NFA) is a fundamental problem, which has many applications in different areas such as graph databases, knowledge compilation, and information extraction. Along with this, generating such words uniformly at random is also a relevant problem, particularly in scenarios where returning varied outputs is a desirable feature.

Conjunctive queries are one of the most common class of queries used in database systems, and the best studied in the literature. A seminal result of Grohe, Schwentick, and Segoufin (STOC 2001) demonstrates that for every class G of graphs, the evaluation of all conjunctive queries whose underlying graph is in G is tractable if, and only if, G has bounded treewidth. In this work, we extend this characterization to the counting problem for conjunctive queries. Specifically, for every class C of conjunctive queries with bounded treewidth, we introduce the first fully polynomial-time randomized approximation scheme (FPRAS) for counting answers to a query in C, and the first polynomial-time algorithm for sampling answers uniformly from a query in C. As a corollary, it follows that for every class G of graphs, the counting problem for conjunctive queries whose underlying graph is in G admits an FPRAS if, and only if, G has bounded treewidth (unless BPP is different from P). In fact, our FPRAS is more general, and also applies to conjunctive queries with bounded hypertree width, as well as unions of such queries.

In this work, we study two simple yet general complexity classes, based on logspace Turing machines, that provide a unifying framework for efficient query evaluation in areas such as information extraction and graph databases, among others. We investigate the complexity of three fundamental algorithmic problems for these classes: enumeration, counting, and uniform generation of solutions, and show that they have several desirable properties in this respect. Both complexity classes are defined in terms of non-deterministic logspace transducers (NL-transducers). For the first class, we consider the case of unambiguous NL-transducers, and we prove constant delay enumeration and both counting and uniform generation of solutions in polynomial time. For the second class, we consider unrestricted NL-transducers, and we obtain polynomial delay enumeration, approximate counting in polynomial time, and polynomial-time randomized algorithms for uniform generation. More specifically, we show that each problem in this second class admits a fully polynomial-time randomized approximation scheme (FPRAS) and a polynomial-time Las Vegas algorithm (with preprocessing) for uniform generation. Remarkably, the key idea to prove these results is to show that the fundamental problem # NFA admits an FPRAS, where # NFA is the problem of counting the number of strings of length n (given in unary) accepted by a non-deterministic finite automaton (NFA). While this problem is known to be P-complete and, more precisely, SpanL -complete, it was open whether this problem admits an FPRAS. In this work, we solve this open problem and obtain as a welcome corollary that every function in SpanL admits an FPRAS.

We present pumping lemmas for five classes of functions definable by fragments of weighted automata over the min-plus semiring, the max-plus semiring and the semiring of natural numbers. As a corollary we show that the hierarchy of functions definable by unambiguous, finitely-ambiguous, polynomially-ambiguous weighted automata, and the full class of weighted automata is strict for the min-plus and max-plus semirings.

Complex event recognition (CER) has emerged as the unifying field for technologies that require processing and correlating distributed data sources in real time. CER finds applications in diverse domains, which has resulted in a large number of proposals for expressing and processing complex events. Existing CER languages lack a clear semantics, however, which makes them hard to understand and generalize. Moreover, there are no general techniques for evaluating CER query languages with clear performance guarantees. In this article, we embark on the task of giving a rigorous and efficient framework to CER. We propose a formal language for specifying complex events, calledcomplex event logic(CEL), that contains the main features used in the literature and has a denotational and compositional semantics. We also formalize the so-called selection strategies, which had only been presented as by-design extensions to existing frameworks. We give insight into the language design trade-offs regarding the strict sequencing operators of CEL and selection strategies. With a well-defined semantics at hand, we discuss how to efficiently process complex events by evaluating CEL formulas with unary filters. We start by introducing a formal computational model for CER, calledcomplex event automata(CEA), and study how to compile CEL formulas with unary filters into CEA. Furthermore, we provide efficient algorithms for evaluating CEA over event streams using constant time per event followed by output-linear delay enumeration of the results.

Complex event processing (CEP) has gained a lot of attention for evaluating complex patterns over high-throughput data streams. Recently, new algorithms for the evaluation of CEP patterns have emerged with strong guarantees of efficiency, i.e. constant update-time per tuple and constant-delay enumeration. Unfortunately, these techniques are restricted for patterns with local filters, limiting the possibility of using joins for correlating the data of events that are far apart. In this paper, we embark on the search for efficient evaluation algorithms of CEP patterns with joins. We start by formalizing the so-called partition-by operator, a standard operator in data stream management systems to correlate contiguous events on streams. Although this operator is a restricted version of a join query, we show that partition-by (without iteration) is equally expressive as hierarchical queries, the biggest class of full conjunctive queries that can be evaluated with constant update-time and constant-delay enumeration over streams. To evaluate queries with partition-by we introduce an automata model, called chain complex event automata (chain-CEA), an extension of complex event automata that can compare data values by using equalities and disequalities. We show that this model admits determinization and is expressive enough to capture queries with partition-by. More importantly, we provide an algorithm with constant update time and constant delay enumeration for evaluating any query definable by chain-CEA, showing that all CEP queries with partition-by can be evaluated with these strong guarantees of efficiency.

Complex Event Recognition (CER for short) has recently gained attention as a mechanism for detecting patterns in streams of continuously arriving event data. Numerous CER systems and languages have been proposed in the literature, commonly based on combining operations from regular expressions (sequencing, iteration, and disjunction) and relational algebra (e.g., joins and filters). While these languages are naturally first-order, meaning that variables can only bind single elements, they also provide capabilities for filtering sets of events that occur inside iterative patterns; for example requiring sequences of numbers to be increasing. Unfortunately, these type of filters usually present ad-hoc syntax and under-defined semantics, precisely because variables cannot bind sets of events. As a result, CER languages that provide filtering of sequences commonly lack rigorous semantics and their expressive power is not understood. In this paper we embark on two tasks: First, to define a denotational semantics for CER that naturally allows to bind and filter sets of events; and second, to compare the expressive power of this semantics with that of CER languages that only allow for binding single events. Concretely, we introduce Set-Oriented Complex Event Logic (SO-CEL for short), a variation of the CER language introduced in [Grez et al., 2019] in which all variables bind to sets of matched events. We then compare SO-CEL with CEL, the CER language of [Grez et al., 2019] where variables bind single events. We show that they are equivalent in expressive power when restricted to unary predicates but, surprisingly, incomparable in general. Nevertheless, we show that if we restrict to sets of binary predicates, then SO-CEL is strictly more expressive than CEL. To get a better understanding of the expressive power, computational capabilities, and limitations of SO-CEL, we also investigate the relationship between SO-CEL and Complex Event Automata (CEA), a natural computational model for CER languages. We define a property on CEA called the *-property and show that, under unary predicates, SO-CEL captures precisely the subclass of CEA that satisfy this property. Finally, we identify the operations that SO-CEL is lacking to characterize CEA and introduce a natural extension of the language that captures the complete class of CEA under unary predicates.

We present the theoretical foundations of a new approach in centrality measures for graph data. The main principle of our approach is very simple: the more relevant subgraphs around a vertex, the more central it is in the network. We formalize the notion of "relevant subgraphs" by choosing a family of subgraphs that, give a graph G and a vertex v in G, it assigns a subset of connected subgraphs of G that contains v. Any of such families defines a measure of centrality by counting the number of subgraphs assigned to the vertex, i.e., a vertex will be more important for the network if it belongs to more subgraphs in the family. We show many examples of this approach and, in particular, we propose the all-subgraphs centrality, a centrality measure that takes every subgraph into account. We study fundamental properties over families of subgraphs that guarantee desirable properties over the corresponding centrality measure. Interestingly, all-subgraphs centrality satisfies all these properties, showing its robustness as a notion for centrality. Finally, we study the computational complexity of counting certain families of subgraphs and show a polynomial time algorithm to compute the all-subgraphs centrality for graphs with bounded tree width.

Descriptive Complexity has been very successful in characterizing complexity classes of decision problems in terms of the properties definable in some logics. However, descriptive complexity for counting complexity classes, such as FP and #P, has not been systematically studied, and it is not as developed as its decision counterpart. In this paper, we propose a framework based on Weighted Logics to address this issue. Specifically, by focusing on the natural numbers we obtain a logic called Quantitative Second Order Logics (QSO), and show how some of its fragments can be used to capture fundamental counting complexity classes such as FP, #P and FPSPACE, among others. We also use QSO to define a hierarchy inside #P, identifying counting complexity classes with good closure and approximation properties, and which admit natural complete problems. Finally, we add recursion to QSO, and show how this extension naturally captures lower counting complexity classes such as #L.

Regular expressions and automata models with capture variables are core tools in rule-based information extraction. These formalisms, also called regular document spanners , use regular languages to locate the data that a user wants to extract from a text document and then store this data into variables. Since document spanners can easily generate large outputs, it is important to have efficient evaluation algorithms that can generate the extracted data in a quick succession, and with relatively little precomputation time. Toward this goal, we present a practical evaluation algorithm that allows output-linear delay enumeration of a spanner’s result after a precomputation phase that is linear in the document. Although the algorithm assumes that the spanner is specified in a syntactic variant of variable-set automata, we also study how it can be applied when the spanner is specified by general variable-set automata, regex formulas, or spanner algebras. Finally, we study the related problem of counting the number of outputs of a document spanner and provide a fine-grained analysis of the classes of document spanners that support efficient enumeration of their results.

Complex Event Processing (CEP) has emerged as the unifying field for technologies that require processing and correlating distributed data sources in real-time. CEP finds applications in diverse domains, which has resulted in a large number of proposals for expressing and processing complex events. However, existing CEP languages lack from a clear semantics, making them hard to understand and generalize. Moreover, there are no general techniques for evaluating CEP query languages with clear performance guarantees. In this paper we embark on the task of giving a rigorous and efficient framework to CEP. We propose a formal language for specifying complex events, called CEL, that contains the main features used in the literature and has a denotational and compositional semantics. We also formalize the so-called selection strategies, which had only been presented as by-design extensions to existing frameworks. With a well-defined semantics at hand, we discuss how to efficiently process complex events by evaluating CEL formulas with unary filters. We start by studying the syntactical properties of CEL and propose rewriting optimization techniques for simplifying the evaluation of formulas. Then, we introduce a formal computational model for CEP, called complex event automata (CEA), and study how to compile CEL formulas with unary filters into CEA. Furthermore, we provide efficient algorithms for evaluating CEA over event streams using constant time per event followed by constant-delay enumeration of the results. Finally, we gather the main results of this work to present an efficient and declarative framework for CEP.

In this work, we study two simple yet general complexity classes, based on logspace Turing machines, which provide a unifying framework for efficient query evaluation in areas like information extraction and graph databases, among others. We investigate the complexity of three fundamental algorithmic problems for these classes: enumeration, counting and uniform generation of solutions, and show that they have several desirable properties in this respect. Both complexity classes are defined in terms of nondeterministic logspace transducers (NL transducers). For the first class, we consider the case of unambiguous NL transducers, and we prove constant delay enumeration, and both counting and uniform generation of solutions in polynomial time. For the second class, we consider unrestricted NL transducers, and we obtain polynomial delay enumeration, approximate counting in polynomial time, and polynomial-time randomized algorithms for uniform generation. More specifically, we show that each problem in this second class admits a fully polynomial-time randomized approximation scheme (FPRAS) and a polynomial-time Las Vegas algorithm for uniform generation. Interestingly, the key idea to prove these results is to show that the fundamental problem #NFA admits an FPRAS, where #NFA is the problem of counting the number of strings of length n accepted by a nondeterministic finite automaton (NFA). While this problem is known to be #P-complete and, more precisely, SpanL-complete, it was open whether this problem admits an FPRAS. In this work, we solve this open problem, and obtain as a welcome corollary that every function in SpanL admits an FPRAS.

Regular expressions and automata models with capture variables are core tools in rule-based information extraction. These formalisms, also called regular document spanners, use regular languages in order to locate the data that a user wants to extract from a text document, and then store this data into variables. Since document spanners can easily generate large outputs, it is important to have good evaluation algorithms that can generate the extracted data in a quick succession, and with relatively little precomputation time. Towards this goal, we present a practical evaluation algorithm that allows constant delay enumeration of a spanner's output after a precomputation phase that is linear in the document. While the algorithm assumes that the spanner is specified in a syntactic variant of variable set automata, we also study how it can be applied when the spanner is specified by general variable set automata, regex formulas, or spanner algebras. Finally, we study the related problem of counting the number of outputs of a document spanner, providing a fine grained analysis of the classes of document spanners that support efficient enumeration of their results.

Rule-based information extraction has lately received a fair amount of attention from the database community, with several languages appearing in the last few years. Although information extraction systems are intended to deal with semistructured data, all language proposals introduced so far are designed to output relations, thus making them incapable of handling incomplete information. To remedy the situation, we propose to extend information extraction languages with the ability to use mappings, thus allowing us to work with documents which have missing or optional parts. Using this approach, we simplify the semantics of regex formulas and extraction rules, two previously defined methods for extracting information. We extend them with the ability to handle incomplete data, and study how they compare in terms of expressive power. We also study computational properties of these languages, focusing on the query enumeration problem, as well as satisfiability and containment.

We present three pumping lemmas for three classes of functions definable by fragments of weighted automata over the min-plus semiring and the semiring of natural numbers. As a corollary we show that the hierarchy of functions definable by unambiguous, finitely-ambiguous, polynomially-ambiguous weighted automata, and the full class of weighted automata is strict for the min-plus semiring.

Probabilistic automata are a computational model introduced by Michael Rabin, extending nondeterministic finite automata with probabilistic transitions. Despite its simplicity, this model is very expressive and many of the associated algorithmic questions are undecidable. In this work we focus on the emptiness problem, which asks whether a given probabilistic automaton accepts some word with probability higher than a given threshold. We consider a natural and well-studied structural restriction on automata, namely the degree of ambiguity, which is defined as the maximum number of accepting runs over all words. We observe that undecidability of the emptiness problem requires infinite ambiguity and so we focus on the case of finitely ambiguous probabilistic automata. Our main results are to construct efficient algorithms for analysing finitely ambiguous probabilistic automata through a reduction to a multi-objective optimisation problem, called the stochastic path problem. We obtain a polynomial time algorithm for approximating the value of finitely ambiguous probabilistic automata and a quasi-polynomial time algorithm for the emptiness problem for 2-ambiguous probabilistic automata.

Descriptive Complexity has been very successful in characterizing complexity classes of decision problems in terms of the properties definable in some logics. However, descriptive complexity for counting complexity classes, such as FP and #P, has not been systematically studied, and it is not as developed as its decision counterpart. In this paper, we propose a framework based on Weighted Logics to address this issue. Specifically, by focusing on the natural numbers we obtain a logic called Quantitative Second Order Logics (QSO), and show how some of its fragments can be used to capture fundamental counting complexity classes such as FP, #P and FPSPACE, among others. We also use QSO to define a hierarchy inside #P, identifying counting complexity classes with good closure and approximation properties, and which admit natural complete problems. Finally, we add recursion to QSO, and show how this extension naturally captures lower counting complexity classes such as #L.

In this paper, we experimentally compare the efficiency of various database engines for the purposes of querying the Wikidata knowledge-base, which can be conceptualised as a directed edge-labelled graph where edges can be annotated with meta-information called qualifiers. We take two popular SPARQL databases (Virtuoso, Blazegraph), a popular relational database (PostgreSQL), and a popular graph database (Neo4J) for comparison and discuss various options as to how Wikidata can be represented in the models of each engine. We design a set of experiments to test the relative query performance of these representations in the context of their respective engines. We first execute a large set of atomic lookups to establish a baseline performance for each test setting, and subsequently perform experiments on instances of more complex graph patterns based on real-world examples. We conclude with a summary of the strengths and limitations of the engines observed. These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Cost register automata (CRA) and its subclass, copyless CRA, were recently proposed by Alur et al. as a new model for computing functions over strings. We study structural properties, expressiveness, and closure properties of copyless CRA. We show that copyless CRA are strictly less expressive than weighted automata and are not closed under reverse operation. To find a better class we impose restrictions on copyless CRA, which ends successfully with a new robust computational model that is closed under reverse and other extensions.

In this paper, we propose a simple and expressive framework for adding metadata to CSV documents and their noisy variants. The framework is based on annotating parts of the document that can be later used to read, query, or exchange the data. The core of our framework is a language based on extended regular expressions that are used for selecting data. These expressions are then combined using a set of rules in order to annotate the data. We study the computational complexity of implementing our framework and present an efficient evaluation algorithm that runs in time proportional to its output and linear in its input. As a proof of concept, we test an implementation of our framework against a large number of real world datasets and show that it can be efficiently used in practice.

We study the problem of bounded repairability of a given restriction tree language R into a target tree language T . More precisely, we say that R is bounded repairable with respect to T if there exists a bound on the number of standard tree editing operations necessary to apply to any tree in R to obtain a tree in T . We consider a number of possible specifications for tree languages: bottom-up tree automata (on curry encoding of unranked trees) that capture the class of XML schemas and document type definitions (DTDs). We also consider a special case when the restriction language R is universal (i.e., contains all trees over a given alphabet). We give an effective characterization of bounded repairability between pairs of tree languages represented with automata. This characterization introduces two tools—synopsis trees and a coverage relation between them—allowing one to reason about tree languages that undergo a bounded number of editing operations. We then employ this characterization to provide upper bounds to the complexity of deciding bounded repairability and show that these bounds are tight. In particular, when the input tree languages are specified with arbitrary bottom-up automata, the problem is coNExp-complete. The problem remains coNExp-complete even if we use deterministic nonrecursive DTDs to specify the input languages. The complexity of the problem can be reduced if we assume that the alphabet, the set of node labels, is fixed: the problem becomes PS pace -complete for nonrecursive DTDs and coNP-complete for deterministic nonrecursive DTDs. Finally, when the restriction tree language R is universal, we show that the bounded repairability problem becomes E xp -complete if the target language is specified by an arbitrary bottom-up tree automaton and becomes tractable (P-complete, in fact) when a deterministic bottom-up automaton is used.

It is highly desirable for a computational model to have a logic characterization like in the seminal work from Buchi that connects MSO with finite automata. For example, weighted automata are the quantitative extension of finite automata for computing functions over words and they can be naturally characterized by a subframent of weighted logic introduced by Droste and Gastin. Recently, cost register automata (CRA) were introduced by Alur et al. as an alternative model for weighted automata. In hope of finding decidable subclasses of weighted automata, they proposed to restrict their model with the so-called copyless restriction. Unfortunately, copyless CRA do not enjoy good closure properties and, therefore, a logical characterization of this class seems to be unlikely. In this paper, we introduce a new logic called maximal partition logic (MPL) for studying the expressiveness of copyless CRA. In contrast from the previous approaches (i.e. weighted logics), MPL is based on a new set of "regular" quantifiers that partition a word into maximal subwords, compute the output of a subformula over each subword separately, and then aggregate these outputs with a semiring operation. We study the expressiveness of MPL and compare it with weighted logics. Furthermore, we show that MPL is equally expressive to a natural subclass of copyless CRA. This shows the first logical characterization of copyless CRA and it gives a better understanding of the copyless restriction in weighted automata.

Integrity constraint management concerns both checking whether data is valid and taking action to restore correctness when invalid data is discovered. In XML the notion of valid data can be captured by schema languages such as Document Type Definitions (DTDs) and more generally XML schemas. DTDs have the property that constraint checking can be done in streaming fashion. In this paper we consider when the corresponding action to restore validity -- repair -- can be done in streaming fashion. We formalize this as the problem of determining, given a DTD, whether or not a streaming procedure exists that transforms an input document so as to satisfy the DTD, using a number of edits independent of the document. We show that this problem is decidable. In fact, we show the decidability of a more general problem, allowing a more general class of schemas than DTDs, and requiring a repair procedure that works only for documents that are already known to satisfy another class of constraints. The decision procedure relies on a new analysis of the structure of DTDs, reducing to a novel notion of game played on pushdown systems associated with the schemas.

While monadic second-order logic is a prominent logic for specifying languages of finite words, it lacks the power to compute quantitative properties, e.g. to count. An automata model capable of computing such properties are weighted automata, but logics equivalent to these automata have only recently emerged. We propose a new framework for adding quantitative properties to logics specifying Boolean properties of words. We use this to define Quantitative Monadic Second-Order Logic (QMSO). In this way we obtain a simple logic which is equally expressive to weighted automata. We analyse its evaluation complexity, both data and combined complexity, and show completeness results for combined complexity. We further refine the analysis of this logic and obtain fragments that characterise exactly subclasses of weighted automata defined by the level of ambiguity allowed in the automata. In this way, we define a quantitative logic which has good decidability properties while being resonably expressive and enjoying a simple syntactical definition.

We consider the problem of repairing unranked trees (e.g., XML documents) satisfying a given restriction specification R (e.g., a DTD) into unranked trees satisfying a given target specification T. Specifically, we focus on the question of whether one can get from any tree in a regular language R to some tree in another regular language T with a finite, uniformly bounded, number of edit operations (i.e., deletions and insertions of nodes). We give effective characterizations of the pairs of specifications R and T for which such a uniform bound exists, and we study the complexity of the problem under different representations of the regular tree languages (e.g., non-deterministic stepwise automata, deterministic stepwise automata, DTDs). Finally, we point out some connections with the analogous problem for regular languages of words, which was previously studied in [6].

What do you do if a computational object (e.g. program trace) fails a specification? An obvious approach is to perform repair: modify the object minimally to get something that satisfies the constraints. In this paper we study repair of temporal constraints, given as automata or temporal logic formulas. We focus on determining the number of repairs that must be applied to a word satisfying a given input constraint in order to ensure that it satisfies a given target constraint. This number may well be unbounded; one of our main contributions is to isolate the complexity of the "bounded repair problem", based on a characterization of the pairs of regular languages that admit such a repair. We consider this in the setting where the repair strategy is unconstrained and also when the strategy is restricted to use finite memory. Although the streaming setting is quite different from the general setting, we find that there are surprising connections between streaming and non-streaming, as well as within variants of the streaming problem.

In the last few years, a lot of attention has been paid to the specification and subsequent manipulation of schema mappings, a problem which is of fundamental importance in metadata management. There have been many achievements in this area, and semantics have been defined for operators on schema mappings such as composition and inverse. However, little research has been pursued towards providing formal tools to compare schema mappings, in terms of their ability to transfer data and avoid storing redundant information, which has hampered the development of foundations for more complex operators as many of them involve these notions.

The inversion of schema mappings has been identified as one of the fundamental operators for the development of a general framework for metadata management. In fact, during the last years three alternative notions of inversion for schema mappings have been proposed (Fagin-inverse [10], quasi-inverse [14] and maximum recovery [2]). However, the procedures that have been developed for computing these operators have some features that limit their practical applicability. First, these algorithms work in exponential time and produce inverse mappings of exponential size. Second, these algorithms express inverses in some mappings languages which include features that are difficult to use in practice. A typical example is the use of disjunction in the conclusion of the mapping rules, which makes the process of exchanging data much more complicated. In this paper, we propose solutions for the two problems mentioned above. First, we provide a polynomial time algorithm that computes the three inverse operators mentioned above given a mapping specified by a set of tuple-generating dependencies (tgds). This algorithm uses an output mapping language that can express these three operators in a compact way and, in fact, can compute inverses for a much larger class of mappings. Unfortunately, it has already been proved that this type of mapping languages has to include some features that are difficult to use in practice and, hence, this is also the case for our output mapping language. Thus, as our second contribution, we propose a new and natural notion of inversion that overcomes this limitation. In particular, every mapping specified by a set of tgds admits an inverse under this new notion that can be expressed in a mapping language that slightly extends tgds, and that has the same good properties for data exchange as tgds. Finally, as our last contribution, we provide an algorithm for computing such inverses.

A schema mapping is a specification that describes how data from a source schema is to be mapped to a target schema. Once the data has been transferred from the source to the target, a natural question is whether one can undo the process and recover the initial data, or at least part of it. In fact, it would be desirable to find a reverse schema mapping from target to source that specifies how to bring the exchanged data back. In this article, we introduce the notion of a recovery of a schema mapping: it is a reverse mapping, M′ for a mapping M, that recovers sound data with respect to M. We further introduce an order relation on recoveries. This allows us to choose mappings that recover the maximum amount of sound information. We call such mappings maximum recoveries. We study maximum recoveries in detail, providing a necessary and sufficient condition for their existence. In particular, we prove that maximum recoveries exist for the class of mappings specified by FO-to-CQ source-to-target dependencies. This class subsumes the class of source-to-target tuple-generating dependencies used in previous work on data exchange. For the class of mappings specified by FO-to-CQ dependencies, we provide an exponential-time algorithm for computing maximum recoveries, and a simplified version for full dependencies that works in quadratic time. We also characterize the language needed to express maximum recoveries, and we include a detailed comparison with the notion of inverse (and quasi-inverse) mapping previously proposed in the data exchange literature. In particular, we show that maximum recoveries strictly generalize inverses. We finally study the complexity of some decision problems related to the notions of recovery and maximum recovery.

A schema mapping is a specification that describes how data from a source schema is to be mapped to a target schema. Once the data has been transferred from the source to the target, a natural question is whether one can undo the process and recover the initial data, or at least part of it. In fact, it would be desirable to find a reverse schema mapping from target to source that specifies how to bring the exchanged data back.In this paper, we introduce the notion of a recovery of a schema mapping: it is a reverse mapping M' for a mapping M that recovers sound data with respect to M. We further introduce an order relation on recoveries. This allows us to choose mappings that recover the maximum amount of sound information. We call such mappings maximum recoveries. We study maximum recoveries in detail, providing a necessary and sufficient condition for their existence. In particular, we prove that maximum recoveries exist for the class of mappings specified by FO-to-CQ source-to-target dependencies. This class subsumes the class of source-to-target tuple-generating dependencies used in previous work on data exchange. For the class of mappings specified by FO-to-CQ dependencies, we provide an exponential-time algorithm for computing maximum recoveries, and a simplified version for full dependencies that works in quadratic time. We also characterize the language needed to express maximum recoveries, and we include a detailed comparison with the notion of inverse (and quasi-inverse) mapping previously proposed in the data exchange literature. In particular, we show that maximum recoveries strictly generalize inverses. We study the complexity of some decision problems related to the notions of recovery and maximum recovery. Finally, we report our initial results about a relaxednotion of maximal recovery, showing that it strictly generalizes the notion of maximum recovery.