The proceedings of FCT 2023 contains papers focusing on algorithms, complexity, formal and logical methods and database theory.

We consider the problem of constructing a Shape Expression Schema (ShEx) that describes the structure of a given input RDF graph. We employ the framework of grammatical inference, where the objective is to find an inference algorithm that is both sound i.e., always producing a schema that validates the input RDF graph, and complete i.e., able to produce any schema, within a given class of schemas, provided that a sufficiently informative input graph is presented. We study the case where the input graph is typed i.e., every node is given with its types. We limit our attention to a practical fragment ShEx0 of Shape Expressions Schemas that has an equivalent graphical representation in the form of shape graphs. We investigate the problem of constructing a canonical representative of a given shape graph. Finally, we present a sound and complete algorithm for shape graphs thus showing that ShEx0 is learnable from typed graphs.

We introduce the notion of high-order deterministic top-down tree transducers (HODT) whose outputs correspond to single-typed lambda-calculus formulas. These transducers are natural generalizations of known models of top-tree transducers such as: Deterministic Top-Down Tree Transducers, Macro Tree Transducers, Streaming Tree Transducers... We focus on the linear restriction of high order tree transducers with look-ahead (HODTR lin), and prove this corresponds to tree to tree functional transformations defined by Monadic Second Order (MSO) logic. We give a specialized procedure for the composition of those transducers that uses a flow analysis based on coherence spaces and allows us to preserve the linearity of transducers. This procedure has a better complexity than classical algorithms for composition of other equivalent tree transducers, but raises the order of transducers. However, we also indicate that the order of a HODTR lin can always be bounded by 3, and give a procedure that reduces the order of a HODTR lin to 3. As those resulting HODTR lin can then be transformed into other equivalent models, this gives an important insight on composition algorithm for other classes of transducers. Finally, we prove that those results partially translate to the case of almost linear HODTR: the class corresponds to the class of tree transformations performed by MSO with unfolding (not closed by composition), and provide a mechanism to reduce the order to 3 in this case.

In a context where most of our data is online, information comes from a great diversity of actors, and is now stored in a great variety of formats. This ranges from highly structured formats such as traditional databases to simple texts, and includes many intermediary formats - such as XML, HTML, JSON or HTML for instance - usually grouped under the term semi-structured data. This diversity in format, but also the wide range of tools that are needed to manipulate those data, coupled with the fact that their structuration is often underspecified is a problem for the end-user as it can be hard for a non-expert to simply extract the data he truly needs. Machine learning can provide solutions to automatically design tools that helps the user to query or transform semistructured data. In this work, we investigate how technics issued from grammatical inference can be adapted in this framework.

HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés. Machine Learning Techniques for Semistructured Data Aurélien Lemay

Massive graph data sets are pervasive in contemporary application domains. Hence, graph database systems are becoming increasingly important. In the experimental study of these systems, it is vital that the research community has shared solutions for the generation of database instances and query workloads having predictable and controllable properties. We present the design and engineering principles of gMark, a domain- and query language-independent graph instance and query workload generator. A core contribution of gMark is its ability to target and control the diversity of properties of both the generated instances and the generated workloads coupled to these instances. Further novelties include support for regular path queries, a fundamental graph query paradigm, and schema-driven selectivity estimation of queries, a key feature in controlling workload chokepoints. We illustrate the flexibility and practical usability of gMark by showcasing the framework’s capabilities in generating high quality graphs and workloads, and its ability to encode user-defined schemas across a variety of application domains.

We study the problem of how to learn tree transformations on a given regular tree domain from a finite sample of input-output examples. We assume that the target tree transformation can be defined by a deterministic top-down tree transducer with regular domain inspection (DTOPi:reg). An RPNI style learning algorithm that solves this problem in polynomial time and with polynomially many examples was presented at Podsu00272010, but restricted to the case of path-closed regular domains. In this paper, we show that this restriction can be removed. For this, we present a new normal form for DTOPi:reg by extending the Myhill-Nerode theorem for DTOP to regular domain inspections in a nontrivial manner. The RPNI style learning algorithm can also be lifted but becomes more involved too.

Graph data management tools are nowadays evolving at a great pace. Key drivers of progress in the design and study of data intensive systems are solutions for synthetic generation of data and workloads, for use in empirical studies. Current graph generators, however, provide limited or no support for workload generation or are limited to fixed use-cases. Towards addressing these limitations, we demonstrate gMark, the first domain-and query language-independent framework for synthetic graph and query workload generation. Its novel features are: (i) fine-grained control of graph instance and query workload generation via expressive user-defined schemas; (ii) the support of expressive graph query languages, including recursion among other features; and, (iii) selectivity estimation of the generated queries. During the demonstration, we will showcase the highly tunable generation of graphs and queries through various user-defined schemas and targeted selectivities, and the variety of supported practical graph query languages. We will also show a performance comparison of four state-of-the-art graph database engines, which helps us understand their current strengths and desirable future extensions.

Les outils de gestion de bases de donnees orientees graphes sont actuellement en evolution permanente. Dans ce contexte, les outils permettant une generation de donnees ainsi que de scenarios de travail sont des elements cles dans les etudes empiriques. Neanmoins, les generateurs actuels de graphe fournissent un support limite, voire inexistant, dans la generation de scenarios d'usage, ou s'en tiennent a un nombre fixe de cas d'etudes. Afin de depasser ces limitations, nous presentons gMark, le premier environnement de generation de graphes synthetiques et de scenarios d'utilisation independant a la fois du domaine et du langage de requete. Ses principales innovations sont : (i) un controle precis de l'instance de graphe genere et des requetes correspondantes a partir de schemas definis par l'utilisateur ; (ii) le support d'un langage de requete expressif, incluant notamment la recursion ; et (iii) une estimation de la selectivite des requetes generees. Durant cette demonstration, nous illustrerons la capacite hautement parametrable des graphes et des requetes generes a travers differents schemas et differentes selectivites choisies, et la variete des langages de requetes supportes par le systeme. Nous montrerons egalement une comparaison des performances de quatre moteurs de requetes de l'etat de l'art, et comment les scenarios generes par gMark nous permettent de comprendre leurs forces respectives, et les evolutions que l'on pourrait souhaiter.

We study the problem of how to learn tree transformations on a given regular tree domain from a finite sample of input-output examples. We assume that the target tree transformation can be defined by a deterministic top-down tree transducer with regular domain inspection (DTOPi:reg). An RPNI style learning algorithm that solves this problem in polynomial time and with polynomially many examples was presented at Pods'2010, but restricted to the case of path-closed regular domains. In this paper, we show that this restriction can be removed. For this, we present a new normal form for DTOPi:reg by extending the Myhill-Nerode theorem for DTOP to regular domain inspections in a nontrivial manner. The RPNI style learning algorithm can also be lifted but becomes more involved too.

We investigate the problem of learning graph queries by exploiting user examples. The input consists of a graph database in which the user has labeled a few nodes as positive or negative examples, depending on whether or not she would like the nodes as part of the query result. Our goal is to handle such examples to find a query whose output is what the user expects. This kind of scenario is pivotal in several application settings where unfamiliar users need to be assisted to specify their queries. In this paper, we focus on path queries defined by regular expressions, we identify fundamental difficulties of our problem setting, we formalize what it means to be learnable, and we prove that the class of queries under study enjoys this property. We additionally investigate an interactive scenario where we start with an empty set of examples and we identify the informative nodes i.e., those that contribute to the learning process. Then, we ask the user to label these nodes and iterate the learning process until she is satisfied with the learned query. Finally, we present an experimental study on both real and synthetic datasets devoted to gauging the effectiveness of our learning algorithm and the improvement of the interactive approach.

Massive graph data sets are pervasive in contemporary application domains. Hence, graph database systems are becoming increasingly important. In the study of these systems, it is vital that the research community has shared benchmarking solutions for the generation of database instances and query workloads having predictable and controllable properties. Similarly to TPC benchmarks for relational databases, benchmarks for graph databases have been important drivers for the Semantic Web and graph data management communities. In this paper, we present the design and engineering principles of gMark, a domain- and query language-independent graph benchmark exhibiting flexible schema and workload chokepoints. A core contribution of gMark is its ability to target and control the diversity of properties of both the generated graph instances and the generated query workloads coupled to these instances. A further novelty is the support of recursive regular path queries, a fundamental graph query paradigm. We illustrate the flexibility and practical usability of gMark by showcasing the frameworku0027s capabilities in generating high quality graphs and workloads, and its ability to encode user-defined schemas across a variety of application domains.

We present an efficient algorithm for testing approximate dtd validity modulo the strong tree edit distance. Our algorithm inspects xml documents in a probabilistic manner. It detects with high probability the nonvalidity of xml documents with a large fraction of errors, measured in terms of the strong tree edit distance from the dtd. The run time depends polynomially on the depth of the xml document tree but not on its size, so that it is sublinear in most cases (because in practice XML documents tend to be shallow). Therefore, our algorithm can be used to speed up exact dtd validators that run in linear time.

We introduce a class of deterministic sequential top-down tree-to-word transduc- ers (STWs) and investigate a number of its fundamental properties and related problems. First, we investigate the problem of normalization of STWs: we identify a subclass of earliest STWs (eSTWs) that is as expressive as stws and present an effective procedure for converting an arbitrary STW into an equivalent eSTW. We then present a Myhill-Nerode characterization of the class of the transformations definable with STWs which also shows that every transformation defined with an stw has a unique canonical representative eSTW. This canonical eSTW is the minimal eSTW defining the same transformation, and consequently, we present a polynomial minimization procedure for eSTWs, thus giving an effective procedure for constructing the canonical representative of any transformation definable with STWs. Finally, we use the Myhill-Nerode characterization to devise an algorithm for inference (learning) of eSTWs from examples of transformation given by the user. Available at http://chercheurs.lille.inria.fr/~niehren/learning-stw/0.pdf

Graph databases are becoming pervasive in several application scenarios such as the Semantic Web, social and biological networks, and geographical databases, to name a few. However, specifying a graph query is a cumbersome task for non-expert users because graph databases (i) are usually of large size hence difficult to visualize and (ii) do not carry proper metadata as there is no clear distinction between the instances and the schemas. We present GPS, a system for interactive path query specification on graph databases, which assists the user to specify path queries defined by regular expressions. The user is interactively asked to visualize small fragments of the graph and to label nodes of interest as positive or negative, depending on whether or not she would like the nodes as part of the query result. After each interaction, the system prunes the uninformative nodes i.e., those that do not add any information about the user's goal query. Thus, the system also guides the user to specify her goal query with a minimal number of interactions.

Massive graph data sets are pervasive in contemporary application domains. Hence, graph database systems are becoming increasingly important. In the study of these systems, it is vital that the research community has shared benchmarking solutions for the generation of database instances and query workloads having predictable and controllable properties. Similarly to TPC benchmarks for relational databases, benchmarks for graph databases have been important drivers for the Semantic Web and graph data management communities. Current benchmarks, however, are either limited to fixed graphs or graph schemas, or provide limited or no support for generating tailored query workloads to accompany graph instances. To move the community forward, a benchmarking approach which overcomes these limitations is crucial. In this paper, we present the design and engineering principles of gMark, a domain- and query language-independent graph benchmark addressing these limitations of current solutions. A core contribution of gMark is its ability to target and control the diversity of properties of both the generated graph instances and the generated query workloads coupled to these instances. A further novelty is the support of recursive regular path queries, a fundamental graph query paradigm. We illustrate the flexibility and practical usability of gMark by showcasing the framework's capabilities in generating high quality graphs and workloads, and its ability to encode user-defined schemas across a variety of application domains.

We study the problem of learning sequential top-down tree-to-word transducers stws. First, we present a Myhill-Nerode characterization of the corresponding class of sequential tree-to-word transformations ${\mathcal{STW}}$ . Next, we investigate what learning of stws means, identify fundamental obstacles, and propose a learning model with abstain. Finally, we present a polynomial learning algorithm.

Specifying a database query using a formal query language is typically a challenging task for non-expert users. In the context of big data, this problem becomes even harder as it requires the users to deal with database instances of big sizes and hence difficult to visualize. Such instances usually lack a schema to help the users specify their queries, or have an incomplete schema as they come from disparate data sources. In this paper, we propose a novel paradigm for interactive learning of queries on big data, without assuming any knowledge of the database schema. The paradigm can be applied to different database models and a class of queries adequate to the database model. In particular, in this paper we present two instantiations that validated the proposed paradigm for learning relational join queries and for learning path queries on graph databases. Finally, we discuss the challenges of employing the paradigm for further data models and for learning cross-model schema mappings.