Tomoyuki Uchida

Discovery of Frequent Tree Structured Patterns in Semistructured Web Documents

Many documents such as Web documents or XML files have no rigid structure. Such semistructured documents have been rapidly increasing. We propose a new method for discovering frequent tree structured patterns in semistructured Web documents. We consider the data mining problem of finding all maximally frequent tag tree patterns in semistructured data such as Web documents. A tag tree pattern is an edge labeled tree which has hyperedges as variables. An edge label is a tag or a keyword inWeb documents, and a variable can be substituted by any tree. So a tag tree pattern is suited for representing tree structured patterns in semistructured Web documents. We present an algorithm for finding all maximally frequent tag tree patterns. Also we report some experimental results on XML documents by using our algorithm.

Polynomial Time Algorithms for Finding Unordered Tree Patterns with Internal Variables

Many documents such as Web documents or XML files have tree structures. A term tree is an unordered tree pattern consisting of internal variables and tree structures. In order to extract meaningful and hidden knowledge from such tree structured documents, we consider a minimal language (MINL) problem for term trees. The MINL problem for term trees is to find a term tree t such that the language generated by t is minimal among languages, generated by term trees, which contain all given tree structured data. Firstly, we show that the MINL problem for regular term trees is computable in polynomial time if the number of edge labels is infinite. Next, we show that the MINL problems with optimizing the size of an output term tree are NP-complete. Finally, in order to show that our polynomial time algorithm for the MINL problem can be applied to data mining from real-world Web documents, we show that regular term tree languages are polynomial time inductively inferable from positive data if the number of edge labels is infinite.

Polynomial Time Matching Algorithms for Tree-Like Structured Patterns in Knowledge Discovery

Graphs have enough richness and flexibility to express discrete structures hidden in a large amount of data. Some searching methods utilizing graph algorithmic techniques have been developed in Knowledge Discovery. A term graph, which is one of expressions for graph-structured data, is a hypergraph whose hyperedges are regarded as variables. Although term graphs can represent complicated patterns found from structured data, it is hard to do pattern match and pattern search in them. We have been studying subclasses of term graphs, called regular term trees, which are suited for expressing tree-like structured data. In this paper, we consider a matching problem for a regular term tree t and a standard tree T, which decides whether or not there exists a tree T′ such that T′ is isomorphic to T and T′ is obtained by replacing variables in t with some trees. First we show that the matching problem for a regular term tree and a tree is NP-complete even if each variable in the regular term tree contains only 4 vertices. Next we give a polynomial time algorithm for solving the matching problem for a regular term tree and a tree of bounded degree such that the regular term tree has only variables consisting the constant number of vertices greater than one. We also report some computational experiments and compare our algorithm with a naive algorithm.

Designing Views in HypothesisCreator: System for Assisting in Discovery

We discuss the significance of designing views on data in a computational system assisting scientists in the process of discovery. A view on data is considered as a particular way to interpret the data. In the scientific literature, devising a new view capturing the essence of data is a key to discovery. A system HYPOTHESISCREATOR, which we have been developing to assist scientists in the process of discovery, supports users designing views on data and have the function of searching for good views on the data. In this paper we report a series of computational experiments on scientific data with HypothesisCreator and analyses of the produced hypotheses, some of which select several views good for explaining given data, searched and selected from over ten millions of designed views. Through these experiments we have convinced that view is one of the important factors in discovery process, and that discovery systems should have an ability of designing and selecting views on data in a systematic way so that experts on the data can employ their knowledge and thoughts efficiently for their purposes.

Toward Genomic Hypothesis Creator: View Designer for Discovery

Software tools for genomic researches like homology search are very useful and have contributed on the progress of the genomic researches. However, these tools are not designed directly toward scientific discovery and more discovery-oriented software tools are strongly expected to assist scientific discovery in genomic researches. We have designed and developed a multistrategic and discovery-oriented system Genomic Hypothesis Creator by introducing two notions: view on data and view space on data. With these newly defined notions, we describe a View Designer, a component of Genomic Hypothesis Creator, which dynamically creates new views on data and searches a view space for more appropriate views. A good view obtained from Genomic Hypothesis Creator makes it possible for us to understand the data and eventually attain to the goal of discovery. Genomic Hypothesis Creator can be extended by adding users own views on data and hypothesis generators into the system with plug-in interfaces. Therefore it would be feasible to apply this system to other problems than genomic researches.

Discovering New Knowledge from Graph Data Using Inductive Logic Programming

We present a method for discovering new knowledge from structural data which are represented by graphs in the framework of inductive logic programming. A graph, or network, is widely used for representing relations between various data and expressing a small and easily understandable hypothesis. Formal Graph System (FGS) is a kind of logic programming system which directly deals with graphs just like first order terms. By employing refutably inductive inference algorithms and graph algorithmic techniques, we are developing a knowledge discovery system KD-FGS, which acquires knowledge directly from graph data by using FGS as a knowledge representation language. n nIn this paper we develop a logical foundation of our knowledge discovery system. A term tree is a pattern which consists of variables and treelike structures. We give a polynomial-time algorithm for finding a unifier of a term tree and a tree in order to make consistency checks efficiently. Moreover we give experimental results on some graph theoretical notions with the system. The experiments show that the system is useful for finding new knowledge.

KD-FGS: A Knowledge Discovery System from Graph Data Using Formal Graph System

A graph is one of the most common abstract structures and is suitable for representing relations between various objects. The analyzing system directly manipulating graphs is useful for knowledge discovery. Formal Graph System (FGS) is a kind of logic programming system which directly deals with graphs just like first order terms. We have designed and implemented a knowledge discovery system KD-FGS, which receives the graph data and produces a hypothesis by using FGS as a knowledge representation language. The system consists of an FGS interpreter and a refutably inductive inference algorithm for FGSs. We report some experiments of running KD-FGS and confirm that the system is useful for knowledge discovery from graph data.

A New Framework for Discovering Knowledge from Two-Dimensional Structured Data Using Layout Formal Graph System

We present a new framework for discovering knowledge from two-dimensional structured data by using Inductive Logic Programming. Two-dimensional graph structured data such as image or map data are widely used for representing relations and distances between various objects. First, we define a layout term graph suited for representing two-dimensional graph structured data. A layout term graph is a pattern consisting of variables and two-dimensional graph structures. Moreover, we propose Layout Formal Graph System (LFGS) as a new logic programming system having a layout term graph as a term. LFGS directly deals with graphs having positional relations just like first order terms. Second, we show that LFGS is more powerful than Layout Graph Grammar, which is a generating system consisting of a context-free graph grammar and positional relations. This indicates that LFGS has the richness and advantage of representing knowledge about two-dimensional structured data. n nFinally, we design a knowledge discovery system, which uses LFGS as a knowledge representation language and refutably inductive inference as a learning method. In order to give a theoretical foundation of our knowledge discovery system, we give the set of weakly reducing LFGS programs which is a sufficiently large hypothesis space of LFGS programs and show that the hypothesis space is refutably inferable from complete data.

Learning Block-Preserving Outerplanar Graph Patterns and Its Application to Data Mining

An outerplanar graph is a planar graph which can be embedded in the plane in such a way that all of vertices lie on the outer boundary. Many chemical compounds are known to be expressed by outerplanar graphs. We proposed a block preserving outerplanar graph pattern (bpo- graph pattern, for short) as a graph pattern common to a set of outerplanar graphs like a dataset of chemical compounds. In this paper, firstly we give a polynomial time algorithm for finding a minimally generalized bpo- graph pattern explaining a given set of outerplanar graphs. Secondly we give a pattern mining algorithm for enumerating all maximal frequent bpo- graph patterns in a set of outerplanar graphs. Finally, in order to show the performance of the pattern mining algorithm, we report experimental results on chemical datasets.

A Context-Aware Fitness Function Based on Feature Selection for Evolutionary Learning of Characteristic Graph Patterns

We propose a context-aware fitness function based on feature selection for evolutionary learning of characteristic graph patterns. The proposed fitness function estimates the fitness of a set of correlated individuals rather than the sum of fitness of the individuals, and specifies the fitness of an individual as its contribution degree in the context of the set. We apply the proposed fitness function to our evolutionary learning, based on Genetic Programming, for obtaining characteristic graph patterns from positive and negative graph data. We report some experimental results on our evolutionary learning of characteristic graph patterns, using the context-aware fitness function and a previous fitness function ignoring context.