Machine Learning is the study of methods for programming computers to learn. Computers are applied to a wide range of tasks, and for most of these it is relatively easy for programmers to design and implement the necessary software. However, there are many tasks for which this is difficult or impossible. These can be divided into four general categories. First, there are problems for which there exist no human experts. For example, in modern automated manufacturing facilities, there is a need to predict machine failures before they occur by analyzing sensor readings. Because the machines are new, there are no human experts who can be interviewed by a programmer to provide the knowledge necessary to build a computer system. A machine learning system can study recorded data and subsequent machine failures and learn prediction rules. Second, there are problems where human experts exist, but where they are unable to explain their expertise. This is the case in many perceptual tasks, such as speech recognition, hand-writing recognition, and natural language understanding. Virtually all humans exhibit expert-level abilities on these tasks, but none of them can describe the detailed steps that they follow as they perform them. Fortunately, humans can provide machines with examples of the inputs and correct outputs for these tasks, so machine learning algorithms can learn to map the inputs to the outputs. Third, there are problems where phenomena are changing rapidly. In finance, for example, people would like to predict the future behavior of the stock market, of consumer purchases, or of exchange rates. These behaviors change frequently, so that even if a programmer could construct a good predictive computer program, it would need to be rewritten frequently. A learning program can relieve the programmer of this burden by constantly modifying and tuning a set of learned prediction rules. Fourth, there are applications that need to be customized for each computer user separately. Consider, for example, a program to filter unwanted electronic mail messages. Different users will need different filters. It is unreasonable to expect each user to program his or her own rules, and it is infeasible to provide every user with a software engineer to keep the rules up-to-date. A machine learning system can learn which mail messages the user rejects and maintain the filtering rules automatically. Machine learning addresses many of the same research questions as the fields of statistics, data mining, and psychology, but with differences of emphasis. Statistics focuses on understanding the phenomena that have generated the data, often with the goal of testing different hypotheses about those phenomena. Data mining seeks to find patterns in the data that are understandable by people. Psychological studies of human learning aspire to understand the mechanisms underlying the various learning behaviors exhibited by people (concept learning, skill acquisition, strategy change, etc.).

Machine learning

This paper studies the polynomial-time learnability of the classes of ordered gapped tree patterns (OGT) and ordered gapped forests (OGF) under the into-matching semantics in the query learning model of Angluin. The class OGT is a model of semi-structured database query languages, and a generalization of both the class of ordered/unordered tree pattern languages and the class of non-erasing regular pattern languages. First, we present a polynomial time learning algorithm for µ-OGT, the subclass of OGT without repeated tree variables, using equivalence queries and membership queries. By extending this algorithm, we present polynomial time learning algorithms for the classes µ-OGF of forests without repeated variables and OGT of trees with repeated variables using equivalence queries and subset queries. We also give representation-independent hardness results which indicate that both of equivalence and membership queries are necessary to learn µ-OGT.

Efficient Learning of Semi-structured Data from Queries

Parallelism constraints are logical descriptions of trees. Parallelism constraints subsume dominance constraints and are equal in expressive power to context unification. Parallelism constraints belong to the constraint language for lambda structures (CLLS) which serves for modeling natural language semantics. In this paper, we investigate the extension of parallelism constraints by tree regular constraints. This canonical extension is subsumed by the monadic second-order logic over parallelism constraints. We analyze the precise expressiveness of this extension on basis of a new relationship between tree automata and logic. Our result is relevant for classifying different extensions of parallelism constraints, as in CLLS. Finally, we prove that parallelism constraints and context unification remain equivalent when extended with tree regular constraints.

Parallelism and tree regular constraints

The paper studies the evaluation of regular path queries on semi-structured data, i.e. path queries of the form nd all objects reachable by path whose labels form a word in p where p is a regular expression. We use local information expressed in the form of path inclusions in the optimization of path queries. These constraints are of the form L v where L is a regular language and v is a word; L v means that the answer to L from a given site is included in the answer to v. We call these inclusions RW-path inclusions. The boundedness problem for regular path queries p with respect to a nite set of path inclusions E is to decide whether the answer to p is included in the answer to a nite path query. We prove that the boundedness problem is decidable for regular path queries w.r.t. a set of RW-path inclusions, extending a result of Abiteboul and Vianu in 1].

On decidability of boundedness property for regular path queries

We advocate to analyze the average complexity of learning problems. An appropriate framework for this purpose is introduced. Based on it we consider the problem of learning monomials and the special case of learning monotone monomials in the limit and for on-line predictions in two variants: from positive data only, and from positive and negative examples. The well-known Wholist algorithm is completely analyzed with respect to its average-case behavior with respect to the class of binomial distributions. We consider di erent complexity measures: the number of mind changes, the number of prediction errors, and the total learning time. Tight bounds are obtained implying that worst case bounds are too pessimistic. On the average learning can be achieved exponentially faster. Furthermore, we study a new learning model, stochastic nite learning, in which, in contrast to PAC learning, some information about the underlying distribution is given and the goal is to nd a correct (not only approximatively correct) hypothesis. We develop techniques to obtain good bounds for stochastic nite learning from a precise average case analysis of strategies for learning in the limit and illustrate our approach for the case of learning monomials.

Analyzing the Average-Case Behavior of Conjunctive Learning Algorithms

The Web of data with meaning in the sense that a computer program can learn enough about what the data means to process it. During the last decade, the World Wide Web has evolved into the most important public data store on world. An important challenge for computer science today is to develop accurate information extraction and question answering mechanisms for the Web. Berners-Lee points out the difficulty of that task, and that it might even require more adequate formats of Web data representation. Information must be structured, and structure should reflect semantic information, so that machines can learn enough about what the data means. The standard document formats of the Web today, HTML and XML, rely on tree structures that encompass textual information. In this project we want to incorporate novel approaches for modeling tree structure and emerging techniques for machine learning into adaptive information extraction systems for the Web. In the future, we might also have to account for semantic information. The research team is a joint project team with the lifl (CNRS and Lille 1 University) and the Grappa Group (Lille 3 University). The project will be located in Lille 3 University.

Modeling Tree Structures, Machine Learning, and Information Extraction

The evolution of XML into the major document exchange format has reawaken a strong interest in modeling tree structures. The main objectives are to query for nodes in trees ( XPath ), to validity of documents, i.e., membership to a set of valid trees ( DTD ), and to transform trees into others ( XSLT ). Modelling approaches rely on tree automata, monadic second-order logics, modal logics, pattern languages, attribute grammars, or tree transducers. The main goal of the project is to design adaptive information extraction systems that fully exploit the tree structures of XML or HTML documents. In our approach we want to combine novel models of tree structures and machine learning techniques. Appropriate languages for modeling node queries in tree satisfy a number of properties required for adaptive information extraction. Possible trade-offs between expressiveness, learnability, and efﬁciency are to be understood. We thus propose new query languages in trees and investigate them from these perspectives. We consider tree automata for ranked and unranked trees, that may be ordered or unordered. We design efﬁcient algorithms for querying semi-structured data by automata. We investigate logical query languages such as monadic Datalog, monadic second-order logic, and modal logical languages. of is are mediator services, in information retrieval and in mining is wrappers are in Business Intelligence and Knowledge Management. we have proved decidable. Future works include a study of learnability of n-ary queries. In [17], we present node selecting tree transducer (NSTT) for representing monadic queries. NSTTs are the query formalism underlying our Squirrel tool. We have shown that deterministic NSTTs can be identiﬁed with monadic queries by unambiguous tree automata. In [18], we propose stepwise tree automata for querying unranked trees equally to binary trees. Stepwise tree automata are traditional tree automata. Stepwise tree automata simplify the often-needed query transfer ranked to unranked trees. This comes with their algebra nature. We have investigated related logics and constraints in ordered trees that are relevant to subtyping in programming languages [12][24] and for underspeciﬁed semantics of natural language [14]. In [15], we investigate an alternative modal logical querying formalism, the spatial logic TQL proposed by Cardelli and Ghelli in the context of the ambient calculus. We exhaustively study combined and data complexity of the model checking problem of TQL and its fragments, and equally investigate satisﬁability. We characterize fragments of TQL that have the same expressive power as MSO. In [27][26], we link TQL to languages with counting constraints such as Presburger monadic second-order logic (PMSO) introduced by Seidl and Schwentick or counting monadic second-order logic (CMSO) introduced by Courcelle in a uniform tree automata framework (see also [11]). sets C of path inclusions p (cid:22) u where p is a regular set of paths and u is a singleton path, which have a ﬁnite exact model. We present an effective way of computing such a model when it exists. (see also [19]). When considering heterogeneous HTML and XML documents as inputs of information extraction tasks – that may be either scattered into pieces or hidden in large parts of purely textual data – different tree wrappers need to be combined with textual wrappers. For classiﬁcation tasks, some machine learning algorithms realize such combinations. We have started to examine how such methods for information extraction. In a ﬁrst step, we reformulate information extraction as classiﬁcation tasks. Second we want to apply known techniques and combine them with structural wrapper induction. The CafeIn framework ( [31],[23]) improve previous approaches based on supervised classiﬁcation for information extraction ( [30], [29]). CafeIn is a single-slot wrapper induction framework for text data or semi-structured data. It combines an adaptive attribute-value representation of documents and a supervised classiﬁcation algorithm. For text data, the classiﬁcation task consists in deciding whether a tuple of two positions in a text are respectively the beginning and the end of data to extract. The document representation use only textual features. For semi-structured data, like XML or HTML trees, data to extract are contained in leaves. Thus the classiﬁcation task consists in deciding whether a leaf is to be extracted or not. In CafeIn, the attribute-value representation of the data is adaptive because it is based on different feature-sets, and it allows the integration of domain knowledge easily. Any supervised classiﬁcation algorithm can be used in the CafeIn framework.Asmentioned above, we have developed a CafeIn prototype and examined its performance with different learning algorithms for classiﬁcation (C4.5 [33], GloBoost [25]) and with different models of data representation. CafeIn is competitive with the others wrapper induction systems based on specialized learning algorithms. It will be continued with the combination of textual and structural informations, and with multi-slots wrappers induction.

Project-Team MOSTRARE Modeling Tree Structures, Machine Learning, and Information Extraction

Handle everyday research tasks with reliable, citation-backed results

Your personal Research Agent to handle research tasks with citation-backed results

Popular Tasks used by Researchers

How can I help with your research?

Meet SciSpace

Get more enhanced response by uploading the PDFs you want me to reference.

No relevant PDFs in your library

SciSpace is the AI research assistant for academics. Run systematic literature reviews on 280M+ papers, and write papers with cited sources. Trusted by 1M+ students, PhDs & researchers.

SciSpace | AI for Research

Analyze PDFs

Code & Manuscripts

Funding & Grants

Literature & Patents

Medical & Clinical Data

Systematic Review

Visualize & Present

Web & Data

Build a Google Scholar-like website for your research.

Build a website

Create charts and images for your research

Create a Chart

Write a paper for submission to a journal

Draft a manuscript

Patent Search

Design eye-catching scientific posters in minutes.

Scientific Poster Generation

Systematic Literature Review

One task is running at the moment. Your messages will be shown right after.

Drag and drop or click here to browse

Loved by <highlight>1 million+</highlight> researchers

Extract a list of specific topics and their sources from unstructured text

Topics

Compare and analyze relevant papers that matches with your search

Papers

Get insights from PDFs and bookmarked papers from your library

My library

Recent searches

Try searching for:

Catch AI-generated content in scholarly and non-scholarly content

{ai} Detector

Ai Writer

Get PDF Summaries, highlighted text explanations 

Chat with PDF

Effortlessly create in-text citations and bibliographies in APA and 2,500 other formats

Citation generator

Get explanations, summaries, and answers on academic papers

Ease up your research workflow with {scispace}'s cohort of exciting AI tools

Elevate your academic writing skills and convey your ideas the way you want

Paraphraser

Explore our range of reading and writing tools

Your file is being prepared and should be ready in a few minutes. If it's a large file, it might take a bit longer. You can close this window, and we'll email you the file when it's done.

You have reached a maximum limit of <strong>{limit}</strong> columns in the table. Remove at least <strong>1</strong> column to add or create another one.

Karine Lewandowski

Author Tools

Chat about Author

Papers

Modeling Tree Structures, Machine Learning, and Information Extraction

Project-Team MOSTRARE Modeling Tree Structures, Machine Learning, and Information Extraction