X This focused retrieval strategy is believed to be XML Retrieval of particular benefit for information repositories containing long documents, or documents cov- Mounia Lalmas1 and Andrew Trotman2 ering a wide variety of topics (e.g., books, user 1Yahoo! Inc., London, UK manuals, legal documents), where the user’s ef- 2University of Otago, Dunedin, New Zealand fort to locate relevant content within a document can be reduced by directing them to the most relevant parts of the document. Synonyms Content-oriented XML retrieval; Focused re- Historical Background trieval; Structured document retrieval; Struc- tured text retrieval Managing the enormous amount of information available on the web, in digital libraries, in in- tranets, and so on requires efficient and effective Definition indexing and retrieval methods. Although this information is available in different forms (text, Text documents often contain a mixture of image, speech, audio, video, etc.), it remains structured and unstructured content. One way widely prevalent in text form. Textual informa- to format this mixed content is according to tion can be broadly classified into two categories, the adopted W3C standard for information structured and unstructured. repositories and exchanges, the eXtensible Mark- Unstructured information has no fixed prede- up Language (XML). In contrast to HTML, fined format and is typically expressed in natural which is mainly layout-oriented, XML follows language. For instance, much of the information the fundamental concept of separating the logical available on the web is unstructured. Although structure of a document from its layout. This this information is mostly formatted in HTML, logical document structure can be exploited to thus imposing some structure on the text, the allow a more focused sub-document retrieval. structure is only for presentation purposes and XML retrieval breaks away from the tradi- carries essentially no semantic meaning. Correct tional retrieval unit of a document as a single nesting of the HTML structure (i.e., to form large (text) block and aims to implement fo- an unambiguous document logical structure) is cused retrieval strategies aiming at returning doc- not imposed. Accessing unstructured information ument components, i.e., XML elements, instead is through flexible but mostly simplistic means, of whole documents in response to a user query. © Springer Science+Business Media LLC 2017 L. Liu, M.T. Özsu (eds.), Encyclopedia of Database Systems, DOI 10.1007/978-1-4899-7993-3_474-2 2 XML Retrieval such as a simple keyword matching or bag-of- strategies have been developed to obtain a rep- words techniques. resentation not only of the content of XML docu- Structured information is usually represented ments, but their structure. Ranking strategies have using XML, a markup language similar to HTML been developed to determine the best elements for except that it imposes a rigorous structure on the a given query. document. Moreover, unlike HTML, XML tags are used to specify semantic information about the stored content and not the presentation. A Query Languages document correctly marked up in XML has a fixed document structure in which semantically In XML retrieval, the logical document structure separate document parts are explicitly identified – is additionally used to determine which document and this can be exploited to provide powerful and components are most meaningful to return flexible access to textual information. as query answers. With appropriate query XML has been accepted by the computing languages, this structure can be specified by community as a standard for document markup, the user. For example, “I want a paragraph and an increasing number of documents are being discussing penguins near to a picture labeled made available in this format. As a consequence Otago Peninsula.” Here, “penguins” and “Otago numerous techniques are being applied to access Peninsula” specify content (textual) constraints, XML documents. The use of XML has gener- whereas “paragraph” and “picture” specify ated a wealth of issues that are being addressed structural constraints on the retrieval units. by both the database and information retrieval Query languages for XML retrieval can be communities [3]. This entry is concerned with classified into content-only and content-and- content-oriented XML retrieval [2, 5]asinvesti- structure query languages. Content-only queries gated by the information retrieval community. have historically been used as the standard Retrieval approaches for structured text form of input in information retrieval. They are (marked up in XML-like languages such as suitable for XML search scenarios where the SGML) were first proposed in the late 1980s. user does not know (or is not concerned with) In the late 1990s, the interest in structured text the logical structure of a document. Although retrieval grew due to the introduction of XML only the content aspect of an information need in 1998. Research on XML information retrieval can be specified, XML retrieval systems must was first coordinated in 2002 with the founding still determine the best granularity of elements to of the Initiative for the Evaluation of XML return to the user. Retrieval (INEX). INEX provides a forum for the Content-and-structure queries provide a evaluation of information retrieval approaches means for users to specify conditions referring specifically developed for XML retrieval. Since both to the content and the structure of the 2012 INEX has been run as part of CLEF, the sought elements. These conditions may refer Conference and Labs of the Evaluation Forum. to the content of specific elements (e.g., the returned element must contain a section about a particular topic) or may specify the type of the Foundations requested answer elements (e.g., sections should be retrieved). There are three main categories of Within INEX, the aim of an XML retrieval sys- content-and-structure query languages [1]: tem is “to exploit the logical structure of XML documents to determine the best document com- 1. Tag-based queries allow users to annotate ponents, i.e., best XML elements, to return as words in the query with a single tag name that answers to queries” [7]. Query languages have specifies the type of results to be returned. been developed in order to allow users to specify For example, the query “section:penguins” the nature of these best components. Indexing requests section elements on “penguins.” XML Retrieval 3 2. Path-based queries are based upon the syntax from the text of the element and all its descen- of XPath. They encapsulate the document dants. structure in the query. An example in the An alternative is to derive the term statistics NEXI language is “//document[about(.,Otago through the aggregation of term statistics of the Peninsula)]//section[about(.//title, penguins)].” element’s own text and those of each of its This query asks for sections that have a title children. A second alternative is to only index about “penguins” and that are contained in a leaf elements and to score non-leaf elements document about “Otago Peninsula.” through propagation of the score of their children 3. Clause-based queries use nested clauses to elements. Both alternatives can include additional express information needs, in a similar way to parameters incorporating, for instance, element SQL. The most prominent clause-based lan- relationships or special behavior for some ele- guage for XML retrieval is XQuery. A second ment types. example is XQuery Full-Text, which extends It is not uncommon to discard elements XQuery with text search predicates such as smaller than some given threshold. A single proximity searching and relevance ranking. italicized word, for example, may not be a meaningful retrieval unit. A related strategy, The complexity and the expressiveness of selective indexing, involves building separate content-and-structure query languages increase indexes for those element types previously from tag-based to clause-based queries. This seen to carry relevant information (sections, increase in expressiveness and complexity often subsections, etc., but not italics, bold, etc.). With means that content-and-structure queries are selective indexing the results from each index viewed as too difficult for end users (because, must be merged to provide a single ranked result they must, e.g., be intimate with the document list across all element types. structure). Nonetheless they can be very useful It is not yet clear which indexing strategy is the for expert users in specialized scenarios and also best. The best approach appears to depend on the have been used as an intermediate between a collection, the types of elements (i.e., the DTD), graphical query language (such as Bricks [12]) and their relationships. In addition, the choice of and an XML search engine. the indexing strategy currently has an effect on the ranking strategy. More details about indexing strategies can be found in the entry on Indexing Indexing Strategies Units. Classical indexing methods in information re- trieval make use of term statistics to capture Ranking Strategies the importance of a term in a document and consequently for discriminating between relevant XML documents are made of XML elements, and nonrelevant content. Indexing methods for which define the logical document structure. XML retrieval require similar term statistics, but Thus sophisticated ranking strategies can be for each element. In XML retrieval there are no a developed to exploit the various additional priori fixed retrieval units. The whole document, (structural) evidences not seen in unstructured one of its sections, or a single paragraph within (flat) text documents. a section, all constitute potential answers to a Element Scoring Many of the retrieval mod- single query. The simplest approach to allow the els developed for flat document retrieval have retrieval of elements at any level of granularity been adapted for XML retrieval. These models X is to index each element separately (as a separate have been used to estimate the relevance of an document in the traditional sense).