SGF - An integrated model for multiple annotations and its application in a linguistic domain

Abstract

Seamless integration of various, often heterogeneous linguistic resources (in terms of their output formats) and merging of the respective annotation layers are crucial tasks for linguistic research. After a decade of concentration on the development of formats in order to structure single annotations for specific linguistic issues, a variety of specifications to store multiple annotations over the same primary data has been developed in the last years. Among these approaches three main architectures can be identified: Prolog-based architectures, XML-related approaches and graph-based models that follow the XML syntax. However, these architectures are not free of disadvantages when used in real world applications. In the Sekimo project the XML-based Sekimo Generic Format (SGF) was developed for the purpose of storing multiple annotations on the same primary data and examine relationships between elements of different annotation layers without prepended conversion. SGF is based on the design principles of graph-based approaches but makes use of the XML-inherent tree structures whenever possible to reduce processing costs. Analysing data stored in SGF can be done via standard XML-related specifications such as XPath, XSLT or XQuery and is done in our project in the linguistic application domain of anaphora resolution.

Integrated Linguistic Annotation Models and Their Application in the Domain of Antecedent Detection

“…Apart from defining an appropriate search window, the position of an antecedent candidate within a paragraph gives hints as to how likely that candidate is chosen as the correct one: 50.2% of the antecedents in the corpus are located paragraph-initial and 29.1% are located paragraph-final whereas only 20.2% are located in the middle of the paragraph. Thus, information on LDS might give information regarding the search window and for selecting the correct antecedent from a set of candidates (see also [47]). …”

“…A more detailed description of SGF can be found in [47], for a discussion of its use in the Anaphoric Bank project cf. [36] (in this volume).…”

“…For the process of anaphora resolution each anaphora-candidate-pair is interpreted as a feature vector which is used for training a classifier (see also [41,46,58]). A detailed description of the candidate list creation process as well as of the XSLT processing script is given in [47].…”

Different Views on Markup

“…For a more detailed description of the Sekimo Generic Format and performance measures on a per-file basis in a real-world application (cf. Stührenberg and Goecke 2008).…”

In pursuit of streamable stylesheet functions in XSLT 3.0

“…XStandoff is the successor of the Sekimo Generic Format (SGF, see [Stührenberg and Goecke 2008]). As such it is a meta markup language that uses a generic annotation format and XML namespaces in combination with a standoff approach to support overlapping hierarchies.…”