|
DEep Linguistic Processing with HPSG - INitiative (DELPH-IN) is a collaboration where computational linguists worldwide develop natural language processing tools for deep linguistic processing of human language.〔(DELPH-IN: Open-Source Deep Processing )〕 The goal of DELPH-IN is to combine linguistic and statistical processing methods in order to computationally understand the meaning of texts and utterances. The tools developed by DELPH-IN adopts two linguistics formalisms for deep linguistic analysis, viz. head-driven phrase structure grammar (HPSG) and minimal recursion semantics (MRS).〔Ann Copestake, Dan Flickinger, Carl Pollard and Ivan A. Sag. 2005. (Minimal Recursion Semantics: An Introduction ). In Proceedings of Research on Language and Computation.〕 All tools under the DELPH-IN collaboration are developed for general use of open-source licensing. Since 2005, DELPH-IN has held an annual Summit. This is a loosely structured unconference where people update each other about the work they are doing, seek feedback on current work, and occasionally hammer out agreement on standards and best practice. == DELPH-IN Technologies and Resources == The DELPH-IN collaboration has been progressively building computational tools for deep linguistic analysis such as the: * LKB system (Lexical Knowledge Builder): a grammar engineering environment where linguists can build unification grammar with the Head-driven Phrase Structure Grammar formalism * PET parser (Platform for Experimentation with efficient HPSG processing Techniques): an open source parser which produces HPSG parse trees with Minimal Recursion Semantics (MRS) outputs 〔(PET Parser website )〕 * ACE processor (Answer Constraint Engine): an efficient system to process DELPH-IN grammars that provide HPSG syntactic parses with MRS outputs. The latest version of ACE is able to generate natural language sentences.〔(ACE parser/generator homepage )〕 * LOGON infrastructure is a collection of software and DELPH-IN grammars to provide transfer-based machine translation. The LOGON approach to machine translation has proven to provide quality oriented hybrid (rule-based and stochastic) translations.〔Stephan Oepen, Erik Velldal, Jan Tore Lønning, Paul Meurer, Victoria Rosén, and Dan Flickinger. 2007.(Towards hybrid quality-oriented machine translation. On linguistics and probabilities in MT ). In Proceedings of the 10th International Conference on Theoretical and Methodological Issues in Machine Translation, pp.144–153. Skövde, Sweden.〕 Other than deep linguistic processing tools, the DELPH-IN collaboration supplies computational resources for Natural Language Processing such as computational HPSG grammars and language prototypes e.g.: * DELPH-IN grammars: a catalogue of computational HPSG grammar hand-crafted to capture deep linguistics analysis specific to the respective languages 〔(DELPH-IN catalog of grammars )〕 * LinGO Grammar Matrix: an open-source starter-kit for rapid prototyping of precision broad-coverage grammars compatible with the LKB. It contains a library of common language phenomena that computational grammarians can inherit for their HPSG grammars.〔Fokkens, Antske, Emily M. Bender and Varvara Gracheva. 2012. (Grammar Matrix Customization System Documentation ). Online resource.〕 * CLIMB libraries (Comparative Libraries of Implementations with Matrix Basis): an extended language library built on the Grammar Matrix. The objective of the CLIMB library is to maintain alternative analyses of the same phenomenon across different languages to test their impact on long-term grammar development.〔Fokkens, A., Avgustinova, T., and Zhang, Y. 2012. (Climb grammars: three projects using metagrammar engineering ). In Proceedings of the Eight International Conference on Language Resources and Evaluation (LREC’12),Istanbul, Turkey.〕 Another range of DELPH-IN resources are not unlike the data use for shallow linguistic processing, such as Text_corpus and treebanks: * MRS Test Suite: a short but representative set of sentences designed to capture some minimal recursion semantics phenomena. The test suites are available in Bulgarian, English, French, German, Greek, Japanese, Mandarin, Norwegian, Portuguese, Russian and Spanish.〔(MRS Test Suite page )〕 * Wikiwoods: WikiWoods is a parsed corpus that provides rich syntacto-semantic annotations for the English Wikipedia.〔Dan Flickinger, Stephan Oepen, and Gisle Ytrestøl. 2010. (WikiWoods: Syntacto-semantic annotation for English Wikipedia ). In Proceedings of LREC-2010, pages 1665–1671.〕 * DeepBank: an ongoing project to annotate the one million words of 1989 Wall Street Journal text (the same set of sentences annotated in the original Penn Treebank project) with the English Resource Grammar, augmented with a robust approximating PCFG for complete coverage.〔Dan Flickinger, Valia Kordoni and Yi Zhang. 2012. (DeepBank: A Dynamically Annotated Treebank of the Wall Street Journal ). In Proceedings of TLT-11, Lisbon, Portugal.〕〔(DeepBank homepage )〕 * Cathedral and the Bazaar: a compilation of an early essay on Open Source by Eric Raymond with translations into multiple languages. It was proposed as a multilingual shared test suite to enable us to compare parses across different grammars.〔(DELPH-IN CatB page )〕〔(Official Cathedral and the Bazaar webpage )〕 The open-source culture of the DELPH-IN collaboration provides the Natural Language Processing community with an array of deep linguistic processing tools and resources. However, the usability of DELPH-IN tools has been an issue with users and application developers new to the DELPH-IN ecology. The DELPH-IN developers are aware of these usability issues and there are ongoing attempts to improve documentation and tutorials of DELPH-IN technologies.〔(DELPH-IN 2013 Summit: Special Interest Group in Useability )〕 抄文引用元・出典: フリー百科事典『 ウィキペディア（Wikipedia）』 ■ウィキペディアで「DELPH-IN」の詳細全文を読む スポンサード リンク
|