Java Libraries for Accessing the Princeton Wordnet: Comparison and Evaluation

作者: Mark Finlayson

DOI:

关键词: Source codeDSPACEApplication programming interfaceDatabase serverInformation retrievalMultiple-criteria decision analysisSoftwareComputer scienceWorld Wide WebJavaWordNet

摘要: Java is a popular programming language for natural processing. I compare and evaluate 12 libraries designed to access the information in original Princeton Wordnet databases. From this comparison emerges set of decision criteria that will enable user pick library most suited their purposes. identify five deciding features: (1) availability similarity metrics; (2) support editing; (3) via Maven; (4) compatibility with retired versions; (5) Enterprise Java. also provide other features each library, exposed by API, versions supports, speed various retrieval operations. In case user’s application does not require one features, show my JWI, MIT Interface, highest-performance, widest-coverage, easiest-to-use available. A developer seeking faced bewildering array choices: there are no fewer than off-the-shelf data, combinations performance. addition these libraries, at least additional libraries1 that, while providing direct data themselves, functions such as metrics deployment database servers. paper compare, contrast, libraries2 so help developers find See Table 6 list all URLs. have made best effort be complete possible identifying Wordnet. It possible, however, missed some more obscure especially whose primary purpose but function. right application. To knowledge first attempt thorough any libraries. proceed follows. First present bottom line, which commonly encountered when using then discuss distinguish from others. an assessment what accessible compatible versions. performance on nine different metrics, well time initialize in-memory dictionaries those suport The code reproducing evaluation (including required source code, copies described Wordnet) available online.3 While software evaluated exclusively Java, limited writing accessing Wordnet, work should helpful who seek interfaces (APIs) interacting data. particular identified here use. 1 Deciding Library Before discussing feature detail lay out line: choose if your falls into common situations below. First, outline constraints. Next, needs feature, deVia DSpace repository CSAIL Work Product: http://hdl.handle.net/1721.1/81949

参考文章(1)
Rion Snow, Daniel Jurafsky, Andrew Y. Ng, Semantic Taxonomy Induction from Heterogenous Evidence meeting of the association for computational linguistics. pp. 801- 808 ,(2006) , 10.3115/1220175.1220276