EVALITA (http://www.evalita.it/) is the reference evaluation campaign of both Natural Language Processing and Speech Technologies for the Italian language. The objective of the shared tasks proposed at EVALITA is to promote the development of language technologies for Italian, providing a common framework where different systems and approaches can be evaluated and compared in a consistent manner. This volume collects the final and extended contributions presented at EVALITA 2011, the third edition of the evaluation campaign. The 36 revised full papers were carefully reviewed and selected from a total of 87 submissions. The papers are organized in topical sections roughly corresponding to evaluation tasks: parsing - dependency parsing track, parsing - constituency parsing track, domain adaptation for dependency parsing, named entity recognition on transcribed broadcast news, cross-document coreference resolution of named person entities, anaphora resolution, supersense tagging, frame labeling over italian texts, lemmatisation, automatic speech recognition - large vocabulary transcription, forced alignment on spontaneous speech.The AM are gender-dependent and speaker adaptive trained (SAT). ... The AM were trained on about 120h of detailed manual transcriptions mainly from previous European or national projects and ... trained and use probabilistic features based on bottleneck multi-layer perceptrons (MLP) and modified TRAP- DCT features.
|Title||:||Evaluation of Natural Language and Speech Tool for Italian|
|Author||:||Bernardo Magnini, Francesco Cutugno, Mauro Falcone, Emanuele Pianta|
|Publisher||:||Springer - 2013-01-03|