This paper presents the description of 12 systems submitted to the WMT16 IT-task, covering six different languages, namely Basque, Bulgarian, Dutch, Czech, Portuguese and Spanish. All these systems were developed under the scope of the QTLeap project, presenting a common strategy. For each language two different systems were submitted, namely a phrasebased MT system built using Moses, and a system exploiting deep language engineering approaches, that in all the languages but Bulgarian was implemented using TectoMT. For 4 of the 6 languages, the TectoMT-based system performs better than the Moses-based one.
In this paper we describe a collection of publicly available data sets for Portuguese that are suitable for the evaluation of distributional semantics models in lexical similarity tasks and in conceptual categorization tasks. These data sets were adapted from English gold-standard test sets, allowing any Portuguese distributional semantics model to be evaluated and also to be compared to mainstream results that have been obtained for this language. We also present an online service that showcases some functionalities of the distributional semantics models.
Machine translation (MT) from English to Portuguese has not typically received much attention in existing research. In this paper, we focus on MT from English to Portuguese for the specific domain of information technology (IT), building a small in-domain parallel corpus to address the lack of IT-specific and publicly-available parallel corpora and then adapted an existing hybrid MT system to the new language pair (English to Portuguese). We further improved the initial version of the EN-PT hybrid system by adding various modules to address the most frequently occurring errors in the initial system. In order to assess the improvements achieved by each of these dedicated modules, we compared all versions of our MT system automatically. In addition, we conduct and report on a detailed error analysis of the initial and final versions of our system.
Abstract:In this paper we present the annotation of a corpus with named entities that are classified into semantic types and disambiguated by linking them to their corresponding entry in the Portuguese DBpedia. This corpus, QTLeap Corpus, is a multilingual collection of question and answer pairs from a chat-based helpdesk service for Information and Communication Technologies. The resulting annotated corpus is a gold-standard named entity annotated lexical resource that is useful in supporting the training and evaluation of named entity annotation and disambiguation tools for Portuguese.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.