BUSCADOR RECOLECTA

This folder contains files for bilingual glossary creation from factored phrase tables that include part of speech tagged text for EL-EN language pair. The tables are firstly filtered using part of speech tag sequences for each language so that entries with unsuitable part of speech sequences are filtered out. Then, feature scores from the phrase table are combined in a log-linear model to score each entry. The user specifies how large the output glossary should be (relative to the input) and the bottom ranking entries are discarded to produce the desired size glossary.

-

-

This data set contains Spanish word n-grams and Spanish word/tag/lemma n-grams in the "Environment" (ENV) domain. N-grams are accompanied by their observed frequency counts. The length of the n-grams ranges from unigrams (single words) to five-grams. The data were collected in the context of PANACEA (http://www.panacea-lr.eu), an EU-FP7 Funded Project under Grant Agreement 248064. The n-gram counts were generated from crawled Web pages that were automatically detected to be in the Spanish language and were automatically classified as relevant to the ENV domain. The ENV domain collection used consisted of approximately 49.86 million tokens. Data collection took place in the summer of 2011.

-

This is a domain-specific lexicon for Spanish subcategorization frames for environment (ENV) domain. This lexicon has been automatically created using the PANACEA web service named tpc_subcat_inductive (http://registry.elda.org/services/223) and the crawled data for this domain and language, previously annotated with Spanish Malt Parser web service (http://registry.elda.org/services/249).

We present a set of English gold-standards for different noun classes created in PANACEA to train and test automatic classifiers. To create these gold-standards we used we the data from the SemEval 2007 workshop Task 07: Coarse Grained English All-Words (Navigli et al., 2007). The words used in this task were first automatically tagged with an automatic clustering method (Navigli, 2006) using senses based on the WordNet sense inventory and later manually validated by expert lexicographers. For our experiments, we extracted all of the words from this inventory that contained as their first sense a sense that corresponded to the lexical semantic classes, i.e. “people” in the case of the class HUMAN. These gold-standards were created in the context of PANACEA http://www.panacea-lr.eu), an EU-FP7 Funded Project under Grant Agreement 248064.

-

How2Sign consists of a parallel corpus of 80 hours of sign language videos (collected with multi-view RGB and depth sensor data) with corresponding speech transcriptions and gloss annotations. In addition, a three-hour subset was further recorded in a geodesic dome setup using hundreds of cameras and sensors, which enables detailed 3D reconstruction and pose estimation and paves the way for vision systems to understand the 3D geometry of sign language.

This is a domain-specific gold-standard for Spanish subcategorization frames, in the case, for environment (ENV) domain. This gold-standard was manually developed, choosing a set of 30 verbs and 200 senteces for each verb. For each sentence, the SCFs present for the studied verb were manually annotated. The sentences were selected from crawled Web pages that were automatically detected to be in the Spanish language and were automatically classified as relevant to the ENV domain. Data collection took place in the summer of 2011. This gold-standard was created in the context of PANACEA (http://www.panacea-lr.eu), an EU-FP7 Funded Project under Grant Agreement 248064.