Travelling Object. The Travelling Object (TO) is the data object that is being exchanged between the platform components. The TO can be different depending on the situation and the components connected and has been chosen following already existing standards. As presented in Xxxx and Bel (2011), there have been relevant proposals from the Language Resources community to reach a consensus about formats to represent annotated corpora. The aim of the PANACEA developers has been to use already defined and used standards for each scenario that required a common data format.
Travelling Object. This section deals with the conversion of the different input/output formats of the aligners from and to the Travelling Object (TO) format (defined in Deliverable 3.1, section 6.1). 7 xxxx://xxxx.xxxxxx.xxx/p/berkeleyaligner/ 8 xxxx://xxxxxxxxxx.xxx/ aligner2to is a webservice that can convert the output of any11 of the aligners that have been integrated into PANACEA to the TO. It calls the FormatConverter tool, originally developed by ILSP (see section 4 of Deliverable 4.2), which has been extended with additional processors in order to handle the formats of the aforementioned aligners. sentalg_tok_to2word_alg is a webservice that takes three inputs in TO format (source and target corpus, both sentence-split and tokenised, and sentence alignment) and outputs two files (the subset of sentences in the source and target corpus that have 1-to-1 sentence alignments) in the plain sentence-split and tokenised format that the sub-sentential alignment webservices take as input. sentsplit_tok2to is a webservice that was required by sentalg_tok_to2word_alg; it converts sentence-split and tokenised files to the TO format. 4 Workflows This section presents several workflows of webservices developed in Taverna (Hull et al., 2006),12 showing possible usages of aligners in the platform, the interactions that arise, etc. The first workflow, depicted in Figure 1, presents a pipeline that performs sentence alignment on bilingual text crawled from the web. It is made up of two sequential workflows: • bicrawler_to_text processes the output of the Focused Bilingual Crawler (see section 3.2 of Deliverable 4.2 for more details) for a pair of languages A and B producing three outputs: plain text versions of the source and target corpora acquired and the bilingual header in XCES format.13 • sentence_alignment takes as input the output from bicrawler_to_text and performs sentence alignment using Hunalign. In order to do so, the text of each language is preprocessed with a sentence splitter and a tokeniser. Finally, the workflow outputs three objects whose format is compliant with the TO: the sentence alignment and the source and target corpora sentence-split and tokenised. 11 The only exception is Anymalign, whose output is not straightforward convertible into the TO. This aligner has been integrated anyway, as it is considered useful for deriving bilingual dictionaries (work-package 5.2). 12 xxxx://xxx.xxxxxxx.xxx.xx/ 13 xxxx://xxx.xxxx.xxx/schema/#align Figure 1. Sentence alignme...
Travelling Object. Corpus and data format: This section aims to describe existing data formats that could be used in PANACEA.
Travelling Object. TO-GR-01: Define/choose simple TO - Design of the first version of the travelling object (simple TO). - Inline annotation. TO-GR-02: Conversion tools (simple TO) - Develop scripts or tools to convert in-house formats to TO format (simple TO). TO-GR-03: Define/choose complex TO - Design of a second version of the travelling object (complex TO). - Stand-off annotation? - New features and improvements. TO-GR-04: Conversion tools (complex TO) - Develop scripts or tools to convert in-house format to complex format.
Travelling Object
