Abstract Meaning Representation (AMR) Annotation Release 3.0

AbstractIntroductionAbstract Meaning Representation (AMR) Annotation Release 3.0 was developed by the Linguistic Data Consortium (LDC), SDL/Language Weaver, Inc., the University of Colorado's Computational Language and Educational Research group and the Information Sciences Institute at the University of Southern California. It contains a sembank (semantic treebank) of over 59,255 English natural language sentences from broadcast conversations, newswire, weblogs, web discussion forums, fiction and web text. This release adds new data to, and updates material contained in, Abstract Meaning Representation 2.0 (LDC2017T10), specifically: more annotations on new and prior data, new or improved PropBank-style frames, enhanced quality control, and multi-sentence annotations. AMR captures "who is doing what to whom" in a sentence. Each sentence is paired with a graph that represents its whole-sentence meaning in a tree-structure. AMR utilizes PropBank frames, non-core semantic roles, within-sentence coreference, named entity annotation, modality, negation, questions, quantities, and so on to represent the semantic structure of a sentence largely independent of its syntax. LDC also released Abstract Meaning Representation (AMR) Annotation Release 1.0 (LDC2014T12), and Abstract Meaning Representation (AMR) Annotation Release 2.0 (LDC2017T10).DataThe source data includes discussion forums collected for the DARPA BOLT AND DEFT programs, transcripts and English translations of Mandarin Chinese broadcast news programming from China Central TV, Wall Street Journal text, translated Xinhua news texts, various newswire data from NIST OpenMT evaluations and weblog data used in the DARPA GALE program. New source data to AMR 3.0 includes sentences from Aesop's Fables, parallel text and the situation frame data set developed by LDC for the DARPA LORELEI program, and lead sentences from Wikipedia articles about named entities. The following table summarizes the number of training, dev, and test AMRs for each dataset in the release. Totals are also provided by partition and dataset: Dataset Training Dev Test Totals BOLT DF MT 1061 133 133 1327 Broadcast conversation 214 0 0 214 Weblog and WSJ 0 100 100 200 BOLT DF English 7379 210 229 7818 DEFT DF English 32915 0 0 32915 Aesop fables 49 0 0 49 Guidelines AMRs 970 0 0 970 LORELEI 4441 354 527 5322 2009 Open MT 204 0 0 204 Proxy reports 6603 826 823 8252 Weblog 866 0 0 866 Wikipedia 192 0 0 192 Xinhua MT 741 99 86 926 Totals 55635 1722 1898 59255 Data in the "split" directory contains 59,255 AMRs split roughly 93.9%/2.9%/3.2% into training/dev/test partitions, with most smaller datasets assigned to one of the splits as a whole. Note that splits observe document boundaries. The "unsplit" directory contains the same 59,255 AMRs with no train/dev/test partition.