AESOP

From UNL Wiki
Revision as of 21:21, 30 July 2014 by Martins (Talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

AESOP is an experimental corpus used to refine the initial versions of the grammar for sentence-based UNLization and NLization, using IAN and EUGENE, respectively. It comprises 6 very short texts in English and their corresponding graphs in UNL.

Contents

Goal

The project AESOP has two main goals:

  1. To provide a translation memory from UNL to natural language, in order to be used for inducing UNL>NL grammars; and
  2. To provide standards for fully-automatic sentence-driven NLization, to be used as the parameter for evaluating the precision of UNL>NL grammars.

Repository

AESOP consists of 6 texts, which are translations of Aesop's fables to English. Most of them have been derived from the standard version by George Fyler Townsend (available at The Project Gutenberg), but they have suffered slight changes in order to become more suitable for natural language processing.

Project Title English* UNL** Number of sentences
AESOP-A1 The Hare and the Tortoise aa1_eng.txt aa1_unl.txt 13
AESOP-A2 The Bat and The Weasels aa2_eng.txt aa2_unl.txt 10
AESOP-B1 The Father and his Sons ab1_eng.txt ab1_unl.txt 11
AESOP-B2 The Ants and the Grasshopper ab2_eng.txt ab2_unl.txt 10
AESOP-C1 The Man and the Lion ac1_eng.txt ac1_unl.txt 11
AESOP-C2 ac2_eng.txt ac2_unl.txt 11

*To be manually translated to your target language in order to be used as the input for UNLization (IAN)
**To be used as the input for NLization (EUGENE)

Instructions

In AESOP, users are expected to map UNL graphs into natural language sentences. This process must take into consideration the following:

  • The NLization is the generation, to the target language, of the information conveyed by the UNL graph. It defines the expected output of UNL in natural language, and will be used to measure the precision of UNL>NL grammars. The NLization must comply with the principles below:
  • The NLization must convey all and only the information available in the UNL graph, i.e., the NLization must not add or suppress any information;
  • The NLization must be a grammatical sentence of the target language, i.e., it should be syntactically and semantically well-formed;
  • The NLization must belong to the standard variety of the target language, i.e., it should not contain slang, jargon, archaisms, SMS language and other non-standard structures;
  • The NLization must contain punctuation signs only if absolutely necessary or explicitly stated in the UNL graph;
  • A single graph may lead to differnt NLizations, to be provided in separate lines. These may convey different order of constituents, if possible in the target language.

Notes

Software