# Translation of Pronominal Anaphora between English and Spanish: Discrepancies and Evaluation

**Authors:** A. Ferrandez, J. Peral  
**arXiv:** [1106.4862](https://arxiv.org/abs/1106.4862)  
**Date:** 2011-06-24  
**Summary page:** https://www.summarizepaper.com/en/paper/1106.4862v1/

## Key points

- The paper evaluates the tasks involved in translating pronominal anaphora in a machine translation (MT) system
- The study focuses on the MT interlingua approach called AGIR, which addresses several issues overlooked by other systems
- AGIR is capable of translating intersentential anaphors, detecting co-reference chains, and accurately translating Spanish zero pronouns into English
- Various types of knowledge, including lexical, morphological, syntactic, and semantic information, are used to resolve and evaluate these anaphora problems in AGIR
- The translation of third-person personal pronouns in English and Spanish (including Spanish zero pronouns) is assessed using unrestricted corpora
- Results show a precision rate of 80.4% for Spanish pronoun translation and 84.8% for English pronoun translation
- The authors suggest that their approach can be extended to other languages such as Portuguese, Italian, or Japanese
- This research contributes to improving machine translation systems by addressing pronominal anaphora challenges and providing effective solutions for accurate translation across different languages

## Summary

This paper, titled "Translation of Pronominal Anaphora between English and Spanish: Discrepancies and Evaluation," evaluates the tasks involved in translating pronominal anaphora in a machine translation (MT) system. The study focuses on the MT interlingua approach called AGIR (Anaphora Generation with an Interlingua Representation), which addresses several issues that have been overlooked by other systems. AGIR is capable of translating intersentential anaphors, detecting co-reference chains, and accurately translating Spanish zero pronouns into English. The authors present the resolution and evaluation of these anaphora problems in AGIR using various types of knowledge, including lexical, morphological, syntactic, and semantic information. The translation of third-person personal pronouns in English and Spanish (including Spanish zero pronouns) into the target language is assessed using unrestricted corpora. The results show a precision rate of 80.4% for Spanish pronoun translation and 84.8% for English pronoun translation. The authors suggest that their approach can be easily extended to other languages such as Portuguese, Italian, or Japanese. Overall, this research contributes to improving machine translation systems by addressing pronominal anaphora challenges and providing effective solutions for accurate translation across different languages.