Using N-Gram Models to Combine Query Translations in Cross-Language Question Answering

This paper presents a method for cross-language question answering. The method combines multiple query translations in order to improve the answering precision. The combination of translations is based on their pertinence to the target document collection rather than on their grammatical correctness. The pertinence is measured by the translation perplexity with respect to the collection language model. Experimental evaluation on question answering demonstrates that the proposed approach outperforms the results obtained by the best translation machine.

• Type: Conference
• Year: 2006
• Where: CICLING
• Authors: Rita M. Aceves-Pérez, Luis Villaseñor Pineda, Manuel Montes-y-Gómez
• Keywords: CICLING 2006, Cross-Language Question Answering, Natural Language Processing, Question Answering, Question Answering Demonstrates