FromTo-CLIR: Web-Based Natural Language Interface for Cross-Language Information Retrieval

This paper proposes a method that uses a semantic category tree and collocation to resolve the ambiguity of query translation and uses a hybrid translation engine that combines a pattern-based module with a rule-based translator and includes pre- and post-fail softeners.

Abstract

This paper describes the implementation of FromTo-CLIR, a Web-based natural language interface for cross-language information retrieval, which consists of four parts: query translator, meta-search engine, front-end and back-end processors, and a robust hybrid machine translator. We propose a method that uses a semantic category tree and collocation to resolve the ambiguity of query translation. We use the front-end and back-end processors in which domain identifier determines the domain of the retrieved documents, machine translators translate the retrieved documents and tag manager deals with the HTML tags to preserve the layout of the original web document. To obtain a robust translation within the Web environment, we use a hybrid translation engine that combines a pattern-based module with a rule-based translator and includes pre- and post-fail softeners.