The inextricable and interdependent link between language identity has long been acknowledged (e.g. Joseph 2004; Norton 2000). Language is, after all, the means through which we identify oursel...

Abstract We present mGENRE, a sequence-to- sequence system for the Multilingual Entity Linking (MEL) problem—the task of resolving language-specific mentions to multilingual Knowledge Base (KB). For mention in given language, mGENRE predicts name target entity left-to-right, token-by-token an autoregressive fashion. The formulation allows us effectively cross-encode string and names capture mor...

Language encoders encode words and phrases in ways that capture their local semantic relatedness, but are known to be globally inconsistent. Global inconsistency can seemingly corrected for, part, by leveraging signals from knowledge bases, previous results partial limited monolingual English encoders. We extract a large-scale multilingual, multi-word analogy dataset Wikidata for diagnosing cor...

This paper proposes a novel method to represent user models in a multilingual manner which caters for multilingual Web search users. Furthermore, an evaluation is presented which examines a result re-ranking algorithm that is based upon that model.

Although there has been substantial research in knowledge management, there has been limited work in the area of multilingual knowledge management. The purpose of this chapter is to review and summarize some of the existing and supporting literature surrounding the emerging field of multilingual knowledge management. It does that by reviewing recent applications from multiple fields and the pre...

Multilingual facilities are indispensable to make the World−Wide Web really worldwide. However, most of the current implementations of the WWW support only Latin−1 (ISO 8859−1) character set. This paper describes a method for exchanging multilingual information through the WWW, based on our experience on development of Multi−Localization Enhancement of NCSA Mosaic for X. Our experimental enhanc...

This paper describes our recent eeort in developing the Glob-alPhone recognizer for multilingual large vocabulary continuous speech. Turkish. Based on ve languages we developed a global phoneme set and built multilingual speech recognizer by variing the method of acoustic model combination. Context dependent phoneme models are created using questions about languages and language groups. Results...