SLICE: Spoken Language Identification from Commonly-occuring Words&Context Evaluation

United States Army personnel, as part of their overseas operations, must regularly interact with the local populations and interpret local audio feeds (such as radio transmission, news reports, etc.). While human translators are sometimes available for this purpose, their numbers are limited and their use often requires exposure to risk. Furthermore, many scenarios require urgent translation of speech, thereby necessitating automated, efficient translation. The first step in such an automated system, determining which foreign language translation engine is to be used, remains an important unresolved issue. SSCI's proposed approach to an Automatic Spoken Language Recognition (ASLR) will be based on a probabalistic approach to matching an incoming speech transmission to a database of frequently occurring words in a language. In addition, SSCI will also make use of additional contextual information, such as location, previous (recent) identifications and user feedback to ensure a robust language identification algorithm.