MEMPHIS defines, develops and validates a new system enabling the operation of commercially significant multilingual premium services, focused on wearable thin clients such as mobile phones, PDAs, etc. MEMPHIS extracts multilingual information from various content sources in different formats, e.g. from the Internet. Content acquisition is based on dynamic user profiles. Acquired content is transformed into an XML meta-representation which will enable content storage independent both of source and destination format. Meta-processing steps include recognition of the source language and automatic translation of the source document into a meta-language. The utilization of information extraction provides compressed documents, e.g. summaries. On user request, a translation choice from meta-language to destination language is provided. Destination format conversion is related to the thin client output format requirements, such as SMS, WAP, WWW, e-mail and so on. This includes conversion into documents of different length but extracted from the same source.
The Language Technology Lab at the DFKI is responsible for multilingual analysis of acquired content including language identification, categorization, extraction of relevant information and language independent XML-based meta-representation, as well as generating text/summaries according to the user profile, destination device and target language.
