Objective
This project seeks to create a set of pronunciation lexicons of European names (city and town names, street names, family names, company names, product names) in a machine-assisted fashion whereby expert linguists (phoneticians/lexicographers) will carry out editorial preparation of the lexicons using customised workstation software.
The objective of the project is to make available, for widescale exploitation, quality controlled pronunciation lexicons in machine readable form (CD-ROM) for use in automated language systems, of primary interest to European companies in the telecommunications sector and in the European (dictionary) publishing industry.
An important sub-goal of the project will be the preparation of a set of letter-to-sound rules specific to the names of each language. The growing pronunciation lexicon will be used to extend the rule set and will also be used to train self-learning software.
A total of nine current languages of the European Community will be addressed on the project: Danish, Dutch, English, French, German, Greek, Italian, Portuguese and Spanish. The aim over the 2-year project is to derive pronunciation dictionaries for up to 1000000 names per language and to investigate the problems of exchanging national names amongst the partners to create a matrix of 'native-ised' pronunciations for each (thereby) foreign name in each other language.
The approach in the project is to define a dictionary consisting of names and their phonetic representation. To create this dictionary a lexicographer will initially select from a names list (provided by the relevant industrial particpant - see below), the most frequently used 20000 to 50000 names for the language. The dictionary would then be generated directly by hand, by an expert phonetician transcribing the conventional pronunciations of the names.
On the basis of the lexicon represented by this initial sample of 20000-50000 names, it will be possible to write an initial set of grapheme-to-phoneme conversion rules, and evaluate these rules in preparation for (semi)-automation of the further lexicographic work. An alternative preprocessor based on neural computing methods will also be investigated. Definition of the conversion standards and linguistic systems to be used for grapheme-to-phoneme conversion in each language, to ensure compatibility, will be one of the early tasks for the project.
The primary deliverables from this project will be a set of multi-language, machine readable CD-ROM pronunciation dictionaries for European city and town names, street names, family names, company names and product names. These dictionaries could be made available to the European Language Industry for commercial exploitation on a royalty basis.
The results from the project in the form of machine readable lexicons prescribing the pronunciation of names will constitute a valuable linguistic resource which allows natural language products to handle names correctly. Benefits will be felt in systems such as future map information systems which can recognize and synthesize names accurately; future automated directory enquiry systems such as future map information systems which can recognize and synthesize names accurately; future automated directory enquiry systems which can provide telephone numbers using advanced machine dialogues, recognising the desired name and address; and systems such as talking newspapers and books (for the blind) which can accommodate occurrences of names without pronunciation errors.
The project has been designed to be operated by a group of nine academic partners with nine Associated Partners from the European telecommunications industry. In themselves this group represents a major user group for potential downstream exploitation of the results of the project.
Fields of science
Topic(s)
Data not availableCall for proposal
Data not availableFunding Scheme
Data not availableCoordinator
EH1 1HN Edinburgh
United Kingdom