The PROMT engine was initially developed for rule-based machine translation (RBMT), but since 2011, PROMT has also offered enterprise solutions based on statistical machine translation (SMT) as well as a PROMT DeepHybrid translation engine.
PROMT RBMT
RBMT is built on the linguistic description of two natural languages (bilingual dictionaries and other databases containing morphological, grammatical, and semantic information), formal grammar, and translation algorithms. Translation quality depends on the volume of linguistic databases (dictionaries) and on the depth of the natural languages description; i.e., it requires consideration of the maximum grammatical structure features.
PROMT has been developing rule-based engines for 20+ years and this is a proprietary technology that combines translation algorithms and dictionaries created by PROMT. Currently, PROMT dictionaries and algorithms support hundreds of inflexion types and thousands of syntactic and semantic characteristics.
Key Components:
- PROMT Dictionaries:
- General (vocabulary with common lexis)
- Specialized (vocabulary with terminology for certain subject domains)
- User (vocabulary with client-specific terminology)
- PROMT Translation Algorithms:
- Algorithms for analysis and synthesis (Transfer Rules)
- User-defined options (Virtual Style Guides)
Key Advantages:
- Full control over terminology. PROMT Dictionaries are the key objects for controlling terminology. They can be created in two modes: simple and advanced. PROMT also supports automatic import and linking dictionaries.
- Simple mode is designed for users without expertise in computational linguistics; the system defines all advanced parameters automatically.
- Advanced mode allows for the full customization of linguistic characteristics available in the PROMT engine; this mode requires certain linguistic skills as well as knowledge of both source and target languages.
- Automatic import allows for the creation of an MT dictionary from an existing glossary; this feature supports import from tab-delimited text files as well as from TBX and OLIF.
- Linking dictionaries makes the maintenance of reversed pairs very easy; when this feature is enabled, every dictionary entry from, for example, the English-Russian dictionary appears in the Russian-English dictionary automatically.
- Virtual Style Guide. A Virtual Style Guide is a powerful tool to customize MT in the desired style of the target language. A Virtual Style guide covers syntax, punctuation, and regional settings as well as many other aspects. For example, a Virtual Style Guide ensures the proper translation of the pronoun “you” from English to European languages (polite or informal) as well as converting number formats (9.2 to 9,2) or quotation marks (“” to «»).
- Predictably. As rule-based technology works on human-defined rules, MT output is deterministic and predictable and does not depend on statistical deviations.
- Profiling. A rule-based engine allows for the creation of multiple profiles for different text translation projects within one translation engine (there is no need to set up a dedicated server for every trained engine).