Natural Language text Processing Technology

Machine Translation is needed often by everybody. Machine Translation from one language into another is never perfect, because we use rigid, structured, mathematically organized programming language to program Natural Language. Natural Language is not a programming language. Natural Language is full of ambiguities on all levels: orthographical, grammatical, morphological, syntactical, semantical, pragmatical. To mention just a few: polysemy, metaphors, metonyms, collocations, co-references, homophones, homographs. Langsoft has found own solutions to these difficulties and has made these solutions public in a number of academic publications and software products.
Machine Translation was always regarded as means to understand, roughly, what was written or said in a language we do not know. Therefore, the translated text needs always post-editing. The myth, that Machine Translation can be perfect, was created by those, who market Machine Translation software programs. Has somebody seen a translation error on a jewel box displaying sample translation? Some companies produce better Machine Translation than other companies. Therefore, only fair competition can bring the best Machine Translation software on the market.

We have produced the following Machine Translation software programs:

- German-English
- English-German
- German-Russian
- French-Russian
- English-Russian
- German-English text-to-voice
- English-German text-to-voice - includes version of SEMCOR
- German-English text-to-video-and-voice
- English-German text-to-video-and-voice
- German-Russian text-to-video-and-voice
- English-Russian text-to-video-and-voice

SEMCOR was developed to correct automatically errors based on meaning, when the word is spelled correctly, but its contextual meaning is incorrect, for example homophones and homographs. This is very important for voice-to-voice Machine Translation, because Natural Language is full of homophones - words that sound the same, but have different meaning in context, for example bred - bread, knight - night, cellar - seller, meat - meet, made - maid, plate - plait, death - deaf, shone - shown, etc., for instance, in sentences such as
Charles nights the courageous soldiers.
We spent the knight at home.
We spent two knights with them.
we arrived at knight.
In deep see.

If we add to the existing homophones all mispronounced words ending in -g, -k (sting - stink, log - lock), -t, -d (made - mate, brought - broad), etc., voice-to-voice Machine Translation will be less than 50% correct!
 
We work, publish books and develop NLP software since 1970