Language is imperfect, so we cannot create a perfect dictionary system.

The long term goal is to have the ability to do semantic operations on language using algorithms. So, the simple initial goal would be to be able to:

  • Automatically replace a word in a sentence with another word, keeping the basic meaning of the sentence the same. For that we would like to use synonyms. Synsets would also be beneficial here.
  • Determine equivalence of two phrases algorithmically.

So whilst we are not clear what semantic algorithms that we want to use, we need to be specific in the dictionaries without creating extra work.

Relationships between words:

  • A superset covers all of the subset. All apes are mammals, so ape is a subset of mammal. Red is an subset of colours. When asked what colour crimson is, we would say it is a type of red. If we wanted to be more precise, we would say it has a strong relationship with red and a weak relationship with blue. Or it has a strong relationship with red and a strong relationship with purple. So a blended colour may have relationships to other colours.

  • Synonyms are words that, in most cases (say 90% of the time) will not change the meaning in an arbitrary sentence when you replace one word with another. So Wednesday (English) and Mittwoch (German) would be synonyms to each other (assuming the reader understands both languages). In general, it is beneficial to have at least one Synonym for each word. So where there are multiple related words, we would like to elect the word with the closest relationship as a synonym if it is still has the same meaning.

krisvan 2017-04-29 05h31