Any Web form input field can be equipped with a "forecaster" that is capable of automatically typing in a complete word based on a few starting letters. (This is similar to predictive text inputs in mobile phones.)
The input field can also display a list of possible words or phrases (hints) based on one or a few letters the user has typed in.
The user can select from hints (or she/he can simply hit the END button to enter the actually selected hint from the list to the input field).
Besides faster and simpler typing, the other benefit of using a predictive text input field is that this system ensures that users use words and form sentences in a "standardised" (controlled) way. This makes services like automated translation easier to implement.
Using a grammar to "power" the predictive text input
In order to provide word or phrase hints to the user a simple vocabulary (e.g. a database) can be used.
It is, however, also possible that simple natural language grammar rules are used to join the words from the vocabulary to form complete sentences or sentence fragments
This way users can write "standardised" sentences very fast. (This technique is used e.g. in writing technical documentation.) See my second demo [1] for a working example.
Grammar rules can be very simple, they can include words from a natural language, and they can also refer to database (e.g. thesaurus) items or structures.