NEXT_STEPS

## NEXT STEPS for Curveship ##

<> Implement syllepsis, narration by category, which will apply very well to
recounting:
"You wandered to many places: From X to Y, from there back to X, ..."
"You picked up several things. In Y, you acquired ..."
"You did some other things. In Y, you ..."
Each of these can be determined through filters on actions.

<> Replace the current parser (hacked together in a few days) with a new one
that is more expandable.

<> See if the use of '_' can indicate discourse entities that are not
simulated, including a "don't worry about that" message for each.

<> Allow multiple direct object inputs ("take keys and lamp") by serializing 
them into commands that each have a single object.

<> Add the "all" quantifier so that "take all" is translated into (for
instance) "take keys; take lamp; take food; take bottle."

<> Disambiguate "drink water" and "drink water" (and all similar commands)
by elaborating the command to "drink water from the bottle" and "drink water 
from the stream"

<> Do recognition of items in phases. For the specific case of TAKE:

1. NOTHELD & ACCESSIBLE
2. NOTHELD & PROXIMATE
3. NOTHELD & VISIBLE
4. HELD
5. KNOWN

For EXAMINE:

1. VISIBLE
2. PROXIMATE
3. KNOWN

Step 1 recognizes all items such that actions can succeed. That is, TAKE cannot 
work unless an item is NOTHELD & ACCESSIBLE. So the rest is really just to 
improve the "error message" explaining that the PC cannot take an item. The 
next step finds the most likely items that cannot be taken: Those in the room 
(PROXIMATE) but in closed containers (so they are not ACCESSIBLE). A bottle in 
a closed cupboard, whether visible or not, would be in this category. (Maybe 
with plannig or better higher-level commands the PC *would* be able to take it, 
but not in the current system.) The next step finds items in adjacent rooms 
that can be seen (but not taken, because they are far away). The next includes 
the PC's own possessions, which cannot be taken, but perhaps the player's 
mistake in typing a command can be explained with refereance to what is HELD. 
Finally, the search expands to all known items -- anything that has been 
mentioned. This could lead to a helpful message such as "The adventurer hasn't 
seen the ax since he was in the Valley."

<> Allow "say" to match multiple words. A hack is in place now that has a 
grammar line for one word, two words, three words, or four words.

<> Add real support for pronoun resolution. Currently object pronouns
are added as referring nouns, so "take it" will produce a disambiguation
message. A simple but real model of pronoun resolution could be added
with little trouble and would work well, because the occurrence of nouns 
is tracked in this system during text generation via slots (simulated items
such as [@buildind\o]) and non-simuated discourse entities (_the_road).
However, the full syntactical structure of the sentence is not represented,
so some useful information will be missing.

<> A scoring system.

<> An encumberance system.

<> A way to allow things like the scoring system and encumberance system to
be added modularly, without patching the main code base.

<> An idea; Implement postures, based on one's ability to SIT and LIE,
although this is often just a matter of being ON a supporter, not really a
posture model.

    stand
    sit
    squat
    crouch
    lie, possibly broken down to:
        lie prone (lie prostrate, prostrate myself, lie face down)
        lie supine (lie on my back)
        lie accumbent (lie on my side)
    kneel
    hang from X
    get on all fours (crawl)

<> A suggestion: Allow the player (in some modes) to enter narrative statements
that are similar to commands (and aren't general natural language) but could
refer to any character: "I go east" "The punk looks at the trash." Perhaps this
would map to the commanded actor being a set of some or all actors. Different
tenses and aspects probably have to be recognized.

<> Implement different levels of "explicit I" and "explicit you" as composable 
output filters. (These are variations in narrating I noted during my 
disseration work, having to do with how many signs of narrator and narratee
are shown.)