/* Subject 0440949. AI department, UvA. Assignment 4, knowledge systems: CSI * Andreas van Cranenburgh. * * Project history: * csi-1.pl started: Fri Nov 4 14:39:29 CET 2005. * -- enormous procastrination/postponement * csi-2.pl started: Fri Nov 15, somewhere 06:00:00 CET 2005. * -- manic brainstorming; mindmaps -- * csi-3.pl started: Thu Nov 17 16:48:45 CET 2005 * -- coding. alas, more debugging! * 1. time and place, actors, actions and objects * time is a point. an interval is two points. time is stored in microSec, or in any case sth very big. * */ %Convertors: % Seconds to BigTime timeformat( [ _Day,_Hour,_Min, Sec, BigTime ] ) :- var(BigTime), BigTime is Sec * 1000000. /* as you may see here, there is a list with Day, Hour, Min and Sec. This is the standard time object in this project. Another standard one is BigTime */ % BigTime back to Seconds timeformat( [ _Day, _Hour, _Min, Sec, BigTime ] ) :- var(Sec), Sec is BigTime / 1000000. /* qualitative is mapped onto quantitative data, using conversion: */ % generate an appropriate seconds value: quantime(Quant, BigTime) :- ( parse_abs_quant(Quant, Sec) ; parse_rel_quant(Quant, Sec) ), timeformat( [ _D, _H, _M, Sec], BigTime). % eg. 29 Oktober 00.45 uur parse_abs_quant( [ DayOfMonth, Month, Time, Unit], ResultSec ], Sec) :- parse_time( Time, [ _Day, Hour, Minute, Sec ] ), %re-use "words" kb. words(Unit, 3600) ( match_month(Month, MontNo) ; ( Month >= 1, Month is =< 12, MonthNo = Month ) ), parse_abs_quant( [ DayOfMonth, Month, Time], ResultSec ], Sec) :- parse_abs_quant( [ DayOfMonth, Month, Time, 'uur'], ResultSec ], Sec). %the months, in dutch. gosh, why am I doing this?? match_month(januarie, 1). match_month(februarie, 2). match_month(maart, 3). match_month(mei, 5). match_month(juni, 6). match_month(juli, 7). match_month(augustus, 8). %both match_month(april, 4). match_month(september, 9). match_month(oktober, 10). match_month(november, 11). match_month(december, 12). %english match_month(january, 1). match_month(february, 2). match_month(march, 3). match_month(may, 5). match_month(june, 7). match_month(july, 8). match_month(august, 9). match_month(october, 10). % first argument is word, second is conversion factor to seconds. words(minute, 60). words(minuten, 60). words(minutes, 60). words(min, 60). words(m, 60). words(uur, 3600). %words(hour, 3600). %words(hr, 3600). %words(h, 3600). %screw this (FIXME, std lib! or state automata): parse_time(TimeStr, TimeObj) :- %convert TimeStr to list?? ( %Define seperators: Sep = '.' ; Sep = ':' ) ( TimeStr = [ Hour, Sep, Minute, _Unit ] ; TimeStr = [ Hour, Sep, Minute ] ), TimeObj = [ _Day, Hour, Sep, Minute, _Sec ]. %parse_rel_quant( [ DayOfMonth, MonthStr, DotTime ], ResultSec ) :- % ...? qualtime(Quant, BigTime) :- approxquant(Quant, Sec), timeformat( [ _D, _H, _M, Sec ] , BigTime). approxquant(Quant, Sec), str2list(Quant, List), ( words(Word, Factor) member(Word, List), %N = [the word before that word] N = 0, Sec is N * Factor ). %qualquant(Qual, Quant) :- var(Quant), % let's forget this for now.. /* * absolute gets converted to relative, or the other way around, whatever is better in code. original value is preserved, since consistency check is forced after EVERY addition. */ /* Data representation * * Timeline * A branchable timeline is an instance of a linked list that can have an arbitrary number of links on each node, both has head and tail. As far as I know it's better not to express it as a tree, since then you end up storing the end twice and usually the crime itself is pretty sure (eg. dead body). * * Objects (in OO sense): * all objects have one name/id, and a list of aliases (maybe empty). Also, a * list of relations (either to annother object's name or alias). Objects are: * - proper objects (so literally objects, no CS-speak). No extra properties etc. * - actors: objects that have a role, listed with (some of) their respective * actions: * - the accused [or criminal, I'd rather not make the distinction here!] * killing, lying, possesing, etc. * - victim * dying, getting_hurt, etc. * - witness * lying, fact-reporting, opinion-speaking * - none (anyone else) * - relation: any id to any other id. this relationship is an equivalence * relation, eg. it's symmetric. But for clarity facts are written with the * earlier event first, if there's any difference at all. For objects order * doesn't matter in the relationship. * * Trust: * The so called "certainty" attribute is actually better called "trust" (more * general), so that it applies to anything. * * This specific model is loosely inspired by GPG's ownertrust. */ 2 weights as much as 1, only in the reasoning it won't be considered that it might be false. */ event-time(melding1, [28, oktober, om, '21.30']). event-whowhatwhere(melding1, ellieb, pistoolscot, alkmaar(elandstraat(12)), witness) alias(melding1, 'De eerste melding'). trust(melding1, 3). relation(melding1, evidence, moord1). relation(melding1, similar, melding2). %eg: relation(melding1, opposes, something). ? %etc:..... event(melding1, [28, oktober, om, '21.30']). whowhatwhere(ellieb, pistoolscot alkmaar(elandstraat(12)), witness) alias(melding1, 'De eerste melding'). trust(melding1, 3). event(melding1, [28, oktober, om, '21.30']). whowhatwhere(ellieb, pistoolscot alkmaar(elandstraat(12)), witness) alias(melding1, 'De eerste melding'). trust(melding1, 3). % in practice. do case. % Let python filter do that? /* Interface, timeline: * - printf the list, one event per line. * - branches: interactively ask which branches to act upon! * - [but: visually show branches, eg. by indenting.] */ #we're good to go! go :- writeln("This is uCSI with hacky prolog CLI interface; so end input with a dot, even when you\'re only expected to enter a single letter"). writeln("(web interface maybe soon to solve this??)"), nl, loop. loop :- write("Main: a. = Evidence-based tasks | b. = Events based task."), write(| c. = Close | e. = Exit | q. = Quit "), nl, input(Letter), ( ( Letter = 'a', evidence_tasks ) ; ( Letter = 'b', events_tasks ) ; ( ( Letter = 'c' ; Letter = 'e' ; Letter = 'q' ), true ) ; loop ) ). %probably reduntant parentheses etc. (we're not getting to LISP levels yet!) evidence_tasks :- write("Evidence menu: a. = Add | c. = Conflicts. | i. = Inspect all evidence"), write(" | l. = leave menu / go back"), nl, input(Letter), ( ( Letter = 'a', add_event ) ; ( Letter = 'c, list_conflicts ) ; ( Letter = 'i', inspect_evidence ) ; ( Letter = 'l', true ) ; evidence_tasks ). events_tasks :- writeln("Events menu: e. = list Evidence for an event | t. = Travers the Timeline"), input(Letter), ( ( Letter = 'e', evidence_tasks ) ; ( Letter = 'e', events_tasks ) ; ( Letter = 'l', true ) ; evidence_tasks. ). ask_trust (Number):- writeln("How much do you trust this assertion about X:"), writeln("-1: I do not trust it"), writeln(" 0: I will not answer"), writeln(" 1: I have done careful checking (beyond reasonable doubt)"), writeln(" 2: I trust it ultimately (scientific)"), write("[0]: "), input(Number), ( ( nonvar(Number), Number is >= -1, Number is =< 2 ) ; ask_trust(Number) ).