/* Drops terms that contain other terms (prefixes, suffixes, etc.) and trims the questionlist to the target size.*/
private ArrayList<MemoryQuestion> dropPrefixSuffixfromSortedList(
List<MemoryQuestion> questionlist, int targetSize) {
ArrayList<MemoryQuestion> resultList = new ArrayList<MemoryQuestion>();
Set<Integer> badIndexSet = new LinkedHashSet<Integer>();
int terms = questionlist.size();
for (int first = 0; first < terms; first++) {
MemoryQuestion mq1 = questionlist.get(first);
String name1 = mq1.correctAnswer;
if (badIndexSet.contains(first)) continue;
resultList.add(mq1);
if (resultList.size() >= targetSize) return resultList;
for (int second = first + 1; second < terms; second++) {
MemoryQuestion mq2 = questionlist.get(second);
String name2 = mq2.correctAnswer;
if (name1.contains(name2) || name2.contains(name1)) {
badIndexSet.add(second);
}
}
}
return resultList;
}
/**
* lookup doesn't have to be synchronized
*
* @throws ClassNotFoundException
* @throws GeneralSecurityException
* @throws IOException
*/
public static synchronized UserStats lookup(String userCode)
throws IOException, GeneralSecurityException, ClassNotFoundException {
List<UserStats> users = readUsersFile();
if (codes == null) return null;
for (int i = 0; i < codes.size(); i++) {
if (codes.get(i).equals(userCode))
return users.get(
i); // simple logic: user is providing code #1. therefore s/he must be user #i. what's
// not to like?.
}
return null;
}
/** Generates person names tests from the given archive. @throws IOException */
public void generatePersonNameQuestions(
Archive archive,
NERModel nerModel,
Collection<EmailDocument> allDocs,
Lexicon lex,
int numClues)
throws IOException, GeneralSecurityException, ClassNotFoundException, ReadContentsException,
ParseException {
this.archive = archive;
questions = new ArrayList<>();
ArchiveCluer cluer = new ArchiveCluer(null, archive, nerModel, null, lex);
tabooCluesSet = new LinkedHashSet<>();
archive.assignThreadIds();
List<ClueEvaluator> evaluators = getDefaultEvals();
List<Document> docs = archive.getAllDocs();
Multimap<Contact, EmailDocument> contactToMessages = LinkedHashMultimap.create();
Multimap<Contact, Long> contactToThreadIds = LinkedHashMultimap.create();
// sort by date
Collections.sort(docs);
Date earliestDate = null, latestDate = null;
Map<Contact, Date> contactToLatestDate = new LinkedHashMap<>();
// compute earliest and latest date across all messages in corpus
for (Document doc : docs) {
EmailDocument ed = (EmailDocument) doc;
if (earliestDate == null || ed.date.before(earliestDate)) earliestDate = ed.date;
if (latestDate == null || ed.date.after(latestDate)) latestDate = ed.date;
}
JSPHelper.log.info(
"===================\nStarting to generate person names memory questions from "
+ docs.size()
+ " messages with "
+ numClues
+ " questions"
+ ", earliest date = "
+ edu.stanford.muse.email.CalendarUtil.formatDateForDisplay(earliestDate)
+ " latest date = "
+ edu.stanford.muse.email.CalendarUtil.formatDateForDisplay(latestDate));
Set<Integer> tabooSentenceHashes = new LinkedHashSet<>();
// create hashes of all sentences seen at least twice (case insensitive, lower cased)
{
Set<Integer> hashesSeen = new LinkedHashSet<>();
for (Document d : docs) {
String contents = archive.getContents(d, true);
String cleanedContents = EmailUtils.cleanupEmailMessage(contents);
SentenceTokenizer st = new SentenceTokenizer(cleanedContents);
while (st.hasMoreSentences()) {
String sentence = st.nextSentence();
sentence = canonicalizeSentence(sentence);
int hashCode = sentence.hashCode();
if (hashesSeen.contains(hashCode)) {
tabooSentenceHashes.add(hashCode);
log.info("Marking sentence as taboo: " + sentence);
} else hashesSeen.add(hashCode);
}
}
}
// compute contactToLatestDate that contact has been seen on
for (Document doc : docs) {
EmailDocument ed = (EmailDocument) doc;
// discard doc if it is not a sent mail
if ((ed.sentOrReceived(archive.addressBook) & EmailDocument.SENT_MASK) == 0) continue;
for (Contact c : ed.getParticipatingContactsExceptOwn(archive.addressBook)) {
Date currentLatestDate = contactToLatestDate.get(c);
if (currentLatestDate == null || currentLatestDate.before(ed.date))
contactToLatestDate.put(c, ed.date);
contactToMessages.put(c, ed);
contactToThreadIds.put(c, ed.threadID);
}
}
log.info("We are considering " + contactToLatestDate.size() + " contacts");
Date currentDate = new Date();
List<Pair<Date, Date>> intervals =
computeDateIntervals(earliestDate, currentDate); // go back from current date
// intervals[0] is the most recent.
JSPHelper.log.info("done computing " + intervals.size() + " intervals");
for (Pair<Date, Date> p : intervals)
JSPHelper.log.info(
"Interval: "
+ edu.stanford.muse.email.CalendarUtil.formatDateForDisplay(p.getFirst())
+ " - "
+ edu.stanford.muse.email.CalendarUtil.formatDateForDisplay(p.getSecond()));
int cluesPerInterval =
(numClues > 0 && intervals.size() > 0)
? (numClues + intervals.size() - 1) / intervals.size()
: 0;
JSPHelper.log.info(
"Will try to generate " + Util.pluralize(cluesPerInterval, "questions") + " per interval");
Multimap<Integer, Contact> intervalToContacts = LinkedHashMultimap.create();
// nSent is the number of sentences allowed in a clue text
int nSent = 2;
for (Contact c : contactToLatestDate.keySet()) {
Date lastSeenDate = contactToLatestDate.get(c);
// which interval does this date belong to? we'll assign this contact in that interval in the
// intervalToContacts map
int interval = -1;
Date intervalStart = null, intervalEnd = null;
{
int i = 0;
for (Pair<Date, Date> p : intervals) {
intervalStart = p.getFirst();
intervalEnd = p.getSecond();
if ((intervalStart.before(lastSeenDate) && intervalEnd.after(lastSeenDate))
|| intervalStart.equals(lastSeenDate)
|| intervalEnd.equals(lastSeenDate)) {
interval = i;
break;
}
i++;
}
}
if (interval < 0 || interval == intervals.size()) {
JSPHelper.log.info(
"What, no interval!? for "
+ edu.stanford.muse.email.CalendarUtil.formatDateForDisplay(lastSeenDate));
continue;
}
intervalToContacts.put(interval, c);
}
log.info("Interval information (interval 0 is the most recent):");
for (int interval = 0; interval < intervals.size(); interval++) {
Collection<Contact> contacts = intervalToContacts.get(interval);
int nContactsForThisInterval = (contacts == null) ? 0 : contacts.size();
log.info(
"In interval "
+ interval
+ " there are "
+ Util.pluralize(nContactsForThisInterval, "candidate contact")
+ " who were last seen in this interval");
}
for (int interval = 0; interval < intervals.size(); interval++) {
Date intervalStart = intervals.get(interval).getFirst();
Date intervalEnd = intervals.get(interval).getSecond();
Collection<Contact> candidateContactsForThisInterval = intervalToContacts.get(interval);
if (candidateContactsForThisInterval == null) {
log.info("Skipping interval " + interval + " because there are no contacts");
continue;
}
Map<Clue, Contact> clueToContact = new LinkedHashMap<>();
log.info("=======\nGenerating questions for interval " + interval);
outer:
for (Contact c : candidateContactsForThisInterval) {
String name = c.pickBestName();
if (name.length() < 2) // could also check if alphanumberic only
continue outer;
// ignore contact if name does not contain all alphabets. Even a period is not allowed. only
// space is allowed.
for (char ch : name.toCharArray()) {
if (!Character.isAlphabetic(ch) && !Character.isSpaceChar(ch)) continue outer;
}
Clue clue =
cluer.createPersonNameClue(
c,
evaluators,
nerModel,
intervalStart,
intervalEnd,
nSent,
archive,
tabooSentenceHashes);
if (clue != null) clueToContact.put(clue, c);
}
List<Clue> clueList = new ArrayList(clueToContact.keySet());
Collections.sort(clueList);
List<Clue> selectedClues = new ArrayList<>();
for (int i = 0; i < cluesPerInterval && i < clueList.size(); i++) {
selectedClues.add(clueList.get(i));
}
log.info(
"For interval "
+ interval
+ " selected "
+ selectedClues.size()
+ " contacts out of "
+ clueList.size()
+ " possible candidates.");
// for (Clue c: clueList)
// log.info ("Clue candidate for " + clueToContact.get(c).pickBestName() + "
// score = " + c.clueStats.finalScore+ " clue is " + c );
// for (Clue c: selectedClues)
// log.info ("Selected clue: " + clueToContact.get(c).pickBestName() + " score = "
// + c.clueStats.finalScore+ " clue is " + c);
for (Clue selectedClue : selectedClues) {
Contact c = clueToContact.get(selectedClue);
String name = c.pickBestName();
List<Integer> lengthList = Crossword.convertToWord(name).getSecond();
String lengthDescr = "";
if (lengthList.size() > 1) lengthDescr += Integer.toString(lengthList.size()) + " words: ";
for (Integer i : lengthList) {
lengthDescr += Util.pluralize(i, "letter") + ", ";
}
lengthDescr =
lengthDescr.substring(0, lengthDescr.length() - 2); // subtract the extra comma.
ClueInfo ci = new ClueInfo();
ci.lastSeenDate = contactToLatestDate.get(c);
ci.nMessages = contactToThreadIds.get(c).size();
ci.nThreads = contactToThreadIds.get(c).size();
questions.add(new MemoryQuestion(this, name, selectedClue, 1, lengthDescr));
}
}
log.info(questions.size() + " questions generated");
log.info("Top candidates are:");
// sort q's by clue score
Collections.sort(questions);
// log.info("Based on clue score, top answers:");
// for (MemoryQuestion mq: questions)
// log.info (mq.correctAnswer + " times= clue=" + mq.clue.clue);
int count = 0;
for (MemoryQuestion mq : questions) {
mq.setQuestionNum(count++);
}
// log the questions as well, just in case we don't get to the final point due to user fatigue
// or crashes
logStats("questions.final", false);
}
