public static void GetUsersTopicDistribution(QuestionList ql, Dictionary<long, TopicIdAndConfidenceInItTuple[]> dict, out Dictionary<string, TopicIdAndConfidenceInItDictionary> questioners, out Dictionary<string, TopicIdAndConfidenceInItDictionary> answerers)
{
questioners = new Dictionary<string, TopicIdAndConfidenceInItDictionary>();
answerers = new Dictionary<string, TopicIdAndConfidenceInItDictionary>();
foreach (var e in dict)
{
var question = ql.GetQuestion(e.Key);
foreach (TopicIdAndConfidenceInItTuple q in e.Value)
{
// добавить пользователя,
if (!questioners.ContainsKey(question.AuthorEmail))
questioners.Add(question.AuthorEmail, new []{q} .ToDictionary(w => w.Item1, w => w.Item2 ));
else
questioners[question.AuthorEmail].UpdateOrAdd(q.Item1, v => v + q.Item2, q.Item2);
foreach (var answer in question.GetAnswers())
{
if (!answerers.ContainsKey(answer.AuthorEmail))
answerers.Add(answer.AuthorEmail, new[] { q }.ToDictionary(w => w.Item1, w => w.Item2));
else
answerers[answer.AuthorEmail].UpdateOrAdd(q.Item1, v => v + q.Item2, q.Item2);
}
}
}
}
private TrigramIndex(QuestionList questionList)
{
// Frequencies dictionary based on Mail.Ru corpus
var frequentWords = GetFrequentWords(questionList);
int i = 0;
IdToWord = new SortedDictionary<int, string>(frequentWords.ToDictionary(q => i++, q => q));
Trigrams = GetDefaultTrigramIndex(questionList);
vocabulary = new HashSet<string>(IdToWord.Values);
}
public AnswerQuality(QuestionList questionList)
{
this.questionList = questionList;
userQuality = new UserQuality(questionList);
foreach (var answer in questionList.GetAllAnswers().Where(answer => !answers.ContainsKey(answer.Id)))
{
answers.Add(answer.Id, answer);
}
}
private TrigramIndex(QuestionList questionList, string wordsDictionaryFileName)
{
// External frequencies dictionary (Google: "ruscorpora"):
Console.WriteLine("TrigramIndex: Geting Words from " + wordsDictionaryFileName);
IdToWord = new SortedDictionary<int, string>(LoadFromFile(wordsDictionaryFileName));
Console.WriteLine("TrigramIndex: Geting Trigram Index now");
Trigrams = GetDefaultTrigramIndex(questionList);
vocabulary = new HashSet<string>(IdToWord.Values);
}
private Dictionary<string, HashSet<int>> GetDefaultTrigramIndex(QuestionList questionList)
{
var getDataFunction = new Func<Tuple<string, HashSet<int>>[]>(
() => CalculateTrigramIndex().Select(pair => Tuple.Create(pair.Key, pair.Value)).ToArray());
return DataActualityChecker.Check
(
new Lazy<Tuple<string, HashSet<int>>[]>(getDataFunction),
t => t.Item1 + "\x2" + String.Join("\x2", t.Item2),
s =>
{
var q = s.Split('\x2');
return Tuple.Create(q[0], new HashSet<int>(q.Skip(1).Select(int.Parse)));
},
new FileDependencies(
String.Format("TrigramIndex_{0}.txt", questionList.GetHashCode()),
Program.QuestionsFileName,
Program.AnswersFileName)
)
.ToDictionary(item => item.Item1, item => item.Item2);
}
public static string[] GetFrequentWords(QuestionList questionList)
{
var getDataFunction = new Func<string[]>(
() =>
{
var statistics = new Statistics.Statistics(questionList);
return statistics.WordFrequencyDistribution(new EmptyStemmer())
.Where(item => item.Value > 10)
.Select(item => item.Key)
.ToArray();
});
return DataActualityChecker.Check
(
new Lazy<string[]>(getDataFunction),
t => t,
s => s,
new FileDependencies(String.Format("FrequentWords_{0}.txt", questionList.GetHashCode()),
Program.QuestionsFileName,
Program.AnswersFileName)
).ToArray();
}
public static TrigramIndex CreateFromDefaultDictionaryAnd(QuestionList questionList)
{
return new TrigramIndex(questionList, Program.FilesDirectory + "1grams-3.txt");
}
private static IEnumerable<Tuple<string, string, double>> GetAllEdges(QuestionList ql)
{
var edges = new Dictionary<Tuple<string, string>, double>();
foreach (var question in ql.GetAllQuestions())
{
foreach (var answer in question.GetAnswers())
{
var from = answer.AuthorEmail;
var to = question.AuthorEmail;
edges.UpdateOrAdd(Tuple.Create(from,to), v => v + 1d, 1d);
}
}
return edges.Select(q => Tuple.Create(q.Key.Item1,q.Key.Item2, q.Value));
}
public Statistics(QuestionList questionList)
: base(questionList)
{
}
protected LDADataGenerator(QuestionList questionList, string vocabularyStorePath, string documentsStorePath)
{
QuestionList = questionList;
VocabularyStorePath = vocabularyStorePath;
DocumentsStorePath = documentsStorePath;
}
public InferFormatLDAGenerator(QuestionList questionList, string vocabularyStorePath, string documentsStorePath)
: base(questionList, vocabularyStorePath, documentsStorePath)
{
}
public ExpertUsers(QuestionList ql)
{
statistics = new Statistics(ql);
this.ql = ql;
}
public void TestId()
{
var ql = new QuestionList(QuestionsFileName, AnswersFileName);
var hasIdenticId = false;
foreach (var question in ql.GetAllQuestions())
{
foreach (var answer in ql.GetAllAnswers())
{
hasIdenticId = true;
if (answer.Id == question.Id)
Console.WriteLine("BAD ID!!!!!!!!! " + answer.Id);
}
//Console.WriteLine(question.Id);
}
Assert.AreEqual(true, hasIdenticId);
}
public TopicsStatistics(QuestionList questionList, string docIdsFile = Program.DocIdsFileName, string topicsFile = Program.TopicsFileName, int allTopicsNumber = Program.TopicsCount)
: base(questionList)
{
questionIds = File.ReadAllLines(docIdsFile).Select(Int64.Parse).ToArray();
questionsToDocs = questionIds.Select((id, i) => new {docId = i, questionId = id})
.ToDictionary(doc => doc.questionId, doc => doc.docId);
topics = ReadTopicsFrom(topicsFile, allTopicsNumber);
}
public UserStatistics(QuestionList questionList)
: base(questionList)
{
var parser = new MailUserPageParser(Program.MailUsersDirectory);
var questionUsers = questionList.GetAllQuestions().Select(q => q.AuthorEmail);
var answerUsers = questionList.GetAllAnswers().Select(a => a.AuthorEmail);
var questionListUsers = new HashSet<string>(questionUsers.Union(answerUsers));
users = parser.GetObjects().Where(u => questionListUsers.Contains(u.Email)).ToList();
TopicStatistics = new TopicsStatistics(questionList);
}
public void Init()
{
var questionList = new QuestionList(Program.QuestionsFileName, Program.AnswersFileName);
statistics = new Statistics(questionList);
}
public UserQuality(QuestionList questionList)
{
this.questionList = questionList;
userInfos = CalculateUserInfos();
}
private static IEnumerable<string> GetAllVertices(QuestionList ql)
{
return ql.GetAllQuestions().Select(q => q.AuthorEmail).Concat(ql.GetAllAnswers().Select(q => q.AuthorEmail)).Distinct();
}
public UserTopicsWalking(QuestionList questionList, ICollection<string> categories, TopicsStatistics topicStatistics)
{
this.topicStatistics = topicStatistics;
UserQuestions = questionList.GetAllQuestions().Where(q => categories.Contains(q.Category))
.GroupBy(q => q.AuthorEmail, (userEmail, hisQuestions) => new { userEmail, hisQuestions })
.Where(u => u.hisQuestions.Count() >= 2)
.ToDictionary(u => u.userEmail, u => u.hisQuestions);
foreach (var questions in UserQuestions.Values)
{
foreach (var question in questions)
{
var topic = topicStatistics.GetTopicByQuestionId(question.Id, Threshold);
if (topic == null) continue;
if (topicDistribution.ContainsKey(topic.Item1))
{
topicDistribution[topic.Item1]++;
}
else
{
topicDistribution[topic.Item1] = 1;
}
}
}
userQuestionsCount = topicDistribution.Sum(it => it.Value);
topicMoves = GetTopicSequenceFrequences(2);
}
public GibbsFormatLDAGenerator(QuestionList questionList, string documentIdsFilePath, string documentsFilePath)
: base(questionList, "", documentsFilePath)
{
this.documentIdsFilePath = documentIdsFilePath;
}
public GibbsFormatLDAGenerator(QuestionList questionList, string documentIdsFilePath, string documentsFilePath, params string[] categories)
: base(questionList, "", documentsFilePath)
{
this.documentIdsFilePath = documentIdsFilePath;
this.categories = categories;
}
public static void ModifyTyposCorpus(QuestionList ql)
{
var detector = new SpellChecker(TrigramIndex.CreateFrom(ql));
Console.WriteLine("I am Modifying");
var start = DateTime.Now;
foreach (var question in ql.GetAllQuestions())
{
question.Text = String.Join(" ", question.Text.SplitInWordsAndStripHTML().Select(detector.Fix));
question.Title = String.Join(" ", question.Title.SplitInWordsAndStripHTML().Select(detector.Fix));
}
Console.WriteLine("Questions modified in {0}", (DateTime.Now - start).TotalSeconds);
start = DateTime.Now;
foreach (var answer in ql.GetAllAnswers())
{
answer.Text = String.Join(" ", answer.Text.SplitInWordsAndStripHTML().Select(detector.Fix));
}
Console.WriteLine("Answers modified in {0}", (DateTime.Now - start).TotalSeconds);
File.WriteAllLines(Program.QuestionsNoTyposFileName, ql.GetAllQuestions().Select(Question.FormatStringWrite));
File.WriteAllLines(Program.AnswersNoTyposFileName, ql.GetAllAnswers().Select(Answer.FormatStringWrite));
}
public BodyCalculator(QuestionList questions, BodyPart body)
{
this.questions = questions;
this.body = body;
}
public static TrigramIndex CreateFrom(QuestionList questionList)
{
return new TrigramIndex(questionList);
}
public void DistributionInit()
{
ql = Program.DefaultQuestionList;
statistics = new Statistics(ql);
Console.WriteLine("Preparations have been done");
}
public UserGraph(QuestionList ql)
: base(GetAllVertices(ql), GetAllEdges(ql))
{
}