/// <summary>
/// Classifies a list of tweets using both of our approaches.
/// </summary>
/// <param name="tweetList">List of tweets</param>
/// <returns>An AnalysisResultObj which contains both results</returns>
private AnalysisResultObj ClassifyTweet(List <Tweet> tweetList)
{
Classifier c = new Classifier();
AnalysisResultObj result = TweetAnalyzer.Instance.AnalyzeAndDecorateTweetsThreaded(tweetList);
result.MIResult = c.RunNaiveBayes(tweetList);
return(result);
}
/// <summary>
/// Splits the tweets int sub lists and analyzes them using tasks
/// </summary>
/// <param name="tweetList"></param>
/// <returns></returns>
public AnalysisResultObj AnalyzeAndDecorateTweetsThreaded(List <Tweet> tweetList)
{
Log.Debug("Spliting " + tweetList.Count + " tweets");
int tweets = tweetList.Count;
List <Task <AnalysisResultObj> > tasks = new List <Task <AnalysisResultObj> >();
List <Tweet> copyTweetList = new List <Tweet>();
copyTweetList.AddRange(tweetList);
int e = tweetList.Count / Constants.TWEET_LIST_AMOUNT;
List <List <Tweet> > splittedList = new List <List <Tweet> >();
int tweetsSplitted = 0;
// Splits to the max number of lists
for (int i = 0; i < Constants.TWEET_LIST_AMOUNT; i++)
{
tweetsSplitted += e;
splittedList.Add(copyTweetList.Take(e).ToList());
copyTweetList = copyTweetList.Skip(e).ToList();
}
//If some tweets are not covered, adds another list such that all tweets are analyzed
if (tweetsSplitted < tweets)
{
splittedList.Add(copyTweetList);
}
//Starts the tasks
foreach (var item in splittedList)
{
Task <AnalysisResultObj> t = new Task <AnalysisResultObj>(() => AnalyzeAndDecorateTweets(item));
t.Start();
tasks.Add(t);
}
Task.WaitAll(tasks.ToArray());
AnalysisResultObj res = new AnalysisResultObj();
//Combines the result
foreach (var task in tasks)
{
res.Count += task.Result.Count;
res.PolCount += task.Result.PolCount;
res.KeywordBias += task.Result.KeywordBias;
res.MediaBias += task.Result.MediaBias;
res.PositiveSentiment += task.Result.PositiveSentiment;
res.NegativeSentiment += task.Result.NegativeSentiment;
res.PositiveTweetsCount += task.Result.PositiveTweetsCount;
res.NegativeTweetsCount += task.Result.NegativeTweetsCount;
}
Log.Debug("Combining tweets");
Log.Debug("Result " + res.GetAlgorithmResult());
return(res);
}
/// <summary>
/// Post method to be used by the TweetRetriever.
/// </summary>
/// <param name="resultObj">The result object returned by algorithm</param>
/// <param name="user">The user</param>
/// <returns>Returns true if the post request succeeded</returns>
private bool PostResultToDB(AnalysisResultObj resultObj, User user)
{
//Create the post request
bool succes = webHandler.DatabaseSendDataRequest(Constants.DB_SERVER_IP + "twitter", "POST",
"twitter_name=" + user.ScreenName, "twitter_id=" + user.Id, "analysis_val=" + resultObj.GetAlgorithmResult().ToString(CultureInfo.InvariantCulture),
"media_val=" + resultObj.GetMediaResult().ToString(CultureInfo.InvariantCulture), "mi_val=" + resultObj.MIResult.ToString(CultureInfo.InvariantCulture),
"sentiment_val=" + resultObj.GetSentiment().ToString(CultureInfo.InvariantCulture), "tweet_count=" + resultObj.Count, "protect=" + Convert.ToInt32(user.IsProtected));
if (!succes)
{
Log.Error("Could not post the user to the database");
return(false);
}
return(succes);
}
/// <summary>
/// Post method to be used in the analysis of the friends of a user in the TweetRetriever
/// </summary>
/// <param name="resultObj">The result object returned by algorithm</param>
/// <param name="user">The user</param>
/// <returns>Returns true if the post request succeeded</returns>
private bool PostResultToDBAndLink(AnalysisResultObj resultObj, User user)
{
//Uses the post method
bool succes = PostResultToDB(resultObj, user);
//Then it tries to find the posted user and use that users record id to link using the add follower method
long tempRecordId = -1;
succes = GetUsersRecordIdOnDb(user, ref tempRecordId);
if (!succes)
{
Log.Error("Could not find the posted user");
return(false);
}
AddFollower(tempRecordId);
return(succes);
}
static void Main(string[] args)
{
List <Tweet> tweetList = new List <Tweet>();
List <Tweet> tweetList2 = new List <Tweet>();
tweetList = FileHelper.ReadObjectFromFile <List <Tweet> >("\\TestData\\MaggieNYT-Left.json");
tweetList2 = FileHelper.ReadObjectFromFile <List <Tweet> >("\\TestData\\MaggieNYT-Left.json");
AnalysisResultObj result = TweetAnalyzer.Instance.AnalyzeAndDecorateTweetsThreaded(tweetList);
AnalysisResultObj result2 = TweetAnalyzer.Instance.AnalyzeAndDecorateTweets(tweetList2);
Console.WriteLine(tweetList[10].Text);
Console.WriteLine(result.GetPolPercent() + " " + result.GetAlgorithmResult() + " " + result.GetMediaResult() + " " + result.Count + " " + result.PolCount + " " + result.KeywordBias + " " + result.MediaBias);
Console.WriteLine(result2.GetPolPercent() + " " + result2.GetAlgorithmResult() + " " + result2.GetMediaResult() + " " + result2.Count + " " + result2.PolCount + " " + result2.KeywordBias + " " + result2.MediaBias);
Console.WriteLine("?");
Console.ReadLine();
}
/// <summary>
/// Analyzes a list of Tweets, and returns the following values as a double-array:
/// Keyword-Bias: Determined political value of the words used in the tweets
/// Media-Bias: Determined political value of news media linked to in the tweets
/// Count: Number of tweets analtyzed
/// Pos: Number of positive words found
/// Neg: Number of negative words found
/// </summary>
/// <param name="tweetList">The list of tweets</param>
/// <returns>AnalysisResultObj</returns>
public AnalysisResultObj AnalyzeAndDecorateTweets(List <Tweet> tweetList)
{
AnalysisResultObj output = new AnalysisResultObj();
foreach (Tweet tweet in tweetList)
{
tweet.HasQuotes = CheckForQuotationMarks(tweet);
if (!tweet.HasQuotes)
{
var puncturation = tweet.Text.Where(Char.IsPunctuation).Distinct().ToArray();
List <String> wordList = tweet.Text.Split(' ').Select(x => x.Trim(puncturation)).ToList <String>();
SentimentAnalysis(wordList, tweet);
KeywordAnalysis(wordList, tweet);
MediaAnalysis(tweet);
if (tweet.GetSentiment() > 0)
{
output.PositiveTweetsCount++;
}
else if (tweet.GetSentiment() < 0)
{
output.NegativeTweetsCount++;
}
output.NegativeSentiment += tweet.NegativeValue;
output.PositiveSentiment += tweet.PositiveValue;
if (tweet.KeywordBias != 0 || tweet.MediaBias != 0)
{
output.PolCount++;
output.KeywordBias += tweet.KeywordBias;
output.MediaBias += tweet.MediaBias;
}
}
}
output.Count = tweetList.Count;
return(output);
}
/// <summary>
/// The threaded tweet retrieve method which is used in the task
/// </summary>
/// <param name="user">The user</param>
/// <param name="auth">The auth object</param>
/// <param name="checkDB">The method to check the database for previous results</param>
/// <param name="classifyMethod">The method to classify tweets</param>
/// <param name="postToDB">The method to post the result to the database</param>
/// <returns>A list of tweets </returns>
private List <Tweet> TweetThreadMethod(User user, AuthObj auth, Func <User, bool> checkDB = null, Func <List <Tweet>, AnalysisResultObj> classifyMethod = null, Func <AnalysisResultObj, User, bool> postToDB = null)
{
bool alreadyExist = false;
AnalysisResultObj tempResult = new AnalysisResultObj();
List <Tweet> temp = new List <Tweet>();
if (checkDB != null)
{
alreadyExist = checkDB(user); //Uses the supplied method to find out if the result already exist on the database
if (alreadyExist)
{
Log.Debug(String.Format("{0,-30} {1,-20} {2,-11}", user.Name, user.Id, "Already exist in db"));
}
}
if (!alreadyExist)
{
temp = RetrieveTweets(user, auth);
if (user.IsProtected)
{
Log.Debug(String.Format("{0,-30} {1,-20} {2,-11}", user.Name, user.Id, "Protected"));
}
else
{
Log.Debug(String.Format("{0,-30} {1,-20} {2,-11}", user.Name, user.Id, temp.Count));
}
if (classifyMethod != null && postToDB != null)
{
tempResult = classifyMethod(temp); //Use the supplied method to classify the list of tweets
postToDB(tempResult, user); //Use the supplied method to post the result to the database
temp = new List <Tweet>();
}
}
return(temp);
}
