#System 1 for Aicyber
##A vector space model approach.
python system_1_baseline.py
##Preprocessing
The text data is first being processed by tweet tokenizer, emoticons are preserved as tokens, some emoticons are replaced by words , eg :) -> good , :( -> bad
##Features
Bag-of-ngram feature is extracted and filtered by a TF-IDF (Salton, 1991) selection.
Resulting feature dimension is around 3800, it is then reduced to 400 by truncated singular value decomposition (SVD) (Klema and Laub, 1980; Halko et al., 2009).
This process is also known as Latent Semantic Analysis (LSA) or Vector Space Model (Turney and Pantel, 2010).
##Classifier
Finally a Linear Discriminant Analysis (LDA) classifier (Hastie et al., 2009) is trained to classify the test data.
##Training and verification
The SemEval 2016 training dataset which contains 3887 tweets are selected to train the TF-IDF,SVD and LDA. Development dataset is used for tuning parameters and develop-test dataset are used for local testing.
##Results:
SVD DIM=400
Twitter2016 0.4048
SVD DIM=600
Twitter2016 0.4261
Please noted this result is slightly better than the system we submitted (0.4025) during evaluation period, due to an upgrade in tokenizer.
#System 2 an i-vector based system
#FAQ: ##When you join the evaluation? One week before evaluation submission deadline. ##Did you download all the data? No, we only managed to download part of SemEval 2016 tweet data, eg 3887 out of 6000 in the training set. ##Why you can't download more tweet data before submission deadline? Tweet is blocked , we are behind the Great Firewall , it take efforts to download data from tweet.