| //SRSMTC: SRS-based MTC Algorithm |
| //the -grams |
| Int ; |
| //the simple random sampling rate |
| Float ; |
| //the number of categories |
| Int ; |
| //the token level memory |
| TLM ; |
| //extract tokens based on overlapping word-level -grams model |
| String Tokenizer(Document ) |
| //sample tokens based on the simple random sampling rate |
| String SimpleRandomSampler(String ) |
| //compute conditional probability P() for each category |
| Float BayesianPredictor(String token) |
| (1) Float := new Float; |
| (2) Loop: For Each Int Do: |
| (2.1) := .TF(, token)/.DF(); |
| (3) Float sum:= Sum(); //add the floats to a sum |
| (4) Loop: For Each Int Do: |
| (4.1) := /sum; |
| (5) Output . |
| //SRSMTC.T: Training Procedure of SRSMTC |
| SRSMTC.T(Document ; Category ) |
| (1) Int := D.size; //get the number of training documents |
| (2) Loop: For Each Int Do: |
| (2.1) Document := ; |
| (2.2) Category := ; |
| (2.3) String := Tokenizer(); |
| (2.4) String := SimpleRandomSampler(); |
| (2.5) .DF():= .DF() + 1; |
| (2.6) Loop: For Each Do: |
| (2.6.1) If .contain() Then: .TF():= .TF() + 1; |
| (2.6.2) Else: |
| (2.6.2.1) .TF():= 1; |
| (2.6.2.2) .TF():= 0; // means all other categories |
| (2.6.2.3) .put(). |
| //SRSMTC.P: Predicting Procedure of SRSMTC |
| Category SRSMTC.P(Document D) |
| (1) Int := .size; //get the number of testing documents |
| (2) Category := new Category; |
| (3) Loop: For Each Int Do: |
| (3.1) Document := ; |
| (3.2) String := Tokenizer(); |
| (3.3) Float ep:= new Float; |
| (3.4) Loop: For Each String Do: |
| (3.4.1) Float := BayesianPredictor(); |
| (3.4.2) Loop: For Each Int Do: |
| (3.4.2.1) ep[]:= ep[] + ; |
| (3.5) Float sum:= Sum(ep); //add the floats to a sum |
| (3.6) Loop: For Each Int Do: |
| (3.6.1) ep[]:= ep[]/sum; |
| (3.7) Int index:= Math.max(ep).getIndex; |
| (3.8) := ; |
| (4) Output C. |
