func main() {
tokenizer, err := goseg.NewTokenizerFromFile(dictDataFilePath)
if err != nil {
log.Fatal(err)
}
for _, item := range allData {
fis, err := ioutil.ReadDir(item.folder)
termFreq := make(map[string]int)
if err != nil {
log.Fatal(err)
}
for _, fi := range fis {
if fi.IsDir() {
continue
}
filePath := filepath.Join(item.folder, fi.Name())
mail := mailfile.NewPOP3Mail(filePath)
if err = mail.Parse(); err != nil {
log.Fatal(err)
}
post, err := mailpost.Parse(mail)
mail.Close()
if err != nil {
log.Fatalf("Err: %v, Mail:%s", err, mail.Path())
}
words := tokenizer.Cut([]rune(post.Subject + " " + post.Content))
for _, word := range words {
key := strings.Trim(word, cutset)
if len(key) > 1 {
termFreq[key] = termFreq[key] + 1
}
}
}
pairList := sortMapByValue(termFreq)
output, err := os.OpenFile(item.output, os.O_WRONLY|os.O_TRUNC|os.O_CREATE, 0644)
writer := bufio.NewWriter(output)
offset := int(float64(len(pairList)) * topTermRatio)
for _, pair := range pairList[offset:] {
if len([]rune(pair.Key)) > termMinLength {
writer.WriteString(pair.Key + " " + strconv.Itoa(pair.Value) + "\n")
}
}
writer.Flush()
output.Close()
}
}
func main() {
classifier := bayesian.NewClassifier(Good, Bad)
tokenizer, err := goseg.NewTokenizerFromFile(dictDataFilePath)
if err != nil {
log.Fatal(err)
}
for _, item := range trainingData {
log.Printf("Traning %s", item.folder)
totalNum := 0
var totalSize int64
startTime := time.Now()
fis, err := ioutil.ReadDir(item.folder)
if err != nil {
log.Fatal(err)
}
for _, fi := range fis {
if fi.IsDir() {
continue
}
totalSize += fi.Size()
filePath := filepath.Join(item.folder, fi.Name())
mail := mailfile.NewPOP3Mail(filePath)
if err = mail.Parse(); err != nil {
log.Fatal(err)
}
post, err := mailpost.Parse(mail)
mail.Close()
if err != nil {
log.Fatalf("Err: %v, Mail:%s", err, mail.Path())
}
words := analyzer.Normalize(tokenizer.Cut([]rune(post.Subject+" "+post.Content)), cutset)
classifier.Learn(words, item.class)
totalNum += 1
}
elapsed := time.Now().Sub(startTime)
fmt.Printf("TotalNum: %d\n", totalNum)
fmt.Printf("Elapsed: %s, TPS(Mail): %f, TPS(FileSize): %s\n",
time.Now().Sub(startTime),
float64(totalNum)/(float64(elapsed)/float64(time.Second)),
fileutil.Humanize(uint64(float64(totalSize)/(float64(elapsed)/float64(time.Second)))))
}
classifier.WriteToFile(output)
}
func NewBayesianAnalyzer(traningDataFilePath string, dictDataFilePath string) (*BayesianAnalyzer, error) {
tokenizer, err := goseg.NewTokenizerFromFile(dictDataFilePath)
if err != nil {
return nil, err
}
coordinator := &sync.RWMutex{}
classifier, err := bayesian.NewClassifierFromFile(traningDataFilePath)
if err != nil {
return nil, err
}
return &BayesianAnalyzer{tokenizer, classifier, nil, coordinator}, nil
}
func NewBayesianAnalyzerWithUpdater(traningDataFilePath string, dictDataFilePath string, updateDelay time.Duration) (*BayesianAnalyzer, error) {
tokenizer, err := goseg.NewTokenizerFromFile(dictDataFilePath)
if err != nil {
return nil, err
}
coordinator := &sync.RWMutex{}
classifier, err := bayesian.NewClassifierFromFile(traningDataFilePath)
if err != nil {
return nil, err
}
updater := NewDelayedUpdater(classifier, traningDataFilePath, updateDelay, coordinator)
return &BayesianAnalyzer{tokenizer, classifier, updater, coordinator}, nil
}
func main() {
classifier := bayesian.NewClassifier(Good, Bad)
tokenizer, err := goseg.NewTokenizerFromFile(dictDataFilePath)
checkErr(err)
goodWords, err := getWords("goodwords.txt", tokenizer)
checkErr(err)
fmt.Printf("Normalized Good Words:\n%s\n", goodWords)
fmt.Println("")
badWords, err := getWords("badwords.txt", tokenizer)
checkErr(err)
fmt.Printf("Normalized Bad Words:\n%s\n", badWords)
classifier.Learn(goodWords, Good)
classifier.Learn(badWords, Bad)
classifier.WriteToFile(output)
}