Implement tokenize().

cmd/ir-tfidf/main.go

@@ -4,15 +4,15 @@ import (
 	"fmt"
 
 	"github.com/gslin/go-ir-playground/internal/artifact"
-	"github.com/gslin/go-ir-playground/internal/ngram"
+	"github.com/gslin/go-ir-playground/internal/tokenize"
 )
 
 func main() {
 	articles := artifact.Read("data/articles.json")
 
 	for _, article := range articles {
-		title_bag := ngram.Bigram(article.Title)
+		title_bag := tokenize.Tokenize(article.Title)
-		body_bag := ngram.Bigram(article.Body)
+		body_bag := tokenize.Tokenize(article.Body)
 
 		fmt.Printf("title_bag = %v\n", title_bag)
 		fmt.Printf("body_bag = %v\n", body_bag)

internal/ngram/ngram.go

@@ -21,6 +21,10 @@ func Bigram(s string) []string {
 	return r
 }
 
+func Unigram(s string) []string {
+	return split(s)
+}
+
 func split(s string) []string {
 	bag := make([]string, 0)
 	for _, w := range re1.FindAllStringSubmatch(s, -1) {

internal/tokenize/tokenize.go

@@ -0,0 +1,14 @@
+package tokenize
+
+import (
+	"slices"
+
+	"github.com/gslin/go-ir-playground/internal/ngram"
+)
+
+func Tokenize(s string) []string {
+	bag := append(ngram.Unigram(s), ngram.Bigram(s)...)
+	slices.Sort(bag)
+	slices.Compact[[]string, string](bag)
+	return bag
+}