A lowercase-only version of clojure.string/split with begin/end position of each word in original string. Beautiful, but less efficient than a loop-based implementation.

tokenizer.clj

(def legal-char (set (map char (concat (range 48 58) (range 97 123)))))

(defn tokenize [s]
  (->> (map-indexed list (string/lower-case s))
       (partition-by (comp not legal-char second))
       (filter (comp legal-char second first))
       (mapv #(hash-map :s (apply str (map second %))
                        :begin (first (first %))
                        :end   (first (last %))))))