更简单的解决方案：

DataStream<String> nwords = input.filter(s -> startsWith("N"));
DataStream<String> others = input.filter(s -> !startsWith("N"));

我相信这比使用侧输出的解决方案效率略低，但它仍然可以在单个任务中运行，使用操作符链接，因此它也不需要ser/de开销，也不需要网络。
别误会——一般来说，边输出是分割流的方法。

我认为您可以利用侧输出来实现这一点。只需使用processfunction发出实际收集器中的匹配元素和带有side output标记的不匹配元素，然后从主流中获取side output元素。
例如，你的代码可以这样修改，

package datastream;

import org.apache.flink.api.java.tuple.Tuple2;
import org.apache.flink.streaming.api.datastream.DataStream;
import org.apache.flink.streaming.api.datastream.SingleOutputStreamOperator;
import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
import org.apache.flink.streaming.api.functions.ProcessFunction;
import org.apache.flink.util.Collector;
import org.apache.flink.util.OutputTag;

public class WordStream {

    public static void main(String[] args) throws Exception {

        final StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();

        DataStream<String> inputData = env.socketTextStream("localhost", 9999);

        // Initialize side-output tag to collect the un-matched elements 
        OutputTag<Tuple2<String, Integer>> unMatchedSideOutput = new OutputTag<Tuple2<String, Integer>>("unmatched-side-output") {};

        SingleOutputStreamOperator<Tuple2<String, Integer>> tokenize = inputData
                .process(new ProcessFunction<String, Tuple2<String, Integer>>() {
                    @Override
                    public void processElement(String value, Context ctx, Collector<Tuple2<String, Integer>> out) {
                        if (value.startsWith("N")) {
                            // Emit the data to actual collector
                            out.collect(new Tuple2<>("Matched=" + value, Integer.valueOf(1)));
                        } else {
                            // Emit the un-matched data to side output
                            ctx.output(unMatchedSideOutput, new Tuple2<>("UnMatched=" + value, Integer.valueOf(1)));
                        }
                    }
                });

        DataStream<Tuple2<String, Integer>> count = tokenize.keyBy(0).sum(1);

        // Fetch the un-matched element using side-output tag and process it
        DataStream<Tuple2<String, Integer>> unMatchedCount = tokenize.getSideOutput(unMatchedSideOutput).keyBy(0).sum(1);

        count.print();

        unMatchedCount.print();

        env.execute("WordStream");

    }
}

注意：我用前缀稍微更改了发射的值 Matched= 以及 UnMatched= 在输出中得到清晰的理解。
对于以下输入，

Hello
Nevermind
Hello

我得到以下输出，

3> (UnMatched=Hello,1)
4> (Matched=Nevermind,1)
3> (UnMatched=Hello,2)