人工操作机器转换成自动程序_为什么将我的应用程序转换为聊天机器人

人工操作机器转换成自动程序

by P. Daniel Tyreus

由P.Daniel Tyreus

我将移动应用程序变成了聊天机器人。 这就是为什么。 (I turned my mobile app into a chatbot. Here’s why.)

Just under a year ago, I embarked on a quest to better understand my seasonal allergies. In support of this quest, I spent a few weekends writing an iOS and Android app called Hayfever. With the app, I could log my symptoms and it would collect data about the allergens in the area. My grand idea was that if enough people logged symptoms using the app, we could crunch the numbers and start to learn from the trends.

不到一年前，我开始寻求更好地了解我的季节性过敏症 。 为了支持该任务，我花了几个周末编写了一个名为Hayfever的iOS和Android应用程序。 使用该应用程序，我可以记录我的症状，并且它将收集有关该地区过敏原的数据。 我的宏伟想法是，如果有足够多的人使用该应用程序记录了症状，我们就可以计算数字并开始从趋势中学习。

Early users liked the idea behind Hayfever. Despite that, it did not get downloaded very much. After looking at the options, I decided to ditch the traditional app interface for a conversational one and to convert Hayfever to a chatbot. This post explains the reasoning that lead me down this path.

早期用户喜欢Hayfever背后的想法。 尽管如此，它并没有下载太多。 在查看了选项之后，我决定放弃传统的应用界面来进行对话，并将Hayfever转换为聊天机器人 。 这篇文章解释了导致我走这条路的原因。

分析趋势 (Analyzing the Trends)

I’m not completely surprised that Hayfever had trouble getting downloads. There are some competing apps with good ranking in the app stores. Most of these are made by the same companies that are trying to sell OTC allergy medications. I have my doubts about the actual motives behind them. But these are big companies that have a lot of advertising budget to spend promoting their apps.

对于Hayfever遇到下载困难，我并不感到完全惊讶。 在应用商店中，有一些竞争性应用具有良好的排名。 其中大多数是由试图出售OTC过敏药物的同一家公司生产的。 我对它们背后的实际动机感到怀疑。 但是这些都是大公司，他们有大量广告预算可用于推广其应用。

On top of that, there are some trends in the industry that make things more difficult for independent developers. In September of 2016, Recode reported that “half of U.S. smartphone users download zero apps per month.” That’s also not a new trend, as we have been seeing similar numbers since 2014.

最重要的是，该行业中的一些趋势使独立开发人员的工作更加困难。 2016年9月， Recode报告说：“美国一半的智能手机用户每月下载零个应用程序。” 这也不是新趋势，因为自2014年以来我们一直看到类似的数字 。

This doesn’t mean people aren’t using apps anymore. In fact, in its year-end retrospective, App Annie said time spent in apps grew more than 25 percent in 2016 in the United States. But there is more to that story. Consumers are spending over 84 percent of their time on their smartphones using using just five non-native apps they’ve installed from the App Store. In summary, consumers are using mobile apps, but it’s primarily existing apps from a handful of companies.

这并不意味着人们不再使用应用程序。 实际上，在年终回顾中，App Annie表示，2016年美国在应用上花费的时间增长了25％以上 。 但是这个故事还有更多。 消费者仅使用从App Store安装的五个非本机应用程序，就将超过84％的时间花费在智能手机上。 总而言之，消费者正在使用移动应用程序，但主要是少数公司现有的应用程序。

Indeed, according to Sensor Tower data, in Q3 2016, the top 5 iOS app downloads were Pokémon GO, Uber, Facebook Messenger, WhatsApp, and Facebook. Three of these are owned by the same company and you may notice that two of the top 5 are messaging apps. Similarly, App Annie reports that top five apps by downloads in 2016 on iOS and Google play were Facebook, WhatsApp Messenger, Facebook Messenger, Instagram, and Snapchat.

实际上，根据Sensor Tower的数据，在2016年第三季度，排名前5的iOS应用下载为PokémonGO，Uber，Facebook Messenger，WhatsApp和Facebook。 其中三个由同一公司拥有，您可能会注意到前五名中的两个是消息传递应用程序。 同样，App Annie 报告称，2016年iOS和Google Play上下载量排名前五的应用程序是Facebook，WhatsApp Messenger，Facebook Messenger，Instagram和Snapchat。

Here is a summary of what we know from these data:

以下是我们从这些数据中了解的内容的摘要：

1. We know that people are spending a lot of time using apps on their phones.
   我们知道人们花大量时间在手机上使用应用程序。
2. We know that people mostly use about five non-native apps.
   我们知道人们通常使用大约五个非本地应用程序。
3. We know app download numbers reflect this pattern. The average user downloads zero new apps per month.
   我们知道应用下载数量反映了这种模式。 平均用户每月下载零个新应用。
4. Of the apps that people do download, a select few rack up most of the installs.
   人们下载的应用程序中，只有少数几个安装了大多数应用程序。
5. Of the top apps downloaded in 2016, roughly half are messaging platforms.
   在2016年下载的顶级应用程序中，大约有一半是消息传递平台。

I’m willing to speculate that it’s easier to acquire a user if the user doesn’t have to download a new app to use a service. I’m also willing to speculate that users are more likely to continue using a service that’s integrated into an app they already use. Finally, I’m willing to gamble that more and more people are comfortable chatting with a bot on a messaging platform.

我愿意推测， 如果用户不必下载新的应用程序来使用服务，则更容易获得用户 。 我还愿意推测用户更有可能继续使用已集成到他们已经使用的应用程序中的服务 。 最后，我愿意赌博，越来越多的人愿意在消息传递平台上与机器人聊天 。

适应趋势 (Adapting to the Trends)

After studying the trends outlined above, it’s clear to me that the slow adoption Hayfever saw in it’s first year was no fluke. This is the reality of the current mobile app development environment. Yet, I still believe that Hayfever addresses a real concern and offers a valuable service. After all there are 50 million nasal allergy sufferers in the United States and rising. So after some consideration, I was able to come up with three options for moving forward:

在研究了上面概述的趋势之后，对我来说很明显，海弗佛在第一年的缓慢采用并不是偶然。 这是当前移动应用程序开发环境的现实。 但是，我仍然相信Hayfever能够解决真正的问题并提供有价值的服务。 毕竟，在美国有五千万的鼻过敏患者正在增加。 因此，经过一番考虑，我能够提出三种前进的选择：

1. Spend money and time on paid downloads and growth initiatives.
   花时间和金钱进行付费下载和增长计划。
2. Introduce a viral component to the app to encourage organic growth.
   向应用程序引入病毒组件，以鼓励有机增长。
3. Pivot to reduce installation friction and go where there are already users.
   旋转以减少安装摩擦并转到已有用户的地方。

My budget for buying advertising and paid downloads is very limited. Hayfever is a side project and I can’t pour money into it. While I would love to introduce a viral component, I haven’t been able to come up with a good hook yet. So when I look at this from a pure user acquisition angle, converting Hayfever to a chatbot seems like a promising choice. But user acquisition is only the first piece of the puzzle. To keep those users their experience with the conversational UI has to be positive.

我购买广告和付费下载的预算非常有限。 海弗佛(Hayfever)是一个附属项目，我无法投入资金。 虽然我想介绍一种病毒成分，但我还没有找到一个好的钩子。 因此，从纯粹的用户获取角度来看，将Hayfever转换为聊天机器人似乎是一个有前途的选择。 但是，获得用户仅仅是难题的第一步。 为了使这些用户保持对会话UI的体验，必须积极参与。

会话用户界面的争论 (Arguments Against a Conversational User Interface)

There are also some arguments against chatbots and conversational user interfaces. Despite a tremendous amount of hype in 2016, there have not been any breakout chatbot successes. There has been nothing even close to the scale of Pokemon GO. In fact, the most successful apps of the on-demand economy seem to show that a conversational UI is less useful than a well-designed graphical UI. Uber, one of the most downloaded apps of 2016, is a perfect example. Uber has shown us that millions of users would rather tap a button on a phone than call a taxi company and talk to a dispatcher.

还有一些反对聊天机器人和会话用户界面的论点。 尽管在2016年进行了大量宣传，但没有突破性的聊天机器人取得成功。 口袋妖怪GO的规模几乎没有。 实际上，按需经济中最成功的应用似乎表明，对话UI的使用不如精心设计的图形UI有用。 Uber是2016年下载次数最多的应用程序之一，就是一个很好的例子。 Uber向我们展示了数百万用户宁愿点击电话上的按钮，也不愿致电出租车公司与调度员交谈。

Also, the user’s location can be a key piece of information for a lot of mobile apps. With traditional mobile apps, as long as the user grants permission, the app can access the user’s current location. This persistent access to location is not available to most chatbot developers yet. On Facebook Messenger the chatbot can request the user’s location, but it must do so each time it needs it. This is a big deal for Hayfever. Location is critical to look up the allergens in the area when a user logs symptoms. No direct access to location is the strongest argument against converting Hayfever to a chatbot at this time.

同样，用户的位置可能是许多移动应用程序的关键信息。 使用传统的移动应用程序，只要用户授予许可，该应用程序就可以访问用户的当前位置。 对于大多数聊天机器人开发人员而言，这种对位置的持久访问尚不可用。 在Facebook Messenger上，聊天机器人可以请求用户的位置，但是每次需要时都必须这样做。 对于海弗佛来说，这是大事。 用户记录症状时，位置对于查找该区域的过敏原至关重要。 目前，禁止直接访问位置是反对将Hayfever转换为聊天机器人的最有力依据。

Finally, developing a chatbot app requires a different skill set than a traditional mobile app. I found myself spending more time writing copy than I ever have for an app. One essentially has to write a script, which is not something most developers are comfortable doing. Also, the tools for writing, editing, and testing chatbot interactions are hard to find. I will write a follow up post about the tools I used to create the chatbot.

最后，与传统的移动应用程序相比，开发聊天机器人应用程序需要不同的技能。 我发现自己花在写作上的时间比花在应用程序上的时间要多。 从本质上讲，必须编写一个脚本，这对大多数开发人员来说都不是一件容易的事。 而且，很难找到用于编写，编辑和测试聊天机器人交互的工具。 我将撰写有关用于创建聊天机器人的工具的后续文章。

聊天机器人可以做什么 (What a Chatbot Can Do Well)

When a technology gets a lot of hype, eager developers will try to use that technology to solve every type of problem. These are still early times for conversational user interfaces. We haven’t really seen best practices emerge yet. But there are several things that conversational UI’s can do very well and are a great fit for an app like Hayfever.

当一项技术大肆宣传时，热心的开发人员将尝试使用该技术来解决各种类型的问题。 对于会话用户界面，这些还处于早期。 我们还没有真正看到最佳实践。 但是，会话UI可以在很多方面做得很好，非常适合Hayfever之类的应用。

In the past I have used several apps to log information like what I’m eating, what workouts I’m doing, or how I’m feeling. My experience with most of these is that interaction with the GUI can be tedious and dry. This means I get bored and stop logging. A conversational UI can add a little personality to an otherwise bland interaction. Instead of checking boxes on a form, I can get the feeling of discussing my symptoms with a sympathetic friend or doctor. A chatbot can express concern on days when my symptoms are bad, offer advice or remedies, and convey happiness when my symptoms subside.

过去，我曾使用过多个应用程序来记录信息，例如我在吃什么，正在做哪些锻炼或感觉如何。 我对大多数这些工具的经验是，与GUI的交互可能很枯燥乏味。 这意味着我无聊并停止记录。 对话式UI可以为原本平淡无奇的互动增加一点个性 。 我可以不用与表格上的复选框打勾，而是可以与同情的朋友或医生讨论我的症状。 当我的症状恶化时，聊天机器人可以表达担忧，提供建议或补救措施，并在症状消退时传达快乐。

A GUI also has the challenge of displaying a large number of possible options on the screen. This can be overwhelming and distracting for users. I acknowledge that a more talented designer could do a much better job than I in laying out the data entry form and hiding unnecessary features. But that’s definitely an art and not a trivial skill. With a conversational UI it’s easier to surface only the questions relevant based on the current user’s history and progression through the conversation. This is a balancing act because leading a user through endless yes/no questions is not a good experience either. I’ve found that once Hayfever knows a little bit about my symptom history, things don’t change too much and logging becomes pleasant.

GUI还具有在屏幕上显示大量可能选项的挑战。 对于用户而言，这可能是不堪重负的，而且会分散用户的注意力。 我承认，在布局数据输入表单和隐藏不必要的功能方面，更有才华的设计师可以比我做得更好。 但这绝对是一门艺术，而不是琐碎的技能。 使用对话式用户界面，可以更轻松地根据当前用户的历史记录和整个对话过程只显示相关的问题 。 这是一种平衡的行为，因为通过无休止的是/否问题引导用户也不是一种好的体验。 我发现，一旦Hayfever对我的症状历史有所了解，事情就不会改变太多，并且伐木变得令人愉快。

Finally, chatbots can handle re-engagement very naturally. With Hayfever, knowing that someone is not experiencing symptoms is as valuable as knowing that they are. But when someone is symptom-free, logging is not top of mind. A chatbot can send a natural language prompt like, “Yesterday you reported mild allergy symptoms. Are you still sneezing?” Traditional apps can do this with push notifications, but not without added complexity. First, your user must allow the push notification to come through. Second, the app must navigate to the correct point in the data entry flow when the user taps the notification. With a chatbot, the user drops directly into that point in the conversation after opening the messaging client.

最后，聊天机器人可以非常自然地处理重新参与。 使用Hayfever，知道某人没有症状与知道自己有病一样有价值。 但是，当某人没有症状时，记录日志就不是首要考虑。 聊天机器人可以发送自然语言提示，例如“昨天您报告了轻微的过敏症状。 你还在打喷嚏吗？” 传统的应用程序可以通过推送通知来做到这一点，但并非没有复杂性。 首先，您的用户必须允许推送通知通过。 其次，当用户点击通知时，应用程序必须导航到数据输入流中的正确点。 使用聊天机器人，用户在打开消息传递客户端后即可直接进入对话中的该点。

加起来 (Summing Up)

For the last few months I have been looking for a way to get better adoption on an app I wrote to let people log and analyze allergy symptoms. I reviewed data showing that users are not installing new apps often on their phones. Some of the top apps that are being installed are messaging apps. So I converted Hayfever into a chatbot on Facebook Messenger. I discovered that creating a chatbot required me to use some different skills, copywriting in particular. I also found that chatbots are not ideal for use cases that rely on a user’s current location. But a conversational user interface was able to simplify repeated data entry. It also made logging information more engaging. I will find out during this allergy season if the chatbot version of Hayfever gets more usage than the app version.

在过去的几个月中，我一直在寻找一种方法来更好地利用我编写的让人们记录和分析过敏症状的应用程序。 我查看了表明用户并非经常在手机上安装新应用程序的数据。 某些正在安装的顶级应用程序是消息传递应用程序。 因此，我将Hayfever转换为Facebook Messenger上的聊天机器人 。 我发现创建聊天机器人需要我使用一些不同的技能，特别是文案写作。 我还发现，聊天机器人对于依赖于用户当前位置的用例不是理想的选择。 但是对话用户界面能够简化重复的数据输入。 这也使日志记录信息更具吸引力。 我将在此过敏季节中找出Hayfever的聊天机器人版本是否比应用程序版本获得更多使用。

If you liked this and want me to write more about my chatbot experiences, click the? below.

如果您喜欢这个，并且希望我写更多有关我的聊天机器人体验的信息，请单击“开始”。 下面。

翻译自: https://www.freecodecamp.org/news/why-i-converted-my-app-to-a-chatbot-96355596725c/

人工操作机器转换成自动程序