<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Mehfuz Hossain on Medium]]></title>
        <description><![CDATA[Stories by Mehfuz Hossain on Medium]]></description>
        <link>https://medium.com/@mehfuzh?source=rss-9a56267236af------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*aA2t2jOyVr5k33hwxmxbSg.jpeg</url>
            <title>Stories by Mehfuz Hossain on Medium</title>
            <link>https://medium.com/@mehfuzh?source=rss-9a56267236af------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 10:33:17 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@mehfuzh/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Natural Language Understanding with Smartloop NLP]]></title>
            <link>https://medium.com/smartloopai/natural-language-understanding-with-smartloop-nlp-42b0b92f56f6?source=rss-9a56267236af------2</link>
            <guid isPermaLink="false">https://medium.com/p/42b0b92f56f6</guid>
            <category><![CDATA[tensorflow]]></category>
            <category><![CDATA[bidirectional-lstm]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[nlp]]></category>
            <category><![CDATA[smartloop]]></category>
            <dc:creator><![CDATA[Mehfuz Hossain]]></dc:creator>
            <pubDate>Sun, 30 Jul 2023 17:55:31 GMT</pubDate>
            <atom:updated>2023-09-26T16:57:44.203Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XWoiZt9KF1P8Oqtza1UCww.png" /></figure><p>Smartloop NLP framework powers the <a href="https://dashboard.smartloop.ai/">platform</a> for processing natural language inputs. Smartloop, over the years, has become a community-driven platform; it is used by hotels, healthcare providers, and name brands worldwide with little or no intervention.</p><p>In that direction, we are open-sourcing the NLP framework that powers most of our platform, which you can fork or contribute using the following link:</p><p><a href="https://github.com/SmartloopAI/sl-core">GitHub - SmartloopAI/sl-core: Natural Language Understanding Framework</a></p><p>In this post, I will cover using Smartloop NLU to train and test natural language queries.</p><p>To use the NLP, you will need Tensorflow &gt;= 2.12.0 installed and configured in your local machine. If you are using <strong>M1</strong> mac then you will need the equivalent tensorflow-macos and metal plugin installed to take advantage of the GPU processing.</p><p>I have used miniconda to configure my system, once you have everything configured, you should see the following:</p><pre>➜  conda activate slp<br>(slp) ➜  nlu-demo python <br>Python 3.10.10 | packaged by conda-forge | (main, Mar 24 2023, 20:12:31) [Clang 14.0.6 ] on darwin<br>Type &quot;help&quot;, &quot;copyright&quot;, &quot;credits&quot; or &quot;license&quot; for more information.<br>&gt;&gt;&gt; import tensorflow as tf<br>&gt;&gt;&gt; tf.__version__<br>&#39;2.12.0&#39;<br>&gt;&gt;&gt; tf.test.is_gpu_available()<br>WARNING:tensorflow:From &lt;stdin&gt;:1: is_gpu_available (from tensorflow.python.framework.test_util) is deprecated and will be removed in a future version.<br>Instructions for updating:<br>Use `tf.config.list_physical_devices(&#39;GPU&#39;)` instead.<br>Metal device set to: Apple M2 Max<br><br>systemMemory: 32.00 GB<br>maxCacheSize: 10.67 GB<br><br>True<br>&gt;&gt;&gt;</pre><p>I am using <strong>macOS</strong> for his post, but GPU training is much more accurate/efficient with nvidia + windows combination, I will cover more detail in the next post on how to optimize the RNN model using TensorRT.</p><p>Now that you have configured your environment, the next step is to install smarloop NLP using pip ,</p><pre>pip install smartloop==1.1.0</pre><p>The next step is to define the model. Copy and paste the following model.json to your root directory of main.py :</p><pre>{<br>    &quot;examples&quot;: {<br>        &quot;intents&quot;: [<br>            {<br>                &quot;text&quot;: &quot;about&quot;,<br>                &quot;intent&quot;: &quot;about&quot;<br>            },<br>            {<br>                &quot;text&quot;: &quot;about&quot;,<br>                &quot;intent&quot;: &quot;about&quot;<br>            },<br>            {<br>                &quot;text&quot;: &quot;learn more&quot;,<br>                &quot;intent&quot;: &quot;about&quot;<br>            },<br>            {<br>                &quot;text&quot;: &quot;more about the platform&quot;,<br>                &quot;intent&quot;: &quot;about&quot;<br>            },<br>            {<br>                &quot;text&quot;: &quot;hello&quot;,<br>                &quot;intent&quot;: &quot;start&quot;<br>            },<br>            {<br>                &quot;text&quot;: &quot;start&quot;,<br>                &quot;intent&quot;: &quot;start&quot;<br>            },<br>            {<br>                &quot;text&quot;: &quot;hi&quot;,<br>                &quot;intent&quot;: &quot;start&quot;<br>            }<br>        ]<br>    },<br>    &quot;lang&quot;: &quot;en&quot;<br>}</pre><p>Next, in your main.py import the following libraries:</p><pre>from smartloop.core import FileConfig<br>from smartloop.core import Project<br>from smartloop.core.text_classifier import TextClassifier</pre><p>Now, let&#39;s create a train Function:</p><pre>def train():<br>cfg = FileConfig()<br>  <br> model_id = int(time.time())<br><br> proj = Project(data_dir=data_dir, project_id=project_id, model_id=str(model_id))<br> cls = TextClassifier(proj=proj, cfg=cfg)<br><br> with open(&#39;model.json&#39;, &#39;r&#39;) as f:<br>  data = json.loads(f.read())<br><br> cls.fit(data)</pre><p>A few things are happening here:</p><ul><li>Loading the config (default config.yaml)</li><li>Creating the project with project_id , data_dir(where the model will be saved)</li><li>Using the model.json for training the model</li></ul><p>Copy and pasteconfig.yaml to your project:</p><pre># number of epochs<br>epochs: 100<br><br># Use tensorboard callback<br>logs: True<br><br># classifier parameters<br>embedded_intent_classifier:<br>    # base neurons to be used by LSTM model<br>    neurons: 64<br>    # length of input len(&quot;hello how are you&quot;) = 4<br>    input_length: 100<br>    # learning rate<br>    learning_rate: 1e-2<br>    # flatten<br>    flatten: False<br>    # number of hidden layer<br>    hidden_layers: 1<br>    # drop rate to avoid overfitting<br>    drop_rate: 0.5<br>    # early stop training in case of not improving<br>    early_stopping: False</pre><p>If you are into machine learning, you can tweak this configuration to fine tune the model, but this should be good enough for our use case and this post.</p><p>One last step before you can train the model is to install nltk corpus:</p><pre>python -m nltk.downloader stopwords  </pre><p>Now, let&#39;s train the model:</p><pre>(misc) ➜  nlu-demo python main.py train<br>Metal device set to: Apple M2 Max<br><br>systemMemory: 32.00 GB<br>maxCacheSize: 10.67 GB<br><br>WARNING:absl:At this time, the v2.11+ optimizer `tf.keras.optimizers.RMSprop` runs slowly on M1/M2 Macs, please use the legacy Keras optimizer instead, located at `tf.keras.optimizers.legacy.RMSprop`.<br>WARNING:absl:There is a known slowdown when using v2.11+ Keras optimizers on M1/M2 Macs. Falling back to the legacy Keras optimizer, i.e., `tf.keras.optimizers.legacy.RMSprop`.<br>Model: &quot;sequential&quot;<br>_________________________________________________________________<br> Layer (type)                Output Shape              Param #   <br>=================================================================<br> embedding (Embedding)       (None, 32, 2)             14        <br>                                                                 <br> bidirectional (Bidirectiona  (None, 64)               8960      <br> l)                                                              <br>                                                                 <br> dropout (Dropout)           (None, 64)                0         <br>                                                                 <br> dense (Dense)               (None, 64)                4160      <br>                                                                 <br> dense_1 (Dense)             (None, 2)                 130       <br>                                                                 <br>=================================================================<br>Total params: 13,264<br>Trainable params: 13,264<br>Non-trainable params: 0<br>_________________________________________________________________<br>2023-07-29 19:57:37.610164: W tensorflow/tsl/platform/profile_utils/cpu_utils.cc:128] Failed to get CPU frequency: 0 Hz<br>Epoch: 0, accuracy:0.0000, loss:0.8271, val_accuracy:0.0000, val_loss:1.1254, <br>########## Epoch: 10, accuracy:0.7500, loss:0.6238, val_accuracy:0.0000, val_loss:1.4519, <br>########## Epoch: 20, accuracy:1.0000, loss:0.1624, val_accuracy:0.0000, val_loss:1.7909, <br>########## Epoch: 30, accuracy:1.0000, loss:0.0670, val_accuracy:0.0000, val_loss:1.1486, <br>########## Epoch: 40, accuracy:1.0000, loss:0.0441, val_accuracy:0.0000, val_loss:1.2058, <br>########## Epoch: 50, accuracy:1.0000, loss:0.0290, val_accuracy:0.0000, val_loss:1.1323, <br>########## Epoch: 60, accuracy:1.0000, loss:0.0249, val_accuracy:0.0000, val_loss:1.1935, <br>########## Epoch: 70, accuracy:1.0000, loss:0.0152, val_accuracy:0.0000, val_loss:1.2751, <br>########## Epoch: 80, accuracy:1.0000, loss:0.0079, val_accuracy:0.0000, val_loss:1.2548, <br>########## Epoch: 90, accuracy:1.0000, loss:0.0057, val_accuracy:0.0000, val_loss:1.5172, <br>########## Epoch: 99, accuracy:1.0000, loss:0.0029, val_accuracy:0.0000, val_loss:1.2377, <br><br>1690678657For this particular model, we may not need 100 epochs , in that case, it would be good enough just with 50 epochs in the config file, I leave it to the reader to tune it further.</pre><p>Here, around 50 epochs will be good enough for this model but I will leave that to the reader to fine-tune</p><p>Next, the step to parse and see that we are getting the correct response:</p><pre>def test(project_id, model_id, query):<br> cfg = FileConfig()<br><br> proj = Project(data_dir=data_dir, project_id=project_id, model_id=model_id)<br> cls = TextClassifier(proj=proj, cfg=cfg)<br><br> return cls.transform(query)</pre><p>Running test should give you the following response:</p><pre>(misc) ➜  nlu-demo python main.py test <br>Metal device set to: Apple M2 Max<br><br>systemMemory: 32.00 GB<br>maxCacheSize: 10.67 GB<br><br>2023-07-29 19:57:48.376778: W tensorflow/tsl/platform/profile_utils/cpu_utils.cc:128] Failed to get CPU frequency: 0 Hz<br>1/1 [==============================] - 0s 293ms/step<br>{&#39;name&#39;: &#39;start&#39;, &#39;confidence&#39;: 0.9848259091377258}</pre><p>Putting it all together:</p><pre>import os<br>import time<br>import json<br>import argparse<br><br>from smartloop.core import FileConfig<br>from smartloop.core import Project<br>from smartloop.core.text_classifier import TextClassifier<br><br><br>data_dir = os.getenv(&#39;data_dir&#39;, &#39;data&#39;)<br><br>def train(project_id):<br> cfg = FileConfig()<br> <br> model_id = str(int(time.time()))<br><br> proj = Project(data_dir=data_dir, project_id=project_id, model_id=model_id)<br> cls = TextClassifier(proj=proj, cfg=cfg)<br><br> with open(&#39;model.json&#39;, &#39;r&#39;) as f:<br>  data = json.loads(f.read())<br><br> cls.fit(data)<br><br> return model_id<br><br>def test(project_id, model_id, query):<br> cfg = FileConfig()<br><br> proj = Project(data_dir=data_dir, project_id=project_id, model_id=model_id)<br> cls = TextClassifier(proj=proj, cfg=cfg)<br><br> return cls.transform(query)<br><br><br>if __name__ == &#39;__main__&#39;: <br> project_id = &#39;demo&#39;<br><br> parser = argparse.ArgumentParser(description=&quot;Smartloop NLP&quot;)<br> parser.add_argument(<br>  &quot;command&quot;,<br>  metavar=&quot;&lt;command&gt;&quot;,<br>  choices=[&quot;train&quot;, &quot;test&quot;],<br>  help=&quot;argument: train|parse&quot;,<br> )<br> parser.add_argument(&quot;-i&quot;, &quot;--id&quot;, dest=&quot;id&quot;)<br><br> args = parser.parse_args()<br><br> if args.command == &#39;train&#39;:<br>  print(train(project_id))<br> elif args.command == &#39;test&#39;:<br>  id = &#39;1&#39;<br><br>  if args.id is not None:<br>   id = args.id<br><br>  resp = test(project_id, id, [&#39;hello world&#39;])<br>  print(resp[&#39;topIntent&#39;])</pre><p>I have used a slightly different expression from what I used to train the chatbot to validate that it is processed by the NLP correctly</p><p>Finally, if you need help in trying to set up the NLP for your project or you want to contribute, drop me a line at <strong>mehfuz at smartloop.ai</strong></p><p>Happy Hacking!</p><h3>References</h3><ul><li><a href="https://pypi.org/project/smartloop/">https://pypi.org/project/smartloop/</a></li><li><a href="https://docs.conda.io/en/latest/miniconda.html">https://docs.conda.io/en/latest/miniconda.html</a></li><li><a href="https://github.com/SmartloopAI/sl-core">https://github.com/SmartloopAI/sl-core</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=42b0b92f56f6" width="1" height="1" alt=""><hr><p><a href="https://medium.com/smartloopai/natural-language-understanding-with-smartloop-nlp-42b0b92f56f6">Natural Language Understanding with Smartloop NLP</a> was originally published in <a href="https://medium.com/smartloopai">Smartloop</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Creating and configuring a GPU-enabled cluster in Microsoft Azure]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/smartloopai/creating-and-configuring-a-gpu-enabled-cluster-in-microsoft-azure-5441141c6ac9?source=rss-9a56267236af------2"><img src="https://cdn-images-1.medium.com/max/1788/1*h0DXa-hSlRL-LGEqq9jpwg.png" width="1788"></a></p><p class="medium-feed-snippet">I have recently created a GPU-enabled cluster (Nvidia Tesla T4) to run a GPU-enabled workload on a k8s instance.</p><p class="medium-feed-link"><a href="https://medium.com/smartloopai/creating-and-configuring-a-gpu-enabled-cluster-in-microsoft-azure-5441141c6ac9?source=rss-9a56267236af------2">Continue reading on Smartloop »</a></p></div>]]></description>
            <link>https://medium.com/smartloopai/creating-and-configuring-a-gpu-enabled-cluster-in-microsoft-azure-5441141c6ac9?source=rss-9a56267236af------2</link>
            <guid isPermaLink="false">https://medium.com/p/5441141c6ac9</guid>
            <category><![CDATA[terraform]]></category>
            <category><![CDATA[k8s]]></category>
            <category><![CDATA[nvidia]]></category>
            <category><![CDATA[cluster]]></category>
            <category><![CDATA[azure]]></category>
            <dc:creator><![CDATA[Mehfuz Hossain]]></dc:creator>
            <pubDate>Fri, 21 Jul 2023 23:54:49 GMT</pubDate>
            <atom:updated>2023-07-22T00:22:04.122Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Create your own Viber Chatbot in Minutes]]></title>
            <link>https://medium.com/recime/create-your-own-viber-chatbot-in-minutes-with-zero-coding-1a622accedcc?source=rss-9a56267236af------2</link>
            <guid isPermaLink="false">https://medium.com/p/1a622accedcc</guid>
            <category><![CDATA[bots]]></category>
            <category><![CDATA[chatbots]]></category>
            <category><![CDATA[viber]]></category>
            <category><![CDATA[development]]></category>
            <category><![CDATA[recime]]></category>
            <dc:creator><![CDATA[Mehfuz Hossain]]></dc:creator>
            <pubDate>Tue, 07 Nov 2017 19:50:10 GMT</pubDate>
            <atom:updated>2021-02-28T18:37:07.230Z</atom:updated>
            <content:encoded><![CDATA[<blockquote>There there is a new version of the blog post available, please read it here: <a href="https://blog.smartloop.ai/createviber-chatbot-in-minutes-b3d6ae175484">https://blog.smartloop.ai/createviber-chatbot-in-minutes-b3d6ae175484</a></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tUZKCr6wD6Pw9FF-6DJcFQ.png" /><figcaption>Builder Dashboard</figcaption></figure><p>Viber was founded in Israel in 2010 and today it has over 900 million users. According to<a href="https://www.lifewire.com/best-mobile-messaging-apps-2654839"> sources</a> it is ranked as #3 after Facebook and WhatsApp with a large user base, mostly in Europe and Asia.</p><p>In this post, I am going to show how to quickly create a Viber chatbot with dynamic data. I am creating a very simple weather bot that will pull data from <a href="https://openweathermap.org/">openweather.org</a> for a given city with the forecast.</p><p>First, create your free <a href="https://smartloop.ai/?utm_source=blog&amp;utm_campaign=viberht">smartloop.ai</a> account. Then create a Weather Bot by clicking the <strong>+Create New Bot</strong> button.</p><p>Once you hit “Submit”. Smartloop will do the following:</p><ol><li>Create a container specific to the subscription to run the bot.</li><li>Create the basic blocks like (start, about, default) to give you an idea of the flow.</li><li>Configure the underlying AI.</li></ol><p>Now click on your weather bot to go into the build tab, you should see the following once everything is ready:</p><p>Next, I am going to use the built-in JSON API plugin. This is one of the core plugins that comes out of the box with our new <a href="https://blog.recime.io/announcing-recime-v2-code-free-chatbot-creation-w-new-customization-options-4ba892fe532a">chatbot builder</a>. There is also a “Script Block” option to bring in your own custom logic to the conversation flow but that is a topic for another post, therefore, stay tuned! The JSON API plugin quickly lets you embed data from an external source. Think of it as <a href="https://www.getpostman.com/">POSTMAN</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fNLuqvUsEi0h1eedAa1Yuw.png" /></figure><p>There are a few different things that you can do with the plugin, including setting query string parameters, customer headers, and POST body. You can pass results to the conversation flow by setting a user variable as shown above.</p><p>Before, I dive more into details, first create the intent for the weather bot. Let’s call it “location” and configure it for a common weather question. Add “How is the weather in Paris?” in the Define Expression tab.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*MUTSuz8H7319xaIUFFa1Hg.png" /></figure><p>Highlight **Paris** as an entity and mark it as “location” this will give AI the pattern that will be used to resolve similar expressions. This part is commonly referred to as training.</p><p>Next step is to make a request to the weather API and pull the current forecast based on geo-location. Here I have to do the following:</p><ol><li>Use lat / lng from the resolved location entity. It is available as a variable for use from the conversation context.</li><li>Make a request to the weather API with lat/lng from #1 to get the weather forecast and set it as a user variable to use in the next step.</li></ol><p>The first step is to set the entity lat/lng as a query parameter.</p><p>Here, I have used {{nlp.entities.location}} to get the location object from resolved entities.</p><p>Check out the resources link below to find out more about context variables.</p><blockquote>Tip: You can easily debug the context variables from within “Cloud Code” and you can access them throughout your conversation flow as a template variable by using *double braces* syntax.</blockquote><p>Next, make a GET request to <a href="https://openweathermap.org/"><em>openweathermap.org</em></a> API to retrieve the current forecast:</p><p><a href="https://api.openweathermap.org/data/2.5/weather">https://api.openweathermap.org/data/2.5/weather</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*mVIVkZ5Ol-FUBEKLTktCaA.png" /></figure><p>Copy and paste the following Query string JSON to your JSON API plugin as shown above:</p><pre>{<br>    &quot;lat&quot;: &quot;{{nlp.entities.location.0.lat}}&quot;,<br>    &quot;lon&quot;: &quot;{{nlp.entities.location.0.lng}}&quot;,<br>    &quot;appid&quot;: &quot;APP_ID&quot;,<br>    &quot;units&quot;: &quot;imperial&quot;<br>}</pre><p>Complete the step by replacing the APP_ID with yours from <a href="https://openweathermap.org/">openweathermap.org</a> and defining the {{result}} variable.</p><p>Configure the conversation UI by adding the following elements:</p><p>1. Add Text response for current weather:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*nx60GytzC6oCnXGQ_iIuHg.png" /></figure><pre>Currently in {{nlp.entities.location.0.formatted}}  it&#39;s {{result.main.temp}}°F.</pre><p>3. Add Text response for the forecast:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*PErBrcJUFjl3-AGAMQXBHQ.png" /></figure><pre>Today&#39;s forecast: {{result.weather.0.description}} and with high of {{result.main.temp_max}}°F  and a low of {{result.main.temp_min}}°F.</pre><p>Lastly, test your bot making sure everything is working correctly:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vQ2Yxk3_qr2ckykVia1-ug.png" /></figure><p>Click on the publish icon from the left navigation bar and connect your bot. Follow the instructions in “Create a Bot” wizard to create a new Viber chatbot. Please refer to the <a href="https://developers.viber.com/docs/general/get-started/#get-started-bots">viber developer documentation</a> on how to create a bot account.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QywzdqvWpeuu00EUDuMkmQ.png" /></figure><p>Go to your Viber desktop or mobile app and start chatting with the bot.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/400/1*__OFq0JBaqzzg965Dczhrg.jpeg" /><figcaption>Weather Bot Demo</figcaption></figure><p>We would love to hear what bots you are creating in Viber. If you have feedback and want to share it with us then drop us a line at <a href="/service/mailto:hello@smartloop.ai">hello@smartloop.ai</a>. We would be glad to assist you.</p><p>Finally, if you liked reading the post, please share it and we always love to hear from you for feedback in improving the product and adding features that make you more productive. Also, don’t forget to check out the resources below and reach to <a href="/service/mailto:support@smartloop.ai">support@smartloop.ai</a> for more information.</p><h3>Resources</h3><p><a href="https://blog.recime.io/announcing-recime-v2-code-free-chatbot-creation-w-new-customization-options-4ba892fe532a">Announcing Recime v2 ✌ Code-Free Chatbot Creation w/ New Customization Options</a></p><p>Hope this helps!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1a622accedcc" width="1" height="1" alt=""><hr><p><a href="https://medium.com/recime/create-your-own-viber-chatbot-in-minutes-with-zero-coding-1a622accedcc">Create your own Viber Chatbot in Minutes</a> was originally published in <a href="https://medium.com/recime">Smartloop</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI and Chatbot to Improve your Customer Experience]]></title>
            <link>https://medium.com/recime/ai-and-chatbot-to-improve-your-customer-experience-3bfeb1bdf22c?source=rss-9a56267236af------2</link>
            <guid isPermaLink="false">https://medium.com/p/3bfeb1bdf22c</guid>
            <category><![CDATA[bots]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[customer-experience]]></category>
            <category><![CDATA[chatbots]]></category>
            <category><![CDATA[facebook-messenger]]></category>
            <dc:creator><![CDATA[Mehfuz Hossain]]></dc:creator>
            <pubDate>Thu, 17 Aug 2017 18:48:45 GMT</pubDate>
            <atom:updated>2017-08-17T21:24:29.680Z</atom:updated>
            <content:encoded><![CDATA[<p>Chatbots are quickly becoming the new way of brand engagement and connecting your users in a more personal way is now even easier. Chatbots are not only good for marketing of products and services but chatbots fueled by AI can greatly improve customer experience.</p><p>AI can increase overall customer satisfaction by quickly answering frequently asked questions, directing a user to a support agent, and make suggestions by analyzing a user’s intent, confidence, and sentiments.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*UTZWjK8PbtgMo3KUj8L7bQ.png" /></figure><p>Live chat made it simple to connect a brand to their customers, but there had been times when you might have been on hold for minutes to get a simple question answered. AI in customer support, not only makes it instant but reduces friction thus saving time and cost on both sides. Especially, when a support expert has to answer the same question over and over again throughout the day and still has to maintain a high level of service and customer satisfaction.</p><p>Moreover, you might need to wait for the next business day and adjust your schedule to match a fixed support window before you know that you can update an existing order.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BPQSah-1j0FzwXMF8VYGfg.png" /></figure><p>Marketing automation tools like intercom.io make the support better by integrating live chat with a notification flow. But it is still driven by humans and you are not guaranteed to get a response instantly. Overall the factor of error or miscommunication is just there by nature.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Gz3_9FfCMg7Wtj5E3sSGUQ.png" /></figure><p>AI is not going to replace support but it can help you take some things off your plate, giving you more time to focus on rather critical queries and can drastically reduce cost and errors in live support. For a business, it can make your staff happier as they have more time to solve knowledgeable problems.</p><p>Fast forward, AI chatbots are available 24/7. It helps you bring some automation to your existing flow and correctly implemented AI can save you countless hours and cost.</p><p>As a bot platform our goal is to bring automation to your daily lives and today we are excited to launch the <strong>HeadChat-Beta</strong>. HeadChat is a chatbot for your website. It allows you to have the same functionality of a Messenger chatbot for your customers via a browser interface. Similar to Intercom.io or any other live chat solution.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CY0aptmuNFhntFBBS13jrg.png" /></figure><p>We have created a fun chatbot to learn <a href="https://recime.io/about/?utm_source=mh-headchat">about us</a>. Please check it out. This will give you the flavor of what it can do.</p><p>Please <a href="https://docs.recime.io/headchat.html?utm_source=mh-headchat">check out the documentation</a> on how to copy a Facebook bot to your website site in just a few steps.</p><p>As always we love your feedback and please do share your love by clicking that ❤ (now clap) icon at the bottom.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3bfeb1bdf22c" width="1" height="1" alt=""><hr><p><a href="https://medium.com/recime/ai-and-chatbot-to-improve-your-customer-experience-3bfeb1bdf22c">AI and Chatbot to Improve your Customer Experience</a> was originally published in <a href="https://medium.com/recime">Smartloop</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Step-by-Step Instructions for Building a Facebook Messenger Bot without Coding]]></title>
            <link>https://medium.com/recime/facebook-messenger-bot-step-by-step-instructions-for-building-a-basic-facebook-messenger-chatbot-26296b6c2409?source=rss-9a56267236af------2</link>
            <guid isPermaLink="false">https://medium.com/p/26296b6c2409</guid>
            <category><![CDATA[chatbots]]></category>
            <category><![CDATA[facebook-messenger]]></category>
            <category><![CDATA[messenger-bots]]></category>
            <category><![CDATA[facebook]]></category>
            <category><![CDATA[bots]]></category>
            <dc:creator><![CDATA[Mehfuz Hossain]]></dc:creator>
            <pubDate>Mon, 31 Jul 2017 18:18:35 GMT</pubDate>
            <atom:updated>2019-06-10T21:01:26.354Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5hWdOwTXWABDvYvm75V8gQ.png" /></figure><p>Facebook Messenger is quickly becoming the de-facto home of chatbots. More than 300K bots already published on Facebook. Businesses are quickly moving to this platform to increase engagement and benefit from the social aspects of it making it topmost destination for bot developers.</p><p>In this post, I’ll show you how to build your own <strong>Facebook Messenger Bot. </strong>I’m going to use <a href="https://smartloop.ai/?utm_source=mh-h2fbb">Smartloop</a> for this to automate some of the things that I would otherwise have to do manually making my development flow straight forward.</p><h4>Step #1 : Create the Bot</h4><p>Provided you have created a free account with <a href="https://smartloop.ai/?utm_source=mh-h2fbb">smartloop</a>. Create a Facebook bot by clicking on the “ + ” thumbnail.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yD4JupiipwKj9nBGeLuPng.png" /></figure><p>This will bring up the wizard to create the bot.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0iw5XVqPQo0ZRwncZwGQYw.png" /><figcaption>Create a new Bot dialog</figcaption></figure><p>Enter the bot title, select a platform. In this case, “facebook” and choose a language which is “English” for this post.</p><p>Press “Submit”</p><p>This will create the bot and take you to the builder once completed.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kXFNik9wYpke0eh0_cGU0w.png" /><figcaption><a href="https://smartloop.ai/?utm_source=mh-h2fbb">Smartloop Chatbot Builder</a></figcaption></figure><p>This will create the basic building blocks for you to give you an head-start. Next, modify the start block to something meaningful to your project this will be the first message a user will see when they first press “Get Started” in your bot.</p><h4>Step #2: Create a Facebook Page</h4><p>If you do not have a Facebook page then you need to <a href="https://www.facebook.com/pages/create/">Create a Facebook Page</a>. This Facebook page is the identity of your bot including the icon and cover that will be shown as someone starts chatting with your bot.</p><p>If you are just creating the page for testing the bot, then you can skip most of the steps and you can always update it later as you make your bot public.</p><h4>Step #3: Create a Facebook App</h4><p>Because, I am using a tool this actually takes care of the app creation, token generation and manually configuring your webhook events.</p><p>You will just need to go to the publish page and connect to the page that you have just created by clicking the connect button (Step #8: Connect your Bot to the Facebook Page).</p><h4>Step#4: Build the Bot Flow</h4><p>In this step, we will create a <strong>Chuck Norris</strong> bot. First, modify the greetings message with a confirmation following a <a href="https://docs.recime.io/building-a-bot.html#user-input">user input</a> (either text input or quick reply) as shown below:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ynXyBUqcFjaq0WvbXG81og.png" /></figure><p>Based on the input ether take the user to the main or alternate flow. To do so first capture the user input:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wKjMY4RRmly1LAk6LPtpKw.png" /><figcaption><a href="https://docs.recime.io/building-a-bot.html#user-input">https://docs.recime.io/builditng-a-bot.html#user-input</a></figcaption></figure><p>Next, send the user to an appropriate block of your bot based on the input:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*lc2z1GRpZifo7KT-fDuYcw.png" /><figcaption><a href="https://docs.recime.io/building-a-bot.html#go-to-block">https://docs.recime.io/building-a-bot.html#go-to-block</a></figcaption></figure><p>Here in this case, create a block named “jokes” and send the user to it based on the input value.</p><p>Inside the block, use JSON API to pull a random joke and store it in a local variable in the following way (The concept of JSON API is pretty much standard among platforms):</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fDoNcv2E--AwQxOg4EFbiQ.png" /><figcaption><a href="https://docs.recime.io/developer-features.html#json-api">https://docs.recime.io/developer-features.html#json-api</a></figcaption></figure><p>Send the result value using a text element in the following way:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*csZTFG_ZHyLewrsVIOXdpw.png" /></figure><p>For alternate flow, give out a specific instruction on how to trigger the “jokes” block.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*NpfLbAGjuwiD0NegFiAzAw.png" /></figure><h4>Step #8: Connect your Bot to the Facebook Page</h4><p>Go to the publish tab by clicking the configure icon from the navigation page and connect to your page by clicking on the “+ Connect” button.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*TIWx34QIk7u30--gURCOUQ.png" /><figcaption><a href="https://docs.recime.io/publish.html#facebook">https://docs.recime.io/publish.html#facebook</a></figcaption></figure><h4>Step #9: Test your Bot</h4><p>Now at this point, your bot is successfully configured and connected to your page. Go to your page, click on “+ Add a Button” and choose <strong>Contact us/ Send a Message. </strong>In the next step, as where to send the users, choose “Facebook Messenger” to respond.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0rBOPUfeF5dU477IYCQm5w.png" /></figure><p>Once the button has been added, hover over it and select “Test Button”. this will bring up the bot and click on “Get started” to chat with the bot.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yAKfWWPPx-IfevMZ41AJeA.png" /></figure><p>If you have completed the last step then you are have successfully created and published your first Facebook bot. If this article was helpful, please don’t forget to recommend in order to help others find it and learn to create their own Facebook bot.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=26296b6c2409" width="1" height="1" alt=""><hr><p><a href="https://medium.com/recime/facebook-messenger-bot-step-by-step-instructions-for-building-a-basic-facebook-messenger-chatbot-26296b6c2409">Step-by-Step Instructions for Building a Facebook Messenger Bot without Coding</a> was originally published in <a href="https://medium.com/recime">Smartloop</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>