<?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 Riddhi Nisar on Medium]]></title>
        <description><![CDATA[Stories by Riddhi Nisar on Medium]]></description>
        <link>https://medium.com/@riddhinn?source=rss-f8ce5745a5e9------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*9-ZwHyyjqT1BrDgIeyin5g.jpeg</url>
            <title>Stories by Riddhi Nisar on Medium</title>
            <link>https://medium.com/@riddhinn?source=rss-f8ce5745a5e9------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 18:12:27 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@riddhinn/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[Create a “text-differentiator” using longest-common-subsequence.]]></title>
            <link>https://riddhinn.medium.com/create-a-text-differentiator-using-longest-common-subsequence-afc431b105dd?source=rss-f8ce5745a5e9------2</link>
            <guid isPermaLink="false">https://medium.com/p/afc431b105dd</guid>
            <category><![CDATA[javascript]]></category>
            <category><![CDATA[algorithms]]></category>
            <category><![CDATA[subsequence]]></category>
            <category><![CDATA[text]]></category>
            <category><![CDATA[dynamic-programming]]></category>
            <dc:creator><![CDATA[Riddhi Nisar]]></dc:creator>
            <pubDate>Fri, 04 Dec 2020 15:28:04 GMT</pubDate>
            <atom:updated>2020-12-04T15:28:04.533Z</atom:updated>
            <content:encoded><![CDATA[<h4>ALGORITHMS IN PRACTICE</h4><h4>.. the problem, use, solution and implementation!</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7sgVz_nPkCDZdEQpyJVOKw.png" /><figcaption><a href="https://riddhinn99.github.io/text-diff/">https://riddhinn99.github.io/text-diff/</a></figcaption></figure><h3><strong><em>The Problem:</em></strong></h3><p>Let us first look at the basic problem: Let X and Y be two arbitrary strings of length u and v respectively. <strong><em>What is the longest common subsequence (LCS) between X and Y?</em></strong></p><p>A subsequence is a subset of the original string where all the letters present in the subsequence appear in the same order in which they were present in the original string. Example:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/560/1*FolVCqL-7RpW1oXmQeuDRg.png" /><figcaption>(if the lines cross, it isn’t a subsequence since the order is changed)</figcaption></figure><p>Given two strings, we are supposed to find the longest subsequence which is common between both these strings.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/604/1*f_chuGv2S245DDv65IUKDQ.png" /><figcaption>The letters circled are the final answer</figcaption></figure><h3>The Use:</h3><p>Before jumping to the solution, let us think where and why would you require such a longest-common-subsequence finder? The first and one of the very useful cases is “gene-similarity-finder”. We can compare 2 strands of DNA by finding the longest common subsequence of bases [<strong>adenine</strong> (A), <strong>cytosine</strong> (C), <strong>guanine</strong> (G), or <strong>thymine</strong> (T)] between them, which will tell us how similar the 2 DNA strands are genetically.</p><p>Another important use-case is in finding the exact opposite of the LCS ie. finding what is the difference between 2 texts. So the best solution for this is to find out the LCS and compare it with the 2 texts to retrieve the difference. This is particularly useful in version control systems such as Git, which compare files and the difference between them between different commits.</p><p>In this tutorial/article, I’ll be creating a simple text-differentiator where you can paste 2 different texts and find how much of it is similar and how much different. We’ll use JavaScript to get the text from the inputs, find the difference, highlight them and show it as output. But first, let us see the solution:</p><h3>The Solution:</h3><p>Now lets see what a <em>recursive</em> solution to this problem might look like.</p><p>Consider stringsX and Y of length u and v respectively. Let LCS(Xᵤ , Yᵥ) be the longest common subsequence between X (length 0 to uᵗʰ character) and Y (length 0 to vᵗʰ character).</p><p>We must divide the problem LCS(Xᵤ , Yᵥ) into smaller subproblems. If the uᵗʰ character of X and vᵗʰ character of Y are equal, it means that we may consider this equal character in the answer of the final subsequence. So breaking it into a smaller problem: LCS between X and Y for length u and v respectively is one plus LCS of X and Y for length u-1 and v-1 (since the uᵗʰ character of X and vᵗʰ character of Y are equal).</p><blockquote>LCS(Xᵤ , Yᵥ) = 1 + LCS(Xᵤ₋₁ , Yᵥ₋₁) -&gt; (for equal characters)</blockquote><p>But what if these characters aren’t equal? Then we have 2 choices: either consider the uᵗʰ character of X and ignore the vᵗʰ character of Y, or vice versa. Then, we will have to find whichever among these 2 choices produces the maximum length subsequence. Thus the recursive equation reduces to:</p><blockquote>LCS(Xᵤ , Yᵥ) = max(LCS(Xᵤ , Yᵥ₋₁), LCS(Xᵤ₋₁ , Yᵥ)) -&gt; (for unequal characters)</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/833/1*emJSKTBYDSk1RfnmcKHOWw.png" /><figcaption>BASIC RECURSION TREE</figcaption></figure><p>Next, whenever you see a big recursive problem, always think of using Dynamic Programming(DP) if you can. We can use dynamic programming if a problem satisfies 2 conditions:</p><ol><li>Optimal sub-structure — a problem that can be divided into subproblems whose optimal solutions can make up the most optimal solution of the main problem too. Here, we see that LCS can be divided into smaller LCS subproblems and so LCS satisfies this condition.</li><li>Overlapping subproblems — some subproblem occurs more than once in the recursion tree ie. we are repeatedly calculating the answer for the same subproblem. LCS satisfies this too.</li></ol><p>Thus, we can use DP to store the values of recurring subproblems, so that we only compute them once.. hence converting the time complexity of LCS from exponential recursive O(2ⁿ) (look at the worst case recursion branches.. each problem gives rise to 2 subproblems.. so counting all the nodes in the tree, we get 2ⁿ subproblems to solve) to O(mn) for lengths of X and Y as m and n respectively.</p><h3>Algorithm/ Code:</h3><p>Let us use the bottom-up approach to solve LCS(one can also use the top-down approach, no harm in that, the time complexity remains the same.. but the recursive call stack increases, so I prefer the bottom-up approach here).</p><p>Using the bottom-up approach, we first create an empty matrix of rows= length(X)+1 and cols=length(Y)+1 . You might ask why +1? Thats because the first row and column are for <em>empty characters</em> and will be filled with 0s. The indexing of our letters in each word will begin from 1. Next, we loop through (i, j) for each element of the matrix and fill it according to the formula: (ignore the red boxes for now, just look at how each row is filled one by one using the formula)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/900/1*WkcHfnlraJw1uyXumMM3tQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/379/1*muMi5zhjHXs2wDWj1nDdHA.png" /></figure><p>Heres the JavaScript code snippet:</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/a4f95ceb942a35d80080ce74b542495b/href">https://medium.com/media/a4f95ceb942a35d80080ce74b542495b/href</a></iframe><p>Now, this function only gives us the length of LCS. We want to find the actual LCS string. For that, we need to backtrack our steps in the matrix as shown by the <em>red boxes</em> in the figure above. The basic formula is: if both the characters at the given index (i, j) are equal, it means we’ve traversed along the diagonal to reach this value. So we append the character to the LCS in the reverse order and move back to (i-1,j-1). If the characters are not equal, we backtrack to the element having maximum value between (i,j-1) and (i-1,j) and append nothing to the reverse output LCS. Heres the JS snippet:</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/9d9ad8a603980bbb970f8e0fbf527606/href">https://medium.com/media/9d9ad8a603980bbb970f8e0fbf527606/href</a></iframe><h3>Final Implementation:</h3><p>All this is well and good, but now we want to see this thing in action 😛</p><p>We use 2 basic input text-areas to let the user input the text. Then, we add an event listener on a submit button. Once the user clicks submit, findDifference(text1, text2) function is run.. which gets the text values from the input and finds the LCS between them.</p><p>Now, we need to show the same text in another &lt;div&gt; for both left and right and highlight the common letters of LCS with <strong>green</strong>, and others with <strong>red</strong>. For this, we use a function called highlightText(commonText, originalText).</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/5877aa57ac7ccd1f53c5452035f00cc3/href">https://medium.com/media/5877aa57ac7ccd1f53c5452035f00cc3/href</a></iframe><p>The final step is to append the text returned by highlightText() to the DOM.</p><p>And, voila! We have a created a simple working <strong><em>text differentiator</em></strong> using the <strong><em>longest-common-subsequence</em></strong> algorithm in practice. 😁</p><p>The entire source code including the HTML, CSS and JS can be found <a href="https://github.com/riddhinn99/text-diff"><strong><em>here</em></strong></a>! You can also try out the text-diff I made <a href="https://riddhinn99.github.io/text-diff/"><strong><em>here</em></strong></a> :)</p><p>Thanks for reading! If you found the article useful and/or fun, do give it a clap and <a href="https://medium.com/@riddhinn"><strong><em>follow</em></strong></a> me for more! 😋</p><p><em>And as always, you and I learn together. If you find something incorrect or if there is anything you’d like to add, leave a response and I’ll check it out. Cheers! ✨</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=afc431b105dd" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[(Visually) Interpreting the confusion-matrix:]]></title>
            <link>https://medium.com/analytics-vidhya/visually-interpreting-the-confusion-matrix-787a70b65678?source=rss-f8ce5745a5e9------2</link>
            <guid isPermaLink="false">https://medium.com/p/787a70b65678</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[classification-metrics]]></category>
            <category><![CDATA[visualization]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[confusion-matrix]]></category>
            <dc:creator><![CDATA[Riddhi Nisar]]></dc:creator>
            <pubDate>Sun, 01 Nov 2020 05:00:50 GMT</pubDate>
            <atom:updated>2020-11-04T18:12:16.076Z</atom:updated>
            <content:encoded><![CDATA[<h4>NORMALIZED vs UNNORMALIZED Confusion Matrices</h4><h4>.. how you, as a data scientist, can draw insights from a confusion matrix visually and more effectively.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/954/1*E-L9IpNzPKeY_paeLTPxhg.png" /><figcaption>Normalized vs unnormalized. Which is better for model interpretation?</figcaption></figure><p>But first, what is a confusion matrix? In machine learning, a confusion matrix is a kind-of confusing table used to understand how well our model predictions perform(<em>especially</em> confusing when we have multiple classes and not the classic binary 0/1 problems). However, gradually I figured out that the confusion-matrix is not so confusing and helps me a ton in understanding the model behaviour and interpreting the results. So I’m going to try to do the same here.. <strong><em>make it less confusing, more interesting and easier to interpret</em></strong>!</p><p>While writing this blog, I assume the reader has some prior knowledge &amp; aim to answer:</p><ol><li>How to interpret the confusion matrix using visual tools?</li><li>Why should you use a normalized confusion matrix for interpretation?</li></ol><p>First, let us look at a simple binary classification (0/1) confusion matrix:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/744/1*P75CJXJLFtFvJ7kKHfQrvQ.png" /></figure><p>The columns represent predictions made by our model and the rows represent the actual classes(this is the format of the very popular Python library for ML: sklearn. Some books have it opposite, but I’ll be sticking with this format). Thus cell(0,0) represents true negatives.. ie. the number of samples which were actually negative(belonging to class 0) and we predicted them as negative as well. Similarly, true positives are the number of samples our model correctly classified as positive and were actually positive.</p><p>False positives mean that we predicted these samples as positive(class 1) but they were actually negative(class 0) and vice-versa for false negatives.</p><p>Now, let us look at how the confusion matrix becomes slightly more complicated as the number of classes increase. Lets consider 3 classes: cats, dogs and fishes.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/705/1*b5FxmrSTdcjnre99EdoNBQ.png" /><figcaption>For multi-class classification of 100 samples</figcaption></figure><p>In the above figure, cell A represents the number of cats that were correctly classified by our model. We can see that only 10 out of total 30 cats were classified correctly. Cell E represents 30/30 correctly classified dogs (yaay 🐶 ). Cell H represents 5 fishes that were incorrectly classified as dogs. Thus we can see that our model performs <strong><em>best</em></strong> for dogs, <strong><em>okay-ish</em></strong> for fishes but really <strong><em>bad</em></strong> for cats. So this is how a multi-class confusion matrix looks like.</p><p>Now, lets come to <strong><em>visually interpreting</em></strong> the <strong><em>confusion matrix</em></strong>:</p><p>I have created a dummy confusion matrix to explain this concept. Here, we consider the prediction outputs for a multi-class classification model having 5 labels:label 0 to 4. The classes are imbalanced ie. number of true samples are not the same for each class. In my example, label 0 has 100 true samples and all other labels have 50 true samples each. So there are 300 samples in total.</p><pre>#Create dummy Confusion Matrix</pre><pre>import numpy as np<br>confusion_matrix = np.array([[79, 1,4,10,6],[0,48,1,1,0],[5,6,25,14,0],[0,0,0,50,0],[9,8,9,19,5]])</pre><pre>print(confusion_matrix)<strong><br>OUTPUT:</strong> <br>[[79 1 4 10 6]<br> [ 0 48 1 1 0]<br> [ 5 6 25 14 0]<br> [ 0 0 0 50 0]<br> [ 9 8 9 19 5]]</pre><p>As you can see, the above confusion-matrix array is really tough to interpret and draw conclusions from. Thus, we use visual tools to make it easier to interpret. I tend to use the seaborn library in Python for visualizations since it is quite quick and easy to use. Let us create a heatmap from the above matrix.</p><pre>#Create a Confusion Matrix heatmap from the above data<br>import seaborn as sns<br>sns.heatmap(cm, annot=True, linewidths = 0.01)</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/333/1*PHqSoroxzEYa4fJZ_Rpq4Q.png" /></figure><p>Glancing at the matrix, we come to know that:</p><ul><li>The diagonal elements represent the total correct values predicted per class. Eg: 79 values have been correctly predicted as belonging to class 0 out of 100 values. (But notice how we don’t come to know that the total class size for label 0 is 100)</li><li><strong><em>The lighter the color, the greater the number</em></strong>.. as we can see from the color-bar on the side. So naturally we would assume that — lighter colours on the diagonal elements and darker on all others mean our model is performing well and vice-versa. (However, this is not true, as we shall see from the normalized matrix.)</li><li>Hold your curiosity just a little more, let me give some more interpretations of this heat-map. The diagonal elements for class 1 and class 3 are nearly the same red. So we also assume that these two classes perform equally badly.</li><li>A lighter colour for diagonal element (0,0) implies our model performs really well for class 0.</li><li>And last but not the least, class 2 and class 4 perform the worst.</li></ul><p>Now, let us look at the <strong>normalized version</strong> of this same heat-map. To get the normalized version, we divide each row element by the sum of the entire row. Since each row represents the total number of actual values for each class label, the final normalized matrix will show us the percentage ie. out of all true labels for a particular class, what was the % prediction of each class made by our model for that specific true label.</p><pre>#Create normalized Confusion Matrix<br>cm_normalized = confusion_matrix.astype(&#39;float&#39;) / confusion_matrix.sum(axis=1)[:, np.newaxis]</pre><pre>sns.heatmap(cm_normalized, annot=True, linewidths = 0.01)</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/343/1*FMg9t7T-55sY1Sv3I1yeXQ.png" /></figure><p>Now, each cell (i,j) represents that P % of class i labels were predicted as class j. Hope that made sense. Interpreting this matrix:</p><ul><li>First major difference. Diagonal elements for class 1 and class 3, which we assumed were badly classified… actually went to a lighter color ie. higher number (or percentage if you will). Infact, class 3 is perfectly classified since its value is 1. Next comes class 1 with 0.96 correct predictions.</li><li>Our assumption that class 0 was the best classified class just went astray. It is now the third best class. Of all the true labels for class 0, our model could classify only 79% of them correctly.</li><li>Class 2, which was performing one of the worst according to our un-normalized confusion matrix, now shows us that it performs comparatively okay since our of all true labels for class 2, it classifies 50% correctly.</li><li>But why this difference.. between a normalized and unnormalized confusion matrix? <strong><em>This difference arises due to an imbalance in classes</em></strong>. Not all classes have the same number of total samples, and thus the unnormalized matrix outputs the absolute value, without considering how much proportion of the total class size is predicted correctly, which leads to improper conclusions due to the colorization of the heatmap.</li><li>So drawing the final and correct conclusions from this matrix, our model performs well for class 1 and class 3, moderately for class 0 and badly for class 2 and class 4. Thus, as a data scientist, you can now focus on why your model is performing badly for class 2 and 4.</li></ul><p>Hence, by using the correct visualization techniques for the confusion matrix, we were able to get correct insights about our model and this is the power of confusion matrix visualization. To conclude,<em> always use a normalized confusion matrix to interpret your results.</em></p><p>The above code can be found <a href="https://github.com/riddhinn99/medium-articles-code/blob/main/confusion_matrix_normalized_vs_unnormalized.ipynb"><strong><em>here</em></strong></a>! Fortunately for us, sklearn also provides an inbuilt function to create both these types of confusion matrix. You can find the code and usage in this<strong><em> </em></strong><a href="https://scikit-learn.org/stable/auto_examples/model_selection/plot_confusion_matrix.html"><em>sklearn documentation</em></a><em>.</em></p><p>Thanks for reading! If you found the article insightful and/or fun, do give it a clap and <a href="https://medium.com/@riddhinn"><strong><em>follow</em></strong></a> me for more! 😋</p><p><em>And as always, you and I learn together. If you find something incorrect or I missed something obvious, leave a response and I’ll check it out. Cheers! ✨</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=787a70b65678" width="1" height="1" alt=""><hr><p><a href="https://medium.com/analytics-vidhya/visually-interpreting-the-confusion-matrix-787a70b65678">(Visually) Interpreting the confusion-matrix:</a> was originally published in <a href="https://medium.com/analytics-vidhya">Analytics Vidhya</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Natural Language Processing — Basic Concepts]]></title>
            <link>https://medium.com/analytics-vidhya/natural-language-processing-basic-concepts-a3c7f50bf5d3?source=rss-f8ce5745a5e9------2</link>
            <guid isPermaLink="false">https://medium.com/p/a3c7f50bf5d3</guid>
            <category><![CDATA[basics-machine-learning]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[nlp]]></category>
            <category><![CDATA[basics]]></category>
            <dc:creator><![CDATA[Riddhi Nisar]]></dc:creator>
            <pubDate>Sun, 04 Oct 2020 06:03:26 GMT</pubDate>
            <atom:updated>2020-10-25T16:49:36.892Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Let us learn how to process (basically, make sense of) natural language(large texts of unstructured data) using an easy and effective toolkit in Python called the Natural Language ToolKit </em><strong><em>NLTK. </em></strong><em>(for the most part)</em></p><p>Lets say our fun use-case is processing the scripts of F.R.I.E.N.D.S… This is the starting script of one of my favorite episodes <strong><em>“The One With All The Poker”</em></strong></p><blockquote><strong>Ross</strong>: Uh, Rach, we’re running low on resumes over here.</blockquote><blockquote><strong>Monica</strong>: Do you really want a job with Popular Mechanics?</blockquote><blockquote><strong>Chandler</strong>: Well, if you’re gonna work for mechanics, those are the ones to work for.</blockquote><blockquote><strong>Rachel</strong>: Hey, look, you guys, I’m going for anything here, OK? I cannot be a waitress anymore, I mean it. I’m sick of the lousy tips, I’m sick of being called Excuse me.</blockquote><blockquote><strong>Ross</strong>: Rach, did you proofread these?</blockquote><blockquote><strong>Rachel</strong>: Uh… yeah, why?</blockquote><blockquote><strong>Ross</strong>: Uh, nothing, I’m sure they’ll be impressed with your excellent compuper skills.</blockquote><blockquote><strong>Rachel</strong>: (upset) Oh my Goood! Oh, do you think it’s on all of them?</blockquote><blockquote><strong>Joey</strong>: Oh no, I’m sure the Xerox machine caught a few.</blockquote><p>Nostalgic. Anyway, lets analyze this text, shall we? But first..</p><h4>Downloading/importing the NLTK Library</h4><p>First <a href="https://www.nltk.org/install.html">install NLTK</a> and to check if you have it installed, either import it or see if the command “pip show nltk” gives any output.</p><p>Next, you would want to download the most ‘popular’ list of data + pre-trained models under the NLTK library, so use the following code :</p><pre>import nltk<br>nltk.download(‘popular’) #run this once, no need to run it again</pre><h3><em>1. Tokenize</em></h3><p>Tokenize basically means that converting the STRING of text into LIST of words.</p><h4>CONVERTING TO SENTENCES =&gt; SENTENCE TOKENIZING:</h4><pre>sentences = nltk.sent_tokenize(script)<br>print(sentences)</pre><pre><strong>OUTPUT:</strong> [&quot;Ross: Uh, Rach, we&#39;re running low on resumes over here.&quot;, &#39;Monica: Do you really want a job with Popular Mechanics?&#39;, &quot;Chandler: Well, if you&#39;re gonna work for mechanics, those are the ones to work for.&quot; .. (and so on)</pre><h4>CONVERTING TO WORDS =&gt; WORD TOKENIZING:</h4><p>We can convert each sentence or the entire script into words, however we need to remove the delimiters like fullstops, apostrophes, commas etc. So we use Python’s regular expression library RE to do so.</p><pre>import re<br>def remove_delimiters(sentence):<br>    return re.sub(r&#39;[^\w\s]&#39;, &#39;&#39;, sentence)</pre><pre>trimmed_sentences = [remove_delimiters(sent) for sent in sentences]<br>print(trimmed_sentences)</pre><pre><strong>OUTPUT:</strong> [&#39;Ross Uh Rach were running low on resumes over here&#39;, &#39;Monica Do you really want a job with Popular Mechanics&#39;, &#39;Chandler Well if youre gonna work for mechanics those are the ones to work for&#39;, .. (and so on)</pre><p>Convert the trimmed sentences to words:</p><pre>all_words = [nltk.word_tokenize(sent) for sent in trimmed_sentences]<br>print(all_words)</pre><pre><strong>OUTPUT:</strong> [[&#39;Ross&#39;, &#39;Uh&#39;, &#39;Rach&#39;, &#39;were&#39;, &#39;running&#39;, &#39;low&#39;, &#39;on&#39;, &#39;resumes&#39;, &#39;over&#39;, &#39;here&#39;], [&#39;Monica&#39;, &#39;Do&#39;, &#39;you&#39;, &#39;really&#39;, &#39;want&#39;, &#39;a&#39;, &#39;job&#39;, &#39;with&#39;, &#39;Popular&#39;, &#39;Mechanics&#39;], [&#39;Chandler&#39;, &#39;Well&#39;, &#39;if&#39;, &#39;youre&#39;, &#39;gon&#39;, &#39;na&#39;, &#39;work&#39;, &#39;for&#39;, &#39;mechanics&#39;, &#39;those&#39;, &#39;are&#39;, &#39;the&#39;, &#39;ones&#39;, &#39;to&#39;, &#39;work&#39;, &#39;for&#39;], ... (and so on)</pre><pre>(NOTE: Directly tokenizing into words causes some problems, like <strong>gonna</strong> became <strong>gon na, </strong>so one can use regex according to ones needs. The base here is creating tokens out of text, whether using NLTK or RE)</pre><h4>[OR] CONVERT ENTIRE SCRIPT TO WORDS ALL AT ONCE:</h4><p>However, if you do not want to (convert to sentences, use regex, then word tokenizer)… there is a simpler one-liner regex approach to convert the entire speech into words:</p><pre>all_words_easy = re.findall(r’\w+’, script)<br>print(all_words_easy)</pre><pre><strong>OUTPUT: </strong>[&#39;Ross&#39;, &#39;Uh&#39;, &#39;Rach&#39;, &#39;we&#39;, &#39;re&#39;, &#39;running&#39;, &#39;low&#39;, &#39;on&#39;, &#39;resumes&#39;, &#39;over&#39;, &#39;here&#39;, &#39;Monica&#39;, &#39;Do&#39;, &#39;you&#39;, &#39;really&#39;, &#39;want&#39;, &#39;a&#39;, &#39;job&#39;, &#39;with&#39;, &#39;Popular&#39;, &#39;Mechanics&#39;, &#39;Chandler&#39;, &#39;Well&#39;, &#39;if&#39;, &#39;you&#39;, &#39;re&#39;, &#39;gonna&#39;, &#39;work&#39;, &#39;for&#39;, &#39;mechanics&#39;, &#39;those&#39;, &#39;are&#39;, &#39;the&#39;, &#39;ones&#39;, &#39;to&#39;, &#39;work&#39;, &#39;for&#39;, ... (and so on)</pre><h3>2. Stop-word Removal:</h3><p>Stop-words are words of less importance that occur frequently in any text, like the, and, a etc. Most of the times, they contribute to little or no meaning in our NLP tasks, so we usually remove them..</p><pre>from nltk.corpus import stopwords<br>stop_words=set(stopwords.words(&quot;english&quot;))<br>print(stop_words)</pre><pre><strong>OUTPUT: </strong>&#39;your&#39;, &#39;him&#39;, &quot;that&#39;ll&quot;, &#39;hers&#39;, &#39;an&#39;, &#39;same&#39;, &#39;did&#39;, &#39;yours&#39;, &#39;that&#39;, &#39;as&#39; .. (you get my point)</pre><pre>#Remove stopwords<br>words_without_stopwords = [word for word in all_words_easy if word not in stop_words]<br>print(words_without_stopwords)</pre><pre><strong>OUTPUT: </strong>[&#39;Ross&#39;, &#39;Uh&#39;, &#39;Rach&#39;, &#39;running&#39;, &#39;low&#39;, &#39;resumes&#39;, &#39;Monica&#39;, &#39;Do&#39;, &#39;really&#39;, &#39;want&#39;, &#39;job&#39;, &#39;Popular&#39;, &#39;Mechanics&#39;, &#39;Chandler&#39;, &#39;Well&#39;, &#39;gonna&#39;, &#39;work&#39;, &#39;mechanics&#39;, &#39;ones&#39;, &#39;work&#39;, &#39;Rachel&#39;, &#39;Hey&#39;, &#39;look&#39;, &#39;guys&#39;, &#39;I&#39;, &#39;going&#39;, &#39;anything&#39;, &#39;OK&#39;, &#39;I&#39;, &#39;cannot&#39;, &#39;waitress&#39;, &#39;anymore&#39;, &#39;I&#39;, &#39;mean&#39;, &#39;I&#39;, &#39;sick&#39;, &#39;lousy&#39;, &#39;tips&#39;, &#39;I&#39;, &#39;sick&#39;, &#39;called&#39;, &#39;Excuse&#39;, &#39;Ross&#39;, &#39;Rach&#39;, &#39;proofread&#39;, &#39;Rachel&#39;, &#39;Uh&#39;, &#39;yeah&#39;, &#39;Ross&#39;, &#39;Uh&#39;, &#39;nothing&#39;, &#39;I&#39;, &#39;sure&#39;, &#39;impressed&#39;, &#39;excellent&#39;, &#39;compuper&#39;, &#39;skills&#39;, &#39;Rachel&#39;, &#39;upset&#39;, &#39;Oh&#39;, &#39;Goood&#39;, &#39;Oh&#39;, &#39;think&#39;, &#39;Joey&#39;, &#39;Oh&#39;, &#39;I&#39;, &#39;sure&#39;, &#39;Xerox&#39;, &#39;machine&#39;, &#39;caught&#39;]</pre><h3>3. Word Tagging (part-of-speech tagging):</h3><p>POS Tagging basically means that associating a tag with each word of the sentence. For example:</p><pre>string = ‘You are awesome’<br>words = nltk.word_tokenize(string)<br>pos_tag = nltk.pos_tag(words)<br>print(pos_tag)</pre><pre><strong>OUTPUT: </strong>[(&#39;You&#39;, &#39;PRP&#39;), (&#39;are&#39;, &#39;VBP&#39;), (&#39;awesome&#39;, &#39;JJ&#39;)]</pre><pre>Here, You =&gt; personal pronoun<br>      are =&gt; verb<br>      awesome =&gt; simple adjective</pre><h3>4. Stemming:</h3><p>Stemming means converting the word to its word-root or cutting off its suffixes. Here we use the PorterStemmer class from NLTK, and as you can see, stemming converts words to their base or chops off the suffix(eg: cooking becomes cook)</p><pre>from nltk.stem import PorterStemmer<br>ps = PorterStemmer()<br>print(ps.stem(&#39;cooking&#39;))  -&gt; cook<br>print(ps.stem(&#39;serving&#39;))  -&gt; serv</pre><p>Drawback: Stemming sometimes reduces words to a non-meaningful base. Thus comes <strong>lemmatization</strong> to the rescue.</p><h3><strong>5. Lemmatization</strong>:</h3><p>It converts the word to its (meaningful) base word root or lemma. The only catch here is that we need to mention what kind of a word are we lemmatizing as the parameter (ie its POS Tag). Here both are verbs so we pass the parameter as &#39;v&#39;.</p><pre>from nltk.stem.wordnet import WordNetLemmatizer<br>lm = WordNetLemmatizer()<br>print(lm.lemmatize(&#39;cooking&#39;, &#39;v&#39;)) -&gt; cook<br>print(lm.lemmatize(&#39;serving&#39;, &#39;v&#39;)) -&gt; serve</pre><h3>6. Dependency Parsing</h3><p>POS Tagging will only convert each word to its respective part-of-speech. However, most of the times, we need the algorithm to understand the relationships between words. Dependency parsers do <em>exactly that</em>. Lets see how:</p><pre>import spacy #another open-source library for NLP<br>nlp = spacy.load(‘en_core_web_md’)<br>from spacy import displacy</pre><pre>temp = nlp(&#39;You are awesome&#39;)<br>displacy.serve(temp, style=’dep’)</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/490/1*oG1rxLctg9TAsaHu9xomTA.png" /><figcaption>POS Tagging + Dependency Parsing using spaCy</figcaption></figure><p>Here, <strong>you </strong>is the subject and <strong>awesome</strong> is the adjective complement. Both these are connected to each other by the word <strong>are.</strong></p><h3>7. Named Entity Recognition</h3><p>This is not a mandatory step in the NLP workflow, however, a good way to classify your words. NER is the mapping of each entity (basic subjects and objects in a sentence) to their <strong>common names.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/460/1*cixVZMlC6aarqhYWIPexmQ.png" /><figcaption>Drawn using <a href="https://excalidraw.com/">https://excalidraw.com/</a></figcaption></figure><pre>sentence = &quot;Chandler works at Google&quot;<br>print (nltk.ne_chunk(nltk.pos_tag(nltk.word_tokenize(sentence))))</pre><pre><strong>OUTPUT: </strong>(S (PERSON Chandler/NNP) works/VBZ at/IN (ORGANIZATION Google/NNP))</pre><p>These are the basic concepts about NLP that one needs to understand before diving deep into analyzing text and processing it for further tasks :)</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Ftenor.com%2Fembed%2F8251489&amp;display_name=Tenor&amp;url=https%3A%2F%2Ftenor.com%2Fview%2Fmonica-gif-8251489&amp;image=https%3A%2F%2Fmedia.tenor.com%2Fimages%2Fcae9d937e8d7998e8397a9965448e6d5%2Ftenor.gif&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=tenor" width="600" height="400" frameborder="0" scrolling="no"><a href="https://medium.com/media/4edd82222ea927c53525bf5d5dd706fd/href">https://medium.com/media/4edd82222ea927c53525bf5d5dd706fd/href</a></iframe><p>Oh hey, maybe later you can use these processed scripts and apply ML to analyze what kind of humour Chandler uses. Humour me ;)</p><p>The above code can be found <a href="https://nbviewer.jupyter.org/github/riddhinn99/medium-articles-code/blob/main/nlp_basics.ipynb"><strong><em>here</em></strong></a>! Thanks for reading! If you found the article insightful and/or fun, do give it a clap and <a href="https://medium.com/@riddhinn"><strong><em>follow</em></strong></a> me for more!</p><p><em>And as always, you and I learn together. If I missed something obvious, leave a response and I’ll check it out. Cheers! ✨</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a3c7f50bf5d3" width="1" height="1" alt=""><hr><p><a href="https://medium.com/analytics-vidhya/natural-language-processing-basic-concepts-a3c7f50bf5d3">Natural Language Processing — Basic Concepts</a> was originally published in <a href="https://medium.com/analytics-vidhya">Analytics Vidhya</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[ElasticSearch: the what?]]></title>
            <link>https://riddhinn.medium.com/elasticsearch-the-what-2342ffa9704d?source=rss-f8ce5745a5e9------2</link>
            <guid isPermaLink="false">https://medium.com/p/2342ffa9704d</guid>
            <category><![CDATA[search]]></category>
            <category><![CDATA[indexing]]></category>
            <category><![CDATA[introduction]]></category>
            <category><![CDATA[nosql]]></category>
            <category><![CDATA[elasticsearch]]></category>
            <dc:creator><![CDATA[Riddhi Nisar]]></dc:creator>
            <pubDate>Sun, 27 Sep 2020 04:15:34 GMT</pubDate>
            <atom:updated>2020-09-27T05:16:15.949Z</atom:updated>
            <content:encoded><![CDATA[<h4>ElasticSearch: WHAT, WHY AND HOW?</h4><p>.. a quick intro to WHAT is ElasticSearch all about?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dda94VlGLQNBFjjiKuotwQ.png" /></figure><p>ElasticSearch is a <strong><em>“distributed open-source analytics and full-text search engine.” </em></strong>Sounds like gibberish for the first time, isn’t it? Don’t worry, I’ll clear it out:</p><blockquote>Basically, ElasticSearch is like a modern-day database used for storage (of, lets say product information on Amazon, tweets on Twitter, etc)... on multiple nodes(hence, <strong>distributed</strong>)... which stores data in JSON-document format(hence, no need for a rigid schema). But thats not it…</blockquote><h3>The real-power of ElasticSearch:</h3><ul><li><strong>Full-text search engine ~ </strong>Here comes the part on how its different from other databases. The search is done on the full-text using inverted indexes(explained in the ending notes). So, if you search for &quot;Bridge of London&quot; , the results will also show documents with &quot;London Bridge&quot; Not something easily achievable in other SQL/noSQL databases. Awesome, isn’t it? (more on this in another post: ElasticSearch: the why? ..the next part)</li><li><strong>Accessible using simple REST API Calls ~</strong>The simplicity of using ElasticSearch lies in its ease of accessibility using the very familiar and popular REST API Calls of GET, PUT, POST and DELETE.</li><li><strong>Speed ~</strong>It is <strong>fast<em>-er</em></strong> than most modern-day noSQL databases in terms of its indexing and searching capabilities. SERIOUSLY.</li><li><strong>Distributed and scalable ~ </strong>the indexes are broken into multiple replicable shards and stored across multiple nodes, leading to high availability and scalability.</li><li><strong>Queries supported ~</strong> structured SQL-like queries, full-text queries (JSON-styled query language), complex queries for analysis, and high-performance geo-spatial and numerical queries too!</li><li><strong>Relevancy scoring ~</strong> <strong><em>Oh, the best part !!! </em></strong>ElasticSearch scores all the returned documents by relevancy. So lets say we wanted to search for &quot;car toys” on Amazon. All the documents containing the term car AND toys will have a higher score compared to documents that just contain toys. And so, Amazon can display the higher-scored documents first and later, down-below, the low-scored ones. Amazon use-case goal: accomplished!</li></ul><h4>ENDING NOTES:</h4><ul><li>Although this is diving a little deeper into ES, it is important to clear it out in the first go itself.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/728/1*hyLmObQCHYcKuMxvrkBesg.png" /><figcaption>Drawn with <a href="https://excalidraw.com/">https://excalidraw.com/</a></figcaption></figure><ul><li>For all the NLP folks out there, ES search engine internals use lowercasing -&gt; tokenization -&gt; stop-word removal -&gt; stemming etc, hence resulting it better search capabilities.</li><li>It supports a variety of languages and official clients are available for: Java, JavaScript (Node.js), Go, .NET (C#), PHP, Perl, Python and Ruby. (an abundance of languages)</li><li>ElasticSearch Customers: Netflix, Uber, Slack etc [if you need any more incentive to learn ElasticSearch 😉 ]</li><li>Even though ES is schema-free, you can define your own mappings as per your particular use-case to take full control of how fields are stored and indexed.</li><li>Now, <strong>inverted index:</strong> it lists every unique word that appears in any document and maps the word -&gt; documents the word appears in, thus allowing faster search.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/845/1*nHYdwDM_igWdQ6G_0VvKSQ.png" /><figcaption>Reference Blog: <a href="https://www.elastic.co/blog/found-elasticsearch-from-the-bottom-up">https://www.elastic.co/blog/found-elasticsearch-from-the-bottom-up</a></figcaption></figure><p><strong><em>So, at the end of the day, WHAT is ElasticSearch? </em></strong>Well, as the official tagline says:<strong><em> You know, for search! (and analysis) </em>😁</strong></p><p>PS: Be on the lookout for <strong>ElasticSearch: the why and the how </strong>for more! 📖</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fgiphy.com%2Fembed%2FXFuQ4InwtXBE4DDPHM%2Ftwitter%2Fiframe&amp;display_name=Giphy&amp;url=https%3A%2F%2Fgiphy.com%2Fgifs%2FFriends-season-9-episode-11-the-one-where-rachel-goes-back-to-work-XFuQ4InwtXBE4DDPHM&amp;image=https%3A%2F%2Fmedia2.giphy.com%2Fmedia%2FXFuQ4InwtXBE4DDPHM%2Fgiphy.gif&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=giphy" width="435" height="362" frameborder="0" scrolling="no"><a href="https://medium.com/media/a6b708250703b2a830f406807d34574a/href">https://medium.com/media/a6b708250703b2a830f406807d34574a/href</a></iframe><p>Thanks for reading! If you found the article insightful and/or fun, do give it a clap and <a href="https://medium.com/@riddhinn"><strong><em>follow</em></strong></a> me for more! Or get in touch with me on <a href="https://twitter.com/riddhinn99"><strong><em>twitter</em></strong></a> :)</p><p><em>And you and I learn together. If I missed something obvious, leave a response and I’ll check it out. Cheers! ✨</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2342ffa9704d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[My Journey to 329: A GRE Story]]></title>
            <link>https://riddhinn.medium.com/my-journey-to-329-a-gre-story-9cd0aa213453?source=rss-f8ce5745a5e9------2</link>
            <guid isPermaLink="false">https://medium.com/p/9cd0aa213453</guid>
            <category><![CDATA[study-abroad]]></category>
            <category><![CDATA[gre]]></category>
            <category><![CDATA[education]]></category>
            <category><![CDATA[tips]]></category>
            <dc:creator><![CDATA[Riddhi Nisar]]></dc:creator>
            <pubDate>Tue, 04 Feb 2020 10:13:43 GMT</pubDate>
            <atom:updated>2020-02-05T06:22:32.264Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*RlXVpUxTdUQ-NPuL" /><figcaption>Photo by <a href="https://unsplash.com/@emmamatthews?utm_source=medium&amp;utm_medium=referral">Emma Matthews Digital Content Production</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>Exactly one year ago on this day, I gave my GRE (<em>thinking I would clearly have to give another attempt</em>). I felt like I was so under-prepared after all. I was under the impression that it would turn out to be a <em>superrr</em> dreadful day. But it was far from dreadful. It was quite pleasant, to say the least :)</p><p>About 8 months earlier, I had joined a 5-week GRE crash course. My plan was to study over the entire semester and give the final exam during my vacations in December. The crash course was very useful as I was able to gain a breadth of knowledge regarding all the question types, topics, and the GRE pattern in general. However, I completely miscalculated the probability of my persistence (<em>as always</em>), and I barely practiced any problems/ GRE words during my regular semester.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fgiphy.com%2Fembed%2FhyMFaxhuQkZTq%2Ftwitter%2Fiframe&amp;display_name=Giphy&amp;url=https%3A%2F%2Fmedia.giphy.com%2Fmedia%2FhyMFaxhuQkZTq%2Fgiphy.gif&amp;image=https%3A%2F%2Fi.giphy.com%2Fmedia%2FhyMFaxhuQkZTq%2Fgiphy.gif&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=giphy" width="435" height="217" frameborder="0" scrolling="no"><a href="https://medium.com/media/27ccca3619c9788581329b6276c23763/href">https://medium.com/media/27ccca3619c9788581329b6276c23763/href</a></iframe><p>But hey, my excuses were reasonable: college projects, labs, exams, and what not!</p><p>Then came December, and as I opened my GRE wordlist I realized I had forgotten almost all the words that I learnt during the crash course. And on top of it, I was supposed to learn more… <em>whaaaaat?</em></p><p>Needless to say, I had to postpone my plan of writing GRE in December. I decided to dedicate one full month to proper GRE preparation and went on to practice problems, words and mock-tests for about a month and a half, give or take.</p><p>One day, on a whim, I decided to book my test date. I was unaware that slots generally get filled about 2 months in advance, especially if you are looking for weekend slots. Unfortunately, I was not able to get a weekend slot and thus booked the earliest available and convenient date: February 4th - <strong><em>a</em></strong> <strong><em>Monday.</em></strong> Let this be a cautionary tale to everyone: <em>slots get filled soon, so book your date as quickly as possible.</em></p><p><strong>Fast forward to D-Day:</strong> Like I mentioned before, I was extremely nervous, barely getting any sleep the previous night. But since I was already planning a re-attempt during summer, I had nothing to lose by panicking. I tried to be the calmest possible version of myself.</p><blockquote>*insert zen meme*</blockquote><p><strong>Err 404 :</strong> <em>Zen mode not found in current state of mind..</em></p><p>The GRE testing environment is really pleasant! You are not allowed to carry anything inside the test room other than your passport and locker keys (which they will provide). The test starts with admin access and the timer will go on ticking, waiting for nobody. The test goes on according to the pattern: 2 thirty-minute AWA Essays, 2 more sections, 10-minute break, 3 more sections (sections can be Verbal or Quant, and in any order). I utilized the 10-minute break by re-energizing myself with some food and water. However, if you are in the zone and do not want to take the break, it is completely up to you. If you do take the break, please take care of your break start time and go back inside well within time, since the test <em>will </em>start exactly after 10 minutes.</p><p>I was well-versed with the GRE pattern owing to the abundance of mocks that I attempted. During my final test, I tried to solve each section with calm and logic, especially the Quants sections. At the end of the test, I was getting a bit impatient but I didn’t let it get the better of me. I think I ended the test before time as I had solved everything and felt that this was the most I could do. Super relieved that the test was over and super anxious for the upcoming results, I hit the submit button, only for a pop-up to ask again if I was sure (<em>don’t make me rethink my decisions like this, GRE, please!</em>).</p><p>Never would I have imagined that this would be the best part of the day for me! The scores flashed on the screen:</p><p><strong>Verbal — 159</strong></p><p><strong>Quants — 170</strong></p><p>I could not believe my eyes. I kept staring at the screen for over a minute.</p><blockquote>Finally — excited, overwhelmed but happy, I digested the fact that I had indeed done well.</blockquote><h4>My preparation:</h4><p>I learnt all words from Barrons 333 high-frequency wordlist and nothing more. My crash course notes had about 1400 GRE words, which I (<em>of course</em>) did not learn, and which now seem unnecessary to me. I will not lie, learning merely 333 words might be too little (which is why I was so sure I would have to give the test again). But if you are just starting, or you do not have much time before D-Day, this is the list to begin with!</p><blockquote>For anyone who has a decent vocabulary and a good sense of reading complicated or long-winded sentences, such as this one haha, you really do not need more than 3 months of preparation at a stretch. I say this because today, one year later, I can barely remember any difficult GRE words that I had learnt for the final exam!</blockquote><p>One great tool to remember meanings of difficult words is the <a href="https://mnemonicdictionary.com/">mnemonic dictionary</a>. Just search for a difficult word on it, you’ll know what I am saying.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*9q2bgddnTPE9CEwW" /><figcaption>Photo by <a href="https://unsplash.com/@brett_jordan?utm_source=medium&amp;utm_medium=referral">Brett Jordan</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>For Reading Comprehensions, practice as much as you can and increase your reading speed, which can only be done by reading more and more, and realize that a lot of timed-practice will make you used to the passages and the pattern in general. <strong>DON’T :</strong> Pick one option, think it is right and mark it. <strong><em>Read all options</em></strong><em>. </em>In fact, use the elimination technique to eliminate wrong options till you end up at one correct answer.</p><p>Read as many good sample AWA Essays as you can get your hands on. Keep a timer and <strong><em>practice, practice, practice</em></strong> on a PC with a keyboard. Take pointers from essays graded as 5 or 6. Get your essays reviewed/graded by someone, if possible.</p><p>Try to make a basic formula sheet for all the topics under Quants. You should revise the formulae once everyday and practice as many problems as you can with a timer. They do not ask difficult questions, but they surely ask tricky questions and I experienced quite a few in my final test. I could figure out that if I had gone ahead and made this silly mistake, I would have gotten another option as the answer. Lucky for me, I caught on to the tricks soon enough. So keep your mind calm and logical, and beware of such traps.</p><blockquote><strong>GIVE AS MANY MOCKS AS YOU CAN.</strong></blockquote><p>I gave about 6–7 mocks in the proper GRE style. Since my final test was scheduled at 8am, every alternate day for the last 2 weeks before my test, I would wake up diligently at 8 and give the test, along with the 10-minute break and everything. After analyzing my mock results, I would come to know my weak areas and work on them. It also means writing AWA essays for 1 hour straight, making no excuses to skip them.</p><h4>Useful Resources:</h4><p>(these are some of the general resources I am mentioning here, pick whichever you are comfortable with. There is no “one perfect resource”)</p><ul><li>ETS Official Guides (best way to understand the pattern, topics, and AWA Essay gradings)</li><li>GRE WORDS: Barron’s high-frequency 333 list (a must), Manhattan Essential Words (500), later Barron’s 700, Magoosh Flash cards etc.</li><li>Tons of mobile applications for GRE such as Magoosh apps, Kaplan, Barron’s, VoLT, Quizlet etc, just one search away.</li><li>Manhattan 5lb for problems with a higher difficulty level.</li><li>2 free Powerprep tests by ETS once you book your GRE test date, one free Manhattan test and so on(for mocks)</li></ul><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fgiphy.com%2Fembed%2FlPF1CyJXXcTZmUrP2J%2Ftwitter%2Fiframe&amp;display_name=Giphy&amp;url=https%3A%2F%2Fgiphy.com%2Fgifs%2FFriends-season-1-episode-11-friends-tv-lPF1CyJXXcTZmUrP2J&amp;image=https%3A%2F%2Fmedia2.giphy.com%2Fmedia%2FlPF1CyJXXcTZmUrP2J%2Fgiphy.gif&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=giphy" width="435" height="362" frameborder="0" scrolling="no"><a href="https://medium.com/media/ef4c67b42c9999769a9709c715e9c562/href">https://medium.com/media/ef4c67b42c9999769a9709c715e9c562/href</a></iframe><p><strong><em>Important Note:</em></strong> You are allowed to send your GRE scores to 4 universities for free. This service is available right after your test on your PC in the testing room. Prepare a list beforehand because once you leave the test room, you cannot send the scores via your online portal.</p><p><strong>I hope you enjoyed reading this article as much as i enjoyed writing it! </strong>😋</p><p><strong><em>If you liked the article, show some love by clapping? Maybe more than once if you loved the gifs as well? :P</em></strong></p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fgiphy.com%2Fembed%2FlD76yTC5zxZPG%2Ftwitter%2Fiframe&amp;display_name=Giphy&amp;url=https%3A%2F%2Fgiphy.com%2Fgifs%2Fthe-end-thats-all-folks-lD76yTC5zxZPG&amp;image=https%3A%2F%2Fmedia3.giphy.com%2Fmedia%2FlD76yTC5zxZPG%2Fgiphy.gif&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=giphy" width="435" height="319" frameborder="0" scrolling="no"><a href="https://medium.com/media/de9e39cc8c838f794c45c94e94acda6e/href">https://medium.com/media/de9e39cc8c838f794c45c94e94acda6e/href</a></iframe><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9cd0aa213453" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>