<?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 Jayabalambika R on Medium]]></title>
        <description><![CDATA[Stories by Jayabalambika R on Medium]]></description>
        <link>https://medium.com/@jayabalambikar?source=rss-982aa3fadb43------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*iIPSiUfj8S3-QLNUCEVenw.jpeg</url>
            <title>Stories by Jayabalambika R on Medium</title>
            <link>https://medium.com/@jayabalambikar?source=rss-982aa3fadb43------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 22:18:00 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@jayabalambikar/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[Compress PDF files using GhostScript and Python]]></title>
            <link>https://jayabalambikar.medium.com/compress-pdf-files-using-ghostscript-and-python-34cad81bc02f?source=rss-982aa3fadb43------2</link>
            <guid isPermaLink="false">https://medium.com/p/34cad81bc02f</guid>
            <category><![CDATA[pdf-compressor]]></category>
            <category><![CDATA[python-programming]]></category>
            <category><![CDATA[ghostscript]]></category>
            <category><![CDATA[script]]></category>
            <dc:creator><![CDATA[Jayabalambika R]]></dc:creator>
            <pubDate>Fri, 13 Jun 2025 11:29:48 GMT</pubDate>
            <atom:updated>2025-06-13T11:29:48.226Z</atom:updated>
            <content:encoded><![CDATA[<p>Are you sceptical about using apps or tools to compress your PDF files?</p><p>The solution is simple. Use software tools and technologies to your benefit.</p><p>Write a simple Python script and a Ghost script.</p><p>Follow these steps</p><ol><li>Install GhostScript</li></ol><p>On Windows:</p><p>Download and install from <a href="https://ghostscript.com/releases/gsdnld.html">https://ghostscript.com/releases/gsdnld.html</a></p><p>Add the Ghost script to your PATH. Provide the complete path until the bin folder.</p><p>For example: It could be like this C:\Program Files\gs\gs10.01.2\bin</p><p>2. Check if GhostScript is accessible in the command prompt</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/d88ea7221f8009535412a83ab748628b/href">https://medium.com/media/d88ea7221f8009535412a83ab748628b/href</a></iframe><p>This will print the version if GhostScript is installed properly.If you get &#39;gswin64c&#39; is not recognized thePATH is not correctly set.</p><p>3. Now write the Python code to call the correct subprocess</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/6aa4a64aedae6f408469cba37789d0d5/href">https://medium.com/media/6aa4a64aedae6f408469cba37789d0d5/href</a></iframe><p>4. Run the above python code.</p><p>5. Navigate to the output folder in your code, and the compressed file is generated in that folder. The program prints the above in the command prompt</p><pre>Compressed file size: 1.64 MB<br>Compression successful and under target size.</pre><p>Happy Experimenting!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=34cad81bc02f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Keras Dream booth Event]]></title>
            <link>https://jayabalambikar.medium.com/keras-dream-booth-event-1956a3428594?source=rss-982aa3fadb43------2</link>
            <guid isPermaLink="false">https://medium.com/p/1956a3428594</guid>
            <category><![CDATA[text2image]]></category>
            <category><![CDATA[dreambooth]]></category>
            <category><![CDATA[keras]]></category>
            <category><![CDATA[stable-diffusion]]></category>
            <dc:creator><![CDATA[Jayabalambika R]]></dc:creator>
            <pubDate>Tue, 14 Mar 2023 16:35:15 GMT</pubDate>
            <atom:updated>2023-03-14T16:35:15.100Z</atom:updated>
            <content:encoded><![CDATA[<p>Community event</p><p>Dream booth is a fine-tuning technique used to teach new imaging concepts to Stable diffusion models using just 3–4 images of the concept. With this we can generate funny images or even realistic images of any concept under the Sun.</p><p>I managed to train the Dream booth model for mobile phone images and have published it on the Hugging Face hub.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HEKRDUrjzMvGpZq-yaEzDA.png" /><figcaption>Demo image-1</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Jq8vVAtKn-uKSHO8S827fA.png" /><figcaption>Demo image-2</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*405V2jk6daBnIYIRerRL_w.png" /><figcaption>Demo image-3</figcaption></figure><p>Please visit the Hugging Face Hub and vote for my model if you like it. <a href="https://huggingface.co/spaces/Jayabalambika/my-app-space">Vote here</a></p><p><strong>References</strong></p><ul><li><a href="https://github.com/huggingface/community-events/blob/main/keras-dreambooth-sprint/README.md">community-events/README.md at main · huggingface/community-events (github.com)</a></li><li><a href="https://keras.io/guides/keras_cv/generate_images_with_stable_diffusion/">High-performance image generation using Stable Diffusion in KerasCV</a></li><li><a href="https://keras.io/examples/generative/fine_tune_via_textual_inversion/">Teach StableDiffusion new concepts via Textual Inversion</a></li><li><a href="https://keras.io/examples/generative/finetune_stable_diffusion/">Fine-tuning Stable Diffusion</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1956a3428594" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Excerpts from the two-day webinar series on Introduction to Vedanta]]></title>
            <link>https://jayabalambikar.medium.com/excerpts-from-the-two-day-webinar-series-on-introduction-to-vedanta-5d10dbc31311?source=rss-982aa3fadb43------2</link>
            <guid isPermaLink="false">https://medium.com/p/5d10dbc31311</guid>
            <category><![CDATA[moral-stories]]></category>
            <dc:creator><![CDATA[Jayabalambika R]]></dc:creator>
            <pubDate>Mon, 26 Dec 2022 16:04:05 GMT</pubDate>
            <atom:updated>2022-12-26T16:04:05.118Z</atom:updated>
            <content:encoded><![CDATA[<p>Brief Case Study of Vedanta using story from Kathopanishad</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/512/1*NPe9z_v3PWKlgjqUfib3jg.png" /><figcaption>#stablediffusion: Generated image: Concept art of the Vedic Schools in India</figcaption></figure><p>This weekend was a nostalgic weekend experience for me. I travelled back in time to my childhood days, of attending the Moral Science class in school. This class was an important and fun class for all of me and my friends because, we used to be taught good habits, and discipline using stories from “Soldiers of God”, a small magazine containing stories on moral values and discipline in life.</p><p>Our yoga teacher referred us to attend a two-day webinar series on an introduction to Vedanta using the Kathopanishad conducted by the Advaita Academy. After attending the Webinar series, felt it would be just useful to pen down, my wee bit little understanding of this concept of Vedanta being taught in the webinar for my future reference.</p><p>The entire webinar was spread across two days with 2 hours in the morning and 2 hours in the evening over the weekend.</p><p>Broadly it was divided into the following parts</p><ol><li>Story of Nachiketa, a young boy who spoke to Lord Yama.</li><li>The Chatur Vida Purushartha and the Two Margas</li></ol><p>This is the summary of my learnings from the story:</p><p>Nachiketa was the son of King Vājashravas. The King had acquired and attained all possible materialistic things, and he desired a gift from God, and he also readily offered his five-year-old son Nachiketa to Lord Yama in return for the gift from God.</p><p>Nachiketa went to meet Yama, but Lord Yama was not there at his office. Nachiketa, patiently waited for Lord Yama to appear. The guards of Lord Yama offered Nachiketa some food to eat, as they could not see this young boy hungry. But Nachiketa refused the offer by the guards and said that down on Earth, his Father was his Master. But since his father sent him to Yama’s office, from then on, he considered Lord Yama to be his Master. Hence, he waited patiently fasting for three days. Finally, Lord Yama himself appeared before Nachiketa, pleased with his patience and discipline. He served the young boy good lunch and then told him that he would grant him 3 boons and he could ask him anything under the Sun.</p><p>The intelligent boy asked the following 3 boons to Lord Yama</p><ol><li>Peace of his Father and himself</li><li>Nachiketa noticed that his Father was donating old, blind, lame cows to his Subjects in the Kingdom. Instead, he asked the Lord for him to be offered to them on Earth so that he could be of use to them</li><li>Nachiketa wanted to learn what comes after death to the body. His question to the Lord was what was beyond Dharma and Adharma?</li></ol><p>To the last question, Yama tried to trick him into offering long life, progeny, and wealth on any planet of his choice. But the clever Nachiketa disregarded saying those are finite things with finite endings, and that would not teach me the real essence of what happens after Death. After a series of dialogues between Yama and Nachiketa, finally Lord Yama agreed to grant him the third boon and told him that only the Atman remains after a person&#39;s death.</p><p>In order to understand the Atman, the seeker needs to be able to clearly distinguish between Vidya and Avidya.</p><p>Lord Yama also explained the qualities of a student and Teacher. They are:</p><ol><li>Student: Viveka(discrimination), Vairagya(dispassion)</li><li>Teacher: Command over the subject being taught in the right way. If the teacher does not have a good command over the subject, the students will not be able to understand well which would result in lots of commotion.</li></ol><p>We were explained in detail the Chatur Vida Purushartha. They are Dharma (duties), Artha (expect returns), Kama(pleasure) and Moksha (break cycle of birth and rebirth). These Chatur Vida Purushartha&#39;s are further divided into Vidya and Avidya. Out of the four Purushartha&#39;s, the Dharma, Artha and Kama fall under the category of Avidya. Moksha falls under the category of Vidya.</p><p>There are two Margas (two paths)</p><ol><li>Shreyas — Vidya (Knowledge)</li><li>Preyas- Avidya (Ignorance)</li></ol><p>In order for us to be able to understand Atman we would need to follow the Shreyas Marga. The essence of the Advaita Vedanta is we can be liberated and free and attain Moksha with Jnana Yoga.</p><p>It is important for each one of us to learn this in small baby steps throughout our life, irrespective of our age or the stage of life we are in, under the guidance of an able Teacher. This will add more purpose to our life and conduct our life beautifully.</p><p>References:</p><ol><li>Our Teachers: Mrs. Meera Narayanan and Mr.Narayanan, <a href="http://eternal-yoga.in/">Home (eternal-yoga.in)</a></li><li>Indica Moksha: <a href="https://www.indicamoksha.com/">Home — Indica Moksha</a>, organizers of the webinar series.</li></ol><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5d10dbc31311" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Mars or Deepavali Fireworks]]></title>
            <link>https://jayabalambikar.medium.com/mars-or-deepavali-fireworks-e758c9bcc449?source=rss-982aa3fadb43------2</link>
            <guid isPermaLink="false">https://medium.com/p/e758c9bcc449</guid>
            <category><![CDATA[hobbies-and-interests]]></category>
            <category><![CDATA[astronomy]]></category>
            <category><![CDATA[python]]></category>
            <dc:creator><![CDATA[Jayabalambika R]]></dc:creator>
            <pubDate>Thu, 27 Oct 2022 13:49:33 GMT</pubDate>
            <atom:updated>2022-10-27T13:49:33.846Z</atom:updated>
            <content:encoded><![CDATA[<p>When is Mars closest to the Earth?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/1*mh3yOvSkv9fVfezUUlE0JQ.jpeg" /><figcaption>Mars zoomed view in the sky on 26th October 2022</figcaption></figure><p>On the third day of Deepavali., the October 26th, 2022, I stepped out to our apartment terrace to get good and clear glimpses of the fireworks in the evening.</p><p>I got lucky and found this bright object in the sky. First mistook it to be the pole star. But the distinguishing feature between pole star and visible planets to the Earth are their size and position. In case of stars, the pole stars will be of the dot shaped and brighter and in same position with reference to time. The planets will be spherical shape and brighter and shifts its position in the sky with reference to time.</p><p>We are assuming that the orbits of planets are elliptical laws of trajectory according to Kepler. The Earth is moving twice the speed on its orbit when compared to Mars. At Opposition, the Earth places Mars at the opposite side of the Sun. This distance is called the perihelic distance. It will rise at the East at sunset, reach highest point at midnight and set in the west at sunrise. Mars Close approaches to Earth occurs every 26th months. The Opposition occurs on December 8, 2022, and now is the time to watch the sky for this Red planet and see it slowly rise every day at sunset on the December 8th, 2022.</p><p>Stay Tuned for python program calculating the distance between Mars and Earth and their relative positions each day.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e758c9bcc449" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Thoughts]]></title>
            <link>https://jayabalambikar.medium.com/thoughts-e68a793e29a4?source=rss-982aa3fadb43------2</link>
            <guid isPermaLink="false">https://medium.com/p/e68a793e29a4</guid>
            <category><![CDATA[teacher-stories]]></category>
            <category><![CDATA[yoga-teacher]]></category>
            <dc:creator><![CDATA[Jayabalambika R]]></dc:creator>
            <pubDate>Wed, 19 Oct 2022 09:54:16 GMT</pubDate>
            <atom:updated>2022-10-19T09:54:16.827Z</atom:updated>
            <content:encoded><![CDATA[<p><strong><em>“When there is earnest desire in a student, then Prakriti automatically blesses them with a Guru to lead them in the right direction”- Echoing the voice of my Guru</em></strong></p><p>I have been blessed with beautiful yoga teacher in my life and really fall short of words to thank God for showering blessing on me with beautiful teacher whom I consider my best friend, philosopher and guide.</p><p>In a nutshell she is a super cool 😎😎person to be with. May their tribes increase!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/359/1*4OaZRNNRfrbPm0wKARxYng.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e68a793e29a4" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Design of a Recommender System]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://jayabalambikar.medium.com/recommender-systems-d08c38713ede?source=rss-982aa3fadb43------2"><img src="https://cdn-images-1.medium.com/max/2600/0*jH6XfpPpS7dAGsEq" width="4000"></a></p><p class="medium-feed-snippet">A proof-of-concept study of Feed Recommender systems</p><p class="medium-feed-link"><a href="https://jayabalambikar.medium.com/recommender-systems-d08c38713ede?source=rss-982aa3fadb43------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://jayabalambikar.medium.com/recommender-systems-d08c38713ede?source=rss-982aa3fadb43------2</link>
            <guid isPermaLink="false">https://medium.com/p/d08c38713ede</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[streamlit]]></category>
            <category><![CDATA[recommender-systems]]></category>
            <category><![CDATA[clustering]]></category>
            <dc:creator><![CDATA[Jayabalambika R]]></dc:creator>
            <pubDate>Mon, 19 Sep 2022 15:36:12 GMT</pubDate>
            <atom:updated>2023-01-04T16:40:24.828Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[The Unison of Body and Mind]]></title>
            <link>https://jayabalambikar.medium.com/the-unison-of-body-and-mind-f7c7dbf4ebea?source=rss-982aa3fadb43------2</link>
            <guid isPermaLink="false">https://medium.com/p/f7c7dbf4ebea</guid>
            <category><![CDATA[yoga-and-meditation]]></category>
            <category><![CDATA[mindfulness]]></category>
            <dc:creator><![CDATA[Jayabalambika R]]></dc:creator>
            <pubDate>Sat, 11 Dec 2021 09:20:21 GMT</pubDate>
            <atom:updated>2021-12-11T09:20:21.059Z</atom:updated>
            <content:encoded><![CDATA[<p><em>My personal experiences as a student in learning Yoga.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*obAQqsyHJDuZjCNd" /><figcaption>Photo by <a href="https://unsplash.com/@satishnagapuri?utm_source=medium&amp;utm_medium=referral">satish nagapuri</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>Listening to your peppy music and going out for a long walk is definitely a good workout, but always accompany it with a full-fledged body and the mind work out together, to get everlasting bliss.</p><p>The teacher plays a very important role in arousing interest in the student in a particular subject. I really love to attend yoga classes conducted by my teacher. Its been just about a few months that I have started learning Yoga from my teacher, and have really reaped enormous benefits with respect to my health and overall wellness and being. I have commenced writing in this series, on the benefits and would wish to add more experiences during this long Divine journey.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*p59rkhbVulwodZtb" /><figcaption>Photo by <a href="https://unsplash.com/@jess33mx?utm_source=medium&amp;utm_medium=referral">Jose Vazquez</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*4JgicS8M3YBPmAF-" /><figcaption>Photo by <a href="https://unsplash.com/@kikekiks?utm_source=medium&amp;utm_medium=referral">kike vega</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>To highlight a few here:</p><ul><li>Promotes and restores balance in the body.</li><li>Regulates the hormonal imbalances in the body.</li><li>The mind becomes very calm and promotes positive thinking.</li><li>Anxiety is reduced and feel really energetic throughout the day.</li><li>We are able to function normally irrespective of the happenings in the background.</li><li>Regulate our temper, easily control the temper.</li><li>Attention span increases and able to focus more effectively on our work or studies.</li><li>Begin the journey of self-realisation smoothly.</li></ul><p>The Yoga classes are more or less about an hour and a half on alternate days, but the journey during the class is just great and here I am falling short of words to explain them. The classes start with a brief invocation of blessings of the Supreme power, to make us more mindful in the practice. This also gives us enough enthusiasm on the alternate days to do a few warm-ups and asanas on our own, so that we integrate yoga into our daily life and live a better healthy life naturally.</p><p>The opening session is a warmup consisting of a few stationary walks, jumps, skips, twists all in coordination with the Breath[Prana Shakthi], and mindfulness. The really interesting part of the class is the Surya Namaskar, marking the mid-session. One will really enjoy the chants that mark the beginning of the Surya namaskar practice. The rhythm and the divine touch by my teacher for the chant make it really blissful. She chants the Beeja mantra for each step of the Surya Namaskar. Towards the latter part of the session are the few asanas, she carefully chooses a different asana, giving it a personal touch, a variety and flavour, so that we are fully in the practice. Each of the asanas and practices is explained to us very well. I am really thankful for having blessed me with a great teacher, who understands me, her students and helps all her students in a very big way. I pray to the Almighty to be able to continue receiving guidance and be able to learn from my teacher in this Divine Spiritual journey.</p><p>With the Almighty’s grace, I wish to share more on our positive blissful experiences as we progress on the Divine Spiritual Journey every day. Stay tuned!</p><h3>References</h3><ul><li>Meera Narayanan- An ardent passionate Yoga teacher and practitioner for 25 years.</li><li><a href="http://eternal-yoga.in/">http://eternal-yoga.in/</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f7c7dbf4ebea" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Deep Learning models for Grammatical Error Handling in Low Resource languages]]></title>
            <link>https://medium.com/data-science/deep-learning-models-for-grammatical-error-handling-in-low-resource-languages-2ac2c5d154f5?source=rss-982aa3fadb43------2</link>
            <guid isPermaLink="false">https://medium.com/p/2ac2c5d154f5</guid>
            <category><![CDATA[encoder-decoder]]></category>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[nlp]]></category>
            <category><![CDATA[tensorflow]]></category>
            <category><![CDATA[tensorflow-serving]]></category>
            <dc:creator><![CDATA[Jayabalambika R]]></dc:creator>
            <pubDate>Mon, 11 Oct 2021 14:49:06 GMT</pubDate>
            <atom:updated>2021-10-12T09:30:07.631Z</atom:updated>
            <content:encoded><![CDATA[<h4>Case study on Kannada grammatical error correction</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*9oftURIUuBN3PSnO" /><figcaption>Photo by <a href="https://unsplash.com/@kryptonitenicky?utm_source=medium&amp;utm_medium=referral">Nikhita S</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><ol><li>Introduction</li><li>Business Problem</li><li>Mapping to ML/DL problem</li><li>Understanding the Data</li><li>Data Processing</li><li>Pre-requisites</li><li>Architectures</li><li>Metrics</li><li>Deployments</li><li>Conclusions and Future Work</li><li>Profile</li><li>References</li></ol><p><strong>1. Introduction</strong></p><p><em>Grammar(Linguistics)</em></p><p><em>In </em><a href="https://en.wikipedia.org/wiki/Linguistics"><em>linguistics</em></a><em>, the grammar (from </em><a href="https://en.wikipedia.org/wiki/Ancient_Greek"><em>Ancient Greek</em></a><em> γραμματική grammatikḗ) of a </em><a href="https://en.wikipedia.org/wiki/Natural_language"><em>natural language</em></a><em> is its set of </em><a href="https://en.wikipedia.org/wiki/Structure"><em>structural</em></a><em> constraints on speakers’ or writers’ composition of </em><a href="https://en.wikipedia.org/wiki/Clause_(linguistics)"><em>clauses</em></a><em>, </em><a href="https://en.wikipedia.org/wiki/Phrase"><em>phrases</em></a><em>, and </em><a href="https://en.wikipedia.org/wiki/Words"><em>words</em></a><em>. The term can also refer to the study of such constraints, a field that includes domains such as </em><a href="https://en.wikipedia.org/wiki/Phonology"><em>phonology</em></a><em>, </em><a href="https://en.wikipedia.org/wiki/Morphology_(linguistics)"><em>morphology</em></a><em>, and </em><a href="https://en.wikipedia.org/wiki/Syntax"><em>syntax</em></a><em>, often complemented by </em><a href="https://en.wikipedia.org/wiki/Phonetics"><em>phonetics</em></a><em>, </em><a href="https://en.wikipedia.org/wiki/Semantics"><em>semantics</em></a><em>, and </em><a href="https://en.wikipedia.org/wiki/Pragmatics"><em>pragmatics</em></a><em>.</em>[Wikipedia]</p><p>The languages which lack large monolingual or parallel corpora and manually crafted linguistic resources are called low resource languages. In NLP, these languages are many when compared to the popular languages such as English which have a relatively matured framework and tools for AI. In this case study, we have considered Kannada as a low resource language for our experiments.</p><h4>Kannada Vyakarna (Kannada Grammar)</h4><p>Kannada is a language predominantly spoken by the people of Karnataka, in the southwestern region of India.</p><p>Grammar is a guide to the proper use and universalization of language. Kannada grammar is based on Keshiraja’s “<strong>S</strong>habdamanidarpana”(1260 AD). Apart from this Kannada grammar is mentioned in “<strong>K</strong>avirajamarga” and Nagavarma II’s “<strong>K</strong>avyalocana” and “<strong>K</strong>arnataka Bhasha Bhushan”.</p><p>There are 49 letters in Kannada alphabets. The main parts are</p><ol><li>Swaras + Yogavahagalu(vowels)</li><li>Vyanjana(Consonants)</li><li>Nama pada(Noun)</li><li>Kriyapada(Verb)</li><li>Sarvanama(Pronoun)</li><li>Linga(Gender)</li><li>Vibhakti and Pratyaya(Object, Subject)</li><li>Tatsama and Tadbhava(Sanskrit and native)</li><li>Sandhi(Compounding)</li><li>Samasa(Compounding)</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/917/1*4AnfFPP11ujW0S2dy_Ndwg.png" /><figcaption>Kannada Alphabets(Image by Author)</figcaption></figure><p><strong>2. Business Problem</strong></p><p>The business problem is, detect at least 30% of grammatical errors in the text/s and correct them in a reasonable turnaround time and optimum CPU utilization. A GEC system in a low resource setting can serve as a word processor, post editor and for learners of the language as a learning aid.</p><p><strong>3. Mapping to Machine Learning Problem</strong></p><p>The above business problem can be solved using statistical models, rule-based models and neural machine translation models. In this case study, we have experimented with a neural machine translation approach.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/737/1*GPCUgW0eoaaSb0VC--2pMQ.png" /><figcaption>Block Diagram(Image by Author)</figcaption></figure><p>The loss function that we use here is sparse categorical cross-entropy, which is loss calculation using labels and predictions.</p><p><strong>4. Understanding the Data</strong></p><p>We have collected the dataset from IndicCorp is one of the largest publicly available corpora for Indian languages.</p><p>We extracted the dataset by running the curl wget widget.</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/0796ac7471fbe76d65ee90c32f593251/href">https://medium.com/media/0796ac7471fbe76d65ee90c32f593251/href</a></iframe><p>We examined the first few lines of the text file we have extracted using the head command.</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/fd036aaa992bb748a5b84de48d3bd32e/href">https://medium.com/media/fd036aaa992bb748a5b84de48d3bd32e/href</a></iframe><p><strong>5. Data Pre-Processing</strong></p><p>In data pre-processing, we need to construct real and inflected pairs of sentences, for each record in the data set. For simplicity, we will consider the input text from the dataset as the original correct(real) sentence. From this, we would need to construct the erroneous sentences. To construct erroneous sentences, we would need to tokenize our entire dataset and individually parse each through a POS Tagger, to get the entity of each token.</p><p>We used the IndicNLP[<a href="https://pypi.org/project/indic-nlp-library/">indicnlp</a>] library to tokenize our text. We passed the resulting file into the POS Tagger[<a href="https://aclanthology.org/W11-3603.pdf">postagger</a>]. We got the resulting tokens along with their pos tags.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/478/1*n-COdS7AN75JNIU16gRjVA.png" /><figcaption>Image by Author</figcaption></figure><p>Using regex and dictionary in python we have generated the inflections on each of the sentences according to the parts of speech and stored them in separate pickle files.</p><p><strong>An alternative approach to generate inflection</strong></p><p>We installed the inltk[<a href="https://pypi.org/project/inltk/">inltk</a>] library, then using the joblib library and Parallel processing generated 3 inflections to the original sentences. This is like Data augmentation in image processing. The sentences generated using this approach, some of them were inflected but the others remained unchanged. This method did not cover all the function points from a grammatical error inflection. Hence this method is a failure in generating inflection. In the future sections, we will discuss how we trained the model using this approach and the model failed.</p><p><strong>6. Pre-requisites</strong></p><p>The reader of this article should be familiar with the basics of Python programming, Machine Learning and Deep Learning, TensorFlow and building data-driven applications.</p><p><strong>7. Architectures</strong></p><p>There were many architectures, experimented with 30k dataset on Google Colaboratory with CPU as a hardware accelerator for dataset collection and pre-processing step and GPU as a hardware accelerator for model train and predictions. With those architectures, the grammatic error handling was not at all satisfactory.</p><p>Finally, we chose <a href="https://arxiv.org/pdf/1409.0473.pdf">Bahdanau’s additive attention</a>, as detailed in the TensorFlow Blog for machine translation.</p><p>We have used the BPEmb in the Encoder layer to get the sub-word embeddings. We have trained 3 models separately</p><ul><li>Model1: The number of words in sentence &lt;= 6. The default parameters left as it is for the Adam optimizer. Without any regularization, the model seemed to overfit.</li><li>Model2: The number of words in sentence &lt;= 6. The learning rate of the Adam optimizer was chosen to be 5e-5. We used an l2 regularizer for the Encoder Dense layer. The predictions were average on the validation dataset</li><li>Model3: The number of words in sentence &gt; 6. The learning rate of the Adam optimizer was chosen to be 5e-5. We used the l2 regularizer for the Encoder Dense Layer. The predictions were good on the validation dataset too.</li></ul><h4>Model 1</h4><p>We have trained the model and a quickly taken single batch of input reduces to zero in 50 epochs. Therefore this is an overfit model.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/598/1*pL_9UJbR74augE5j9mY9Hw.png" /></figure><p>We also did random predictions for few records in the training dataset. The errors were handled, but those words that did not change from the input sentence to the output were randomly predicted.</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/7a3b7e0e8cf3c581513c70857040a569/href">https://medium.com/media/7a3b7e0e8cf3c581513c70857040a569/href</a></iframe><figure><img alt="" src="https://cdn-images-1.medium.com/max/682/1*4k5tJQcVxGRQW7e3fKKPnw.png" /><figcaption>Training predictions(Image by Author)</figcaption></figure><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/85134e33330dcda591d2ae9431c059e8/href">https://medium.com/media/85134e33330dcda591d2ae9431c059e8/href</a></iframe><figure><img alt="" src="https://cdn-images-1.medium.com/max/680/1*X1D_R0QAurOF7cq2ADxNQQ.png" /><figcaption>validation predictions(Image by Author)</figcaption></figure><h4>Model2</h4><p>We have trained the model for 500, and quickly taken a single batch of input that does not reduce to zero. Hence this model is not an overfit model. However, the predictions are not satisfactory. This may be mainly is due to the lesser number of inflections in the input dataset, due to its reduced length.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/565/1*Rk8rx18GJ6tSCdA4KZj88w.png" /><figcaption>Model2 single batch loss(Image by Author)</figcaption></figure><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/651e18cd8b07ec2fc9e9c9d788543a85/href">https://medium.com/media/651e18cd8b07ec2fc9e9c9d788543a85/href</a></iframe><figure><img alt="" src="https://cdn-images-1.medium.com/max/773/1*xT_yZfmGAC6UMa4N7Fu_aw.png" /><figcaption>validation dataset predictions(Image by Author)</figcaption></figure><p><strong>Model3</strong></p><p>We have trained the model and loss was reduced to zero in 500 epochs for the entire batch. Therefore this is the best model in our current environment settings. We have limited the max length of each sentence based on the majority of the sentence length to 10 while doing the translation. But the model seems to faithfully correct all of the tokens it is presented at a given point of time in a lesser turnaround time with optimum CPU utilization.</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/36acabcbd9c63bcf762f1b3439c956cc/href">https://medium.com/media/36acabcbd9c63bcf762f1b3439c956cc/href</a></iframe><figure><img alt="" src="https://cdn-images-1.medium.com/max/715/1*y0lI9SbwViwjFFE-hyAAQw.png" /><figcaption>predictions on the validation dataset(Image by Author)</figcaption></figure><p><strong>8. Metrics</strong></p><p>We have chosen to use both the BLEU score and ROUGE-L score as our performance metrics for all the models. Here is a comparison of the scores across all the models.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/833/1*IWmCmMKHhD-U5EbcCu_KVg.png" /><figcaption>BLEU SCORES COMPARISONS(Image by Author)</figcaption></figure><p>Following are the rouge-l scores of all three models for training and validation datasets. The total size of the original dataset is 30k. We are considering random 100 sentences in training and validation sets for the calculation of rouge-l scores. The following table summarizes the F1 scores of the training and validation datasets. Based on the F1Scores we can conclude that Model3 is performing the best amongst the given models.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/263/1*pBvZcALBto3y1uFGu0q3Xw.png" /><figcaption>Scores Comparison</figcaption></figure><p><strong>9. Deployments</strong></p><p>The translate functions were exposed as service in TensorFlow using the TensorFlow API. and can be deployed on the TensorFlow server.</p><p>The following links load the demonstration of the predictions in Google Colaboratory<strong>.</strong></p><ul><li><a href="https://drive.google.com/file/d/1YN-KYZkc-I9wM1-h8i7DABWX7ek9QczB/view?usp=sharing">finetune_less_length.mp4</a></li><li><a href="https://drive.google.com/file/d/1VOV9YmoZlH4JE8Sgm8LEG2jnXjK2dSTA/view?usp=sharing">m3.mp4</a></li></ul><p><strong>10. Conclusions and Future Work</strong></p><p>We can conclude that for the given setup the model 3 performs the best amongst the rest of the models.</p><p>I am pursuing the following ideas on</p><ul><li>Generate more parallel corpus and Dataset for error inflections</li><li>Hyperparameter tuning on more advanced grammar concepts like Phrases and clauses</li></ul><p><strong>11. Profile</strong></p><p>Full source code is available on <a href="https://github.com/jayaBalaR/casestudy2">Githublink2</a></p><p>If you have any questions or suggestions on any of the above please connect with me on <a href="http://www.linkedin.com/in/mypage1">LinkedIn</a></p><p><strong>12. References</strong></p><ul><li><a href="https://www.appliedaicourse.com/">Applied Course (appliedaicourse.com)</a></li><li><a href="https://indicnlp.ai4bharat.org/home/">IndicNLP | AI4Bharat IndicNLP</a></li><li><a href="https://aclanthology.org/W11-3603.pdf">Cross-Language POS Taggers (and other Tools) for Indian Languages: An Experiment with Kannada using Telugu Resources (aclanthology.org)</a>, <a href="http://sivareddy.in/downloads/index.html#kannada_tools">Downloads | Siva Reddy</a></li><li><a href="https://bpemb.h-its.org/">BPEmb (h-its.org)</a></li><li><a href="https://www.tensorflow.org/text/tutorials/nmt_with_attention#define_the_loss_function">Neural machine translation with attention | Text | TensorFlow</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2ac2c5d154f5" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-science/deep-learning-models-for-grammatical-error-handling-in-low-resource-languages-2ac2c5d154f5">Deep Learning models for Grammatical Error Handling in Low Resource languages</a> was originally published in <a href="https://medium.com/data-science">TDS Archive</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Life Lessons learnt the hard way]]></title>
            <link>https://jayabalambikar.medium.com/life-lessons-learnt-the-hard-way-6e3e6c138e5f?source=rss-982aa3fadb43------2</link>
            <guid isPermaLink="false">https://medium.com/p/6e3e6c138e5f</guid>
            <category><![CDATA[moral]]></category>
            <dc:creator><![CDATA[Jayabalambika R]]></dc:creator>
            <pubDate>Mon, 28 Jun 2021 09:16:33 GMT</pubDate>
            <atom:updated>2021-06-28T11:39:07.763Z</atom:updated>
            <content:encoded><![CDATA[<p>“Change is the only constant.”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*spqziII3sdyHPOSw" /><figcaption>Photo by <a href="https://unsplash.com/@rrajputphotography?utm_source=medium&amp;utm_medium=referral">Rajesh Rajput</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><h4>“In the middle of difficulty lies opportunity.” by Albert Einstein.</h4><p>I was sitting on my couch after my lunch this afternoon and rather taken into deep introspection. Probably if not for the pandemic, then we would have not learnt to bring back the childlike behaviour in us. The pandemic has taught us all many life’s lessons but rather in a very hard way. Just listing some of them here.</p><ul><li>Pay attention to our true self and be in the present.</li><li>Always be joyful and positive.</li><li>Do service to people around you and in their happiness feel your happiness.</li><li>Read good books and keep the company of good people.</li><li>Always think of the Almighty who is Omnipresent. Sing the praise of the Lord.</li><li>Be mindful of what you eat. Follow a sattvic diet incorporating lots of green vegetables and fruits.</li><li>Do gentle stretches and exercise daily to be strong both in the body and in the mind.</li><li>Be very mindful always in everything you do.</li><li>Continuously learn new things, thus keep our mind and intellect sharp</li></ul><p>Relating the above points as per our religious scriptures, these were the ancient wisdom given by many Saints of all religion. But we human beings have been totally attached to worldly desires and have totally forgotten the ancient commandments, therefore has been a lot of imbalance in the environment around us. Mother Nature is gently reminding us to follow the ancient Gospel and wisdom, by way of a pandemic. Once every living being in this world follows the ancient wisdom there will be a New Era a new beginning of Life. Positivity, Health, Peace will surface everywhere.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6e3e6c138e5f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Classification of Radar Returns]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/data-science/classification-of-radar-returns-c79fa1ce42eb?source=rss-982aa3fadb43------2"><img src="https://cdn-images-1.medium.com/max/2600/0*JxnR3pubMjOesA29" width="5184"></a></p><p class="medium-feed-snippet">Case Study of Ionosphere Dataset</p><p class="medium-feed-link"><a href="https://medium.com/data-science/classification-of-radar-returns-c79fa1ce42eb?source=rss-982aa3fadb43------2">Continue reading on TDS Archive »</a></p></div>]]></description>
            <link>https://medium.com/data-science/classification-of-radar-returns-c79fa1ce42eb?source=rss-982aa3fadb43------2</link>
            <guid isPermaLink="false">https://medium.com/p/c79fa1ce42eb</guid>
            <category><![CDATA[supervised-learning]]></category>
            <category><![CDATA[binary-classification]]></category>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[gradient-boosting]]></category>
            <dc:creator><![CDATA[Jayabalambika R]]></dc:creator>
            <pubDate>Mon, 14 Jun 2021 13:46:35 GMT</pubDate>
            <atom:updated>2021-06-26T09:24:29.992Z</atom:updated>
        </item>
    </channel>
</rss>