Please help me in improving my CMU MS in Language Technologies essay
There is one attitude, one phrase that I apply to everything I do:
Yes, I can.
This was what I told myself as I wore out the surface of the treadmill losing fifty pounds, and this was what I repeated as I struggled with convolutional neural networks. The fact that I could create a system and train it felt like a revelation. I have spent three years of my career on machine learning, and my desire to pursue it further is one of the primary motivations that drives me to go through the rigors of a master's degree.
We are slowly getting closer to creating a machine that can pass the Turing test. For a machine to exhibit intelligent behavior equivalent to a human, it must understand humans and the way they interact. The language that humans speak and write is the most expressive means of communication we have. If we are successfully able to teach a machine to understand the natural language of humans in both letter and spirit, we would be creating a revolution. After I graduate, I intend to contribute toward solving some of the most pressing problems in the field of machine learning and natural language processing, and ultimately transform my ideas into a business of my own.
My first project with machine learning went through recommender systems. I used an algorithm based on cosine similarity to compute the distance between two nodes in a semantic graph, and it taught me the importance of how features can be used to make predictions. The following semester, I developed a collaborative-filtering-based recommender system and got right into regression analysis, with a goal of predicting the rating a user might give to a song based on their profile. This project edified me about the limitations of classifiers and how approaches like ensemble learners excel. To further my knowledge, I also took up electives in machine learning throughout my undergraduate study. I needed a deeper understanding dimensionality reduction and spent my last semester working on a thesis project on feature extraction.
I did not limit myself to only academics. In addition to that, I also held the position of student placement coordinator, where some of my responsibilities included managing job placements of around 450 students by coordinating with more than 90 companies worldwide. I was also a member of the college literary club, in which I organized several popular events. Even though on a few occasions these made my academic life more difficult, I do not regret them. They have taught me how to manage my time as skillfully as possible.
In my penultimate semester, XXX offered me the position of application developer, where I was a part of a 24-month Global Graduate Program whose aim was to create the technological leaders of the future. As an application developer, I got the opportunity to work in a department called XXX Research, which publishes thousands of financial text publications for clients, and this enormous amount of data was an excellent resource for solving many classic problems in machine learning.
In XXX Research, while I was trying to understand the intricacies of publishing financial research, I observed that my team was using Apache Lucene's text matching capabilities to perform named entity recognition (NER). While Lucene is a fast text indexing solution, it lacks the capabilities of modern NER systems. I hence developed a conditional-random-field (CRF) based NER system that could recognize 20,000 different organizations.
I also worked with Mr. XXX and drew upon his experience as a researcher at Cornell University, to solve a multi-label classification problem on a large text dataset. After a few months, I had the opportunity to work with XXX Innovation Lab based in Tel Aviv, on a global R&D project to develop a question answering system to answer text questions entered by users based on the publications we produce. I contributed toward the development of a knowledge-based question answering system, where I helped in converting thousands of XXX Research publications into RDF using semantic ontology. For this project, I evaluated and worked with multiple vendors, including IBM's Watson, Thomson Reuters Open Calais, Lymba and Amenity's VIP. I dug deep into their products and came to understand how they solve major problems like NER, entity disambiguation, co-reference resolution, conversion of text to an RDF graph, and many more. Here, I got an understanding of how large-scale machine-learning-based software solutions are created, and how can we model them to widely-varying use cases. Presently, I am working on using an information-retrieval-based approach to find answers to Factoid questions. I have also mentored two new joiners in my team and helped them to get on-boarded to their projects.
Continuing my current trajectory, some of the topics that are of greatest interest for graduate studies are deep learning for NLP, finding optimal feature selection and reduction strategy for problems, building scalable machine learning solutions and combining linked open data principles with NLP. Besides these areas, there are many problems that have caught my attention, and one of the most intriguing is robo-writing, where we use machine learning to generate human-like written text. XXX was particularly interested in using this approach to convert SEC filings into text articles. I was deeply inspired by Andrej Karpathy's char-RNN, where he uses a multi-layer recurrent neural network that learns to predict the next character in a sequence. Considering the revolution this will create in the publishing industry, I want to contribute toward an efficient solution to this problem. Word embedding in NLP also fascinates me, and I would like to work with one of its novel extensions, namely Thought Vectors.
I am determined to gain a strong background in computer science and plan to use it to make a dent in the vast field of machine learning and natural language processing, by contributing to academic research. I understand Prof Jaime Carbonell has done significant work in the field of NLP. His work history and current pursuits have a strong convergence with my interests, and I would like to request that you consider me as a graduate research assistant under his direction. I also found Prof Taylor Berg's work, especially in Question Answering systems, highly inspiring.
To excel as a graduate student and in my later career, I possess strong problem-solving skills along with the ability to understand new ideas and implement them using the programming skills I have gained working in the industry. I plan to use the experience I gained as a Student Placement Coordinator to forge strong industry partnerships in the form of research projects and placement opportunities. I am confident that the time I spent mentoring others in college and work, would help me make a good teaching assistant. I believe the resonance of my interests and previous work with that of Language Technologies Institute at Carnegie Mellon University will help me to enrich my knowledge and make a meaningful contribution to my field.
I would like to conclude my thoughts with a timeless creed that sums up the spirit of my life: Yes, I can.
Thank you very much for your time and attention.
There is one attitude, one phrase that I apply to everything I do:
Yes, I can.
This was what I told myself as I wore out the surface of the treadmill losing fifty pounds, and this was what I repeated as I struggled with convolutional neural networks. The fact that I could create a system and train it felt like a revelation. I have spent three years of my career on machine learning, and my desire to pursue it further is one of the primary motivations that drives me to go through the rigors of a master's degree.
We are slowly getting closer to creating a machine that can pass the Turing test. For a machine to exhibit intelligent behavior equivalent to a human, it must understand humans and the way they interact. The language that humans speak and write is the most expressive means of communication we have. If we are successfully able to teach a machine to understand the natural language of humans in both letter and spirit, we would be creating a revolution. After I graduate, I intend to contribute toward solving some of the most pressing problems in the field of machine learning and natural language processing, and ultimately transform my ideas into a business of my own.
My first project with machine learning went through recommender systems. I used an algorithm based on cosine similarity to compute the distance between two nodes in a semantic graph, and it taught me the importance of how features can be used to make predictions. The following semester, I developed a collaborative-filtering-based recommender system and got right into regression analysis, with a goal of predicting the rating a user might give to a song based on their profile. This project edified me about the limitations of classifiers and how approaches like ensemble learners excel. To further my knowledge, I also took up electives in machine learning throughout my undergraduate study. I needed a deeper understanding dimensionality reduction and spent my last semester working on a thesis project on feature extraction.
I did not limit myself to only academics. In addition to that, I also held the position of student placement coordinator, where some of my responsibilities included managing job placements of around 450 students by coordinating with more than 90 companies worldwide. I was also a member of the college literary club, in which I organized several popular events. Even though on a few occasions these made my academic life more difficult, I do not regret them. They have taught me how to manage my time as skillfully as possible.
In my penultimate semester, XXX offered me the position of application developer, where I was a part of a 24-month Global Graduate Program whose aim was to create the technological leaders of the future. As an application developer, I got the opportunity to work in a department called XXX Research, which publishes thousands of financial text publications for clients, and this enormous amount of data was an excellent resource for solving many classic problems in machine learning.
In XXX Research, while I was trying to understand the intricacies of publishing financial research, I observed that my team was using Apache Lucene's text matching capabilities to perform named entity recognition (NER). While Lucene is a fast text indexing solution, it lacks the capabilities of modern NER systems. I hence developed a conditional-random-field (CRF) based NER system that could recognize 20,000 different organizations.
I also worked with Mr. XXX and drew upon his experience as a researcher at Cornell University, to solve a multi-label classification problem on a large text dataset. After a few months, I had the opportunity to work with XXX Innovation Lab based in Tel Aviv, on a global R&D project to develop a question answering system to answer text questions entered by users based on the publications we produce. I contributed toward the development of a knowledge-based question answering system, where I helped in converting thousands of XXX Research publications into RDF using semantic ontology. For this project, I evaluated and worked with multiple vendors, including IBM's Watson, Thomson Reuters Open Calais, Lymba and Amenity's VIP. I dug deep into their products and came to understand how they solve major problems like NER, entity disambiguation, co-reference resolution, conversion of text to an RDF graph, and many more. Here, I got an understanding of how large-scale machine-learning-based software solutions are created, and how can we model them to widely-varying use cases. Presently, I am working on using an information-retrieval-based approach to find answers to Factoid questions. I have also mentored two new joiners in my team and helped them to get on-boarded to their projects.
Continuing my current trajectory, some of the topics that are of greatest interest for graduate studies are deep learning for NLP, finding optimal feature selection and reduction strategy for problems, building scalable machine learning solutions and combining linked open data principles with NLP. Besides these areas, there are many problems that have caught my attention, and one of the most intriguing is robo-writing, where we use machine learning to generate human-like written text. XXX was particularly interested in using this approach to convert SEC filings into text articles. I was deeply inspired by Andrej Karpathy's char-RNN, where he uses a multi-layer recurrent neural network that learns to predict the next character in a sequence. Considering the revolution this will create in the publishing industry, I want to contribute toward an efficient solution to this problem. Word embedding in NLP also fascinates me, and I would like to work with one of its novel extensions, namely Thought Vectors.
I am determined to gain a strong background in computer science and plan to use it to make a dent in the vast field of machine learning and natural language processing, by contributing to academic research. I understand Prof Jaime Carbonell has done significant work in the field of NLP. His work history and current pursuits have a strong convergence with my interests, and I would like to request that you consider me as a graduate research assistant under his direction. I also found Prof Taylor Berg's work, especially in Question Answering systems, highly inspiring.
To excel as a graduate student and in my later career, I possess strong problem-solving skills along with the ability to understand new ideas and implement them using the programming skills I have gained working in the industry. I plan to use the experience I gained as a Student Placement Coordinator to forge strong industry partnerships in the form of research projects and placement opportunities. I am confident that the time I spent mentoring others in college and work, would help me make a good teaching assistant. I believe the resonance of my interests and previous work with that of Language Technologies Institute at Carnegie Mellon University will help me to enrich my knowledge and make a meaningful contribution to my field.
I would like to conclude my thoughts with a timeless creed that sums up the spirit of my life: Yes, I can.
Thank you very much for your time and attention.
