Renato Negrinho describes the Undergraduate Internship at CMU as:
"
Incredibly Intellectually Stimulating"

2015 UIP Renato Negrinho  Renato Negrinho is a young researcher with a passion for machine learning and very clear future goals. After his admission to the Undergraduate Internship Program, he was admitted to the Ph.D. in Machine Learning at CMU, and to an internship at Xerox in Europe. All these opportunities came from his will to do more and to continuously evolve as a person and as a researcher.

Graduated by Institituto Superior Técnico (IST), Renato Negrinho made an internship at Priberam where he had the opportunity to work closely with André Martins, dual degree alumnus from IST and CMU, in the scope of the CMU Portugal Program. When he applied to the UIP, he had as sponsors in Portugal André Martins and Mário Figueiredo, researcher at IST, and was hosted at CMU by Noah Smith. "My expectations, regarding these experiences, but also in general, are to be able to continue to work hard and learn a lot. To remain engaged and excited about what I am doing. To look for challenges that allow me to continue to grow. These have been my objectives for quite some time now. Just be passionate about what you are doing and work hard and good things will come," says Renato Negrinho.

CMU Portugal Program: How do you comment on your experience as an Undergraduate Internship Member, that had the opportunity to be at CMU early this year (Feb.-May)?
Renato Negrinho [RN]: I think that overall the experience was extremely positive. CMU is a incredible place for Computer Science, with top-class faculty and students, and an unparalleled research environment. During my stay, I was able to meet people from the Language Technologies Institute and the Machine Learning Department. There is a lot of collaboration among faculty and students. Collaboration is really emphasized at CMU and it is not constrained by department boundaries. Getting faculty involved sometimes seemed as easy as knocking on their office doors and talking to them. This is something I have not seen anywhere else I was also surprised by the range and quality of Computer Science research done at CMU. While at many places there are just a few faculty that work on some set of topics, CMU truly has a wide coverage of the Computer Science spectrum, from the more theoretical to the more applied. There are also frequent talks and regular seminars on topics like Computer Science Theory, Statistics, Machine Learning, and Natural Language Processing, that bring experts from other top US universities and from other places around the world. It was a surprise to see that both students and faculty are extremely hardworking and friendly, being usually more than happy to discuss research. All these facts combined, create a Computer Science research environment at CMU that is second to no other place.

CMU Portugal Program: Your main goal, as described on your application form, was “to devise methods for the identification of multiword expressions with the goal of improving the performance of natural language processing systems.” Were you able to accomplish this goal?
RN: The goal of the internship was partially accomplished by the time it ended. We devised several interesting methods for multiword expression identification, but in Natural Language Processing, the specification a model is usually not enough. It is a scientific field that is very much methods and results driven, and therefore, proposing a new model involves not only specifying it, but also implementing it and running a large number of experiments and comparisons to validate that the proposed model is actually useful, that is, that it is actually better (according to some desired criteria) than other simpler baselines. This requires designing and decomposing the system into parts, and writing the code that implements them, which are very time consuming tasks. This was specially so in our case, as the work we built on is fairly new and still poorly supported code wise, meaning that were not able to reuse much code and had to rewrite most of it from scratch. At the present moment, there is some code left to be written to finish the experiments.

CMU Portugal Program: What kind of support (in terms of research) did you received from your sponsors in Portugal? And from Noah A Smith, at CMU?
RN:
I cannot do anything but praise the support that I was given. My sponsors in Portugal, Dr. André Martins and Prof. Mário Figueiredo, encouraged my application and were always ready to help during the internship period, by answering my questions, pointing to relevant references, or just calling my attention to things I should be aware of. During this internship I worked mostly with Prof. Noah Smith and with a senior PhD student from his group, Waleed Ammar. Prof. Noah Smith was extremely kind and attentive, taking the time to integrate me in his group and to meet with me regularly to talk about the project progress and to discuss research in general. I thank him for his time and attention, which made my experience at CMU enjoyable and rewarding.

CMU Portugal Program: What did you gain from this internship experience?
RN:
I think that the first and most obvious thing that this internship gave me was just the experience of being in the US at the very best school for Computer Science, doing research within a very good Statistical Natural Language Processing group. I think that describing it this way makes apparent that this is a very rare and valuable opportunity. Besides getting to enjoy the wonderful research environment and meeting some very interesting and hardworking people (both students and faculty) at CMU, being in the US allowed to attend some of the open houses for the PhD programs I got admitted to. In 2014, I prepared and submitted applications to PhD programs, mostly in the US, that I found that could provide an exciting research environment for Machine Learning. Typically, around mid-February, the application decisions come out and you get to know if you have been offered admission or not. Then, somewhere around March, there are usually the so-called open house events, which last for two or three days, where admitted students are invited to visit the university, and meet and talk to the faculty and other admitted and current students. This is a great experience meant to give admitted students a better feel for the university and the program, so that they can make an informed decision on which place to attend.

"One of the things that I like about Machine Learning is that it is has a very diverse community and set of interests interests. Researchers in the community draw from a large number of tools and ideas from a wide range of disciplines, such as Computer Science, Statistics, Mathematics, Signal Processing, and Optimization, to tackle problems that they care about. I also like the fact that problems in Machine Learning drive a lot of research in many other scientific fields these days."


CMU Portugal Program: You have started an internship at Xerox Research, last May, and you will start your Ph.D. at CMU next September. How did these two opportunities came into place?
RN:
I spent the year of 2014 with a scholarship at Priberam and Instituto de Telecomunicações doing research with Dr. André Martins in Machine Learning. This period got me even more excited about research in the field, which led me to apply for a PhD. The field has grown immensely in last few years and with the increasing computational power and data availability, it is safe to assume that Machine Learning it will continue to grow in importance. One of the things that I like about Machine Learning is that it is has a very diverse community and set of interests. Researchers in the community draw from a large number of tools and ideas from a wide range of disciplines, such as Computer Science, Statistics, Mathematics, Signal Processing, and Optimization, to tackle problems that they care about. I also like the fact that problems in Machine Learning drive a lot of research in many other scientific fields these days. Also, companies have a keen interest in the field, with many of the major technology companies following it very closely. Essentially, any company that collects data as part of their operation and that hopes to extract value from it, will have automatically an interest in Machine Learning. In part, this is one of the reasons why I chose the field: besides being very intellectually satisfying, advances in Machine Learning can be really impactful, both academically and technologically. So, after making my mind, I had to prepare the applications, which is a lengthy process requiring a few months of preparation. Getting admitted to a good PhD program in the US is not something easy to accomplish.

CMU Portugal Program: In your opinion, what are the the key factors to be admitted for a Ph.D. in American universities?
RN:
In my opinion, having a strong application is a combination of several factors: doing and publishing relevant research in the same (or in a related) field of the program that you applying to; having good recommendations letters from people you have worked with and that are known in community; writing a good motivation letter that tells the admissions committee what are your research interests, what is the work you have done so far, why do you want to pursue such a degree, and why the program that you applying to is right for this. Typically grades and standardised test scores are not strong deciding factors, but may be used for weeding out candidates. Coming from and doing well in a good university helps, but this is hard to gauge for international PhD students because, unless you come from one of the few well known international universities, the committee will not know very much about it and therefore, it will look instead for other indicators of success. A great letter of recommendation from someone that you have worked with previously and is well known in the field is much more valuable. Another of the reasons why I am grateful to the CMU Portugal Program is because I believe that this internship at CMU strengthened my application, providing an international reference that is more familiar to the admissions committee. After submitting the applications, the only thing left to do is hope for the best.

"This opportunity at Xerox is great and I urge other people to go for an international experience. In academia, it is important not be confined to one's local research community, but rather to strive to be part of the global community."


CMU Portugal Program: And what about the internship at Xerox?
RN:
About the internship at Xerox, after finishing my internship at CMU, I would still have about three months until the start of my PhD, so that got me thinking about applying for another internship. At first, my intention was to stay in the US, but there were several problems (mainly because US federal regulations and company policies) and because I was starting to run out of time, I decided to turn myself to Europe and try my luck there. I talked to Prof. Noah Smith about wanting to do an internship and about the problems I was having so far, and he named some researchers in academic and industry settings that maybe could be interested in having an intern and told me he could write a reference letter if needed. I emailed these researchers with my CV and explained my situation, and eventually some of them showed interest. Dr. Ariadna Quattoni and Dr. Xavier Carreras, which are both at Xerox Research Centre Europe in Grenoble, were looking to take an intern and were interested in my profile. We discussed several possibilities and came up with a project that we were both excited about, and soon after they offered me the position, which I gladly accepted. This opportunity at Xerox is great and I urge other people to go for an international experience. In academia, it is important not be confined to one's local research community, but rather to strive to be part of the global community. Attending CMU for a PhD in Machine Learning is as much as one can hope to ask for. It is has an amazing research environment and I think it will provide me an unparalleled opportunity to learn and to take my research further. I intend to be worthy to these opportunities. I want to finish by thanking Dr. André Martins (Priberam), and Prof. Pedro Aguiar (IST) and Prof. Mário Figueiredo (IST) as they encouraged and guided me through the process of applying to PhD programs. I am very happy to have them as mentors.

July 2015