I taught an online guest course at the Munich Graduate School of Economics in October 2022, called “NLP for Economists”. I taught it in October 2020 and wrote a post sharing my reflections about it back then. This post is a reflection on the recent course, although it is coming several months late (I just realized today, 17 Feb 2023, that I did not commit the post!)

In our group’s reading group, we recently discussed the following paper:

I taught an online guest course at my almamater, Eberhard Karls University of Tuebingen last month (January 2021). It is called “NLP without a readymade annotated dataset”. This post is a reflection on how it went, along with some comments from student feedback.

I attended COLING 2020 two weeks ago and started blogging about day to day learnings. After 2 days, I gave up due to conference fatigue and some family situation. However, I decided to wrap up the conference related posts today with those few papers I checked in the last 2 days of the main event. Here they are:

I am attending COLING 2020 and here are some notes about what I found interesting in Day 2 (9th December 2020) program. Note that I did not read them thoroughly. I just used titles to choose what talks to attend and either watched the video or followed the slides, and then browsed through the paper if I found them interesting. I skipped some of these initial picks in the first few minutes, once I realized it does not interest me. So, this is what remains after a highly ad-hoc and subjective selection process.

I am attending COLING 2020 and here are some notes about what I found interesting in today’s program. Note that I did not read them thoroughly. I just used titles to choose what talks to attend and either watched the video or followed the slides, and then browsed through the paper if I found them interesting. So, this is all highly a highly ad-hoc and subjective selection process. I categorized the papers I followed into two groups: Model understanding and analysis and NLP Applications.

I saw Claire Cardie’s keynote talk at EMNLP 2020. This blog post is a collection of my thoughts about it. The talk video with slides can be accessed here.

I taught an online guest course at the Munich Graduate School of Economics last month (October 2020), called “NLP for Economists”. The last real course I taught was in Spring 2018, and my professional experiences since then substantially influenced my views on what to include when teaching NLP to a non-CS audience. This post is a reflection on my course, along with some comments from student feedback.

September 2nd was my alma mater IIIT-H’s foundation day anniversary. As I connected to youtube early in the morning to watch the live stream from Canada, I started wondering about the institute’s role in my personal and professional growth. This post contains some of my personal reflections on IIIT-H. I spent 2006-08 in IIIT-H, working on my masters at the Information Retrieval and Extraction Lab there (although this new webpage does not list me as an alumnus!!). Looking back, I feel made criminal misuse of my time there, by not having any focus on the present or future. However, things are not so bleak as they seemed then now.

Today is the 5th anniversary of my PhD defense. I was wondering if I should write some kind of reflections post, I started and deleted it over the weekend. However, while on call with a project collaborator this morning, we ended up chatting about our personal journeys while waiting for others to join - that prompted me to write again. Usual disclaimer: I am writing it for myself, and not intending to be too critical about myself.

All previous posts in this series here
(Days 17 and 18 are included in one post with 4 papers)

In 2018, Taraka Rama and I published a paper“Experiments with Universal CEFR Classification”. At that time, we shared the paper’s code, along with the result files we obtained during our experiments on github. When we learnt about REPROLANG 2020, aimed at “Reproduction of Research Results in Science and Technology of Language”, we volunteered our paper, thinking this should be a easy job, as we shared everything and the dataset we used is public too (we shared processed version). When the time came for me to review the submissions of those who reproduced our paper, I learnt about how many possible roadblocks exist in replication and reproduction of NLP and ML work.

All previous posts in this series here

All previous posts in this series here

All previous posts in this series here

All previous posts in this series here

All previous posts in this series here

All previous posts in this series here

All previous posts in this series here

All previous posts in this series here

All previous posts in this series here

All previous posts in this series here

All previous posts in this series here

All previous posts in this series here

All previous posts in this series here

All previous posts in this series here

All previous posts in this series here

My parental leave ends in a month, and so, I was considering getting back into some regular reading habit. I felt committing myself to writing quick notes on the papers I came across that day will be a good way to bring some working discipline back after months of a different form of busy-ness. Note that these notes are for my own reference, I don’t have any greater good in mind. Another important thing is that I am not reading each and every paper as if I am writing a review or I am writing my thesis on it. I am just skimming through, focusing on breadth rather than depth.

Several months ago, I started writing on automatic keyphrase extraction, but couldn’t continue. I was building a keyphrase extractor for legal documents at that time. Today, I came across a ArXiv paper (soon to appear in NAACL 2019), which is making me post on the topic again.

After about 4 months of break, I feel compelled to write a blog post on some readings again! I was working with fastText over the past few weeks, as a part of preparing some tutorials, and I ended up reading some of the original papers behind it. This post consists of some of my notes/comments on three papers (first one has the longest summary, as I consider the rest as in some way building on top of this).

A few months back, I wrote two posts about Named Entity Recognition (Part 1, Part 2), which is the language processing task that is concerned with identifying “entities” i.e., person/location/organization names, specific expressions such as time and money, and other pre-defined miscellaneous set of categories in a given piece of text. Months later, I came across two talks by Joel Grus, who is a research engineer with AllenNLP - one called “Why I don’t like Notebooks”, and the other, more recent one given with others from AllenNLP called “Writing code for NLP Research”. Both of these are on my favorite recent rant-topics, and I started taking a look at AllenNLP as a consequence.

I was planning on writing more on Key phrase extraction and document tagging in NLP, but got occupied with work. In the meanwhile, I had a few discussions with people on one of the basic pre-processing tasks in NLP - tokenization (i.e., the task of breaking of a piece of text into individual tokens where a token can refer to a word, punctuation mark etc.). This is essential to do anything else further down in the NLP pipeline for any application (e.g., information extraction, sentiment analysis, machine translation, chatbots etc).

For the past few weeks, I have been working on automatic keyphrase extraction from documents. This post is the first in (hopefully) a series of posts to note down my observations on the topic.

Over the past few weeks, I spent quite a bit of time reading and thinking about the difference between academic and industry R&D for NLP and Machine Learning systems. This post is a quick summary of my thoughts, and readings. The disclaimer: this is from someone who switched very recently (2 months) from years in academia, into industry. All views are my own, not my employer’s.

In the last post, I wrote about the general idea of Named Entity Recognition, the issues I noticed in the way researchers discuss the problem in contemporary research articles. My point for this post is to show - we don’t need complex models to report state-of-the-art accuracies on standard datasets. I took a freely available dataset (CONLL-03) and trained three NER models:

I have been spending quite sometime in the past 2-3 weeks reading and working on Named Entity Recognition aka NER, which is a sub-problem of information extraction from textual documents, and deals with automatically identifying names of people, organizations, locations, mentions of events, date, time etc. Comparing early and recent work, developing a few NER models and evaluating them, I noticed a few issues with the way this is done, and how it affects general understanding of the problem, and its application in real-world. This series of posts are my reflections. I want this first post to be more like a background, and in the next post, I will talk about evaluating existing NER models.

This post is my comments/notes about the following paper:

I gave a tutorial at the Mid-west Big Data Summer School 2018, held at Iowa State University, USA. It is titled “Natural Language Processing with Python” and this post is just a quick summary of what I did, what I wanted to do etc.

Having 5 hours to kill in the airport, and not wanting to browse, I started checking the old issues of Computational Liguistics journal, which was then called “American Journal of Computational Linguistics”. The ones I checked out were either book reviews or short communications (There is was one review of a conference on Computational Linguistics in Medicine, from 1980!).

I gave a guest lecture today in a research methods class of comprising of graduate students from Rhetoric and Professional Communication. When Prashant Rajan, who teaches this course, asked me about this a few months back, I, for the first time, asked myself- “What exactly are the research methods of computational linguistics?” I never had a class dedicated to research methods - we learnt methods of NLP/CL in those classes which taught NLP/CL. I read about scientific method, specific stuff like - methods for social science research, methods for psychology research etc, and I think many of my friends who came from these backgrounds had specific methods classes. But, I don’t remember any CS friends I had, with or without NLP, talking about a research methods class.

This post is a short summary of the following research paper

This post is a continuation from my previous post on teaching a course “Language as Data” for introducing text processing and analysis methods to liberal arts majors. The posts in this series can be accessed here.

I teach a course “Language as Data” for introducing text processing and analysis methods to liberal arts majors. The course is a new experimental course first taught in Spring 2017, and now in its second round, in Spring 2018. I use R as the programming language, and this series of posts are my notes and observations about teaching in R for non-programmers.

Several months back, in July 2017, I wrote about Universal Dependencies and wondered about not seeing anything for Telugu. Thanks to @phylostar, I ended up in successfully collaborating on a UD-Telugu project during July-October 2017 and we successfully released the first freely available dependency treebank for Telugu in November 2017.

Today, I came across a tweet by François Chollet:

This are some notes I am making about the following article I came across on ArXiv.org:

This post is a continuation from my previous posts on teaching a 100-level undergraduate course called Language and Computers. As mentioned earlier, it is a very diverse class, and I use this textbook: Language and Computers by Marcus Dickinson, Chris Brew and Detmar Meurers. I was quite busy over the past few weeks, and I almost thought of stopping this series, but decided to give it a completion. The last topic in my class was “Machine Translation”.

This post is a continuation from my previous posts on teaching a 100-level undergraduate course called Language and Computers. As mentioned earlier, it is a very diverse class, and I use this textbook: Language and Computers by Marcus Dickinson, Chris Brew and Detmar Meurers.

This post is a continuation from my previous posts on teaching a 100-level undergraduate course called Language and Computers. As mentioned earlier, it is a very diverse class, and I use this textbook: Language and Computers by Marcus Dickinson, Chris Brew and Detmar Meurers.

This post is a continuation from my previous posts on teaching a 100-level undergraduate course called Language and Computers. As mentioned earlier, it is a very diverse class, and I use this textbook: Language and Computers by Marcus Dickinson, Chris Brew and Detmar Meurers.

This post is a continuation from my previous posts on teaching a 100-level undergraduate course called Language and Computers. As mentioned earlier, it is a very diverse class, and I use this textbook: Language and Computers by Marcus Dickinson, Chris Brew and Detmar Meurers.

This post is a continuation from my previous posts on teaching a 100-level undergraduate course called Language and Computers. As mentioned earlier, it is a very diverse class, and I use this textbook: Language and Computers by Marcus Dickinson, Chris Brew and Detmar Meurers. After finishing 4/8 topics in the textbook, I decided to take a break for 2 weeks from its content. One of these weeks was dedicated to introducing the general idea of Natural Language Processing and talking about typical tasks involved, where they are useful etc. 2nd Week mostly consisted of mid-term group presentations (totally enjoyed - more on them in the next post). So, what can we talk about NLP in a 100-level course? This post is my reflections on that.

I got interested in Exploratory Factor Analysis (EFA) recently, thanks to some of the students with whom I work right now. They come from a background of statistical methods in language testing, where EFA is generally used to look for validating the test items, pondering over questions such as whether they are all assessing the same underlying constructs or different ones etc. I saw it as a good method to come up with and validate some hypotheses about data (That, by the way, is not a ground breaking idea. It has been done before.), and to use it as a way to group features based on some idea of a common underlying construct (unlike PCA, whose primary goal is dimensionality reduction) and later use Confirmatory Factor Analysis (CFA) to validate the “theory” about feature groups from EFA.

This post is a continuation from my previous posts - (Part 1, Part 2 and Part 3) on teaching a 100-level undergraduate course called Language and Computers. As mentioned earlier, it is a very diverse class, and I use this textbook: Language and Computers by Marcus Dickinson, Chris Brew and Detmar Meurers. This post is about Chapter 4 of that book, called Searching. The chapter primarily dealt with ideas behind searching through unstructured text.

This post is a continuation from my previous posts - Part 1 Part 2 on teaching a 100-level undergraduate course called Language and Computers. As mentioned earlier, it is a very diverse class, and I use this textbook: Language and Computers by Marcus Dickinson, Chris Brew and Detmar Meurers. This post is about Chapter 3 of that book, called Language Tutoring Systems. The chapter primarily dealt with idea of Computer Assisted Language Learning i.e., designing software to teach and assess learning of a new language.

This post is a continuation from my previous post on teaching a 100-level undergraduate course called Language and Computers. As mentioned earlier, it is a very diverse class, and I use this textbook: Language and Computers by Marcus Dickinson, Chris Brew and Detmar Meurers. This post is about Chapter 2 of that book, called Writers’ Aids. The chapter primarily dealt with the intuitions behind spelling and grammar checkers.

I teach a 100 level course called Language and Computers to a class of undergrads at all stages in their degree programs, and coming from diverse backgrounds. About the half the class are from CS and related disciplines, but there are also students from Journalism and Mass Communication, Physics, Biology, Chemical Engineering, and even Advertising, to name a few. This (and a few more future posts) are mostly my notes on teaching them (and it is my first time with this course), and on how to change for the next iteration, whenever that happens.

Thanks to the recent discussions with @phylostar, I have been reading quite a bit on Universal Dependencies(UD) and on the creation of manual annotations following the UD scheme. While there are elaborately written guidelines on the website, clearly, each time you start working with a new language (especially something not very close to English), you need to figure out how to accomodate certain language specific elements into UD without losing the “universal” part and not losing adequate representation of the language phenomenon. This was the main theme in my last week’s readings about the creation of UD treebanks for various morphologically rich languages. [2] gives a general overview of the idea of UD for the uninitiated researchers. I used UD in Stanford parser before, but never knew it now exists for over 50 languages and several more are getting ready!

This post is a summary of the following paper

I came across this short paper today in the ArXiv digest. This post is a quick summary.

This is a quick summary of this short paper:

I participated in my first Kaggle competition in this month (Sberbank Russian Housing Market) along with @embedsri, and this post is a reflection on that.

This is a reading summary of the following paper:

This post is my quick note-taking about the following 1982 article:

Singer, Harry and Donlan, Dan. Active comprehension: Problem-solving schema with question generation for comprehension of complex short stories.Reading Research Quarterly, pp. 166–186, 1982.

This post is about the following short paper:

Generating Natural Language Questions to Support Learning On-Line

This post is my quick summary of what I understood an article I read, by a team from Microsoft Maluuba.

I have been reading about automatic question generation on and off for about an year now. There has been an increase in stuff that is coming up since last year, thanks to work on machine comprehension. My idea is to read regularly and write my summaries here from now on.

My first post has to be about why I chose to set up another blog when I already have another one for about 11 years now. I used to write about all sorts of things over there, but repeated readings about Jekyll tempted me into trying this out now. So, I hope to write my technical posts on this blog from now on.