Posts

In the era of large language models (LLMs), knowledge has become one of the most intensively studied topics in natural language processing (NLP). In this post, I introduce trends from scientific papers about knowledge and NLP published in 2023. For this post, I analyzed papers from the following conferences. I didn’t target journals due to their longer review cycles. ACL 2023 EMNLP 2023 EACL 2023 theWebConf 2023 SIGKDD 2023 SIGIR 2023 ISWC 2023 CIKM 2023 Extract keyphrases First, I applied keyphrase extraction to the titles of the papers. I calculated the normalized pointwise mutual information (nPMI) values for all the n-grams (up to 3-grams) in the titles. Then, I selected keyphrases with nPMI values above 0.5. ...

Everyone would agree on the importance of keeping a diary, but they would also acknowledge the challenge of maintaining motivation to write one. For me, the biggest issue was the difficulty in recalling what I did during the day when writing a diary at night, which ultimately led to a loss of motivation. Me giving up keeping a diary One day, I had an epiphany that I had been unknowingly maintaining another diary: I consistently use my TODO list app to manag e most stuff of my daily life, and a list of tasks marked as completed in a day well depicts what I did during that day! Additionally, the day’s weather can be obtained from other data sources. I keep my schedule on Google Calendar. I realized that aggregating this information could create a comprehensive diary entry for the day. ...

(Update on Dec 30, 2021: I created a python library luigiflow for the workflow management I describe here. Any feedback is appreciated!) I’m a research engineer of machine learning (ML), exploring ML models for various tasks. Like many ML researchers, I spend a lot of time doing experiments. So improving the efficiency of experiment workflows would yield considerable benefits. MLOps is one of the hottest topics in the field, but I’ve found that the primary goal of MLOps isn’t for me. MLOps aims to improve ML models continuously, whereas I’m responsible for developing new ML models for new tasks. I believe this isn’t the case just for me. I’ve met many people who have the same trouble, from academia to industry. So I started to think about what I don’t need in addition to what I need for managing ML experiments. ...