The SOL project is Twig Farm's LETR team NIA and Multi-campusIt is a program that works hand in hand to support the development of excellent artificial intelligence talents. This is an opportunity for trainees to experience practical business projects and experience the LETR team's research and development culture up close.
Since the first phase of the project last August, 9 prospective developers have been with the LETR team for about 3 months. We were divided into 3 groups to work on a team project so that we could experience the actual scene more closely. I also had time to interact and collaborate with LETR team researchers in the business through mentoring.
I heard the stories of the 9 participants in the first round, which finished a while ago. I think they all felt differently while actually working on the project. He shared an honest story about what he learned, thought, and felt through this experience.
This time, in the third order, 'Open text correction'Here are the stories of the interns who participated in the project. Also, maybe not yet”Natural Korean'and'Classification of profanity and hate speech'If you haven't watched the team's interview, I recommend reading it together.
Artificial intelligence is the first (?) Because
Hallo Please introduce yourself!
Yonghoon: Hallo I'm Kwon Yong-hoon, who participated in the 'Floating Writing Correction' project.
Taeyoung: I am I studied computer science for 2 years in graduate school, and then participated in national research projects in the field of defense for 3 years. I first learned natural language processing along with machine learning and deep learning through a multi-campus curriculum, and I even got to practice like this. Previously, I had been using Python for about a year.
Junho: While majoring in language (Chinese literature) at the undergraduate level, I also chose a convergence software related major as a double major. Since then, I am preparing to enter related fields through various experiences such as participating in application development contests.
Why did you become interested in artificial intelligence, and natural language processing in particular?
Yonghoon: Regardless of fields such as image processing, natural language processing, and voice recognition, I was very interested in artificial intelligence in general. I became more interested in natural language processing when I started studying this time.
Taeyoung: I've always wanted to go into the field of artificial intelligence, but after taking a natural language processing course, I became more interested.
Jinho: Due to the nature of my major, I was able to encounter texts in various styles, from ancient literature to contemporary essays. In the process, I was able to realize that there aren't as many restrictions as I thought on how to express my intentions or feelings. Later, as I experienced programming, I came up with the idea that I wanted to develop natural language processing technology with a sense of humanity through artificial intelligence learning.
After experiencing the SOL project
What led you to apply for the SOL project?
Yonghoon: I was curious about how to use what I learned in the multi-campus natural language processing curriculum, and how others are using it. In particular, I was really curious about how the display correction using artificial intelligence works.
Taeyoung: I was curious about how the theory I learned during the curriculum would be used in practice. At the end of the day, the CEO of Twig Farm and two researchers were mentors in my class, and I was curious about the company and applied because I thought it would be a great opportunity to experience practical projects in the business.
Junho: While studying natural language processing courses on multiple campuses, I came across a recruitment announcement. I actually applied because I wanted to experience the field at a place that actually provides technical services such as language refinement and correction, and participate in practical projects.
What was your experience with the SOL project?
Yonghoon: I looked for code for prior research papers related to spacing, and was responsible for modifying environment settings, model training, and data preprocessing. This time, I gained confidence in using an external model by setting up the environment using BERT, modifying data reflecting the learning results, and analyzing the model.
Taeyoung: The goal was to implement a library that can correct spacing for everyday written words other than jargon. The project was carried out through meetings with team members, and interim progress was checked through weekly team reports and mentoring meetings.
And the actual process began by setting specific goals, the form of deliverables, and an outline progress plan. After that, I started doing literature research, and since spacing correction is a problem limited to the Korean language, I looked for many academic and journal articles in Korean.
After analyzing the paper, we investigated the relevant library code for actual implementation to determine the reference model, confirmed whether the code actually works properly, defined the required data, and went through the process of extracting performance by learning a model that matches the data.
Junho: The project was carried out for about 3 months, starting with case studies and analysis of prior papers, journals, and services, to the discovery of applied datasets and libraries, and performance analysis. Every Tuesday, I received feedback and adjusted research direction through mentoring, and on Friday, I did my own midterm inspection with team members. After that, my weekly routine went back to writing a weekly report on median performance and progress on the Monday. Other presentations were also prepared during the midterm and final presentations where all project teams gathered together.
What did you feel while working on a project with your team members?
Yonghoon: I learned how to write a project report and how to present it. It really helped me a lot!
Taeyoung: I've learned that to keep a project from losing direction, you have to start by drawing the big picture. To do this, the project's goals, deliverables, plans, and timelines must be clear. Until now, I was in a hurry to get the job done right in front of me, so I was often out of direction.
However, this time, I was able to come to an understanding in the process of thinking and deciding through discussions with the team members. If I understand what I'm doing now and do it, I can reduce unnecessary waste of time later.
Finally, I also made a commitment to constantly think and take the initiative to work no matter what I do.
Junho: Team members with diverse backgrounds, from relatively experienced graduate school graduates to friends who have now graduated from high school, joined the 'Spacing Correction' team. I'm in an intermediate position in terms of age and experience, so I think I learned what I didn't know and, conversely, shared what I knew. Thanks to this, I was able to move away from a fixed role and have a valuable experience thinking about solutions from various perspectives.
What was the most memorable thing you did when finishing a project?
Yonghoon: We made relationships that kept in touch even after the project was completed. There were good people left who cheered for each other by sharing what they are preparing now and how things are going.
Taeyoung: This was the day the final announcement was made. I was able to meet the people I met online due to the coronavirus and all my mentors in person. Also, I remember that the CEO always gave me lots of positive feedback and good comments (probably considering that he was an intern).
Junho: I remember the most memorable time when I finished presenting the results of the project and looking back on the past period while writing the final report. Until now, I was focusing on weekly goals, so I worked hard on small things, but when I was able to see the overall context at the end, I felt a different feeling than before.
The SOL Project and beyond
What fields do you want to challenge and what goals do you want to achieve in the future?
Yonghoon: I think people are interested in technology that helps them solve their problems, even minor ones. Also, I don't think such technology can be developed by simply developing technical ability. So after finishing the project, I came up with the idea that I wanted to solve various problems people have by combining web development and artificial intelligence.
Now, many people are interested in it, and they dream of being a machine learning web developer dealing with technology they can develop together.
Taeyoung: There are many fields of natural language processing, so I would like to experience more diverse things in the future. In particular, no matter what you do, I think the most important thing is the data used for learning, and I think it would be a great experience if I had the chance to deal with this kind of data in the future.
Above all, no matter what I do, I want to be a person who never gets tired, never loses interest, and is always open to changes and advice from those around me.
Junho: The goal was to develop a service that identifies even the intentions contained in questions through sentiment analysis research. Image-based services that analyze emotions based on facial expressions have already been commercialized in fields such as AI interviews and forensics. I believe that soon, we will be able to develop a universal service that can satisfy both popularity and expertise even with simple words or sentences.
Finally, please leave a message for the juniors who will experience the project in the future.
Yonghoon: If you work hard on your own, it will help you a lot. If you're working on a space-related project, check out our repository!
Taeyoung: (You won't receive close management every day, but you'll be able to experience the actual work firsthand and feel what kind of place the site is. This is a great opportunity to broaden your horizons, so be sure to try it out!
Junho: This time, I learned that there are more diverse language analysis datasets and open libraries than I thought. If it's not accompanied by clear topic setting and project management, the possibility of unnecessary trial and error is bound to increase. In that regard, the SOL project is a great opportunity to receive professional mentoring and infrastructure, so I hope everyone can make good use of it to create meaningful results.
While finishing
The SOL project interns have had a more intense time than anyone else in the past 3 months. In the midst of such a busy schedule, I wrote reports, released open source, and created a demo site to share results.
The results of the three SOL project teams are shown below, and we ask for your interest and support.
All of the SOL Project interns had a great time during the three months. I'll see you soon!
Natural Korean
https://github.com/twigfarm/letr-sol-koFISH
Classification of abusive hate speech
https://github.com/twigfarm/letr-sol-profanity-filter
https://huggingface.co/dobbytk/KSL-BERT
Spacing correction
https://github.com/twigfarm/letr-sol-spacing
Good content to watch together
Why does LETR, a language processing engine, focus on text languages?Planting the seeds of tomorrow's AI developers, SOL Project Intern Interview 1Planting the seeds of tomorrow's AI developers, SOL Project Intern Interview 2
