Haryo Akbarianto Wibowo

Haryo Akbarianto Wibowo

Builder on Artificial Intelligence Field

About Me

I am an experienced AI Practitioner in Industry for Natural Language Processing and Computer Vision fields. Bachelor of Computer Science with a hobby of experimenting and reading article about Artificial Intelligence in my free time. Have a good knowledge about Artificial Intelligence, Machine Learning, Natural Language Processing, Deep Learning, and a bit of Web Development. Currently, I am going to go deep in NLP field and research. You can see my research interests below.

Aside of that, I am also loves to contribute on building something that is useful for everyone. Currently doing project with my friend to build something useful for application developer.

Passionate to contribute to the society in the world especially with Artificial Intelligence. Write actively at my blog especially about Artificial Intelligence. Love to learn and want to make myself better as a person continuously.

Research Thoughts

I am interested in creating model that can output a controllable text (i.e.: style transfer). To do it, I believe we need a way to train the model that is able to output a robust global representation that can be tweaked (i.e.: latent variable). This problem is a challenge for the NLP research community. I am eager to help solve it.

I am also interested in finding the efficiency and explainability of Transfer Learning. Many papers in NLP have different ways of training a language model. It intrigues me to think about the learning process. I believe that crafting a way to train a model is similar to creating a syllabus for students, where if you have a good syllabus, it will make students learn efficiently. It would be good if we have explainability about the learning itself, so the research community will have better feedback on what to improve.

I come from a country that had under-represented data and models. I could feel the pain of finding them to research NLP. Available resources are in English and another popular language. This problem is the struggle of the researcher that resides in the country. Furthermore, current progress on the model in NLP tends to make a bigger model. For our country, which is not as rich as a developed country, the researchers cannot follow that trend. Thus, we need to create a low resource model that is efficient and still good to be used and explored. I want to create an opportunity for anyone to have a low-barrier entry to contribute to the NLP research society by solving the low resource problems of data and models.

Interests
  • Natural Language Processing - Style Transfer
  • Natural Language Processing - Transfer Learning
  • Natural Language Processing - Low Resource
  • Probing and Get Explainability of a model
Education
  • BSc in Computer Science, 2017

    University of Indonesia

Recent Publications

Quickly discover relevant content by filtering publications.
(2022). Which Student is Best? A Comprehensive Knowledge Distillation Exam for Task-Specific BERT Models.

PDF

(2021). Indocollex: A testbed for morphological transformation of indonesian word colloquialism. ACL.

PDF Dataset

(2021). BERT Goes Brrr: A Venture Towards the Lesser Error in Classifying Medical Self-Reporters on Twitter. SMM4H.

PDF DOI

(2020). Costs to consider in adopting NLP for your business.

PDF

(2020). Semi-supervised low-resource style transfer of indonesian informal to formal language with iterative forward-translation. IALP.

PDF Code DOI

(2018). Towards product attributes extraction in Indonesian e-commerce platform.

PDF

(2018). Cross-lingual and supervised learning approach for Indonesian word sense disambiguation task.

PDF

Experience

 
 
 
 
 
Kata.ai
Research Scientist
May 2019 – Feb 2022 Jakarta, Indonesia
  • Researched NLP to enhance the chatbot performance
  • Researched colloquial problem and style transfer. I explored word-level and char-level seq2seq approaches. My team published two academic papers that are published in reputable conferences such as ACL and IALP
  • Developed a text data synthesis tool that can produce text dataset automatically. It increased the productivity of Kata.ai’s Conversational Designer team significantly.
  • Won first place at a shared task on Social Media Mining for Health (SMM4H) 2021 shared task (an ACL shared-task). My team wrote an academy paper about it.
  • Researched about Model Compression Technique in NLP to benchmark several approaches and compare its performance (quality, time, and space).
  • Researched Image Retrieval as one of the planned company chatbot’s feature.
  • Researched semi-supervised learning on a sequence to sequence problem.
 
 
 
 
 
Deligence.ai
AI Software Engineer
Deligence.ai
Jan 2019 – Mar 2019 Jakarta, Indonesia
  • Worked as an AI Software Engineer developing several applications with Deep Learning
  • Developed an object detection model to identify car’s type
  • Had a secondary role as a Business Developer, planning business
  • Helped on making a business hypothesis and also help the company to create pitch deck.
 
 
 
 
 
University of Indonesia - Faculty of Computer Science
Research and Teaching Assistant
University of Indonesia - Faculty of Computer Science
Jan 2017 – Jan 2018 Depok, Indonesia
  • Worked as a Teaching Assistant at the Faculty of the Computer Science University of Indonesia. Assisted Courses:
    • Artificial Intelligence
    • Data Mining
    • Data Science and Analytics
    • Design and Analytic Algorithm
    • Data Mining and Business Intelligence
    • Math I.
  • Conducted research about Cross-Lingual Word-Sense Disambiguation using Machine Learning supervised approach. Wrote a paper as the conclusion of the published research on the Global Wordnet Conference (GWC)
  • Conducted research about creating the dataset on product attributes extraction on e-commerce platforms. The result of the work is an academic paper.

Contact Me

My primary email