Award-Winning Experimentation on Humanizing AI, with Rachel and the Semantic Research Team
The computer is humming, keyboard clacking, and kitty “Meatball” softly purrs from atop the desk. Rachel has been in home isolation for weeks now, sat at her workspace, competing in a prestigious international AI competition. The competition ranking constantly shifts in real-time as competitors update their entries. As many of the others grow stronger, Rachel maintains adamant focus.
Every day, she stared at her monitors as the rankings shifted and she pondered the possibilities for improvement. If her ranking was overtaken by other competitors, she challenged them at once and worked to discover new solutions. After a month passed, Rachel had won the International Workshop on Semantic Evaluation (SemEval 2020). Here’s a bit of insight into Rachel’s experiences as an AI engineer who innovates by exploring uncharted technological territory.
OPPO won both first and second place in Sentiment Analysis for Code-Mixed Social Media Text in Spanish/English. This is a major feat which demonstrates OPPO's leading capability in the field of mixed language sentiment analysis.
Rachel is an artificial intelligence engineer at OPPO Research Institute, and a representative of OPPO in the 14th International Semantic Evaluation, SemEval 2020. The contest in which she participated specifically pertained to social media text sentiment analysis. The analysis of human emotions is an enormously difficult task for artificial intelligence. To take on this next development of AI, OPPO has established a Speech and Semantic Research Team which works on multi-dimensional analysis and holds repeated comparison experiments on the emotion classification model.
SemEval is the most recognized competition in the field of artificial intelligence, and all of the world’s top technology companies and academic institutes participate in this competition each year. To participate in it is an honor and to rank highly in this competition is an astounding accomplishment.
Rachel is somewhat of an unusual competitor – when matched up among these vastly experienced computer scientists, she looks like the odd one out. She recently graduated and had been with OPPO for less than a year prior to entering SemEval 2020. Her professional background is in civil engineering, which is not exactly the same as AI specialization but a related field. She has made incredible strides since beginning work at OPPO.
Rachel reports the second half of the competition as an extreme difficulty spike, because that’s when the epidemic hit and caused a need to work from home. However, she was able to adapt and overcome this new challenge.
“Psychologically, it is mainly that I can’t communicate with the team in time, but fortunately we are equipped with a variety of software needed for remote office and can overcome this.”
She advises that people who work from home maintain normal work hours and a comfortable home office environment so that you feel you can wholeheartedly invest in your work.
Rachel’s workspace at home is kept neat and free of unnecessary distractions so that she can fully concentrate on her computer research. Her cat Meatball often joins her by nestling up on the desk.
“Meatball crawls across the table and sleeps quietly by the computer, which makes me feel calm as I work.”
Rachel says that she will go on to apply sentiment analysis technology to OPPO products with interactive user services. She aims to make OPPO devices better understand human emotion, which will enable understanding of the needs and desires of users.
“Natural Language Processing” is considered by many to be among the most complex areas in computer research. It goes beyond what we currently know and delves into a field which is sort of poetic in a sense – requiring engineers to approach it with a certain dedication to the unexplored.
OPPO’s Speech and Semantics Research Team of the OPPO Research Institute is committed to expanding upon what AI is capable of and can accomplish for the good of humankind. By combining specific characteristics of task data with advanced model tuning and training, OPPO has established itself as a pioneer and leader in the future of artificial intelligence.
“This is not a field that can be completely solved in a structured, digital way. Rather, it requires us to approach computers in a way that intertwines with and mimics human cognition. This is why I like it – I like its complexity and am fascinated by its elusiveness.”
-Rachel, OPPO AI Engineer