Kakao Enterprise Ranks First in Global English Machine Reading Comprehension Competition

[Asia Economy Reporter Seungjin Lee] Kakao Enterprise has proven its global artificial intelligence (AI) technological prowess.

Kakao Enterprise announced on the 12th that it ranked first in the English machine reading comprehension competition ELI5 (Explain Like I’m 5), surpassing countries such as the United States and China.

ELI5 is a global machine reading comprehension competition hosted by Meta AI’s knowledge-intensive language tasks (KILT) integrated benchmark. It evaluates performance not only in extracting answers from learned documents but also in generating sentence-form answers by referring to multiple documents. Since it targets models that can generate text not present in the documents, the competition is more challenging than simple extraction-type contests.

Generative models have been researched relatively recently compared to other models in the machine reading comprehension field. Kakao Enterprise is the first domestic company to participate in this competition and secured first place.

This model is an improved version of the ‘LittleBird’ model, which ranked first in the Korean machine reading comprehension evaluation ‘KoQuAD,’ adapted for generative models. While most companies submit large models to achieve high performance, Kakao Enterprise won first place with a lightweight model optimized for speed.

The company explained that the model submitted by Kakao Enterprise is easier to apply in actual services compared to large models and that achieving global first-place performance with a small model is advantageous in terms of cost and efficiency.

Kakao Enterprise plans to further advance its services such as the AI contact center ‘Kakao i Connect Center’ and AI chatbot ‘Kakao i Connect Talk’ based on this model. Previously, answers had to be classified one by one for each question, but with this model, it is possible to automate learning documents and directly generate answers. This is suitable for companies where response manuals frequently change and rapid service application is required.

Meanwhile, Kakao Enterprise has previously ranked first twice in the AI Korean reading comprehension evaluation KoQuAD in 2019 and March 2022.

Choi Dong-jin, Chief AI Officer (CAIO) and Vice President of Kakao Enterprise, said, “This first place in ELI5 is significant for the AI industry as it demonstrated global first-place performance with a lightweight model optimized for speed. We will gradually expand this model to create one that combines both performance and efficiency.” He added, “Kakao Enterprise will continue research to ensure that AI contact centers and AI chatbots not only accurately deliver necessary information to users but also generate answers in more natural sentences.”

This content was produced with the assistance of AI translation services.