DAEJEON, Sept. 8 (Korea Bizwire) — The Electronics and Telecommunications Research Institute (ETRI) said Tuesday that it had developed new artificial intelligence (AI) technology that can search electronic documents by understanding questions in a level similar to human performance.
Searching electronic documents requires examining the words contained in document titles and file names and then opening each document, thus lagging behind in efficiency.
The South Korean research lab developed two types of application programming interface (API) that can not only search for the information desired, but also provide answers and grounds for the questions asked by users.
The Administrative Document QA API technology uses a deep learning language model to recognize paragraphs and charts as well as to search answers and grounding sentences.
The Paraphrase Recognition API technology can figure out if sentences with different types of phrases have the same meaning.
The ETRI addressed the robustness problem of existing AI technologies that fail to recognize semantic relation properly in case of slight alterations in sentences.
Kevin Lee (firstname.lastname@example.org)