Naver had 40 patents in digitalization during Q4 2023. Naver Corp’s patents in Q4 2023 include a ranker for neural information retrieval, a method for generating simplified network layouts, a system for training embedding functions, and a confidence estimation system using neural networks. These innovations focus on optimizing document ranking, network layout readability, relevance between different modalities, and confidence estimation in neural networks. GlobalData’s report on Naver gives a 360-degreee view of the company including its patenting strategy. Buy the report here.
Smarter leaders trust GlobalData
Naver grant share with digitalization as a theme is 50% in Q4 2023. Grant share is based on the ratio of number of grants to total number of patents.
Recent Patents
Application: Neural ranking model for generating sparse representations for information retrieval (Patent ID: US20230418848A1)
The patent filed by Naver Corp. describes a computer-implemented ranker for a neural information retrieval model that includes a document encoder with a pretrained language model layer and a query encoder. The document encoder generates sparse representations for documents predicting term importance over a vocabulary, while the query encoder generates representations for queries. These representations are compared to rank the documents based on document scores. The document and query encoders are separate and can be differentiated by various factors such as model architecture, size, and regularization methods. Additionally, the document encoder can expand documents within the vocabulary, while the query encoder is more efficient and may use different regularization techniques.
The method for information retrieval outlined in the patent involves generating sparse representations for documents and query representations using separate encoders, comparing these representations to rank documents, and utilizing different regularizers for the document and query encoders. The document encoder is trained using middle training before fine-tuning for information retrieval, and the pretrained language model may be trained using masked language model training combined with FLOPS regularization. The ranker can also produce additional document scores using a lower-latency retrieval method, merge these scores, and rank the documents accordingly. Overall, the patent focuses on optimizing the neural ranker for efficient information retrieval by utilizing separate encoders, different regularization techniques, and middle training for the language model.
To know more about GlobalData’s detailed insights on Naver, buy the report here.
Premium Insights
From
The gold standard of business intelligence.
Blending expert knowledge with cutting-edge technology, GlobalData’s unrivalled proprietary data will enable you to decode what’s happening in your market. You can make better informed decisions and gain a future-proof advantage over your competitors.
GlobalData, the leading provider of industry intelligence, provided the underlying data, research, and analysis used to produce this article.
GlobalData Patent Analytics tracks bibliographic data, legal events data, point in time patent ownerships, and backward and forward citations from global patenting offices. Textual analysis and official patent classifications are used to group patents into key thematic areas and link them to specific companies across the world’s largest industries.
