nir-vana innovation network

Text to Video (TTV) is a service that automatically makes text data into video content which extracts valuable data from text data containing various information and converts it into video content (converts long text (up to 12,000 B2C texts) to video). TTV service is running through cloud servers. Works in multi-cloud environments such as AWS (Amazon Web Service) and MS Azure (Microsoft Azure). [How the keyword extraction model works] ■ The keyword extraction model performs an understanding of the text that the user enters first. ■ Specify which part of this is determined to be the core, and search for words that are usually repeatedly printed as keywords in the beginning. The more important the words in the text, the more they are repeated determine it as a keyword and proceed with keyword extraction. ■ Select and compare sentences containing the keyword, or select and compare words. ■ The A.I.'s keyword extraction model compares words. Extracts repeated words and compare them with cosine similarities. Each word is embedded as a vector to compare it with cosine similarity. After comparing words, leave the word group with the highest similarity. [How to extract the similarity of cosine] ■ For example, catch 'apple' and compare it with the word that appears afterward. If "cat," "pear," "chair," and "shine muscat" are counted as words that appear afterward, the more similar, the word is embedded in close proximity to each other, and cosine similarity allows you to determine the distance between the vectors If you know, you can find words that have similar meanings. ■ In other words, the words closest to "apple" are most likely "bae" and "shine muscat," and in the case of "cat" and "chair," they are located relatively far away, so they are not viewed as similar keywords.