CASE 5
Is it possible to grasp potential preferences and tastes by analyzing bloggers words, then put that to use in the form of an advertising or product recommendation system?
Case overview
Developed a system in which AI reads and understands blogs created by users, then automatically estimates those blogger's favorite topics.
Background / Challenges
- Although bloggers blog about a variety of topics, could it be possible to figure out what topics they're really interested in?
- Might it be possible to grasp the taste and potential preferences of a blogger by analyzing what they've written?
- Can we improve the accuracy of an algorithm that is used to recommend other articles to those who have read a certain article?
Results / Outlook
AI can now estimate a blogger's favorite topic through the use of natural language processing.
Our solution
- Natural language processing by AI
-
We developed a system to structurally understand sentences by dividing them into subjects, predicates, objects, word context, 5W1H, etc. using the latest SRL (Semantic Role Labeling) technology.
- Creating learning data for AI
-
Using the latest annotation technology, developed a system to create learning data that allows AI to understand sentences.