Applied NLP Challenges

All of the following challenges would make the system better and need to be solved in a multilingual manner:

Difficulty

• Estimating article difficulty level - e.g. CEFR (in progress)
• Personalizing article difficulty - based on the detailed history
• Text simplification
• Automatically detecting learner capability based on their interactions

Better exercises

• Generating textual-knowledge questions
• Detecting particle verbs (and similar particles) reliably so we can highlight them correctly when translating

Content Recommendations

• Fine-grained interest discovery
• Video / audio recommendations
• Social media sources (requires a new kind of interaction with the extension)
• Content-filtering - the student who wanted “not to see articles about people dyeing”
• Main-content extraction from web articles (the last 5%)

Chat agents

• that adapt to the current learner level and interests