O slideshow foi denunciado.
Utilizamos seu perfil e dados de atividades no LinkedIn para personalizar e exibir anúncios mais relevantes. Altere suas preferências de anúncios quando desejar.
ADAPTIVE FILTERING OF
TWEETS WITH MACHINE
LEARNING
Neri Van Otten
Data Scientist, Conversocial
About Conversocial
• Social customer service platform
Intelligent Prioritization
Continual Improvement
• Incorporate feedback from customers
• Constant monitoring performance
How to Process Text
• Create features
• Tf-idf for bag of words
• N-grams
• Other features e.g. length of text, contains u...
Parameter Optimization
• Separate server
• Runs on backup copy of database
Training Models
• Daily training
• Queued and trained when servers are under utilized
• Lower priority as the system still...
Models in Production
• Servers specific for prioritization
• Chef to configure new servers
• Servers download models from ...
Questions
Próximos SlideShares
Carregando em…5
×

Adaptive Filtering of Tweets with Machine Learning by Neri Van Otten

Adaptive Filtering of Tweets with Machine Learning by Neri Van Otten

  • Entre para ver os comentários

Adaptive Filtering of Tweets with Machine Learning by Neri Van Otten

  1. 1. ADAPTIVE FILTERING OF TWEETS WITH MACHINE LEARNING Neri Van Otten Data Scientist, Conversocial
  2. 2. About Conversocial • Social customer service platform
  3. 3. Intelligent Prioritization
  4. 4. Continual Improvement • Incorporate feedback from customers • Constant monitoring performance
  5. 5. How to Process Text • Create features • Tf-idf for bag of words • N-grams • Other features e.g. length of text, contains url, … • Convert to matrix of 0/1 • Singular value decomposition (SVD) • Machine learning models
  6. 6. Parameter Optimization • Separate server • Runs on backup copy of database
  7. 7. Training Models • Daily training • Queued and trained when servers are under utilized • Lower priority as the system still has working models • Store to S3
  8. 8. Models in Production • Servers specific for prioritization • Chef to configure new servers • Servers download models from S3 • Cache as many models in memory as possible • Evict older models • Use the client specific model to classify message
  9. 9. Questions

×