Explore the latest developments in natural language processing for feedback analysis in this 42-minute conference talk from Devoxx. Dive into topic detection and sentiment analysis techniques used at Qualtrics to revolutionize text analytics. Learn about word embedding as a foundational concept for neural network models. Discover popular neural network-based approaches for topic detection and sentiment analysis, and gain insights into productizing research models for real-world applications. Examine the challenges of traditional methods like LDA and explore alternative approaches such as semantic clustering. Understand the improvements in topic coherence and distribution. Investigate sentiment analysis examples, challenges, and neural network-based solutions including Convolutional Neural Networks and Recurrent Neural Networks/LSTM. Gain valuable learnings on opinion extraction and related problems in the field of NLP-driven feedback analysis.
Overview
Syllabus
Intro
Agenda
Feedback collection through surveys
How can NLP help?
What do you expect from the answers?
What options do we have?
LDA result on hotel reviews
Reasons for low coherence of LDA results
Alternative approaches
Semantic relationships in vectors
Semantic Clustering
Topic Coherence Improvements
Topic Distribution Improvements
Improved Results
SENTIMENT ANALYSIS
Examples
Challenges
Neural Network Based Approaches
Convolutional Neural Network
Recurrent Neural Network/LSTM
Sentiment Results
Other related problems Opinion Extraction
Learnings
Taught by
Devoxx
