Aspect Based Sentiment Analysis in 20+ Languages
summary
Describes a grammar-based, multilingual Aspect-Based Sentiment Analysis system from Bitext that uses a language-independent lexical analyzer and a PDA-based GLR parser plus large corpus-derived lexicons and unification grammars to detect topics and polarity in 20+ production languages. Compares statistical ABSA approaches (including probabilistic and symbolic parsing) against a hybrid strategy that uses linguistic/symbolic processing for topic and polarity detection and machine learning for aspect categorization, and explains how domain adaptation is handled via adding a small number of rules without requiring annotated training corpora.
AI-written summary · about
chapters (22)
- 0:00Start
- 0:40Table of contents
- 1:44Polarity-based sentiment analysis
- 1:54Topic-based sentiment analysis
- 2:59Classical dashboard
- 3:20Topic-level granularity
- 4:39Benchmark for ABSA - Key facts
- 8:34Accuracy and effort required
- 10:42Typical pipeline for ML ABSA
- 13:09ABSA - Sentiment at topic level
- 14:53ABSA - Categorized sentiment
- 15:46Bitext Platform architecture
- 16:53Lexical Analysis
- 23:12Syntactic Analysis
- 25:08Step 2 - Parse Treess uses
- 25:50Semantic postprocessing
- 29:19Ideal pipeline for ABSA
- 30:17Penn Treebank sample
- 32:44One last note on parsing
- 34:11Syntactic & semantic functions
- 35:02Text as a multipurpose graph
- 35:33Conclusions
related talks
39mSentiment Analysis, its techniques and applications
43mSentiment Analysis in Action: Building Your Real-time Pipeline
01h 31mAn Introduction to Sentiment Analysis of...
22mSentiment Analysis on Sustainability Reporting
32mUnpaired Sentiment-to-Sentiment Translation: Using Reinforcement Learning
01h 31mNatural Language Understanding: Foundations and State-of-the-Art
46mContext is all you need: FUNdamental linguistics for NLP
43mGetting Sentimental: exploration of iOS sentiment analysis