Pushing the frontiers of Information Extraction
summary
Algorithm extracts sentences that are calls for action and then classifies what is being asked, using a 2‑step ML pipeline built into AmCAT and trained on 5,000 labeled sentences. Features combine tf‑idf n‑grams with 32 linguistically informed signals derived via AmCAT + scikit‑learn + WEKA + Stanford CoreNLP; binary detection reaches about 82.2% accuracy, while nine‑way agenda classification stays low (<50%, ~0.6 when merged), so more labeled data and added tools are needed.
AI-written summary · about
chapters (5)
- 0:00Start
- 23:48NORMALIZATION
- 25:09STOP-WORD REMOVAL, STEMMING, & LEMMATIZATION
- 33:48FEATURES: DATA ORIGIN
- 34:34ALGORITHMS
related talks
16mSoK: Automated TTP Extraction from CTI Reports – Are We There Yet?
31mTaking AI/ML to the next level: Text!
44mContent Classification: NLP vs. AI in the Age of Data Vectorization
23mContact Center Voice AI: Low-Latency Intelligence Extraction from Messy Audio Streams
15mMachine Learning for All
31mWhen humans teach machines
41mAutomatic Text Categorization
30mProcessing Threat Reports at Scale Using AI and ML: Expectations and Reality