Text analytics on annual reports
summary
A working Azure application ingests PDF annual reports, extracts text via pdftotree, runs NLP with comparisons between TextBlob NLTK and Google NLP then moves toward models in spaCy, and surfaces sentiment polarity, subjectivity, and a Gunning Fog index readability score. The analyzer app uses a service bus, labels from internal users build a training set, and version-diffing flags paragraph edits. Processing an ~80–100 page report takes about five minutes, and annotator agreement is cited around 60–65% while models remain under active training.
AI-written summary · about
chapters (11)
- 0:00Annual report analytics intro
- 1:46Auditor challenges in review
- 3:14Data-driven report analysis
- 5:01Text analytics tool architecture
- 7:10Sentiment analysis libraries
- 9:39Switching to spaCy
- 12:36Readability analysis standards
- 14:21Tracking report version changes
- 15:45Tool interface and visualization
- 19:41Word clouds and topics
- 20:26Future improvements
related talks
30mProcessing Threat Reports at Scale Using AI and ML: Expectations and Reality
21mFast and Efficent Text Analysis using Azure's Language Cognitive Service
51mTaking Models to the Next Level with Azure Machine Learning
50mAzure AI, Power new possibilities for every organization
01h 05mSocial Text, Sentiment, and Tone Analysis
45mBuilding Emotionally Intelligent Apps using Azure AI
30mIntelligent document processing: Extract data & insights from documents
19mAI Driven Data Enrichment with Azure Cognitive Search