US Election Web Monitor

The US Election 2020 Web Monitor is a freely available Web intelligence platform to analyze the public debate and track the candidates’ performance on the campaign trail.

www.weblyzard.com/us-election-2020-web-monitor

Its Elasticsearch-powered visual analytics dashboard is an advanced information exploration system to automatically identify opinion leaders, and to classify online coverage along multiple metadata dimensions - e.g., by topic or geographic location. The system is continuously updated with live content feeds, reflecting social perceptions and events that impact the coverage.

Transcript

1. 1 US Election 2020 Web Monitor us2020.weblyzard.com Arno Scharl, Alexander Hubmann-Haidvogel webLyzard technology ▪ Modul University Vienna www.weblyzard.com/us-election-2020-web-monitor
3. 3 Sentiment Share of Voice Brand Reputation Measuring Social Perceptions Success Metrics WYSDOM Success Metric Wheel of Emotions
4. 4 2004 Scharl, A. and Weichselbraun, A. (2008). “An Automated Approach to Investigating the Online Media Coverage of US Presidential Elections”, Journal of Information Technology & Politics, 5(1): 121-132.
5. 5 2008
13. 13 Elasticsearch | Integration History - Started with PostgreSQL/Tsearch2 - Switched to Apache Lucene in 2009 - Adopted Elasticsearch early 2014 Current Cluster - 5 (soon 7) physical machines – XEON E5, 40 cores, 256GB RAM - 3 x 2TB M.2 NVMe Samsung 960 Pro in striped LVM - Multiple Elasticsearch nodes per machine 4 data nodes, separate master nodes, one coordinating-only node per machine - Docker containers, overlay network, discovery using DNS
14. 14 Elasticsearch | Integration Indexing – Custom component based on Vert.x – Data read from CockroachDB (PostgreSQL for legacy data) – Indexer applies transformations/de-normalizations – Enriched with additional metadata e.g., translations, named entities, geographic coordinates – One index per source, language and month e.g., English-language news media: en.4.media.2021-01 – Balance between index/shard size and number of indexes affected by queries and aggregations
15. 15 Elasticsearch | Visualization Example 1: Geographic Map – Hashgrid aggregation for extracted locations – User-selectable precision – Aggregate average document sentiment Example 2: Word Tree – Inner hits on search query – Filtered for sentences matching the query – Maintaining document-level sorting
16. 16 Elasticsearch | Visualization Example 3: Story Clustering – Aggregate top keywords per interval in date histogram – Sub-aggregation of top associations for each keyword – Converted to graph (connect keywords in subsequent timespans) – Community detection using Louvain algorithm – Date histogram aggregation for each story to obtain document count over time
17. 17 Sentiment Share of Voice Brand Reputation Measuring Social Perceptions Success Metrics WYSDOM Success Metric Wheel of Emotions Weichselbraun, A., Steixner, J., Braşovean, A., Scharl, A., Göbel, M. and Nixon, L. (2021). “Automatic Expansion of Domain-Specific Affective Models for Web Intelligence Applications”, Cognitive Computation: Forthcoming. Weichselbraun, A., Gindl, S., Fischer, F., Vakulenko, S. and Scharl, A. (2017). “Aspect-Based Extraction and Analysis of Affective Knowledge from Social Media Streams”, IEEE Intelligent Systems, 32(3): 80-88.
18. 18 Sentiment Share of Voice Brand Reputation WYSDOM Success Metric Wheel of Emotions Measuring Social Perceptions Success Metrics
19. 19 Innovation R&D Funding – Selected Research Projects
20. 20 Newsletter www.weblyzard.com/newsletter Technology Showcases www.weblyzard.com/showcases E-Mail info@weblyzard.com @ SlideShare slideshare.com/weblyzard SpeakerDeck speakerdeck.com/weblyzard YouTube youtube.com/+weblyzard Twitter @weblyzard Facebook facebook.com/weblyzard LinkedIn www.linkedin.com/company/ weblyzard-technology