Ted talks

Web-based tool to explore speeches content, trends in time and delivery style 

Mauro Gentile

March 2, 2018

Metis, New York City


The dataset is made of 2.700 Ted talks spanning from 2004 to 2017

25 "pure" (mono-thematic) clusters have been extracted applying optimized NMF to speeches

Each topic has been analyzed in terms of content, style and trend in time through a web-based tool, introduced in the next section

Further information is available by clicking Further info in the tool page

Insert topic (1-25) Further info

Word-similarity graph (zoom-in and zoom-out is enabled): Further info

Number of speeches in the topic:

Top 10 NMF words:

1. Lorem

2. Lorem

3. Lorem

4. Lorem

5. Lorem

6. Lorem

7. Lorem

8. Lorem

9. Lorem

10. Lorem

Topic speeches (as % of speeches in the year) Further info

Sentiment cluster distribuiton (vs overall in blue) Further info