際際滷

際際滷Share a Scribd company logo

Using TV Metadata to optimise the repurposing and republication of TV Content across online channels @ EBU MDN 2019

ReTV project
ReTV project

The document discusses the ReTV project which aims to optimize TV content repurposing and republication across online channels using TV metadata. It outlines several key services including: 1. Collecting and annotating TV metadata from various sources to power predictive and recommendation services. 2. Developing prediction models for audience metrics, content success, and scheduling based on metadata and external event knowledge. 3. Creating tools to repurpose TV content like clips and summaries for different online channels based on predicted popularity. 4. Developing recommendation services like a chatbot and personalization platform to promote optimized TV content to audiences.

1 of 24
Download to read offline
retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project Using TV Metadata to optimise the re- purposing and re- publication of TV Content across online channels EBU MDN 2019 workshop Geneva, 11 June 2019 Lyndon Nixon, MODUL Technology GmbH 1
retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project https://memad.eu/ai4tv2019/ Submissions open until July 8, 2019. 2
Viewing of linear broadcast TV is decreasing while time spent with digital content on Catchup TV, on-demand OTT or social media rises. Broadcaster audiences are fragmented across digital channels and digital channels are full of competing content offers for their limited attention. The TV industry is still catching up with their online competition in the use of Web technology: interaction, tracking, personalisation and targeting.
retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project TV content needs to be re-born: optimised for each and every channel 4 Global Web Index, The State of Broadcast TV in 2019, Feb 2019. Zenith Media, Media Consumption Forecasts 2018, May 2018.
retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project A Trans-Vector Platform for cross-channel content analysis and publication 5
retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project TV Metadata aggregated from heterogeneous sources & used for data-driven services: - prediction - repurposing of content - recommendation 6
retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project TV data collection EPG data Social media YouTube Twitter Facebook Websites TV/Radio sites Hybrid sites Video archives Europeana 7
retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project TV data annotation Core annotations Person Organization Location ReTV specific profiles TV channel (reference detection) TV actors, characters & presenters Named Entity Linking (NEL) Local identifiers from our own Semantic Knowledge Base Alignment to external Knowledge Graphs, e.g. Wikidata 8
retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project TV data analytics Audience metrics No of viewers by channel / program / topic Content success metrics (by channel) Frequency of mention of a topic Sentiment towards a topic Disagreement around a topic WYSDOM Content success metrics (by source) Reach Impact 9
retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project Video annotation Fragmentation Scene Shot Sub-shot Labeling with visual concepts Training sets (TRECVID-323, ConceptNet) Self-defined (Sandm辰nnchen) Brand and channel logo detection Identify spatial regions with a brand or channel logo (Dis)appearance of channel logos -> ad break detection 10
retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project Event Knowledge Base Available events Significant global events Source: Wikidata by type Examples: sports, political, weather events Public holidays Future recurrences have been calculated Country or region validity is available Local scheduled events Source: iCal Examples: DE/AT/CH league football matches, Cup and European competitions 11
retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project Events in predictive analytics 1. We have analysed what types of events affect TV audience patterns 2. We have associated past events with TV audience variations 3. We can use these associations in training a model for predicting future TV audience variations 1. Future events need to be comparable with past events 2. In audience prediction, we will need knowledge about where & when events are broadcast 12
retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project Keyword-based prediction Weighted list of keywords returned for a date: e.g. News/EN documents since 1/1/19 mentioning October 31 13 e.g. News/EN documents since 1/1/19 mentioning September 20 opening of the Rugby World Cup but also with e.g. film releases:
retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project TV audience prediction 14 1. Extrapolate time series data (audience metrics) 2. Factor in event knowledge
retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project TV content success prediction 15 1. Extrapolate time series data (success metrics) for a topic on a vector 2. Factor in event knowledge 3. Predict future success metric for the topic on the vector
retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project TV content re-purposing 16 1. Video length restrictions (e.g. social media) 2. According to topic(s) (predicted to optimise success) 3. Guided by purpose (e.g. trailer to promote future content, highlights of past content)
retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project TV content recommendation 17 1. Publish on a future date -> which content will have optimal success? 2. Publish selected content -> when and on what channel will it have optimal success? 3. Promote content to consumers -> which content are they most likely to watch? 1. Introduce consumers to content they wouldnt have otherwise watched 2. Keep consumers engaged with the content when they would otherwise not be
retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project Topics Compass: combine analytics and prediction to inform a more successful content publication strategy Topic Compass (scheduling) Visualise topics which are predicted to be popular on future dates Visualises popularity, polarity and communication success of topics on different vectors. ? It will predict future popularity, polarity and communication success. Visualise associations made in online content with a target (such as a TV program)
retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project Content Wizard: re-purpose and re-publish content for optimal reach on every channel Content Wizard (repurposing) Publish on the right channel at the right time for the optimal predicted reach. Selects video clips for combination and re-publication, with recommendations for the vector and time, and adaptations of content to that vector. Select content for publication based on predictions of popularity of that content (Topics Compass). Create content summaries based on (a) channel (e.g. video duration limits), (b) topics of interest and (c) purpose.
retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project 4u2: a chatbot for recommending trending content
retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project Content sWitch: dynamic insertion of personalised in- stream content
retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project @ReTV_EU Facebook: ReTVeuwww.ReTV-Project.eu Instagram: retv_project TV isnt dead, its reborn. https://www.thinkwithgoogle.com/data/millennial-tv- consumption-statistics/ Jan 2018
The ReTV Stakeholder Forum is your opportunity to engage with us, be first to get updates and test the services and tools! Send a mail to: Prototypes of the scenarios can be seen here at EBU MDN during all the demo breaks! Reports on the scenarios and first evaluations with users will be available by end of September 2019 info@retv-project.eu
retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project Dr Lyndon Nixon nixon@modultech.eu MODUL Technology GmbH 24 This project has received funding from the European Unions Horizon 2020 research and innovation programme

More Related Content

Using TV Metadata to optimise the repurposing and republication of TV Content across online channels @ EBU MDN 2019

  • 1. retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project Using TV Metadata to optimise the re- purposing and re- publication of TV Content across online channels EBU MDN 2019 workshop Geneva, 11 June 2019 Lyndon Nixon, MODUL Technology GmbH 1
  • 2. retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project https://memad.eu/ai4tv2019/ Submissions open until July 8, 2019. 2
  • 3. Viewing of linear broadcast TV is decreasing while time spent with digital content on Catchup TV, on-demand OTT or social media rises. Broadcaster audiences are fragmented across digital channels and digital channels are full of competing content offers for their limited attention. The TV industry is still catching up with their online competition in the use of Web technology: interaction, tracking, personalisation and targeting.
  • 4. retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project TV content needs to be re-born: optimised for each and every channel 4 Global Web Index, The State of Broadcast TV in 2019, Feb 2019. Zenith Media, Media Consumption Forecasts 2018, May 2018.
  • 5. retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project A Trans-Vector Platform for cross-channel content analysis and publication 5
  • 6. retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project TV Metadata aggregated from heterogeneous sources & used for data-driven services: - prediction - repurposing of content - recommendation 6
  • 7. retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project TV data collection EPG data Social media YouTube Twitter Facebook Websites TV/Radio sites Hybrid sites Video archives Europeana 7
  • 8. retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project TV data annotation Core annotations Person Organization Location ReTV specific profiles TV channel (reference detection) TV actors, characters & presenters Named Entity Linking (NEL) Local identifiers from our own Semantic Knowledge Base Alignment to external Knowledge Graphs, e.g. Wikidata 8
  • 9. retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project TV data analytics Audience metrics No of viewers by channel / program / topic Content success metrics (by channel) Frequency of mention of a topic Sentiment towards a topic Disagreement around a topic WYSDOM Content success metrics (by source) Reach Impact 9
  • 10. retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project Video annotation Fragmentation Scene Shot Sub-shot Labeling with visual concepts Training sets (TRECVID-323, ConceptNet) Self-defined (Sandm辰nnchen) Brand and channel logo detection Identify spatial regions with a brand or channel logo (Dis)appearance of channel logos -> ad break detection 10
  • 11. retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project Event Knowledge Base Available events Significant global events Source: Wikidata by type Examples: sports, political, weather events Public holidays Future recurrences have been calculated Country or region validity is available Local scheduled events Source: iCal Examples: DE/AT/CH league football matches, Cup and European competitions 11
  • 12. retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project Events in predictive analytics 1. We have analysed what types of events affect TV audience patterns 2. We have associated past events with TV audience variations 3. We can use these associations in training a model for predicting future TV audience variations 1. Future events need to be comparable with past events 2. In audience prediction, we will need knowledge about where & when events are broadcast 12
  • 13. retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project Keyword-based prediction Weighted list of keywords returned for a date: e.g. News/EN documents since 1/1/19 mentioning October 31 13 e.g. News/EN documents since 1/1/19 mentioning September 20 opening of the Rugby World Cup but also with e.g. film releases:
  • 14. retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project TV audience prediction 14 1. Extrapolate time series data (audience metrics) 2. Factor in event knowledge
  • 15. retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project TV content success prediction 15 1. Extrapolate time series data (success metrics) for a topic on a vector 2. Factor in event knowledge 3. Predict future success metric for the topic on the vector
  • 16. retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project TV content re-purposing 16 1. Video length restrictions (e.g. social media) 2. According to topic(s) (predicted to optimise success) 3. Guided by purpose (e.g. trailer to promote future content, highlights of past content)
  • 17. retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project TV content recommendation 17 1. Publish on a future date -> which content will have optimal success? 2. Publish selected content -> when and on what channel will it have optimal success? 3. Promote content to consumers -> which content are they most likely to watch? 1. Introduce consumers to content they wouldnt have otherwise watched 2. Keep consumers engaged with the content when they would otherwise not be
  • 18. retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project Topics Compass: combine analytics and prediction to inform a more successful content publication strategy Topic Compass (scheduling) Visualise topics which are predicted to be popular on future dates Visualises popularity, polarity and communication success of topics on different vectors. ? It will predict future popularity, polarity and communication success. Visualise associations made in online content with a target (such as a TV program)
  • 19. retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project Content Wizard: re-purpose and re-publish content for optimal reach on every channel Content Wizard (repurposing) Publish on the right channel at the right time for the optimal predicted reach. Selects video clips for combination and re-publication, with recommendations for the vector and time, and adaptations of content to that vector. Select content for publication based on predictions of popularity of that content (Topics Compass). Create content summaries based on (a) channel (e.g. video duration limits), (b) topics of interest and (c) purpose.
  • 20. retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project 4u2: a chatbot for recommending trending content
  • 21. retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project Content sWitch: dynamic insertion of personalised in- stream content
  • 22. retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project @ReTV_EU Facebook: ReTVeuwww.ReTV-Project.eu Instagram: retv_project TV isnt dead, its reborn. https://www.thinkwithgoogle.com/data/millennial-tv- consumption-statistics/ Jan 2018
  • 23. The ReTV Stakeholder Forum is your opportunity to engage with us, be first to get updates and test the services and tools! Send a mail to: Prototypes of the scenarios can be seen here at EBU MDN during all the demo breaks! Reports on the scenarios and first evaluations with users will be available by end of September 2019 info@retv-project.eu
  • 24. retv-project.eu @ReTV_EU @ReTVeu retv-project retv_project Dr Lyndon Nixon nixon@modultech.eu MODUL Technology GmbH 24 This project has received funding from the European Unions Horizon 2020 research and innovation programme