Data Science & Artificial Intelligence · Data Science and Artificial Intelligence at the Scholarly Communications Frontier John Sack Founding Director HighWire Press 27 February

Programme• Data Science & Artificial Intelligence

Data Science andArtificial Intelligence at theScholarly Communications Frontier

John SackFounding DirectorHighWire Press27 February 2018

Topics: AI/DS. Authors/Editors

• AI Applications Overview• AI and Data Science in Publishing

• Some Cautions58

ArtificialIntelligence

DataScience

1 2

3 4

Readers / Authors Editors / Publishers

58



• Some Cautions59


DataScience

1 2

3 4


59

AI writesFinancialReports

60

AI ‘Creativity’:Making Music

61

…but AI Creativityhas its limits…

62

AI ‘Creativity’:Making Music

AI in theLab

63

AI-basedPortfolioManagement

• Could an AI ”edit” ajournal as it canmanage a stockportfolio?

• Discuss.

64



• Some Cautions65


DataScience

1 2

3 4


65

AI for Researchers: Readers/Authors

• “Keeping Up”– Which new articles match my fields?– Which articles are about my specific interests?– Which articles contain something related to my

specific interests?• “Writing Up”

– Draft my newest article for me– Where shall I submit my newest article first?

• ”Dreaming Up”– How might Gene A and Protein C be related?

6666

Keeping Up With My Field

6767

Unsilo article topics

68

(UNSILO Example)

68

Finding a Needle in a Haystack

AI WritesResearch Articles

69

AI WritesResearch Articles

70

WhichJournalShould ISubmitMy PaperTo?

7171

“Dreaming Up”:What

assertionsdoes an article

make?

Constructing”Truth Tables”

7272

“DreamingUp”:What

assertionsdoes an

article make?

SourceDataat

EMBO

7373



• Some Cautions74


DataScience

1 2

3 4


74

HighWire Editors’ Workshop on AI:What Editors Asked Us For

• “Make long meetings shorter”

7575

HighWire Editors’ Workshop on AI:What Editors Asked Us For

• Journal Performance– Visualize the performance of a journal, its context and

competitors• Community Engagement

– Tools to attract and retain the best authors, and thebest papers, in advance of a paper’s submission

• Emergence of the New: fields, topics, authors,labs– Tools to see over the horizon

• Workflow Productivity/Efficiency– Deal with repetitive, expertise-burning tasks

7676

Meta’sPublisher

Dashboard

77

Meta’s“BibliometricIntelligence”to evaluatesubmissions

• Is my journal anappropriate venuefor thismanuscript?

• What is thepotential impactof this paper?

• To whom should Isend this paperfor review?

3/17/2018 HighWire Press -- CONFIDENTIAL7878

Fishing in the Preprint Stream forProspective Papers

– Journals can leverage preprints by findingthe best and most appropriate articles to solicit

– Already happeningmanually

3/17/2018 7979

Meta “Horizon Scanning”: Emerging Concepts

8080

UNSILO Article Categorization Dashboard

8181

Evaluationof FigureIntegrity

8282

“We suggest that agreat deal ofscientific fraud willbe, sooner or later,detectable byautomatic means.”



• Some Cautions83


DataScience

1 2

3 4


83

Data-ScienceGuided Decision on

Which journals toSubmit a paper to:

Is this journala club

I aspireto join?

Data-ScienceGuided

Reading

8484



• Some Cautions85


DataScience

1 2

3 4


85

Impact Vizor:Data-driven Insight into Editors’ Big Questions

• Rejected Articles– Who publishes my rejected articles?– Am I rejecting high-impact articles?

• Published Articles– Which articles are driving my impact up, or down?– Which types of articles are resonating?– Are this year’s articles higher impact than last year’s?

• Cohort Articles– What are the trending topics in my fields?– And who is publishing them?

8686

The Different Faces of Impact Vizors’ Viewers


< RejectedArticleTracker

Section >PerformanceAnalyzer

< Hot Article(& Object)Tracker

Advance >Correlator ofCitations &Usage

< CitationDistributionSurveyor

Cohort >Comparator

87

Which Sections of a Journal are Consistently Below Par?What Groups of Articles Are Cited Above and Below Average

Are New Articles Cited More or Less than Previous Ones?

3/17/2018HighWire Press -- CONFIDENTIAL

8888

3/17/2018 8989

Marketing toauthorsbased on theassuranceOf peer attention

JournalPerformance:Competitor

Streamgraph


Q: Who is publishingIn my fields, andHow much? What isMy “market share”?

90

JournalPerformance:

TopicStreamgraph


Q: Which fields thatI publish in aregrowing, and whichare shrinking?

91

Analytics Support forEvidence-Based

DesignHeat maps

Confetti (click) maps




• Some Cautions– The Hype Cycle, and Validation

93


DataScience

1 2

3 4


93

94

“People tend to

overstatethe short-term potential of new technologies and

understate

their long-term consequences.”- Paul Saffo- Ray Amara, ”Amara’s Law”

95

Validation – Calibration - Confidence• HighWire’s Impact Vizor Validating Meta’s Predictions

– Correlate AI-Predicted Cites to Actual Cites– Examine N, R, R-Squared, and P-value, by Journal,

TOC, Collection

96

Programme• Data Science & Artificial Intelligence

Data Science & Artificial Intelligence · Data Science and Artificial Intelligence at the Scholarly Communications Frontier John Sack Founding Director HighWire Press 27 February

Documents