Upload
brianbishop
View
57
Download
0
Embed Size (px)
Citation preview
Crowdsourcing citation suggestions, without needing to ask the crowd
Brian Bishop, Co-founder@mochasteak
Brian BishopBusiness / Product stuff
Formerly VP of Platform Development at Springer,
MBA,14-year scientific
publishing veteran
Pawel KowalskiTech stuff
Tech Principal at iterativ GmbH, full-stack dev, MSc in
Computer Science, the whole deal
Our Team
02
Timeline*
03
Prototype goes live
November 2014
Today
January 2015
Company registered
August 2014
*timeline not to scale
We asked 1,044 authors
“What is the most difficult or frustrating part of writing a paper?”
and this is what they said…
05
reading between the tags:
06
How painful is this problem?
07
836Authors
surveyed
“How long did it take you to gather all the references / find sources for your paper?”
25%“More than a week”
35%“More than a month”
That’s a lot of time spent not doing research
The average scientific paper has 35 citations
09
This is the citation context
Each of those citations is embedded in a sentence
“Both experimental and atomistic simulation results show that when the dimensions of the structures become small then the ‘size effect’ has significant role in the mechanical properties (Ruud et al. 1994).”
10
Now go the other way
Paper A
Paper B
Paper C
11
We can extract out all the citing contexts from all the papers, and group them according to the document being cited.
A database of contexts
A clinical implementation was described first for total mesorectal excision (TME) in the treatment of rectal cancer [4, 5]. | The circumferential rectal margin (CRM) was assessed according to the method of Quirke et al. [13], and a margin of < 1 mm was considered CRM-positive. | Complete mesocolic excision (CME) with central vascular ligation (CVL), according to the sound principles of total mesorectal excision (TME) [6, 7] for rectal cancer | A standardized routine pathology examination was performed using the protocol of Quirke et al. [36]. | The surgical specimens were handled according to standard clinical practice as advocated by Quirke et al. [20] and were pathologically examined in accordance with the Tumor Node Metastasis staging system. | For rectal cancer, the specimen was processed using the slicing technique as described by Quirke et al. 33 | Pathologists were trained to examine the specimens according to the protocol of Quirke et al. regarding the circumferential resection margin (CRM), lymph nodes, and dissection plane.12 | Quirke showed a linear correlation between the development of a local recurrence and an inadequate resection with positive circumferential margins [14, 22]. | The pathologists were trained to identify Circumferential Resection Margin (CRM), positive nodes, and lateral spread of tumor according to the protocol of Quirke et al. [14]. | CRM was measured according to the guidelines of Quirke et al.[37]. | Since 2002, all pathology examinations for rectal cancer have been performed according to the guidelines of Quirke et al. [15]. | Total mesorectal excision (TME) removes the primary tumor with its surrounding mesorectum as an intact package, preventing residual tumor cells in the mesorectum from developing into local recurrence.1,2 | Incomplete resection of the lateral tumor margins is now considered the most important cause of local recurrence [15–17]. | In a study by Quirke et al. [15], 83% of the patients with a positive CRM had local tumor
12
Document X What everyone said about Document X
Example keywords
ana standardized routine pathology examination
preventing residual tumor cells
complete mesocolic excision
total mesorectal excision
embryologic tissue planes
radical oncologic resection
standard clinical practice
circumferential rectal margin
pathology examinations
inadequate resection
0 5 10 15 20 25
13
Where are we today?
230,000 fulltext
articles indexed
3.6 million
articles with a citation
8 million
citing contexts
14
Sources
…and you! (If you’re a publisher)
15
Now for the really fun part
Our Service
PROCESSINGForeCite breaks their text down into sentences (not a trivial task) and compares each sentence against our database
17
UPLOADAn author has written (or is
writing) their paper. They upload it to ForeCite for
analysis
RECOMMENDATIONSFinally, we provide a number of different recommendation options (with scoring) along with the citing context for users to compare against their text
ALGORITHMThere are many smart ways in which texts can be compared. So far we have only scratched
the surface. Hey, it’s early days.
Example Recommendations
Suggested citationsSentence with
suggested citations
What others wrote when they cited the suggested article, inline for comparison purposes
What people said about Article X
One of the features of ForeCite is that
you can look up an individual article and
see all of the citing contexts that we
have for this article
Benefits
20
Improves quality
Saves time
Enables discovery