PLOS ALM 13, day1
Wed Oct 16, 2013
## Cameron Neylon - Introduction & Welcome
Interesting - this is the first PLOS ALM meeting that is a “normal” scheduled presentation. Time is going to be tight.
Pete Binfield ALM: Looking back, moving forward
A large chunk of OA does not select for impact - this is why ALMs are key for this space. PLOS didn’t invent ALMs - Frontiers were doing it a little ahead of PLOS’s launch. Web of science didn’t tell PLOS until 2010 that PLOS one was being tracked for an impact factor.
People are quoting ALMs in their resumes for applying to be on the editorial board on PeerJ!
Steve Pettifer claims ALMs have helped him to get tenure.
The big thing that is missing is that we have not provided mechanisms to provide readers with recommendations. Filtering mechanisms for ALMs have not been well developed yet.
There has been a period of talking to ourselves, but there seems to a broader interest now.
Why have publishers not yet improved on the display of the PLOS ALM data?
ALMs were built on the assumption that there is a pool of data available via APIs, but already in the last few years a number of sources have already hit the deadpool. (We need to own local copies of our own ALM data, as a publisher - really really important).
Filtering tools were on the backlog from day one - but not yet built.
Best Practices and Standards
## Todd Carpenter - NISO altmetrics project
Probably best to see my notes that I posted.
## Scott Chamberlain - Programmatic access for Altmetrics
Works with rOpenSci - creates open software to make science better.
Mentions raml.org - a markdown file for creating an API. I’ve not heard about this before, sounds very interesting.
Kaitlin Thaney Metrics for web-native science
This was a great talk, but I as I’m quite familiar with a lot of the topics that Kaitlin was discussion, I ended up working on my own slides during this talk.
## Geoffrey Bilder - CrossRef metadata and services in the service of ALMs
Geoff has got to the point where he has started to caution people about the use of DOIs, there is a bit of a cargo cult growing around this.
Brand is important - Geoff does the thing with showing a blacked out paper, we all recognize it.
He says that DOIs are beginning to gain that type of branding. By and large, when we encountering DOIs we tend to think of them as scholarly entities - but every DVD in the world now includes a DOI - these are just being used as a registry, DOIs may escape into the wild.
DOIs are not an arbiter of quality - some articles have two DOI’s, one is the author copy and one is the publisher copy.
search.crossref.org has had a nice facelift - it also has a REST api.
Crossref has citations in their metadata, and an increasingly large number of publishers are starting to do this. They are collecting funder information via fundref, the funder taxonomy is available under a CC0 license. They are increasingly including license information in the form of URIs. They are gathering ORCID information too. Abstracts are starting to be distributed via JATS compatible abstracts (we should do this from within eLife).
They are starting to gather patent info for any patent that cites a paper.
They are starting to look at reverse lookup service for referrals via DOI. (some people played with some of this data at the data challenge hackday on the following Saturday, and it looks very promising and interesting).
I really like the distinction between persist-able and persistent identifier. There is nothing intrinsically persistent about a DOI, it’s intrinsically redirect-able. Persistent doesn’t mean that something will be around forever, but rather that it’s stubborn.
Scholarly Research on ALM
William Gunn - Research assessment using Mendeley readership data
William is going to talk about the Mendeley data set. William talks about wanting to measure units of innovation and impact (I think this is one use cases for ALMs, but I think other use cases, such as domain mapping, are also pretty interesting).
One of the more interesting discussions I’ve had so far at the meeting was last night with William talking about the reproducibility initiative that he is working on with Elizabeth Iorns from ScienceExchange. You can read more about that here.
Zohreh Zahedi - What is the impact of the publications read by the different Mendeley users? Could they be considered as signals of future impact?
It is fantastic to see the ALM community starting to bring in researchers from the bibliometrics community. Anything that we can have that can help with us to understand the robustness of ALMs can only help. They ran this study against 20k randomly selected documents, and then with 200k publications.
This talk This is research in progress, but the answer seems to be that there are advantages to using readers over citations, and the trend is that post-docs and phd students read more highly cited papers than professors. Further testing needs to be done.
## Ehsan Mohammadi - Mendeley readership altmetrics for clinical medicine and engineering
He was unable to get a visa to come to the US. Stefanie will give Ehasn’s presentation instead. There are a number of bibliometritians looking at altmetrics, at the recent bibliometrics conference there were two full sessions on this. This is another talk that compares mendeley reader data to citations in clinical medicine.
There is a significant correlation between Mendeley readers and citations. Cancer his the highest correlation of 0.6. Some fields have a lower correlation, but at some point over the conference it was mentioned that a lower correlation in clinical fields might be expected, where doctors do read, but publish less, and so in a field like that Mendeley may be capturing impact that is not visible through citation counts.
Stefanie Haustein - Empirical analyses of scientific papers and researchers on Twitter: Results of two studies
I think this is similar to the talk that Stefaine gave yesterday. They looked at a large scale analysis of tweets on 1.4M articles from PMC, and they also looked at a small scale analysis of 32 astronomers who are active on twitter.
They divided papers into four classes, based on low vs high coverage of documents from within a disciple against high vs low correlation with citations.
The overall correlation of twitter was low - 0.1, with a 9% coverage of the literature.
“Penis” papers get highly tweeted. (on the topic of tweets, the comment from Euan yesterday is interesting to keep in mind - that authors love to read the actual tweets about their papers).
The in-depth study was also interesting. Those who do not publish, tweet regularly. Those who publish frequently don’t tweet frequently. There is a negative correlation between the number of tweets and the number of publications.
What this tells us is that tweets and followers are definitely a different measure to citations and standard citation metrics.
Overlap between phrases in abstracts and twitter is pretty low. This means that 4.1% of tweets are tweeting bits of abstracts - this seems low, but of the 100 most frequently used terms in abstracts, most of these terms do make it to twitter. (So either tweets are having parallel conversations to the abstracts, or a using a very different shorthand to refer to abstracts.)
# Reporting on ALM
## Andrea Michalek - Alternative metrics in practice
I really like the direction that this product is heading in.
## Adam Dinsmore - An exploratory review of PLOS ALM Reports: a funder’s perspective
There is a great example of Wellcome looking at ALM data for evaluating one of their researchers. They used the PLOS ALM reports tool. Great example about looking at the people who tweeted about a paper on alcohol abuse.
Institutional identifiers would help the Trust.
## Juan Alperin - Visualizing ALMs in d3.js
Does not like:
- monotonically increasing numbers. For a reader it’s harder to see what the actual usage is over time.
- a table of a lot of numbers is hard to use.
- lot of prominence to the sources is not so useful.
- the doughnut is not helpful for the reader - you are only guided by the number.
- the line in the total-impact metric only tells you one number.
- badges from impact story are giving the you the same percentile info.
- does not like squeezing the three types of data into one view - you can’t follow the pdf trend-line.
- XML data does not add much value.
- having a reference set.
- likes being able to get more details from altmetric - being able to see the individual tweets.
- does like the non-cumulative views.
His visualization tries to get to the following principles:
- has a number for the source.
- has a sense of history and not giving you a cumulative view, I like the weekly/yearly switch.
- on the axis the only thing that is showing on the y-axis is the maximum value to drive a sense of history.
- being able to rollup into different time levels.
- few labels - no grid lines.
- having a tool-tip with hover over the data point.
- graphs are the same size, but stacked on top of each other, and scaled accordingly.
- different data sources have different granularity, so it would not be simple to make the graphs switch to different granularities simultaneously
- when there is a small number of events you need to be able to turn off the graphs.
- being able to distinguish between a 0 count, and not having any data for that time.
- be like to be able to turn on and turn off sources, rearrange the stacking.
- should be able to go to a page that contains a more detailed article profile.
They are going to roll this out to all of the journals that implement OJS and that sign up for the ALM service. They are going to run the PLOS ALM tool.
# Heather Piwowar - The altmetrics CV: opportunities and challenges (also featuring Jason Priem)
They think that if you give researchers ALMs then you are putting science back in the hands of scientists (I guess rather than having it in the hands of ISI).