Error rates for Google Scholar citation parsing

Another new prodding of Google Scholar, this time from the latest First Monday “Testing Google Scholar bibliographic data: Estimating error rates for Google Scholar citation parsing”…

While data quality is good for journal articles and conference proceedings, books and edited collections are often wrongly described or have incomplete data. We identify a particular problem with material from online repositories [where there appears to be] considerable inhomogeneity in the implementation of data standards [and] a mismatch between repository software and the harvesting protocols employed by Google Scholar.

One of Scholar’s other problems is that it includes Google Books results. While 30% of the time its Google Books inclusions can useful, there is no way to exclude Books results. One might want to exclude because Scholar still can’t seem to determine a proper book from a robot-produced shovelware ebook that assembles public-domain content. Scholar has no ‘edition authority’ which states that the Joshi-edited and annotated Penguin Classics edition of H.P. Lovecraft’s “Dexter Ward” is the gold-standard and that it has a text that has been fully corrected of the many textual errors, omissions and editing mistakes of previous decades. Unlike the public-domain shovelware ebooks that flood Amazon and (often) Google Books.

A basic undergraduate level search, for instance, for Lovecraft “Dexter Ward”, demonstrates the problem on the first page. Joshi is nowhere to be seen, and the searcher is hammered by links to shovelware ebooks (or worse), often with citation counts that suggest they are legitimate.

Advertisements

Google Scholar at 389 million

Michael Gusenbauer, “Google Scholar to overshadow them all? Comparing the sizes of 12 academic search engines and bibliographic databases”, Scientometrics, November 2018.

The findings provide first-time size estimates of ProQuest and EBSCOHost and indicate that Google Scholar’s size might have been underestimated so far by more than 50%. By our estimation Google Scholar, with 389 million records, is currently the most comprehensive academic search engine.

With the later proviso that there are likely to be many duplicates and near-duplicates, with such tools reporting…

the number of all indexed records on a database, not the number of unique records indexed. This means duplicates, incorrect links, or incorrectly indexed records are all included in the size metrics provided by ASEBDs.

As you can see, the article coins the ugly and unreadable “ASEBDs” for “academic search engines and bibliographic databases”. MASTs might be more mellifluous — Massive Academic Search Tools.

News you can loose…

It looks like stories from U.S. news outlets, those that blank UK and European visitors, are now simply being removed from the Google News results. Spotted today, under the Google News results…

Annoying for those inclined to turn on their VPN and see the news story regardless. But, in practice, the publications still blocking overseas visitors are such low-grade regional newspapers that it’s no loss.

Clips and flicks

All U.S. film-makers can now crack anti-copying technologies on content ($ paywalled at law.com), if they need that content for ‘fair use’ use in a new production…

“Digital Millennium Copyright Act (DMCA) exemptions aren’t just for documentary filmmakers any more. The U.S. Copyright Office and Library of Congress last week broadened a DMCA exception to now allow more filmmakers to circumvent anti-copying technology and rip short video clips for purposes of commentary and criticism.”

However, it isn’t a free-for-all. Note that the PDF for the rules states that this new measure is specifically for…

“where the clip is used for parody or its biographical or historically significant nature”.

In a drama movie, the “commentary and criticism” would thus presumably be seen to be implied by the nature of the scene, rather than done in a directly academic or journalistic manner. For instance, I can imagine a dramatised scene of dancing on the beach as the Apollo 11 rocket lifts off behind the dancers. This scene would be a sort of implied commentary on the optimism engendered in the nation by the historically significant moment of sending men to the Moon. And if the high-res source needed for that was only available from Time-Life rather than NASA, then their Blu-ray disc could be cracked and a clip used as the background in the composite. Actually these days it’s probably easier to do it with 3D models and copy of Vue, but some may want the original footage — and historical personages can’t simply be conjured up in the same way.

Also, as the word “clip” is used and video is assumed in the PDF’s text, that leaves hazy the cracking of content protection to obtain a high-res still picture. A film-maker might need such a still for a Ken Burns “pan and scan” type film, and could perhaps argue that the still was required as a irreplaceable source needed to make the film’s video “clip”. But that’s probably something to be clarified in a future round of rule changes.

Retraction Watch Database

Retraction Watch now has a unified database of retracted papers.

Text Cleanup 2.0 – now free

I’m pleased to see that Text Cleanup 2.0 is now freeware. It’s Windows desktop software from 2003 that “fixes” text automatically when you copy-paste it. For instance, by unwrapping a chunk of text that has hard line-breaks. Text Cleanup has a nice balance of power and ease-of-use, can save user presets, and still runs fine on a Windows 8.x desktop.

The Art Institute of Chicago’s CC0 pictures

The Art Institute of Chicago now has 44,000 items from its collection downloadable as pictures under a CC0 licence. I did a test search for cat. What struck me first was the rich range.

My excitement was dampened when I realised that most of these results had no hi-res download. What I should have done was spotted the easy-to-miss faded “filters” button, up top, which when clicked pops out a sidebar. In the sidebar you can tick to filter by “Public domain”, which gives you the results with the downloadable images.

The filtered results are still fairly impressive, but of course lack the nicer “wow” illustrations made after about the 1910s. Some images download without file extensions, possibly because they already have a . in their title (e.g. “Honorable Mr. Cat”)…

The pictures seem to mostly be around 2,000 to 3,000px and 96dpi. There’s no sign-up needed, and access is free and public.

Audioburst Search

A new podcast search is being heavily hyped by the usual auto-parrots.

But it doesn’t do well on my quick test search for Tolkien. On a simple search for the most popular author in the world, Audioburst Search doesn’t yet appear to be any kind of contender in podcast search…

Compared to ListenNotes, which offers over 5,000 sortable results…

More interestingly, Spotify is gearing up to enter the podcast search space.

NAVER Academic

NAVER Academic is an attempt at a Google Scholar-alike for South Koreans. It’s fast and neatly designed, and is interesting because (unlike Scholar) it has a “Free” filter for its search results.

My quick test search suggests it’s mainly interesting for those seeking non-English content from the region. A search for mongolian folk song filtered by “free” had 19 results. Six of these had English titles and were checked…

1. English abstract. No full-text.
2. English abstract. No full-text.
3. Very short English abstract. CORE record. No full-text.
4. Short English abstract. CORE/JSTOR record. No full-text, DOI link led to 20 Euro paywall.
5. Very short English abstract. CORE record. Link led to full-text PDF “Music Classes of Elementary School in Mongolia” in Japanese.
6. English abstract. No full-text.

Of the 19, only one was actually free. Thus the main problem here is that “free” does not = “full-text”, unless you have academic log-ons within South Korea.

Open and Closed Articles in Norway

“Grades of Openness. Open and Closed Articles in Norway” (August 2018)…

Based on the total scholarly article output of Norway, we investigated the coverage and degree of openness according to three bibliographic services: 1) Google Scholar, 2) oaDOI by Impact Story [now called Impactstory], and 3) 1findr [formerly oaFindr]. According to Google Scholar, we find that more than 70% of all Norwegian articles are openly available. However, degrees are profoundly lower according to oaDOI and 1findr, respectively 31% and 52%.

open shares vary considerably by discipline, with … the Humanities at the lower end