React

React, an interesting new academic development in visual search. It works on the reverse-search principle: upload a picture, and see similar pictures in the results.

The prototype limits results to a couple of the UK’s larger national digitized art collections (National Archives, the V&A) and leavens these with the Edinburgh Botanic Garden for some flowers and curious pods and suchlike. An AI assists the “does it look like this…?” sorting.

The COAR of the issue

A useful new analysis today from COAR, “Don’t believe the hype: repositories are critical for ensuring equity, inclusion and sustainability in the transition to open access”. Recent…

publishers’ comments portray gold open access as the only ‘legitimate’ route for open access, and attempt to diminish the repository (or green) route.

According to the author, some publishers are even implying that repositories have no aggregators, or are not present in Google Search or in specialist search-engines such as Scholar and GRAFT. Laughably, they apparently suggest that poor over-worked researchers will instead…

need to search through individual repositories to find the articles.

The publishers are also said to be trying to stop all but a sub-set of elite repositories from being used for data deposit, via…

proposing to define the repository selection criteria for where their authors’ should deposit research data. These criteria, which are very narrowly conceived, threaten to exclude thousands of national and institutional repositories as options for deposit.

Again, this sounds like it is designed to make researchers feel it’s more convenient to publish their article + data via a big publisher.

Subject to change

“Subject indexing in humanities: a comparison between a local university repository and an international bibliographic service”, Journal of Documentation, May 2020.

… the use of subject index terms in humanities journal articles [is] not supported in either the world’s largest commercial abstract and citation database Scopus or the local repository of a public university in Sweden. The indexing policies in the two services do not seem to address the needs of humanities scholars for highly granular subject index terms with appropriate facets; no controlled vocabularies for any humanities discipline are used whatsoever.

Internet Archive Scholar is live

Internet Archive Scholar, formerly the Fat Cat project, now live and purring. Full-marks for having that rarest of sidebar search-filters, “OA”, though “Fulltext” is presumably broader and thus the one most likely to be used most. It’s also great to see there’s now a keyword-based way to search across all those microfilm journal runs that Archive.org has been uploading recently.

I wouldn’t have used the open ISSN ROAD as a source, nor visually implied that it’s a possible quality-marker. But at least it’s being balanced against the more rigorous DOAJ, and there’s a yes/no flag for both services on the article’s record-page…

It’s good that the “Read full-text” button goes to a PDF copy at the WayBack Machine, and yet there is also a live link on the record-page that serves to keep a record of the source URL.

Not all record pages have full-text, though these are very rare. In which case the user is prompted to find and save…

Unfortunately IA Scholar doesn’t appear to respect “quote marks” in search, which is not ideal for a scholarly search engine. For instance a search for “Creationism” defaults to results for “creation”. Nor can it do Google-y stuff like intitle: or anything similar via the sidebar, though I guess such refinements may be yet to come. Update: the command is: title:

A quick test search for Mongolian folk song suggests it’s not wildly astray in terms of relevance. It’s not being led astray by ‘Song’ as a common Chinese author name, for instance, or mongolism as a genetic disease.

How far will Google Search index the fatcat URL? Will they block it from results in due course, for being too verbose and swamping results? Or just tweak the de-duplication algorithm to suppress it a bit? Well, they’re indexing it for now, and as such it’s been experimentally added to JURN. It may well come out again, but I want to test it for a while. If Google Search fully indexes, that should theoretically then give JURN users a way into all the microfilm journal-runs that Archive.org that has recently been uploading.

Report: Equitable access to research in a changing world

Released in June 2020, a new consultancy report titled “Equitable access to research in a changing world: Research4Life Landscape and Situation Analysis”. This surveys the pressures on the Research4Life aid programmes. Established 20 years ago, Research4Life gives developing countries “free or low-cost” online access to journals and books from some 175 publishers. Along with other aid initiatives, this means that African universities often have better free access to journal databases than do some academics in advanced nations. The new report makes no recommendations, but a key point to note is that…

… some of the most relevant and influential research undertaken in low-and-middle income countries happens outside academia: in specialised research institutes, think tanks, or government-backed research agencies. In some countries, research agencies and institutes conduct research in national priority areas and have direct access to and influence on decision-makers” [yet] “these non-governmental organisations have in the past been excluded from open access debates, and may be unable to take advantage of initiatives such as Research4Life.

It could be useful to quantify that “may”, through further research. Do developing nations find roundabout ways to include their research agencies in Research4Life, such as giving off-campus agency researchers special log-ins to access the national university system? Or are such arrangements rather moot, in the age of open-access and Sci-hub? If not, would there be a real benefit if Research4Life were to be extended to bona fide government research agencies and suitable NGOs? How much would such an expansion actually cost, and what could the returns be in such nations?

What to do about DocFetcher?

Update: DocFetcher Pro now available and stable at 31st May 2021, with embedded Java, for $40 via Gumroad.

The freeware desktop file-indexer and keyword searcher DocFetcher has been sporked by the Java runtime update, specifically failing to launch due to an error with the JIntellitype64.dll file. The code archive for this file suggests similar problems for others in the past. And the comments at SourceForge suggest other are finding the latest Java (mid July 2020) repeatedly crashes DocFetcher. Apparently it’s also causing problems for several other bits of software.

The fallback is not the official portable version of DocFetcher, sadly, which has the same problem. Nor is falling back to an earlier version of DocFetcher. Nor is the solution to download and install the latest 64-bit Java for Windows again. It appears that the old 32-bit software just doesn’t play nicely with the latest mid-July Java. This is confirmed by a comment buried on SourceForge from the developer…

“A proper bugfix for DocFetcher won’t be available until 2021, so for now downgrading to Java 8u251 is the only workaround”.

But by that time the software will be “DocFetcher Pro” and $50 paid for a perpetual licence. Ah well. Still, that’s good value compared to dtSearch, and is not a subscription like Copernic Desktop. But… $50. So, an alternative freeware option will soon be needed. I took a look…

1) There is Recoll on Windows, which looks like it’s halfway there, but it costs 5 Euros. That’s not viable if you were wanting to distribute a bit of full-text search freeware with the archive of a large defunct technical forum. Still, by 2021 it might have developed further. (Update: the maker has commented, noting it’s GPL and copies may be freely redistributed).

2) The developer of the freeware AnyTXT Searcher has been knocking the rough edges off it and expanding file types, over the last year. But, while it bills itself a “Google Desktop Search Alternative” is still appears not to have any sort of acceptable in-file preview on its search results. The other problem is that its start-up time is extremely slow. Several minutes, rather than seconds. You expect that of behemoths like Photoshop, but not of a little Windows utility. Plus it appears to be “all or nothing”, and there’s no ability to index just a few folders. Uninstalled.

3) Another possible choice is Exselo, said to be very powerful and yet also free desktop search. But… like DocFetcher it’s Java based. Plus, it’s Registerware and “Invites are sent to friends” (register via Facebook?). It’s a system-hog, and it stops working after 14 days if you don’t accept automatic updates. The developers were obviously hoping to sell it on, and lacking a buyer are now pitching it as a trendy “secure chat environment”? Blugh.

4) The old standby Copernic Desktop has become slightly better. The ‘last good’ wholly free version was 2.30 build 30 (no Deskbar feature on Windows 64-bit, PDF manual here). The current 2020 free version still has no .PDF or Word support, but the 10,000 file limit has now been raised to 25,000. It also has a new $15 “knowledge worker” edition, but that just turns out to be a “per-year subscription”. It’s now Registerware, even to just download the Trial. Also requires big .NET Framework downloads, which the 2.3 doesn’t. Thus it’s not feasible as freeware to distribute along with a large forum archive.

5) The old 2010 Multifind would be a good choice, if only it built an initial index and was thus fast. For some, the lack of a requirement to build an index may be a feature not a drawback. Despite its slowness due to a lack of an index, it can find and display text inside files. And it’s genuine old-school Windows freeware and has a tiny footprint. If you wanted to make something to fill the freeware gap that’s looming with the loss of DocFetcher, you might do worse than buy the rights to this and start developing it again.

So it’s back to DocFetcher. One can’t go back beyond DocFetcher 1.1.20, as that was when it started indexing HTML with no body element (e.g. RSS-feed forum-threads archived in XML and re-named .HTML), and anyway that doesn’t fix the problem. So it looks like the only real solution to get DocFetcher working is the downgrade to Java SE Runtime Environment 8u251 (jre-8u251-windows-x64.exe), which is a security risk unlikely to be welcomed by those who just want a free search tool for use with their forum archives. Perhaps what’s needed is to make a truly portable DocFetcher, which never has to call on the Windows system’s Java runtime?

Towards a meta-pedia browser add-on

An interesting idea: a meta-pedia browser add-on, to consult all public encyclopedias at the same time. Presenting the results as an elegant full-screen dashboard with a strip of side-links to Google Books, Archive.org, Scholar, JURN etc. Ideally with configurable sources…

* Wikipedia
* Current Britannica
* 1911 Britannica
* Specialist public encyclopedias is they exist for the topic, e.g. Philosophy, Catholic, Science-fiction etc.

I’m assuming this would need to be a browser add-on, as a cloud service that did this would face lawsuits and frame-busting scripts. The closest I can find is 2019’s free ResearchKit which shows Wikipedia and the current Britannica side-by-side, above bot-driven auto-summaries of their text. It’s not exactly elegant to look at it, but it works.

Obviously some fuzzy-lookup might be needed to align search topics, though the individual encyclopedias strive to do that on their pages via navigation strips and links.

But rather than jumping straight to a presumed page, perhaps each encyclopedia panel might first show sub-panels with a half-dozen ‘possible’ hits, colour-shaded by order of likely relevance to the search. If such a browser addon was in widespread use, the data gathered from such mass human-driven topic-selection/alignment might be rather useful, over time being judiciously used to augment existing ‘knowledge navigation trees’ that are able to cope at a meta-level with shifting topic titles (e.g. Aetheopia > Abyssinia > Horn of Africa > Eastern Africa > Ethiopia).

Another way to do it might be for the addon to ‘read’ such existing navigation on the encyclopedia pages, make its own deft distillation of such, and then use that to ‘prime’ with keywords the sidebar links to Google Books, Archive.org, Scholar, JURN etc.

New from Google Research

Google Research has launched COVID-19 Research Explorer. This has “a semantic search interface” that enables better search and discovery across “more than 50,000 journal articles and preprints”.

A new group test of search

Which Academic Search Systems are Suitable for Systematic Reviews or Meta-Analyses? Evaluating Retrieval Qualities of Google Scholar, PubMed and 26 other Resources, October 2019.

“Our tests revealed that the help files of numerous search systems promise a Boolean search functionality that our tests could not verify. These findings were especially alarming because users of such systems rely on functionalities that they assume work properly, but that may not be the case.” … “our results contradict systematic review guidance that assumes that “all the search engines in some way [would] permit the use of Boolean syntax operators to expand or restrict the search””.

Regarding… “full Boolean search strategies” the authors also noted that “Google Scholar [does] not offer such functionality”. The word “full” here is the critical word, and indicates that NOT is still a missing operator for Google Scholar.

For open access, this new test concludes that those outside of biomedical research… “are limited to the multidisciplinary system BASE” for discovering open access material, but that unspecified… “other open, or partially open search systems that fail to meet the criteria for query-based search might still be useful for supplementary search methods.”

Review of Cabell’s Predatory Journal Blacklist

A new review of a paywalled up-to-date blacklist of predatory journals, “Cabell’s Predatory Journal Blacklist: An Updated Review”, at the Scholarly Kitchen.