Multiplicity of tags
Sooner or later something will have to be done about the crazy wilderness of tags. First, there are the totally useless ones. "Family Notices" helps nobody; it is far too general. Neither do tags such as "Murders", "Horses", "Deaths", with no date, place or names.
Then there is the question of the multiplicity of tags referring to the same individual. There may be instances when two or three tags might be justified for somebody who fulfilled a number of roles over a long period of time. Mostly, I check to see whether the person I want to enter already has a reasonably sensible tag, then follow the form that has the largest number of entries, unless it seems to be obviously inadequate or misleading.
Just now, I wanted to tag the Rev Samuel Marsden, having found an untagged reference to the first sermon that he preached on arriving in Sydney. There were seven tags for him (including the obviously erroneous reference to him as "Reverand").
Please, can the powers-that-be do SOMETHING to sort out this mess? Maybe a group - either of administrators or experienced contributors - organised with power to alter tags when such a silly situation is notified? Perhaps with a courtesy notification to persons whose tags might be altered or deleted, if they are still working on Trove?
As time passes, the tag situation will only become more and more unmanageable.
I agree with the uselessness and over abundance of some tags to describe the same thing - I too have been guilty of 'useless' tags in the beginning, which I am gradually beginning to delete as I come across them again. Some definite standard form of tagging needs to be or should be implemented - something along the lines of perhaps museum standard formatting such as a three part tag (i.e. a primary, secondary and tertiary tag) - something along the lines of say Government-Education-School etc. Or Book-Reference-Encyclopedia etc. Definitely something that is standardised.
These are just examples.
Maybe Trove can develop a self-generated list of terms or perhaps adopt some appropriate museum standard tagging system. This list may somehow be able to put on the Trove page as a drop down menu, whereby taggers can choose the appropriate three tags to describe the particular newspaper article. This also ensures there will be no mis-spellings as well. Surely the Natinal Library of Australia would have their own certain tagging system that is appropriate for their collections.
This is just my thoughts on the matter and some of this may not be practical on some levels, but surely something needs to be done to eliminate the over abundance of useless and inappropriate tags.
Do you know about the Tagging Guidelines available here?
These provide some guidance but are a bit out of date. There might be a bit more consistency in tagging if more people read these.
The thing with Tags is that they are there to help find things, and there is really no way for you or anyone else to know how the the 'tagger' was intending to use their tags. So a tag that may appear useless to you may well be extremely useful to someone else in the way they intend to use it. Just to take some examples:
- "Family Notices" and "Advertising" are Tags that are used by several correctors (as recommended in the Text Correction Guidelines) in cases where family notices appear, mistakenly, under an advertising category; or when advertising is all that appears, mistakenly, under a family notices category. This will help when specifically searching for family notices, so that you can, for example, search just the Family Notices category and then search the Advertising category, but restrict the search of Advertising to that which has also been tagged with "Family Notices". There are actually a large number of Family Notices mistakenly misclassified by Trove into the Advertising category.
- You say that Tags like "Murders" and "Horses" help no-one. This is simply not true; Tags used consistently in this way can be extremely powerful when used across large numbers of newspapers and a broad range of time.
For a very good example of this check out the Tagging project being run by the 'South-East Australian Recent Climate History' project (SEARCH).
Using a consistent set of Tags like these across a large range of papers and years can produce valuable statistics and research insights. Their aim in this case is to fill the gap in our understanding of climate history before consistent, modern, widespread meteorology was recorded in Australia (only from about 1900). I use their Tag list quite often on anything that is weather/climate related.
In another post we have discussed with Mark the issues around how to Tag people's names.
It is difficult to know what the best way is to Tag peoples names, and as I concluded in the other post it probably depends on who it is you are Tagging, and who you want to find the Tag, as to what is the best approach. There is also the issue of how to Tag people who have used multiple names over time (through marriage, professional names, or deception).
As you point out people are currently Tagging names in all sorts of inconsistent (and sometimes incorrect) ways.
The other thing with Tags is that there are Public Tags and Private Tags. There seems to be a lot of people making Public Tags that probably are truly of no use to anyone else but themselves, and in these cases they should have used a Private Tag.
The catch is that Private Tags are not currently searchable, but Public Tags are. You can restrict a search to Public Tags only by using the 'publictag:' search, e.g.:
But this will only find tags that exactly match "Cyclone Tracey", so you can use a wildcard(*) to find all cyclone Tags, for example:
The following thread provides another useful way to use the 'publictag:' search to help you find things you have 'not' tagged.
I think it would be practically impossible to clean up other people's Tags because, as I said, you cannot know what their intention was for that Tag. I suppose an administrator could possibly fix up obvious spelling errors, but even that opens up a can of worms. What appears to be spelt incorrectly to one person (especially with names), may in fact be correct - the other person may have been better informed than you assume. So the same principal should probably apply as does for newspaper corrections - i.e. leave it as it is, even if you think you know it spelt incorrectly.
All these Tags might look a bit messy, but the aim is to increase the searchability and access to the information hidden away in the newspapers. So more Tags can only be better for searching. More consistency in the Tagging would definitely be better, so Taggers should try to be a bit more consistent with other Tags (as you are doing), but even if they are not it will not be making things worse (just not as better as it could be).
The suggestion for Trove to develop a set of suggested Tags would certainly help in creating more consistency in Tagging. The down-side is that by itself it would never be sufficient to meet the needs of all researchers who could have all sorts of unknown objectives, etc. For example, peoples names - as discussed in the other thread, how you tag them may depend on who they are in themselves as well as what they mean to the researcher (a relative, a politician, a scoundrel?).
Last edited by Spearth; 01-02-2012 at 10:33 PM.
The reason why I say that a tag such as "Horses" is useless: This would retrieve anything from the winners of the Melbourne Cup or the Grand National for as long as they have been running (or in fact the predictions and results of just about every race meeting held in Australia and many from overseas), the breeding of horses for the Indian army, dressage events at the Olympic Games, the training of Lippizaners (spelling?), the making of children's toys, the invasion of Poland, the Charge of the Light Brigade during the Crimean War, the use of cavalry in the Peninsular War, the use of pit ponies in mines, diseases of horses - I just don't see how it would really help anybody. It needs to be more specific. If all entries re horses were tagged that way, they would run into thousands. I think that it is probably possible to limit such a search to what you really want, but then, why bother with "Horses" in the first place?
I get the point re "Family Notices", but in that form I do not at the moment see how it achieves its purpose.
I have some doubt about one of the tags I have been using: "Afghan Wars". I see that somebody has been using "First Afghan War", which is more specific. I had in mind a common thread linking the conflicts in Afghanistan (or Affghanistan as it was written in those days); besides this, I was not sure of the parameters of each war in Afghanistan, to say nothing of the smaller conflicts between the big wars. I dare say that there is nothing really wrong in doubling up on tags such as these.
@ Spearth: Re the climatehistory site you mention: it is not going to help them much if location and/or date are not mentioned in the climate tags. What looks reasonable when you tag it is not always useful when somebody wants to search. That group does not want to know about floods in the Ganges Delta or droughts in East Africa or snow in the Swiss Alps. That is what they could get if the tags that they suggest are used just as they are.
(PS: it might help them if they wrote "lightning" correctly. "Lightening" is what foils and bleach do to hair.)
I totally agree with the idea of a primary, secondary and tertiary tag format. This is basically known as a triple tag, or "machine tag," and is commonly incorporated by sites such as Flickr to make more links, more accurately.
It takes the syntax namecode: predicate=value. So for example, if I wanted to more accurately organise my tags for newspaper articles on cyclones (see Spearth's post—let's take cyclone tracy as an example) by name and geography, I could tag tracy:geo=-12.384354,130.848784. Tagging an article in the Australian Women's Weekly on the impact of tracy on a house in the Darwin suburb of Nightcliffe, I can give much more information about the article to others without having to read the whole thing again. I could even "layer" machine tags on top of one another, such as damage etc.
Even more exciting, using an API such as the unofficial one developed by Tim Sherrat, I could then scrape all of the tags for all the cyclone latitude and longitude values. Why not enter those into Google Maps—hey presto, I now have a beautiful visual map, with piles of selected information, of all the Australian cyclones I'm researching!
At the moment however, Trove strips all of the syntax from such a tag. So we must wait.
As to comments about the "crazy wilderness of tags," surely that's a good thing? I'm with Spearth on this one, tags, as an organisational tool and whether public or private, are most useful to the person who created them in the first place. In terms of searchability they don't really add much unless the tagger is associating a bunch of new words or phrases to the article, such as the full name of an ancestor with a birth notice. On the other hand, "horse" for an article on the Melbourne Cup doesn't do much, but it doesn't really hurt anyone either.
I reckon it would be a mistake to start policing tags, or to create a list of NLA-approved tags. Such steps may discourage tagging in the end.
There's a certain tendency to believe that if we work hard enough at a thing, streamline it, it can be perfected. But this is pleasant myth. Trove will never achieve perfection. It's so organic that any steps to hinder the free expression of its users, its authors, in the pursuit of that ideal, would inevitably be to the detriment of ourselves.
Last edited by mogedon; 02-02-2012 at 11:01 AM.
Good point on the Lightning vs Lightening !! You should contact them and let them know about that one.
I can see that you perceive Tags in a very different way to myself (and the SEARCH project). The way I see it is that broadly speaking there are two ways to use Tags:
1) The way you describe, which is to identify and help find a specific thing, person or event (e.g. by adding date and place to the Tag). Which is obviously a perfectly valid use of Tags, and using them something like the contents list of book, to get you to information about a specific topic; OR
2) As something more akin to an index entry in a large book (extremely large in the case of the newspapers database!). For example, you may look up "drought" and find all references Tagged as drought, regardless of which drought you are talking about.
The use of Tags as in 2), is equally valid when you understand how this can actually be a very powerful tool for research on a broad topic - as opposed to a specific topic. This is exactly what the climate history (SEARCH) project is trying to achieve.
In regards to a Tag such as "drought" being useless on an article about drought - I think to say this is missing some key points:
- It needs to be remembered that people are not necessarily correcting all the articles they are tagging. In reality there are far too many articles to correct in any reasonable time frame. Tags allow people to add 'index words', or 'keywords', to articles and if the aim is to increase the searchability of the newspaper database then this is definitely a much quicker, and more efficient, way to achieve that, rather than correcting every article. The correcting efforts are great and ultimately very useful, but in reality we will be waiting a very, very long time for most articles to be corrected.
- It should not be assumed that the article had the word "drought" in it or was correctly translated/corrected so that it could be found in a search. A drought is often more obvious after the fact, and may not have necessarily been called a drought in the article. Another good example is "tsunami" used as a Tag. In the past everyone called them "Tidal Waves", when in fact there are two quite separate unrelated natural phenomena, one being a true "Tidal Wave" and the other actually a "Tsunami". So with the hindsight of modern understanding a tagger can Tag with the word "Tsunami" when it is obvious that the phenomena was related to an earthquake, or possibly a large landslide event; as opposed to an unusually large tidal phenomena.
- General Tags used like an 'index word' and populated over a large number of newspapers and years can be a very powerful research tool (as the SEARCH project is trying to achieve). Even the word "horse" could be useful if used consistently across a lot of articles. For example, running statistics on the use of the word "horse" across the decades could be quite useful in researching our changing relationship with horses in our society. Tim Sherrat has already done some very interesting analysis on things like this using the translated text as it is, but widespread tagging could refine the results of such analysis, as long as it was done with some consistency.
- When researching, the use of Tags such as those used by the SEARCH project after many years of tagging (in a consistent way), you would probably not just search for "drought", for example, using the standard search. You would use a [ publictag:drought ] or maybe [publictag:drought* ] type of search to look for all those consistent Tags. If you were doing something similar with "horse" then you could search for the "horse" Tags and not necessarily get all the articles that use the word "horse" (e.g. "clothes horse"). Tagging with "horse" would also capture all those articles, for example, that used the word "cavalry" but did not actually use the word "horse". Again it all depends on what the tagger was trying to achieve with their Tags.
- It is useful to think of tags such as "drought", "Afghanistan", etc. like the 'keywords' that are listed at the start of many scientific papers. They are there to help people searching for relevant information to their research. The Keywords can be quite general, e.g. "flood", or quite specific, e.g. "flood Lismore 1889", because they are trying to lead the researcher to the scientific paper and they may have any number of research interests from the very general to the quite specific.
In regards to the "Family Notices" tag and not seeing how it achieves anything - here is an example of how this is quite useful.
Say you have unsuccessfully looked for a specific marriage between a Mr Smith and a Miss Jones by searching (using the Advanced Search) the 'Family Notices' category only, with the search phrase "Smith Jones". [You would restrict your search to the 'Family Notices' category only because the "Smith Jones" combination of names in lists etc. is far too common across all other articles.] Now because we know there are a lot of family notices misclassified by Trove in the Advertising sections, we can then take advantage of the "Family Notices" tag to search for this marriage in only those advertising sections that are known to contain family notices. You could achieve this with a search like the following in a general search:
"smith jones" AND publictag:"Family Notices".
I see the real problem with Tags as not being their "messiness" or the "crazy wilderness", but more about encouraging people to use them in consistent ways and in ways that will maximise the success rate of searches for that particular article that is being tagged.
Last edited by Spearth; 02-02-2012 at 11:22 PM.
OK, Spearth. I'm not twisting anybody's arm.
I think the mis-assumption is that people are tagging for the benefit of others. On the contrary, I think many people are tagging in order to easily find something again for their own projects. So the person who goes around tagging "Horses" is probably not tagging just any old article about horses, they probably have a quite specific interest in horses and simply tagging articles that match their interest so they can retrieve them conveniently later. They use the tag Horses to distinguish it from their other project Murders (or whatever).
I tag articles "NotAlbury". What that means (to me) is that this article is retrieved by a keyword search on Albury but is not an article about my Albury family (because there are many articles about the town Albury as well as unrelated people). It allows me to do keyword searches
Albury NOT publictag:NotAlbury
which avoid revisiting articles that I already know are of no interest to me. Since this "NotAlbury" tag is just a piece of personal process management, you might think I should use a private tag, but unfortunately I cannot use private tags in searches so I am forced to use a public tag.
I have several hundred tags "Alfred John Raymond". If I were to find an article tagged "RAYMOND Alfred John", that's all very fine and logical but no use to me for my personal retrieval needs, so I would tag it again as "Alfred John Raymond" because I only want one tag in *my* tag space (not two).
Again, you might say "why not use a list instead of tags?". The answer is that tags have one huge advantage over lists. You can sort tagged articles by date easily, but the only way to have lists sorted by date is painful. You have to manually put each article into the right position in the list. And as the list gets longer, the task gets more and more tedious. Lists are very unattractive for any purpose in which date order is important.
How about we campaign for a change to Trove so that Private Tags can be used in searches, i.e. something like:
which would only search for your own private tags.
I completely understand why people use Tags like "NotAlbury" in the way that Kerry has, but the Tag is almost useless to everyone else. So being able to search Private Tags may reduce the number of truly useless (to everyone else) tags.
Perhaps, in the interim, we should be trying to make Tags like these more useful to other people. For example, the "NotAlbury" tag could be rendered more useful to others by making it "Not Albury family" instead.
Instances of people tagging names in a format like "JONES William John" is evidence that people are not reading the 'Tagging Guidelines', which clearly state that users should Tag names of people, organisations, etc. in their 'natural language order' - i.e. first-name second-name surname for people. I presume the reasoning for this is that they want the names in Tags to be the same order as the name order most commonly found in the newspaper articles - in this way the results of searches for a particular name would be maximised, whether the name is in the article or an attached Tag or Comment.
I think there are two changes that could be made to Trove that will help 'clean up' Public Tags:
1) provide a way for Private Tags to be used in searches, so reducing the need to use Public Tags that can only be useful to the originator; and
2) provide a link to the Tagging Guidelines (perhaps updated), and the Text Correction Guidelines for that matter, somewhere very obvious on every newspaper article viewing page (and text correction page).