9. Finding NLA digitised content you can download#

On this page

Trove provides a platform for the delivery of digitised content from the National Library of Australia and its partners. This digitised content is easy to find in the Newspapers & gazettes and Magazines & newsletters categories – they’re full of it! But it’s not so easy to find digitised content in other categories where it’s mixed with works aggregated from a range of different sources.

There are a few different approaches to finding digitised content, but they all have potential problems. While I highlight what I think is the most reliable option, it’s really a matter of working out what will best meet your research needs.

9.1. Search for records including nla.obj#

All of the digitised resources in Trove (except for Newspapers & Gazettes) have NLA identifiers of the form nla.obj-[NUMBER]. So a search for "nla.obj" should return all digitised resources. Here’s the results of running this search across all categories:

Hide code cell source 
params = {"q": '"nla.obj"', "category": "all", "encoding": "json", "n": 0}

headers = {"X-API-KEY": YOUR_API_KEY}
totals_nlaobj = []

response = requests.get(
    "https://api.trove.nla.gov.au/v3/result", params=params, headers=headers
)
data = response.json()
for cat in data["category"]:
    total = cat["records"]["total"]
    if total:
        totals_nlaobj.append({"category": cat["code"], "total": total})
df_nlaobj = pd.DataFrame(totals_nlaobj)

df_nlaobj.style.format(thousands=",").hide()
category total
book 640,918
diary 211,747
image 295,691
list 5
magazine 1,807,140
music 58,537
newspaper 1,854
people 38
research 365,685

Try it!

The problem with this approach is that includes a certain amount of ‘noise’. For example, there are no digitised resources in the people and list categories, and the newspaper category uses different identifiers. Any matches in these categories are probably because digitised items are referenced within records or user annotations.

You might also wonder about the large number of results in the Books & Libraries category – are there really that many digitised books? Probably not. If you look closer at the results for the Book & Libraries category you’ll see that most of the works are ‘Not available online’. Digging deeper, it seems that some of these have been digitised by the NLA, but the digitised version has not been published online. These records include a note saying something like ‘Digital master available ; National Library of Australia’ with a nla.obj identifier. However, the majority of the ‘not online’ records are actually empty – ghostly remnants of some past processing failure.

If your aim is to harvest details of all the NLA digitised works that are available online, you’ll probably want to exclude all the ‘not online’ records. One way to do this is to set the l-availability facet to y. Here’s how that affects the number of results:

Hide code cell source 
params = {
    "q": '"nla.obj"',
    "category": "all",
    "l-availability": "y",
    "encoding": "json",
    "n": 0,
}

headers = {"X-API-KEY": YOUR_API_KEY}
totals = []

response = requests.get(
    "https://api.trove.nla.gov.au/v3/result", params=params, headers=headers
)
data = response.json()
for cat in data["category"]:
    total = cat["records"]["total"]
    if total:
        totals.append({"category": cat["code"], "total": total})

pd.DataFrame(totals).style.format(thousands=",").hide()
category total
book 101,795
diary 135,248
image 260,112
list 5
magazine 1,807,140
music 40,403
people 38
research 362,037

Try it!

Another potential problem is that a search for "nla.obj" returns digitised resources as well as born digital works submitted through the National E-Deposit scheme (NED). Some of the NED resources can only be viewed onsite at a participating library, so they’re not really online. To exclude these from your results you can set l-availability to y/f (free access):

Hide code cell source 
params = {
    "q": '"nla.obj"',
    "category": "all",
    "l-availability": "y/f",
    "encoding": "json",
    "n": 0,
}

headers = {"X-API-KEY": YOUR_API_KEY}
totals = []

response = requests.get(
    "https://api.trove.nla.gov.au/v3/result", params=params, headers=headers
)
data = response.json()
for cat in data["category"]:
    total = cat["records"]["total"]
    if total:
        totals.append({"category": cat["code"], "total": total})

pd.DataFrame(totals).style.format(thousands=",").hide()
category total
book 61,843
diary 135,122
image 259,448
list 5
magazine 1,807,140
music 33,030
people 38
research 359,957

While other NED resources are freely accessible, most are only available as PDFs or in an e-book format. If you’re collecting data about digitised works from which you plan to harvest images or text, then you might want to exclude all NED resources. The NED collection has its own NUC identifier, so you can exclude it from your search by adding NOT nuc:"ANL:NED" to the simple search box, or to the q parameter in an API request.

Warning

Trove’s grouping of versions into works can cause unexpected results. It’s possible that by excluding all NED resources, you might lose some of the NLA’s digitised works. This could happen if there were multiple versions of a single work, including one that has been digitised, and another that has been submitted via the National E-Deposit service. In this case, setting NOT nuc:"ANL:NED" would exclude the work and all its versions from your results.

So what’s the best approach? I’d suggest starting broad by using the query "nla.obj" and setting l-availability to y. This gives your best chance of finding all online digitised resources. If this is generating too much noise, you can add further limits – though be aware of unintended consequences!

If you’re using the API to create a dataset of digitised resources, you can inspect metadata records after you’ve downloaded them to decide whether or not to add them to your dataset. This gives you more fine-grained control than you can get by tweaking the search parameters. For example, if an item has been digitised and published online, the identifiers field in the metadata should include a url containing ‘nla.obj’ with the linktype of fulltext. ==Link to accessing data section for more?==

9.2. Other options#

There are a few other ways of finding digitised content, but they’re not as reliable as searching for “nla.obj”. There’s no public documentation about how these indexes are created, so it’s difficult to interpret the results they return. But they might be useful in some circumstances.

Limit results to the ‘Trove Digital Library’#

In the Advanced Search form for the Books & Libraries, Research & Reports, Images, Maps & Artefacts, Diaries, Letters & Archives, and Music, Audio & Video categories, there’s an option to limit the source of the records in your results by selecting from a list of ‘Organisations’. Hidden away in this list is the ‘Trove Digital Library’. Unfortunately, selecting ‘Trove Digital Library’ in Advanced Search doesn’t work at the moment because of a bug in the web interface, but once you know it exists you can manually add it to your searches.

../_images/advanced_search_orgs.png

Fig. 9.1 Selecting ‘Trove Digital Library’ in the Advanced Search form – note the NUC identifier in brackets#

Like other contributing organisations, the ‘Trove Digital Library’ has it’s own NUC identifier: ANL:DL. You can use this to limit your search by adding nuc:"ANL:DL" to the simple search box in the web interface, or to the q parameter in an API request.

Try it!

The results look ok, but the problem is that some digitised publications aren’t included in the ‘Trove Digital Library’ – here’s some examples from the Books & Libraries category. It’s not clear why this is, and what a search for nuc:"ANL:DL" actually includes (or excludes).

fullTextInd:y#

The Trove documentation states that by adding fullTextInd:y to your query you can:

find digitised or born digital items that include full text, particularly digitised books & journal articles

It’s not obvious from this description, but this includes much more than just NLA digitised resources and publications submitted through NED. Using fullTextInd:y without other filters will return a wide range of content aggregated from multiple sources, including full-text publications that require authentication to access. Many of these will be in PDF format. It’s not a reliable way of finding digitised items with downloadable text.

Furthermore, fullTextInd:y doesn’t seem to consistently applied to NLA’s own digitised resources. For example, it seems to exclude a number of periodicals with digitised issues. Compare this search for Platt’s Almanac with and without fullTextInd:y.

Adding fullTextInd:y does seem to exclude NLA digitised resources that have no OCRd text. This might be useful in combination with a “nla.obj” search, but again there might be unintended consequences.

has:correctabletext#

Adding has:correctabletext to your query limits the results to works that have OCRd text you can correct in the Trove web interface.

If the text content of a resource is ‘correctable’ then you’d expect it to be an NLA digitised item with OCRd text you can download. So adding has:correctabletext to your query should limit the results to digitised items with downloadable text. This seems to be the case (though watch out for more ghost records), but again it’s not clear what you are excluding – is every item with OCRd text correctable?

imageInd#

Adding imageInd:thumbnail to your query limits results to works that have a thumbnail image.

Both fullTextInd:y and has:correctabletext filter records based on whether they have accessible text. But there are many digitised resources that either contain no text at all, or have no text that can be extracted by OCR. Adding imageInd:thumbnail to your search can help find these items. However, like fullTextInd this index is applied to aggregated collections as well as digitised resources, so your results can include all sorts of content, from book covers to pictures of politicians. There’s also no guarantee that an item with a thumbnail will provide a larger image for download.