Finding articles in digitised newspapers and more...

 

How does relevance ranking work?

By default search results are sorted by relevance according to the following rules:

Principle Example
1. Inverse document frequency (i.e. rarer words are weighted more than common words) A document with 17 occurrences of ‘on’ and 1 occurrence of ‘smartboard’ ranks lower than one with 17 occurrences of ‘smartboard’ and 1 occurrence of ‘on’.
2. Headings (title/subtitle/author) are weightedmore than full-text If the word ‘Zoo’ appears in the title it will rank higher in the list than if the word ‘Zoo’ appears in the last line of the article.
3. Length of article and density of words A short article with 3 occurrences of a word ranks higher than a longer article with 3 occurrences of the word.
4. Words in close proximity are weighted. A phrase (two words together) ranks higher than where two words occur within 5 words of each other, which ranks higher than where two words occur within 20 words of each other.
5. Exact word matches are weighted higher than ‘fuzzy’ matches, i.e. stemmed or plural words. (The search automatically performs fuzzy searches). If you search for ‘smile’ the results will also contain ‘fuzzy matches’ automatically e.g. smiling, smiley but these will be lower in the list since they don’t match the search term exactly.
6. Category of article. Articles are ranked lower if they are ‘Detailed Lists’ (20% as important), or even lower if they are ‘Advertising’ (10% as important). If three articles had each scored 500 to this point, an article categorised as Article would score 500, an article categorised as Detailed Lists, Results and Guides would score 100, and an article categorised as Advertising would score 50.

Results may also be sorted by Ascending or Descending date, however please be aware that sorting by these options will result in less relevant articles being displayed earlier in the results list.

 

Search for illustrated articles

To search for illustrated articles, put your search term(s) in the simple search box as usual. From the search results screen, select 'Illustrated' on the left hand side of the screen to expand the menu, then select 'Illustrated'. Once the page has refreshed select the type of illustration (e.g. photo) from the 'Illustration type' menu that now appears.

Please note that display advertisements are not included these results, and may need to be searched or browsed for separately.

You may also search for illustrated articles by using the advanced search form and selecting 'search for illustrated articles only'.

 

Refine your search results

The column to the left of the search results is labelled 'Refine your results'. This column is a list of filters that group the articles into predefined categories used to refine your search results. Some of these categories, such as Decade and Illustration, open new menus when an option is selected, permitting further refining by these categories. Once a filter has been selected, you may remove the selected filter, by clicking on the 'x' next to the filter under 'Search limited to:', or next the the selected filter in the left hand column. You may clear all applied filters by clicking the 'clear all' link.

Alternatively you may use the 'Refine Search' link (next to the total number of results) which will take you to the advanced search screen and populate the search box with the search terms used.

 

Using the Advanced Search

To use the advanced search, click on 'advanced search' from the Digitised newspapers and more... homepage, or from your search results for a search in the Digitised newspapers and more... zone.

On the advanced search page you can:

  • Specify an 'all words' search (default for simple search);
  • Specify a phrase search;
  • Specify an 'any word' search;
  • Specify a 'without the word/s' search; or
  • Any combination of the above.

You may also restrict your search to:

  • Anywhere in the article, including tags and comments (default for simple search);
  • Headings only;
  • Headings, author or first 4 lines (first 4 lines in the Article category only); or
  • Search picture captions only.

Note: If you choose to search the headings, author and first 4 lines you will only search the metadata supplied in processing. These searches do not search the Electronically Translated Text as seen in the article view.

In addition to these options, you may restrict your search to:

  • Publication date - must include a year. Day and month are optional (cannot include day and year only), entering a date in the 'from' fields only will search all articles after that date, entering a date in the 'to' fields only will search all articles prior to that date;
  • Newspaper title & Location - select the newspapers you are interested in, if you wish to search across all newspapers leave this box blank;
  • Article category - select the category or categories you are most interested in, a description of the article categories is available from the ANDP website;
  • A specified length of article; and/or
  • Illustrated articles only.

You may also change the order that the search results are sorted.

 

I am trying to find an article which I know contains particular distinctive words. Why doesn't this article appear in the search results?

Searches look across the electronically translated text, produced by the OCR process. If the words have not been translated accurately, then the system will not be able to find these terms, even though they may appear on the original newspaper page. We recommend correcting the electronically translated text in these situations to improve searching for other users.

 

I am using an index to browse for an article. Why isn't the article where the index says it is?

There are four main reasons why an article may not be where the index says it is:

  1. The edition of the newspaper that appears in Trove is different to the edition that was used to make the index.
  2. The page referred to in the citation is missing from Trove.
  3. The page sequence number in Trove is not the page number as it appears in the index, check the surrounding pages.
  4. The index has an incorrect citation.

 

Search for names or phrases containing punctuation, e.g. O'Reilly or 2/14th Battalion?

Punctuation marks are not searched by Trove, as the process of creating the electronically translated text often adds additional punctuation that would decrease the accuracy of the search results. This may, however, result in difficulties when searching for names or phrases that include punctuation, such as O'Reilly and 2/14th Battalion.

To search for terms that would normally include punctuation, omit the punctuation from your search terms, running the terms together (i.e. oreilly and 214th battalion rather than o'reilly and 2/14th Battalion). It is also worth replacing the punctuation with a space, and searching for the term as a phrase (i.e. "o reilly" or "2 14th battalion), as the punctuation may not have been recognsed.

 

Search for names starting with 'Mc'

Names beginning with Mc may be printed as M‘, and processed as an M followed by an apostrophe, M'.

As Trove does not search on punctuation, try searching for the name excluding the 'c', or replacing the 'c' with a space and searching as a phrase (i.e. mhugh or "m hugh" in addition to mchugh).

 

Search for hyphenated names or words

As it is not possible to accurately determine the context in which a hyphen is used, all hyphens (and other dashes) in Trove have been indexed in the same way. This includes hyphenated names, hyphenated words and hyphens where a word has been split across the end of a line.

Hyphenated words in Trove are indexed as both parts of the word, then as the word all run together, for example:

Original word Words indexed in Trove
Venn-Jones Venn, Jones, VennJones
To-day to, day, today
opera-tion opera, tion, operation

To search for hyphenated names, omit the hyphen from the name, running the words together (i.e. vennjones rather than venn-jones) or search for the name as a phrase, replacing the hyphen with a space, to catch the examples where the hyphen has not been recognised by the electronically translated text (i.e. "venn jones" rather than venn-jones).

 

Search for an exact word

When searching, Trove searches for and retrieves those things which exactly match the search terms first, and displays them at the start of the results list (so long as the sort order is set to relevance). Trove also broadens the search by truncating search terms, which is known as fuzzy searching.

As long as the sort order in Trove is set to relevance, articles that have been matched on a fuzzy search term will appear towards the bottom of the search results list. This aids in finding articles when the electronically translated text is not completely accurate.

To stop Trove from doing a fuzzy search, and only search on the exact search terms entered, use the 'fulltext: syntax before your search term.

e.g

 

I'm doing a phrase search, using quotation marks, e.g. “Melbourne Cup” but the results are finding other things as well, why is that?

When you search for an exact phrase in quotations the search engine looks for and retrieves those matches first and displays them at the top of the list (so long as you leave the sort order set to relevance). It also broadens the search by truncating terms and also finding terms not directly next to each other. This is known as fuzzy searching.

As long as the sort order is set to relevance, articles that have matched on a fuzzy search term will appear towards the bottom of the search results list. This aids in finding articles when the electronically translated text is not completely accurate.

If you wish to turn off the fuzzy search, use the 'fulltext:' syntax, as described in Search Tips.

e.g

 

Sometimes my search term is not highlighted - why?

Matched terms in the electronically translated text are always highlighted. This includes both exact matches and matches from the 'fuzzy' search. When searching for a phrase all terms in the search are highlighted, however this does not mean that the article was matched on every occurrence of a highlight (e.g. in a phrase search for "Watchmaker and Jeweller" all occurrences of 'and' will be highlighted, however this is only searched for as part of the phrase).

Terms are highlighted on the article image when a search term has been matched with the data originally supplied. Terms will not be highlighted on the article image if the match is to text that has been corrected.

Where an article has been matched on the originally supplied text, however this has since been corrected out, there will be a highlight on the article image, but no corresponding highlight in the electronically translated text.

 

What are article categories?

Articles are given categories to define their broad subject or type at the time of processing. The aim of this is to enable users to limit their searches to specific types of articles they are interested in, e.g. Family Notices only, and to aid in relevance ranking. A description of categories assigned to newspapers is available here.

It is not currently possible to change article categories.

 

Why does the search hang sometimes when I try to use the truncation symbol?

Sometimes when searching using the truncation symbol (*) the search will hang and appear not to find any results, without bringing up a 'No results match your search' message. This is due to the way that the truncation symbol is searched in Trove, which can be resource intensive and for common word stems will result in Trove being unable to complete the search.

To avoid this, have at least 3 characters before the truncation symbol, unless it is a particularly distinctive starting string. If this search fails, add more characters to your starting string.

 

Why is the electronically translated text sometimes incorrect or different to the text that I can read on the newspaper page?

The electronically translated text is created using the process described here. The computer 'eye' that looks at an image and translates it into searchable text is often not as good as the human eye. On good quality images the text is generally translated accurately and well; however if the source material is of poor quality (e.g. scanned newspapers) the results can vary significantly and translations may include incorrect characters, words, question marks or what looks like 'gobbly gook'. This happens when the computer eye does not see a clear black and white image. More information is available here.

 

I used to see the Article Category of 'News', now I see 'Article', why has this changed?

In order to better reflect the varied content in this category, a decision was made in mid August 2010 to globally convert articles in the 'News' category to 'Article'.