Using the full text search on FamilySearch
Full text search of all the images from the Registry of Deeds on
FamilySearch
All our problems are solved??
Background
Optical character recognition (OCR) has made tremendous strides since its beginnings. Applied to newspapers it has changed the research world. No longer do we painstakenly trawl through paper or microfilm versions of newspapers, we simply enter some search terms and are provided with marvellous results. We find obscure relults including advertisments and court reports.
I have used the following with success:
- Trove (Au)
- British Newspaper Archive (UK Ire)
- Papers Past (NZ)
- Chronicling America (US)
- Chronicling America (US)
- Irish newspaper archive (Ire)
While OCR is wonderful it is not perfect. I published this little article in 2019. OCR and the Bleeding Obvious in The Ancestral Searcher Volume 42 Number 1 pp 22-23.
OCR of typescript is hard enough, but OCR of cursive writing is much harder. It has, however, progressed fantastically with the inclusion of artificial intelligence.
FamilySearch recently released a full text search of the Registry of Deeds images. This collection has over 2500 scanned microfilms. It has some wild card and phrase search facilities. If you have a rare name or combination your search may return some gems.
Examples
My ggggrandfather Thomas Swan Croker (d 26 January 1837 in London) and his son William Swan Croker have a reasonably rare combination of names. So, we can try a key word search entering "swan croker" in the search box.
This simple search returns 33 results seemingly in random order. Most of the earlier references I had found previously. One of the results is from 1903 and is referring to Thomas Swan Croker as a previous occupant. I would not have looked too hard for these late entries without this tool.
This is a reasonable success.
Delving deeper
When using a search facility, we need to understand what the searched data actually is. The basic unit of the memorial transcription books is a memorial. What the results return are occurrences of the search terms on a double page opening of the memorial transcription book. This comes to the fore when we are searching for two names. One name may be in one of the memorials with some words visible at that opening and the other in an unrelated memorial also with some words on that opening.
One issue for me, is that the results have very limited metadata. We need to find the memorial number and the volume number. The memorial number may be on the opening otherwise you may need to scroll a couple pages. With a very long memorial you might have to scroll twenty pages. e.g. George Hutton is on the last page of a memorial that starts on image 424.
OCR errors
While searching for Croker I have found some misinterpretation of Croker as broker. the transcription is " John Croker to make a suitable provision for his children by Sarah broker otherwise Pennefather his then wife". See a result page.
Of course, it is hard to determine whether a letter is an "e" or "i". The search facility has a way around this using the wild card "?". Other letters that might cause confusion are "u" and "n", and "m" and "nn".
Abbreviation
Another issue for searching is the use of abbreviations such as Wm for William, Edw for Edward, Jas for James etc. Abbreviations are also used family names and place names. Thus, you may need to be a bit creative in your search terms.
Conclusions
The text search tool on the Registry of Deeds images is a great addition to the available tools for research in the Registry. Its strong points include the bredth of coverage.
OCR of handwriting is getting better but mistakes will occur.
Areas of improvement include better metadata that includes volume page and memorial numbers. It would also be great if there were a text correction facility as is available if newspaper site like Trove.
There is still a need for an intelligent index with strong metadata. I would encourage users of the full text search facility add their results to our index.