Alma AI Metadata Assistant Feedback by Chingmy Lam

Example:

chingmylam1 — 2024-10-27 22:16:07 UTC

The record has enough metadata elements to describe the resource.

Example:

chingmylam1 — 2024-10-27 22:19:40 UTC

The assigned subject headings are not accurate.

Example:

chingmylam1 — 2024-10-27 22:21:11 UTC

Monographs - Non-fiction

Example:

chingmylam1 — 2024-10-27 22:23:44 UTC

Allow more images to be uploaded.

Example:

chingmylam1 — 2024-10-27 22:27:27 UTC

The captialisation and punctuation are not correct, e.g. the 6XX fields do not have full stop at the end.

chingmylam1 — 2024-10-27 22:35:05 UTC

chingmylam1 — 2024-10-27 23:53:36 UTC

Monograph - First Nations

2024-10-30 23:21:35 UTC

#1 Miwatji Yolŋu - Exhibition Catalogue of Yolngu aritsts.

#1 Miwatji Yolŋu

2024-10-30 23:34:55 UTC

650 - LCSH subject headings

710 - Added Entry-Corporate Name

Overall access points are adequate for discovery at a basic level.

#1 Miwatji Yolŋu

2024-10-30 23:44:49 UTC

300 - No pagination ($a), measurement was incorrect ($c)

008 - lacking place and date of pub, illust.

245 - Wrong punctuation for subtitle

336 - lacking field for still image

520 - inocrrectly transcribed from the copyright statement

500 - Missed adding "1st edition of 1,000 copies"

584/500 - Lacking exhibiton note with dates

650 - Lacking AIATSIS subject headings

Monographs - First Nations

2024-10-31 04:51:27 UTC

#2 Warra Warra Wai

First Nation's accounts of the Endeavour

#2 Warra Warra Wai

2024-10-31 04:56:58 UTC

650 - LCSH subject headings

020 - Added ebook ISBN

100/700 - Author fields with $e

#2 Warra Warra Wai

2024-10-31 05:07:55 UTC

520 - assistant added $$7 Generated by AI which is incorrect

245 - Did not capture subtitle from image of title page, indicator/s incorrect

Most punctuation is absent between subfields and at end of fields

No dewey number

008 does not reflect bib

No AIATSIS headings

LDR missing pos 18 (cataloguing form)

300- No pagination or measurement

Monographs - Fiction (Mark K)

2024-11-01 03:35:05 UTC

The Next Girl (new), The Game Darrells (enhanced), Things Will Calm Down Soon (new), Homecoming (new)

Monographs - Fiction (Mark K)

2024-11-01 03:37:29 UTC

Subject headings seem consistently appropriate, 245 title and 100 author seem to have used correct punctuation.

When values: text, unmediated, volume are added in Content note field 336, 337, 338 are added

Monographs - Fiction

2024-11-01 03:41:56 UTC

Values input in 008, 264, 300 cannot be trusted

Punctuation in 264, and 300 need checking

Some fields need to be manually added, i.e., 082

Monographs - Fiction (Mark K)

2024-11-01 03:48:05 UTC

5 or 6 out of 10 for accuracy. However, it is inaccurate enough to require check on most fields both for values and punctuation.

The 7 subfield in the 520 field is a problem ($$7 Generated by AI) often mostly this is not generated by AI but copy paste from the source or publisher.

Monographs - Fiction (Mark K)

2024-11-01 03:49:51 UTC

5 out of 10 300 consistently needs to be added

2024-11-01 03:51:27 UTC

Adding to this padlet is very time consuming and appears to replicate the test records log work

Monographs - Picture Books (Mark K)

2024-11-01 06:03:12 UTC

How to draw a dragon (enhanced), Are we there yet? (new), Ethel the penguin (new)

Monographs - Picture Books (Mark K)

2024-11-01 06:09:02 UTC

300 values in picture books (unnumbered pages) is pretty good.

Monographs - Picture Books (Mark K)

2024-11-01 06:09:22 UTC

Punctuation in 264 and 300 fields is not correct

Monographs - Picture Books (Mark K)

2024-11-01 06:09:30 UTC

008 needs corrections for each record

Monographs - Picture Books (Mark K)

2024-11-01 06:09:37 UTC

For Picture Books, yes but fields still need checks i.e., LDR, 008, 250, 264, 300 because of punctuation errors.

Thesis (Mark K)

2024-11-05 05:51:17 UTC

Norfolk Island and the Polynesian Narrative: New Evidence from a Marginal Mystery Island (SeS online), An investigation of circuit noise arising from ferromagnetic core elements (RBSC hard bound), Ether anaesthesia in children (SeS online)

Thesis (Mark K)

2024-11-05 05:53:32 UTC

Many details from the image of an old title page (1904) were correctly recorded. Subject heading look good based on title and summary/abstract

Thesis (Mark K)

2024-11-05 05:58:19 UTC

In thesis 991032666425505106 the binder was recorded as the publiser in the 264 field. In the content note I added the value text, computer online. This was recorded correctly in 336, 337 and 338 but it was also added incorrectly to the 300 field. USYD thesis require fields that and values that AI assistant does not include. This resource type needs its own prompt.

Thesis (Mark K)

2024-11-05 06:03:46 UTC

Incomplete, thesis is a resource type that requires a specialised template and the input form needs more fields than current Title; Author; Content note Summary Note

Thesis (Mark K)

2024-11-05 06:04:49 UTC

Subject headings for theses are better than expected

Chingmy:

chingmylam1 — 2024-11-07 00:55:59 UTC

Apart from image files, also accept PDF or WORD docs.

Danni:

chingmylam1 — 2024-11-07 01:36:28 UTC

#1 It would be great if we could use templates to dictate what fields are required for a format. eg: institutional guidelines for certain format, such as boundwiths, scores etc.

Danni:

chingmylam1 — 2024-11-07 01:38:16 UTC

Suggestion that there be a resource type tickbox in the initial AI box to ensure the appropriate fields are used for that resource type - this would then tie into appropriate templates being used

Liz R:

chingmylam1 — 2024-11-07 01:39:05 UTC

Would be good to be able to "highlight" the information you want captured. Maybe even "direct" it to certain MARC fields.

Liz R:

chingmylam1 — 2024-11-07 01:39:44 UTC

Be great if the AI assistant could access external sources (like publishers website) to fill in gaps like pagination and contents notes.

Danni:

chingmylam1 — 2024-11-07 01:48:45 UTC

#1: Created a 700 creator field based on author of second title.

Danni:

chingmylam1 — 2024-11-07 01:49:31 UTC

#1: did not input date or location in 008 field, punctuation wrong in 100 field, second indicator wrong in 245 field, only included 520 for first title. Added second title as additional 245 field, which is non-repeatable. Changed Lim (titlepage) to Ltd. (bib record). No 040 field, which is needed to identify original cataloguing. LDR has "unknown' for descriptive cataloguing for pos. 18. Size was wrong.

Danni:

chingmylam1 — 2024-11-07 01:50:18 UTC

#1: I would say it is a basic record, such as that of a beginner cataloguer might create. It's a step up from a brief bib.

Liz R:

chingmylam1 — 2024-11-07 01:55:48 UTC

Be great if the AI assistant could access external sources (like publishers website) to fill in gaps like pagination and contents notes.

Liz R:

chingmylam1 — 2024-11-07 01:56:17 UTC

Ability to use more than 2 images.

Mark K:

chingmylam1 — 2024-11-07 01:56:42 UTC

Agree with Dannie and Liz, need the ability to add text files and more the 2 images/files.

Monograph - First Nations

lizforwork334 — 2024-11-07 03:37:23 UTC

#3 Ceremony / Adam Goodes

lizforwork334 — 2024-11-07 03:45:12 UTC

Improved record with LCSH, expanded 300 and added contributors 7XX.

Good input of 500 note.

Interestingly when i tested again enriching with publisher's website image instead the subject headings were much better (ie added more 6XXs and subdivsion for juvenile literature)

#3 Ceremony / Adam Goodes

lizforwork334 — 2024-11-07 03:45:58 UTC

Left old 300 field.

Missing AIASIS headings.

Lacked subfields and statement of responsibility in 245.

Didn't input Indig place name in 264 despite being in photo.

Liz R:

lizforwork334 — 2024-11-07 03:47:42 UTC

Would be good for the AI to pick up the Indigenous place names for the 264

Monographs - Non-fiction

lizforwork334 — 2024-11-07 03:55:34 UTC

#1 Analogue Images

Various viewed in Primo VE (Kawei)

kawei_lei — 2024-11-07 04:04:19 UTC

Theses, monographs (fiction and picture books). Only viewed the scores briefly.

kawei_lei — 2024-11-07 04:06:15 UTC

Kawei: it's clearly stated in the summary and notes fields that the record is "AI generated" - good from user perspective and for staff who need to identify between AI and non-AI created records.

Various

kawei_lei — 2024-11-07 04:12:27 UTC

Kawei: for the most part, the records are pretty functional for user who just want to access the item. However, some examples of metadata inaccuracies and/or errors include:

991032396712605106: dollar signs visible in creator/contributor field. It states "hysical description not provided." right after providing item measurements.
991032666317305106: no punctuation in description field
991032666510405106: very basic record, not enough details about the item

Resource/content type errors

kawei_lei — 2024-11-07 04:15:38 UTC

Kawei: all scores and theses generated have been categorised as "book" in Primo. Not sure if AI record is able to assign an accurate resource type?

Referencing: citation feature

kawei_lei — 2024-11-07 04:23:55 UTC

Kawei: although most of the records are functional for those who just want to access the item, where errors/inaccuracies may be noticed or directly impact students is when they use the citation feature from the library catalogue and it's generating a larger volume of "bad" citations (more than normal). Having said that, a warning message is already in place.

Expectations of researchers/scholars/SMEs of a specific field

kawei_lei — 2024-11-07 04:45:11 UTC

Kawei: researchers looking for highly specific historical items such as rare books may find some records lacking if more accurate descriptions are expected.

Monograph - French (Mark K)

2024-11-08 04:30:52 UTC

Gaetana (new), Théatre bizarre (new), le Poete Attius sur la tragedie Latine (enrich), Les quatre jours du pauvre homme (new) Epoques des lettres francaises (enrich)

Monograph - French (Mark K)

2024-11-08 04:36:50 UTC

I rated most French language books as 'Good'. The subject heading were good, 250, 264, 336, 337 and 338 included correct values. If I added pagination in the Content notes field it was added to 300 a subfield.

Monograph - French (Mark K)

2024-11-08 04:39:00 UTC

Issues/problems included The 008 date was 2021, while date in 264 was 1950. The 008 language was eng while the 041 language was (correctly) fre. Also 100 values recorded the author name correctly but did not use the authority heading even if it was available.

Monograph - French (Mark K)

2024-11-08 04:43:00 UTC

Generally the metadata was accurate, however I noticed that sometimes the 300 subfield c value was added which was often incorrect; I did not know where/how this value was sourced or calculated. Maybe a hallucination.

Monograph - French (Mark K)

2024-11-08 04:52:40 UTC

Generally sufficent or complete except for the LDR and 008 values which appeared to be default values. Many records also included a 520 field that reiterated information from other fields, not good or bad just different.

Monograph - French (Mark K)

2024-11-08 04:54:29 UTC

Work on making LDR and 008 better more accurate and not just default values

Monograph - French (Mark K)

2024-11-08 04:56:54 UTC

I thought the quality of records of this resource type was the better than most other types. So non-English resources are on average good (not fair)

Monograph - Non-fiction

lizforwork334 — 2024-11-11 02:24:15 UTC

#2 Horsley and Lang's meetings (law textbook)

#2 Horsley and Lang's meetings

lizforwork334 — 2024-11-11 02:27:42 UTC

Good contents note from my input (got from pub website). Edition statement input correctly. Good LCSH.

#2 Horsley and Lang's meetings

lizforwork334 — 2024-11-11 02:28:28 UTC

No pagination and punctuation issues as usual. 008 not correct. Input Australia as publication place despite Chatswood being in one of the images. Does it "prefer" title page info to TP verso?

Liz R.

lizforwork334 — 2024-11-11 02:33:02 UTC

Would be better if the AI gave preference to more granular places names in 264. ie can we get it to preference Chatswood over Australia when both are present in the images.

Upload local standards/rules

aliespeirs — 2024-11-11 02:42:05 UTC

Be able to upload local cataloguing policies and notes into the tool. e.g. when there is local practice to use certain fields; to not repeat fields that are repeatable; add local notes according to when certain names are mentioned, recognise bookplates and signatures and add desired local notes

Liz R:

lizforwork334 — 2024-11-11 02:53:05 UTC

Enriching records is not as useful as creating new. The only improvement is adding LCSH

Monograph - Non-fiction

lizforwork334 — 2024-11-11 02:56:19 UTC

#3 The last secret agent

lizforwork334 — 2024-11-11 02:58:13 UTC

Only the addition of LCSH and 33X fields improved the brief bib.

#3 The last secret agent

lizforwork334 — 2024-11-11 03:00:36 UTC

Didn't add a more complete 245. Duplicated inadequate fields (300)

Boxed Collections

fionaberry3 — 2024-11-13 02:55:09 UTC

Paul Diehl Collection (new), Margo Lanagan Collection(new), John Horton Young collection (new)

Boxed collections

fionaberry3 — 2024-11-13 02:58:34 UTC

LCSH subject headings correct. Appears to be an OK basic record. Values added in 'Content note' field resulted in addition of 336, 337 and 338.

Boxed collections

fionaberry3 — 2024-11-13 03:02:13 UTC

Publisher incorrectly identified as University of Sydney in all cases. John Horton Young collection: 338 incorrect (volume). Paul Diehl collection: mismatch 008 language code / 041 language. Date ranges of collections not picked up/reflected in record in any way.

Boxed collections

fionaberry3 — 2024-11-13 03:08:56 UTC

If using finding aids as sources of information, more than 2 images would definitely be needed- 2 images is not really enough to include anything other than introductory sections of finding aids (which are usually 1.5-2 pages).

Boxed collections

fionaberry3 — 2024-11-13 05:46:13 UTC

Jane Windeyer Collection (enhance)

Boxed collections

fionaberry3 — 2024-11-13 05:46:35 UTC

Subject headings good. 700 fields added with correspondents (not sure if this is desired but makes sense). 300 field added with correct extent (3 boxes).

Boxed collections

fionaberry3 — 2024-11-13 05:48:21 UTC

Usyd identified as publisher. 338 incorrect ('volume').

Mongraphs German (Mark K)

2024-11-19 03:59:59 UTC

Zeitalter Goethes, Prophetischer Gottesbescheid in Mari und Israel : eine Strukturuntersuchung, Man lebt nur zweimal :ein kriminalroman, Hesiodos' Theogonie als phoinikische Kosmologie

Videorecordings - DVD (Mark K)

2024-11-19 04:02:34 UTC

Gallipoli, In the mood for love, Rashomon, Matrix revolutions, Prick up your ears

Monographs - French (Mark K)

2024-11-19 04:04:36 UTC

Subject headings, 264 was generally accurate, 336, 337, 338 were generally correct.

Monographs - German (Mark K)

2024-11-19 04:06:12 UTC

Subject headings, 264 was generally accurate, 336, 337, 338 were generally correct.

Videorecordings - DVD (Mark K)

2024-11-19 04:15:11 UTC

Subject headings were good/very good, 336, 337, 338 were generally correct. A fair/good DVD record could created with very limited information and 2 images, i.e., a Title 'title name' and Content Note "1 videodisc (DVD) Publisher"

Monographs - German (Mark K)

mark_kosta — 2024-11-19 04:33:17 UTC

Values in 008 especially date is incorrect. 100, 600 and 700 field never used LoC author headings. In the 300 field pagination and dimension was always incorrect.

Mongraph - German (Mark K)

mark_kosta — 2024-11-19 04:36:58 UTC

Generally the metadata was accurate except for 008 field and the 300 field. It appears to hallucinate pagination and dimensions in 300

Videorecordings - DVD

mark_kosta — 2024-11-19 04:38:54 UTC

The LDR, 007 and 008 values need improvement.

Videorecordings - DVD (Mark K)

mark_kosta — 2024-11-19 04:45:13 UTC

A mixed bag, some fields were consistently inaccurate LDR, 008. Field 007 was always missing. The descriptive metadata was generally good to very good.

Monograph - German (Mark K)

mark_kosta — 2024-11-19 04:51:33 UTC

The record needs to be enhanced before it would be acceptable for USYD records. Values in fields are incomplete or inaccurate. However descriptive fields 520 and subject heading are good.

Videorecordings - DVD

mark_kosta — 2024-11-19 04:54:02 UTC

007 field missing. The metadata is often complete but unreliable.

Monograph - German (Mark K)

mark_kosta — 2024-11-19 05:02:54 UTC

Make LDR and 008 values accurate and trustworthy. Work out a way for AI to not hallucinate 264 pagination and dimensions.

First nations resouces (in language)

lizforwork334 — 2024-11-21 03:28:02 UTC

Title: Naatji Nyordo

MMSID: 991032666757705106

Failed to add language (Noongar) in 650 despite it being in the image.

041/008 had eng which is incorrect.

Error message when uploading images other than tp, colophon, index, contents etc (Danni)

danniellewilliams1 — 2024-11-21 03:50:25 UTC

Monograph - latin. Two images uploaded, one of the title page and one of a bookplate. Resulted in error message stating "AI assistant processing for [title] could not be completed". Using just the image of the titlepage resulted in a record. Instead of discarding information from images, it issued a blanket error and did not result in a record.

Multi-volume works

lizforwork334 — 2024-11-21 03:53:27 UTC

Title: Wartime rescue of Jews by the Polish Catholic clergy

MMSID: 991032666757605106

Correctly identified multi-volume work form title pages x 2

Monographs - Non fiction

lizforwork334 — 2024-11-21 04:35:41 UTC

Title: An Introduction to Property Law in Australia

MMSID: 991032666757505106

Completely hallucinated the 300 field. No idea where it got that from!

Subfield punctuation

lizforwork334 — 2024-11-21 04:37:24 UTC

This has definitely improved since testing began!