Include Page | ||||
---|---|---|---|---|
|
Warning | ||
---|---|---|
| ||
|
Background
What makes a good use case?
...
We need to be careful to not put in effort in areas where other projects are already working.
Clustering use cases
The purpose of this table is to group use cases into 5 clusters (columns) to help identify a small set of exemplar implementable use cases that can be the focus of engineering work. Good candidates to exemplar or to be merged to form an exemplar are marked "top".
Goal | bib+curation | bib+person data | leverage external authorities | leverage deeper graph via queries or patterns | leverage usage data | no cluster |
---|---|---|---|---|---|---|
use cases by cluster in priority groupings | top: 3, 11, 24, 35 | | challenges: 18 | top: 6, 9 | top: 22, 23 | med: 19, 25, 28, 42 | challenges: 5, 21, 32 | low: 15 | top: 12, 38 | med: 29 | challenges: 14, 17, 41 | low: 39 | top: 1 | med: 16 | challenges: 10,26 | challenges: 30,31,33,34,37, 27 | low 2,4,7,8,13,40 |
pragmatic | 3 | 22, 42 | 41 | 16 | ||
community added-value | 35 | 1 | ||||
cross-institutional data | 3 | 5, 22, 32 | 38 | 16 | 31 | |
leverage existing external authorities | 23, 19, 25, 42, 21, 15 | 38 | 8 | |||
leverage researcher networking data | 6 | 23 | 7 | |||
leverage existing sources of LOD | 22, 23, 42 | 38, 41 | ||||
integration out into the Web | 22, 19 | |||||
cross-discipline | ||||||
help core missions | 12, 14, 38 | |||||
multi-data | ||||||
highlight unusual data | 5 | 12, 14 | 1, 16 | |||
media "photogenic" | 22, 42 | 12, 17, 41 | ||||
interesting analysis or visualization | 11 | 38, 17 | ||||
take advantage of usage data | 1, 16, 10, 26 |
Bib + Curation CLuster
Build a virtual collection
Use Cases (suggestions, in draft form)
...
Cluster
...
Demonstrates scholarly communities learning from one another across institutions
This divides up by subject; others are possible
...
5.1 usage
...
H5 C5 S0
+ value of non-traditional data
+ cross-lib
- no data
...
Heavily focused on the UI. How central should that be to the use cases.
How much should people be able to say about the collection itself and not just its items. Or info about the items but specific to the collection.
...
H5 S5 C5
+ cross institution
+ pragmatic value
...
Maps are a gnarly type. And how do we get the geolocated items? And this isn't very fresh. Is there a way to show relations among the pins? Or something? - dw May be possible to do autotagging by place names but may not be reliable, and are we tagging author origin, publisher, aboutness? jcr
Perhaps restrict their viewing to a geo location.
We'd need a gazetteer.
Nasty data
Not linked data base
...
H4 C4 S0
If we have the data. (Harvard does.) Also, scale it back.
+ cross-instit.
+ unusual data
- no data?
...
requires a commitment to maintain information on departed faculty in other systems of record, or taking on that task in the library. jcr
we'd be ok with just current data
good one
integrates person data, so good
...
S5 H4 C4
Not very dramatic and may already have been done, but worth supporting
...
S5 H0
-No data.
...
Doesn't demonstrate linked data, and similar to a couple of others - dw
Population across the three institutions is big enough.
...
S0 H3 This is an application of #1.
+ cross-institutional
...
You'd have to log in. (Shelf.io does that)
A top priority for Stanford and something they want to get out of this grant
Privileged or not privileged? My personal list or public?
Great area to explore
A third party app might do this, with the three systems incorporating
...
H4 S5
Lists are a subset of virtual collections (#3)
...
S0 H5
+ Great for the media
+ Discovery tool
+ Shows use of new data
...
S0 H4
+ Discovery
+ Unusual data
-- no data
-- SNAC
-- Data model??
...
As it stands this use case doesn't seem to leverage our work. If the method by which one decided which "John Smith" you wanted used other information about the person, their works, their geography, etc. then it would seem to fit. Still seems a somewhat standalone tool though – sw
...
3.4 leveraging authorities
...
S5 H0
-Does not need our data
...
S0 H4.5
+ useful
+ uses unusual data
(It's another use of #1)
...
I would like to be able to ask questions about the relationships among library data using something like natural language (although I'd be happy to accept some common sense restrictions) and get back interesting results. For example,
"What books were written by Cornell biologists on topics that Harvard and Stanford biologists rarely write books about?"
"What books are available about communication technology that don't use 'Communication' in their title or subtitle?"
"Create a timeline of science books that charts them by how many of those books have illustrations."
"Find all the anthologies that have chapters by both Marshall McLuhan and Neil Postman."
"Create a table that shows how many genetics books were published in England per year versus how many were published in the United States."
"Find me books about Christian Fundamentalism read (or assigned) in Divinity schools in the 1990s but less so after 2000."
"Show me chronologically the usage in medical schools of books under the headings of both vaccination and autism."
"Map by publisher location the clustering of books about American slavery since 1640."
(Library Graph)
...
(We could limit the NLP issue by creating a pick list of operations, relations, and outcomes. Or maybe there's an Open Source NLP library we could use. I dunno. - dw)
I think the NLP part of this is too much to bite off. I think it would be a reasonable goal for this project to use a structure query language to express and then answer these queries, I think the NLP part would be extremely cool but would need NLP expert collaborators – sw
Keep an eye out for an NLP person who might want to hook up an NLP front end to this, for a demo.
...
S0 H5
+Great for the media
(Remember that these are use cases, not things we're saying we'll build. And we have the data – not the NLP – to do this. I.e., we should be aiming at enabling someone else to build this.)
...
S5 H5 but only if it brings in outside sources
...
I would like intelligent term expansion:
- based more on entities and relationships than on purely linguistic analysis
- that predictively disambiguate query strings (e.g., bank)
...
This would rely on some of the ideas of authority control #15 – sw
OCLC Fast?
We could do this for authors etc. but how well by navigating the linked data we have?
...
H4.5 S5
+cross-inst.
+external data LD
+pragmatic
+media
...
S5 H5 depending on data implications
(Similar to #22)
+pragmatic
...
aka LD-powered CuLLR for SearchWorks
...
H5 S5
(Not sure what it implies for the data model)
...
H4 S5
+ useful
(see #15)
...
So I can uncover relationships and knowledge that were otherwise hidden.
Maybe do this more as a specific narrative?
...
SPARQL-based OPAC + directory + circ reporting?
Important
Great if we can come up with some good UI, or subset UI that does interesting things. Is there an existing UI to drop on top of a SPARQL endpoint? We don't know of any.
Make this use case more specific? And we need some concrete examples.
Can we do inferencing?? Make that its own use case?
...
S0 H0 This isn't a use case. These capabilties would support use cases, but these capabilities are what the project overall is about, aren't they?
...
H0 S0
-no data
...
I'd like to be able to query which library resources were used in courses (as noted by syllabi, reading lists or course reserves) and compare that to a collection's holdings. (This may be at my or another institution)
...
more specific case of #8
combine with 8. drop 8
...
H0 S0
-no data
...
H4 S0
If we have the data
(cf. #14)
...
What else do we have that is EAD-related?
#14 would also rely upon decomposition of EAD – sw
...
H0 S0
-hard
-how different from a faculty profiler?
...
H0 S0
-low utility
-part of faculty profiler?
...
H0 S0
-no data
...
I like this one as it uses relatedness along different axes combined with type taxa and and entity – sw
...
4.1 deeper graph
...
H5 S0
+ external data
+ discovery
+ cross-inst.
...
Where is the data linking publications to equipment or facilities? – sw
...
H0 S0
-no data
...
How do we infer dataset use where there is no citation? – sw
...
H0 S0
-no data
...
H5 S0
-useful
-external data
-media photogenic
...
H5 S0
-useful
-external data
-media photogenic
Works: Curation
Virtual Collections (3,11,35)
Build a virtual collection from items across institutions, regardless of their original collection and metadata.
Metadata Assignment (24)
Assign information about an item in a Linked Data way, such as subject/tag. A potential implementation strategy for designating virtual collections.
Works: Discovery
Discovery of Related Items (12,19,21,22,23,25,29,38,42)
Using the graph of relationships between items, improve:
- discovery algorithms
- discovery interfaces
Based on information about the Items:
- geo/temporal provenance and topic
- authorship
- other metadata
Persons
Highlight specific people (6)
Given a set of people, highlight them when they are referenced from Items (eg faculty authors)
Display relationships between people (9)
Given a set of people and their relationships to each other, display the relationships in relevant contexts
Display intersection of metadata about people (42)
Given a set of people and their relationships to other objects, use the metadata from those objects to generate sets of people and display those sets in relevant contexts
Usage
Improve Discovery based on Usage (1,16)
Given usage information about a set of objects, improve discovery and ranking of those objects in appropriate interfaces.