[OPEN-ILS-GENERAL] ***SPAM*** Re: Awesome Box Integration

[Vanya Jauhal]

Hello Everyone

As Kathy mentioned, making the recommendation system configurable would
make the system more flexible and the administrators can decide their
settings depending upon their respective needs. As far as the privacy
issues are concerned, we can initially build the model on people who
volunteer. And yes, we will have to implement it in blocks. And as you
mentioned, collecting the awesome tags and sending them along to the
Awesome Box site can be the starting point. And the ML part can be built
simultaneously.
I am currently working on Detecting and Categorizing Circles in Social
Networks, which also works on a similar principle, so I guess training the
data as it gets collected will not be a very huge overhead.

Also I think Rogan has a very good point. We can store the data independent
of the patron data. As in every user can be given a user id which will be
used for all his further transaction records. This thing will kind of
anonymise the user information. And the access to the table linking every
user with corresponding user id can be restricted as per the needs of the
organization. So, the suggestion system can be built on this without loss
of any information.
Also this will help us to give proper recommendations to users which
otherwise would not have been possible.

What do you all think?



On Fri, Sep 26, 2014 at 9:26 AM, Rogan Hamby <rogan.hamby at yclibrary.net>
wrote:

> I'm concerned with project creep as well as I noted in one of early
> missives.  If this is stored independent of patron data (which actually I
> think it should) then I think we should also track circs since the feature
> was turned on so it could say "3 out of 4" people found it awesome.
>
> Stepping back a bit to recommendations and anonymizing records, we don't
> anonymize historical circs.  We don't expose that data and take staff level
> access to it pretty seriously.  Due to varying state and county regulations
> dictating minimum record retentions we're still at least 2 years out from
> being to safely wipe our oldest records.  Maybe more.
>
> And anonymizing it closes certain opportunities.  Some are mundane like
> addressing old conflicts and billing questions but those can be big in
> their own right.  As the circ manager who talks to the upset patron I may
> have a different point of view on that.  :)
>
> Analyzing circulation patterns is far more interesting though and I am
> long term interested in recommendations.  In the age of Anazin, Netflix and
> everyone else this is not just valuable but expected.  It's perhaps the
> patron request I hear most.
>
> Coupled with some holds features it would be a great great boon for home
> bound services which I feel are a critical function of libraries, at least
> in my state where it's a strong traditional service.  I assume elsewhere as
> well though I know mileage varies.
>
> And it was the building block of several functions that GA PINES
> identified as critical for TBS support during the Loblolly conference.  We
> may never fully support TBS programs in Evergreen but I thought GA PINES
> collected a lot of great ideas and input there and would hate to discard
> that.
>
> On Thursday, September 25, 2014, Kathy Lussier <klussier at masslnc.org>
> wrote:
>
>>  Hi all,
>>
>> Great discussion so far!
>>
>> We had a bit of a discussion about privacy concerns in IRC after Terran
>> sent her original message. One approach we were discussing was storing the
>> awesome tags in an anonymous fashion, except in cases where patrons have
>> opted into saving their circ history. In those cases, the user has already
>> consented to having this information saved and could have a more enhanced
>> experience with the recommendation engine. Others who were part of the
>> discussion could elaborate or correct me if I'm not articulating the ideas
>> correctly. The discussion can be found at
>> http://irc.evergreen-ils.org/evergreen/2014-09-25#i_126632.
>>
>> In relation to genres, Vanya said:
>>
>> Maybe, as a solution to that, we can have a hierarchical algorithm for
>> categorizing. In other words, we can allow the administrator to decide
>> whether the categorization comes all the way down to genres, or just takes
>> into account the overall weight of the user's awesome tag.
>>
>>
>> I like the idea of making this configurable, because there may be systems
>> where data identifying genre is a little more clear cut. Better yet, how
>> about if we allow an Evergreen site to define the categories that are used.
>> Some sites may use the MARC fixed fields for fiction/non-fiction. Other
>> sites may decided that values stored in the 655 MARC field work for them.
>>
>> Is there something already exists in Evergreen that we could leverage for
>> this purpose? My first thought was MVF.
>>
>> I do have one general recommendation speaking with my OPW admin hat on.
>> It really is a  general recommendation for any of the OPW candidates who
>> might be following along. I mentioned in IRC today that I'm not a
>> developer, but I've managed a lot of development projects, and one thing I
>> try to watch out for is project creep. As we continue to talk about the
>> project and think of new configuration options to make it a more flexible
>> project, it can also become a very large project that isn't as easy to
>> manage.
>>
>> Therefore, as you think through how you plan to implement the project, I
>> recommend breaking it up into distinct milestones. You might want to start
>> with smaller tasks as you ease into the project (e.g. collecting the
>> awesome tags and sending them along to the Awesome Box site), and then move
>> on to the larger components once you become more familiar with the system.
>>
>> Kathy
>>
>>
>> Kathy Lussier
>> Project Coordinator
>> Massachusetts Library Network Cooperative
>> (508) 343-0128klussier at masslnc.org
>> Twitter: http://www.twitter.com/kmlussier
>> #evergreen IRC: kmlussier
>>
>> On 9/25/2014 6:40 PM, Tim Spindler wrote:
>>
>> Overall, I really like the ideas talked about but I agree with Terran
>> that something would have to be done with circ data related to patrons.  We
>> use the purge function to anonymize our patron data but I could see other
>> ways of dealing with this.   We also have retention policies related to
>> retaining patron circulation data.
>>
>> On Thu, Sep 25, 2014 at 4:54 PM, Rogan Hamby <rogan.hamby at yclibrary.net>
>> wrote:
>>
>>> I suppose I don't understand the concern on your part as at that level
>>> if someone could access the raw db they could just query someone's
>>> circulation history, fine payments, etc... since those are recorded as
>>> transactions unless you're doing something to anonymize or wipe those as
>>> soon as they're done.  Even then someone could see all current transactions
>>> at that level.
>>>
>>>
>>>
>>> On Thu, Sep 25, 2014 at 4:33 PM, McCanna, Terran <
>>> tmccanna at georgialibraries.org> wrote:
>>>
>>>> This relies on the circulation and rating data still being tied to the
>>>> patron in the system, though - yes, it'd be on the database side and not on
>>>> public view, but it's still creating a picture of a patron's reading
>>>> history that has privacy implications. Of course, this feature should be
>>>> set for systems to enable or disable, so that systems that are concerned
>>>> about privacy simply won't turn it on. (PINES, for example, limits the
>>>> retention of circulation history in the system as much as we can because of
>>>> our privacy policies, so any feature that is linked to a patron's history
>>>> would be unusable for us.)
>>>>
>>>> If ranking data were stored completely independently of the patron,
>>>> then library systems would be able to use it without privacy concerns, and
>>>> patrons wouldn't even need to be logged in to use it  - but then it
>>>> wouldn't be able to give completely customized recommendations to a
>>>> specific patron, either. It's a definite tradeoff.
>>>>
>>>>
>>>> Terran McCanna
>>>> PINES Program Manager
>>>> Georgia Public Library Service
>>>> 1800 Century Place, Suite 150
>>>> Atlanta, GA 30345
>>>> 404-235-7138
>>>> tmccanna at georgialibraries.org
>>>>
>>>> ----- Original Message -----
>>>> From: "Vanya Jauhal" <vanyajauhal at gmail.com>
>>>> To: "Evergreen Discussion Group" <
>>>> open-ils-general at list.georgialibraries.org>
>>>>  Sent: Thursday, September 25, 2014 3:41:02 PM
>>>> Subject: Re: [OPEN-ILS-GENERAL] Awesome Box Integration
>>>>
>>>>
>>>>
>>>> Hello Rogan
>>>>
>>>> This is exactly what I had in mind. All the recommendation processing
>>>> will take place in background, and all the user will see is a
>>>> recommendation and not the information of any other patron. This way his
>>>> experience with Awesome Box will get enhanced.
>>>>
>>>>
>>>> And yes, we can maybe, start off with some broad level genres, like, as
>>>> you mentioned, fiction, non-fiction, documentaries, etc. Then, depending
>>>> upon the infrastructure of the system and the response of that
>>>> categorization, we can build upon the algorithm accordingly.
>>>>
>>>>
>>>> You are right- it would be a big task in itself, but since the number
>>>> of parameters involved are few and explicit, it gets simplified to an
>>>> extent.
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> On Fri, Sep 26, 2014 at 12:50 AM, Rogan Hamby <
>>>> rogan.hamby at yclibrary.net > wrote:
>>>>
>>>>
>>>>
>>>> I don't see an issue with doing analysis of circulation patterns on the
>>>> backend so long as nothing identifying is exposed.
>>>>
>>>>
>>>> For example, if all I saw as a patron was a tab in my opac that said
>>>> "you thought The Yiddish Policeman's Union was Awesome! Some others do did
>>>> also thought this was Awesome .... " I don't see that as different from
>>>> doing the same thing with circulations. It's not telling patrons even what
>>>> the points of comparison were unless they only had a single item in their
>>>> circulation history and even then it doesn't tell them how many other
>>>> patrons, how much, etc....
>>>>
>>>>
>>>> I'm dubious about subject headings also but wouldn't want to dismiss it
>>>> out of hand. It might work. Without doing some experimenting I could see it
>>>> going either way. Some fixed fields I could see working, like fiction and
>>>> non-fiction. Age groups? Well, at least I can tell you I can't rely on
>>>> those in my catalog. :)
>>>>
>>>>
>>>> However, I also worry that reading recommendations based on circulation
>>>> history could easily grow into a much more complicated task, especially
>>>> depending on how we deliver those recommendations. Looking at a single
>>>> boolean value tied to the user and item (circ table?) could still be quite
>>>> a project by itself especially once all the useful bits and pieces are
>>>> built in.
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> On Thu, Sep 25, 2014 at 2:37 PM, McCanna, Terran <
>>>> tmccanna at georgialibraries.org > wrote:
>>>>
>>>>
>>>> Agreed - it's a great idea in theory, but I'm not sure how well it
>>>> would work in actual practice. Even in a single library, genre subject
>>>> headings are usually pretty inconsistent in the MARC records because of
>>>> copy cataloging, and that usually gets even more inconsistent in a
>>>> consortium of libraries. Perhaps it could be partially weighted on genre
>>>> subject headings, but not overly reliant on them? It might be worth
>>>> considering the fixed field values for fiction vs. non-fiction and for age
>>>> groups, too.
>>>>
>>>> I love the idea of providing recommendations based on other people that
>>>> have similar taste ("other people that liked this book also liked these
>>>> books...") but if the data is tied to actual patrons (and I'm not sure how
>>>> it couldn't be) then quite a few library systems would face legal privacy
>>>> issues and wouldn't be able to use it. We're currently using a commercial
>>>> service to pull in reading recommendations because the recommendations
>>>> can't be tied back to any of our patrons.
>>>>
>>>>
>>>> Terran McCanna
>>>> PINES Program Manager
>>>> Georgia Public Library Service
>>>> 1800 Century Place, Suite 150
>>>> Atlanta, GA 30345
>>>> 404-235-7138
>>>> tmccanna at georgialibraries.org
>>>>
>>>>
>>>>
>>>> ----- Original Message -----
>>>> From: "Rogan Hamby" < rogan.hamby at yclibrary.net >
>>>> To: "Evergreen Discussion Group" <
>>>> open-ils-general at list.georgialibraries.org >
>>>> Sent: Thursday, September 25, 2014 2:02:58 PM
>>>> Subject: Re: [OPEN-ILS-GENERAL] Awesome Box Integration
>>>>
>>>>
>>>> I can see some challenges to tracking genre and I'd be hesitant to put
>>>> too much value on it. There are ways to catalog it but in my experience
>>>> actually relying on it being in records (much less being consistent) is
>>>> very unreliable in organizations that do a lot of copy cataloging / don't
>>>> have centralized and controlled cataloging and there quite a few in that
>>>> boat.
>>>>
>>>>
>>>> That concern aside, I've always thought this would be a fun and
>>>> potentially valuable thing to add.
>>>>
>>>>
>>>> On Thu, Sep 25, 2014 at 1:44 PM, Vanya Jauhal < vanyajauhal at gmail.com
>>>> > wrote:
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> Hello everyone
>>>>
>>>> I'm Vanya, from India. I'm a candidate for OPW Round9 internship with
>>>> evergreen.
>>>>
>>>> While discussing the idea of Awesome Box integration with Evergreen,
>>>> Kathy and I discussed the possibility of making the Evergreen support for
>>>> Awesome Box more interpretive using Artificial Intelligence.
>>>>
>>>> What if we could train the system to give weightage to people's
>>>> "awesome" tags on items, depending upon how much their previous tags are
>>>> appreciated by other people.
>>>>
>>>> For example: Let's say you tag a book to be awesome. Now, if 100 other
>>>> people check that book in, and (lets say) 80 of them also tag it to be
>>>> awesome- it will mean that your opinion matches a majority of people. On
>>>> the other hand, if 100 other people check that book in and (say) only 5 of
>>>> them tag it as awesome, this would mean that your awesome tag is not in
>>>> coherence with the majority.
>>>> So, in the former case, your awesome tag can be given more weightage as
>>>> compared to the latter.
>>>>
>>>> Also, the weightage may vary according to genres. So- you may have a
>>>> good taste in mystery books but your taste in classical literature might
>>>> not be the same as the majority crowd. So- the weightage of your awesome
>>>> tag in mystery would be higher than classical literature.
>>>>
>>>> We can even extend it to provide recommendations to users depending on
>>>> their coherence with other users with similar taste.
>>>>
>>>> I am looking forward to your suggestions and feedback on this.
>>>>
>>>> Thank you for your time
>>>>
>>>> Vanya
>>>>
>>>>
>>>>
>>>>
>>>> --
>>>>
>>>>
>>>>
>>>> Rogan Hamby, MLS, CCNP, MIA
>>>> Managers Headquarters Library and Reference Services,
>>>> York County Library System
>>>>
>>>>
>>>> “You can never get a cup of tea large enough or a book long enough to
>>>> suit me.”
>>>> ― C.S. Lewis
>>>>
>>>>
>>>>
>>>>
>>>> --
>>>>
>>>>
>>>>
>>>> Rogan Hamby, MLS, CCNP, MIA
>>>> Managers Headquarters Library and Reference Services,
>>>> York County Library System
>>>>
>>>>
>>>> “You can never get a cup of tea large enough or a book long enough to
>>>> suit me.”
>>>> ― C.S. Lewis
>>>>
>>>
>>>
>>>
>>>  --
>>>
>>> Rogan Hamby, MLS, CCNP, MIA
>>> Managers Headquarters Library and Reference Services,
>>> York County Library System
>>>
>>>  “You can never get a cup of tea large enough or a book long enough to
>>> suit me.”
>>> ― C.S. Lewis <http://www.goodreads.com/author/show/1069006.C_S_Lewis>
>>>
>>
>>
>>
>> --
>> Tim Spindler
>> tjspindler at gmail.com
>>
>>  *P**   Go Green - **Save a tree! Please don't print this e-mail unless
>> it's really necessary.*
>>
>>
>>
>>
>>
>
> --
>
> Rogan Hamby, MLS, CCNP, MIA
> Managers Headquarters Library and Reference Services,
> York County Library System
>
> “You can never get a cup of tea large enough or a book long enough to suit
> me.”
> ― C.S. Lewis <http://www.goodreads.com/author/show/1069006.C_S_Lewis>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://libmail.georgialibraries.org/pipermail/open-ils-general/attachments/20140926/5b3a90d2/attachment-0001.htm>