Microsoft Word - June_ITAL_Owen_final.docx Engine  of  Innovation:     Building  the  High  Performance  Catalog        Will  Owen  and   Sarah  C.  Michalak     INFORMATION  TECHNOLOGY  AND  LIBRARIES  |  JUNE  2015               5   ABSTRACT   Numerous  studies  have  indicated  that  sophisticated  web-­‐based  search  engines  have  eclipsed  the   primary  importance  of  the  library  catalog  as  the  premier  tool  for  researchers  in  higher  education.   We  submit  that  the  catalog  remains  central  to  the  research  process.  Through  a  series  of  strategic   enhancements,  the  University  of  North  Carolina  at  Chapel  Hill,  in  partnership  with  the  other   members  of  the  Triangle  Research  Libraries  Network  (TRLN),  has  made  the  catalog  a  carrier  of   services  in  addition  to  bibliographic  data,  facilitating  not  simply  discovery,  but  also  delivery  of  the   information  researchers  seek.   INTRODUCTION In  2005,  an  OCLC  research  report  documented  what  many  librarians  already  knew—that  the   library  webpage  and  catalog  were  no  longer  the  first  choice  to  begin  a  search  for  information.  The   report  states,   The  survey  findings  indicate  that  84  percent  of  information  searches  begin  with  a  search   engine.  Library  Web  sites  were  selected  by  just  1  percent  of  respondents  as  the  source  used  to   begin  an  information  search.  Very  little  variability  in  preference  exists  across  geographic   regions  or  U.S.  age  groups.  Two  percent  of  college  students  start  their  search  at  a  library  Web   site.1   In  2006  a  report  by  Karen  Calhoun,  commissioned  by  the  Library  of  Congress,  asserted,  “Today  a   large  and  growing  number  of  students  and  scholars  routinely  bypass  library  catalogs  in  favor  of   other  discovery  tools.  .  .  .  The  catalog  is  in  decline,  its  processes  and  structures  are  unsustainable,   and  change  needs  to  be  swift.”2     Ithaka  S+R  has  conducted  national  faculty  surveys  triennially  since  2000.  Summarizing  the  2000– 2006  surveys,  Roger  Schonfeld  and  Kevin  Guthrie  stated,  “When  the  findings  from  2006  are   compared  with  those  from  2000  and  2003,  it  becomes  evident  that  faculty  perceive  themselves  as   becoming  decreasingly  dependent  on  the  library  for  their  research  and  teaching  needs.”3   Furthermore,  it  was  clear  that  the  “library  as  gateway  to  scholarly  information”  was  viewed  as   decreasingly  important.  The  2009  survey  continued  the  trend  with  even  fewer  faculty  seeing  the       Will  Owen  (owen@email.unc.edu)  is  Associate  University  Librarian  for  Technical  Services  and   Systems  and  Sarah  C.  Michalak  (smichala@email.unc.edu)  is  University  Librarian  and  Associate   Provost  for  University  Libraries,  University  of  North  Carolina  at  Chapel  Hill.     ENGINE  OF  INNOVATION:  BUILDING  THE  HIGH-­‐PERFORMANCE  CATALOG  |  OWEN  AND  MICHALAK       doi:  10.6017/ital.v34i2.5702   6   gateway  function  as  critical.  These  results  occurred  in  a  time  when  electronic  resources  were   becoming  increasingly  important  and  large  Google-­‐like  search  engines  were  rapidly  gaining  in   use.4     These  comments  extend  into  the  twenty-­‐first  century  more  than  thirty  years  of  concern  about  the   utility  of  the  library  catalog.  Through  the  first  half  of  this  decade  new  observations  emerged  about   patron  perceptions  of  catalog  usability.  Even  after  migration  from  the  card  to  the  online  catalog   was  complete,  the  new  tool  represented  primarily  the  traditionally  cataloged  holdings  of  a   particular  library.  Providing  direct  access  to  resources  was  not  part  of  the  catalog’s  mission.   Manuscripts,  finding  aids,  historical  photography,  and  other  special  collections  were  not  included   in  the  traditional  catalog.  Journal  articles  could  only  be  discovered  through  abstracting  and   indexing  services.  As  these  discovery  tools  began  their  migration  to  electronic  formats,  the   centrality  of  the  library’s  bibliographic  database  was  challenged.   The  development  of  Google  and  other  sophisticated  web-­‐based  search  engines  further  eclipsed  the   library’s  bibliographic  database  as  the  first  and  most  important  research  tool.  Yet  we  submit  that   the  catalog  database  remains  a  necessary  fixture,  continuing  to  provide  access  to  each  library’s   particular  holdings.  While  the  catalog  may  never  regain  its  pride  of  place  as  the  starting  point  for   all  researchers,  it  still  remains  an  indispensable  tool  for  library  users,  even  if  it  may  be  used  only   at  a  later  stage  in  the  research  process.   At  the  University  of  North  Carolina  at  Chapel  Hill,  we  have  continued  to  invest  in  enhancing  the   utility  of  the  catalog  as  a  valued  tool  for  research.  Librarians  initially  reasoned  that  researchers   still  want  to  find  out  what  is  available  to  them  in  their  own  campus  library.  Gradually  they  began   to  see  completely  new  possibilities.  To  that  end,  we  have  committed  to  a  program  that  enhances   discovery  and  delivery  through  the  catalog.  While  most  libraries  have  built  a  wide  range  of   discovery  tools  into  their  home  pages—adding  links  to  databases  of  electronic  resources,  article   databases,  and  Google  Scholar—we  have  continued  to  enhance  both  the  content  to  be  found  in  the   primary  local  bibliographic  database  and  the  services  available  to  students  and  researchers  via  the   interface  to  the  catalog.   In  our  local  consortium,  the  Triangle  Research  Libraries  Network  (TRLN),  librarians  have   deployed  the  search  and  faceting  services  of  Endeca  to  enrich  the  discovery  interfaces.  We  have   gone  beyond  augmenting  the  catalog  through  the  addition  of  MARCIVE  records  for  government   documents,  by  including  Encoded  Archival  Description  (EAD)  finding  aids  and  selected  (and  ever-­‐ expanding)  digital  collections  that  are  not  easily  discoverable  through  major  search  engines.  We   have  similarly  enhanced  services  related  to  the  discovery  and  delivery  of  items  listed  in  the   bibliographic  database,  including  not  only  common  features  like  the  ability  to  export  citations  in  a   variety  of  formats  but  also  more  extensive  services  such  as  document  delivery,  an  auto-­‐suggest   feature  that  maximizes  use  of  Library  of  Congress  Subject  Headings  (LCSH),  and  the  ability  to   submit  cataloged  items  to  be  processed  for  reserve  reading.     INFORMATION  TECHNOLOGY  AND  LIBRARIES  |  JUNE  2015     7   Both  students  and  faculty  have  embraced  e-­‐books,  and  in  adding  more  than  a  million  such  titles  to   the  UNC-­‐Chapel  Hill  catalog  we  continue  to  blend  discovery  and  delivery,  but  now  on  a  very  large   scale.  Coupling  catalog  records  with  a  metadata  service  that  provides  book  jackets,  tables  of   contents,  and  content  summaries,  cataloging  Geographic  Information  Systems  (GIS)  data  sets,  and   adding  live  links  to  the  finding  aids  for  digitized  archival  and  manuscript  collections  have  further   enhanced  the  blended  discovery/delivery  capacity  of  the  catalog.   We  have  also  leveraged  the  advantages  of  operating  in  a  consortial  environment  by  extending  the   discovery  and  delivery  services  among  the  members  of  TRLN  to  provide  increased  scope  of   discovery  and  shared  processing  of  some  classes  of  bibliographic  records.  TRLN  comprises  four   institutions  and  content  from  all  member  libraries  is  discoverable  in  a  combined  catalog   (http://search.trln.org).  Printed  material  requested  through  this  combined  catalog  is  often   delivered  between  TRLN  libraries  within  twenty-­‐four  hours.   At  UNC,  our  search  logs  show  that  use  of  the  catalog  increases  as  we  add  new  capacity  and  content.   These  statistics  demonstrate  the  catalog’s  continuing  relevance  as  a  research  tool  that  adds  value   above  and  beyond  conventional  search  engines  and  general  web-­‐based  information  resources.  In   this  article  we  will  describe  the  most  important  enhancements  to  our  catalog,  include  data  from   search  logs  to  demonstrate  usage  changes  resulting  from  these  enhancements,  and  comment  on   potential  future  developments.   LITERATURE  REVIEW   An  extensive  literature  discusses  the  past  and  future  of  online  catalogs,  and  many  of  these   materials  themselves  include  detailed  literature  reviews.  In  fact,  there  are  so  many  studies,   reviews,  and  editorials,  it  becomes  clear  that  although  the  online  catalog  may  be  in  decline,  it   remains  a  subject  of  lively  interest  to  librarians.  Two  important  threads  in  this  literature  report  on   user-­‐query  studies  and  on  other  usability  testing.  Though  there  are  many  earlier  studies,  two   relatively  recent  articles  analyze  search  behavior  and  provide  selective  but  helpful  literature   surveys.5     There  are  many  efforts  to  define  directions  for  the  catalog  that  would  make  it  more  web-­‐like,  more   Google-­‐like,  and  thus  more  often  chosen  for  search,  discovery,  and  access  by  library  patrons.   These  articles  aim  to  define  the  characteristics  of  the  ideal  catalog.  Charles  Hildreth  provides  a   benchmark  for  these  efforts  by  dividing  the  history  of  the  online  catalog  into  three  generations.   From  his  projections  of  a  third  generation  grew  the  “next  generation  catalog”—really  the  current   ideal.  He  called  for  improvement  of  the  second-­‐generation  catalog  through  an  enhanced  user-­‐ system  dialog,  automatic  correction  of  search-­‐term  spelling  and  format  errors,  automatic  search   aids,  enriched  subject  metadata  in  the  catalog  record  to  improve  search  results,  and  the   integration  of  periodical  indexes  in  the  catalog.  As  new  technologies  have  made  it  possible  to   achieve  these  goals  in  new  ways,  much  of  what  Hildreth  envisioned  has  been  accomplished.6       ENGINE  OF  INNOVATION:  BUILDING  THE  HIGH-­‐PERFORMANCE  CATALOG  |  OWEN  AND  MICHALAK       doi:  10.6017/ital.v34i2.5702   8   Second-­‐generation  catalogs,  anchored  firmly  in  integrated  library  systems,  operated  throughout   most  of  the  1980s  and  the  1990s  without  significant  improvement.  By  the  mid-­‐2000s  the  search   for  the  “next-­‐gen”  catalog  was  in  full  swing,  and  there  are  numerous  articles  that  articulate  the   components  of  an  improved  model.  The  catalog  crossed  a  generational  line  for  good  when  the   North  Carolina  State  University  Libraries  (NCSU)  launched  a  new  catalog  search  engine  and   interface  with  Endeca  in  January  2006.  Three  NCSU  authors  published  a  thorough  article   describing  key  catalog  improvements.  Their  Endeca-­‐enhanced  catalog  fulfilled  the  most  important   criteria  for  a  “next-­‐gen”  catalog:  improved  search  and  retrieval  through  “relevance-­‐ranked  results,   new  browse  capabilities,  and  improved  subject  access.”7     Librarians  gradually  concluded  that  the  catalog  need  not  be  written  off  but  would  benefit  from   being  enhanced  and  aligned  with  search  engine  capabilities  and  other  web-­‐like  characteristics.   Catalogs  should  contain  more  information  about  titles,  such  as  book  jackets  or  reviews,  than   conventional  bibliographic  records  offered.  Catalog  search  should  be  understandable  and  easy  to   use.  Additional  relevant  works  should  be  presented  to  the  user  along  with  result  sets.  The   experience  should  be  interactive  and  participatory  and  provide  access  to  a  broad  array  of   resources  such  as  data  and  other  nonbook  content.8     Karen  Markey,  one  of  the  most  prolific  online  catalog  authors  and  analysts,  writes,  “Now  that  the   era  of  mass  digitization  has  begun,  we  have  a  second  chance  at  redesigning  the  online  library   catalog,  getting  it  right,  coaxing  back  old  users  and  attracting  new  ones.”9   Marshall  Breeding  predicted  characteristics  of  the  next-­‐generation  catalog.  His  list  includes   expanded  scope  of  search,  more  modern  interface  techniques,  such  as  a  single  point  of  entry,   search  result  ranking,  faceted  navigation,  and  “did  you  mean  .  .  .  ?”  capacity,  as  well  as  an  expanded   search  universe  that  includes  the  full  text  of  journal  articles  and  an  array  of  digitized  resources.10     A  concept  that  is  less  represented  in  the  literature  is  that  of  envisioning  the  catalog  as  a   framework  for  service,  although  the  idea  of  the  catalog  designed  to  ensure  customer  self-­‐service   has  been  raised.11  Michael  J.  Bennett  has  studied  the  effect  of  catalog  enhancements  on  circulation   and  interlibrary  loan.12  Service  and  the  online  catalog  have  a  new  meaning  in  Morgan’s  idea  of   “services  against  texts,”  supporting  “use  and  understand”  in  addition  to  the  traditional  “find  and   get.”13  Lorcan  Dempsey  commented  on  the  catalog  as  an  identifiable  service  and  predicts  new   formulations  for  library  services  based  on  the  network-­‐level  orientation  of  search  and  discovery.14   But  the  idea  that  the  catalog  has  moved  from  a  fixed,  inward-­‐focused  tool  to  an  engine  for   services—a  locus  to  be  invested  with  everything  from  unmediated  circulation  renewal  and   ordering  delivery  to  the  “did  you  mean”  search  aid—has  yet  to  be  addressed  comprehensively  in   the  literature.   ENHANCING  THE  TRADITIONAL  CATALOG   One  of  the  factors  that  complicates  discussions  of  the  continued  relevance  of  the  library  catalog  to   research  is  the  very  imprecision  of  the  term  in  common  parlance,  especially  when  the  chief  point     INFORMATION  TECHNOLOGY  AND  LIBRARIES  |  JUNE  2015     9   of  comparison  to  today’s  ILS-­‐driven  OPACs  is  Google  or,  more  specifically,  Google  Scholar.  From   first-­‐year  writing  assignments  through  advanced  faculty  research,  many  of  the  resources  that  our   patrons  seek  are  published  in  the  periodical  literature,  and  the  library  catalog,  the  one  descended   from  the  cabinets  full  of  cards  that  occupied  prominent  real  estate  in  our  buildings,  has  never  been   an  effective  tool  for  identifying  relevant  periodical  literature.   This  situation  has  changed  in  recent  years  as  products  like  Summon,  from  ProQuest,  and  EBSCO   Discovery  Service  have  introduced  platforms  that  can  accommodate  electronic  article  indexing  as   well  as  MARC  records  for  the  types  of  materials—books,  audio,  and  video—that  have  long  been   discovered  through  the  OPAC.  In  the  following  discussion  of  “catalog”  developments  and   enhancements,  we  focus  initially  not  on  these  integrated  solutions,  but  on  the  catalog  as  more   traditionally  defined.  However,  as  electronic  resources  become  an  ever-­‐greater  percentage  of   library  collections,  we  shall  see  a  convergence  of  these  two  streams  that  will  portend  significant   changes  in  the  nature  and  utility  of  the  catalog.   Much  work  has  been  done  in  the  first  decade  of  the  twenty-­‐first  century  to  enhance  discovery   services  and,  as  noted  above,  North  Carolina  State  University’s  introduction  of  their  Endeca-­‐based   search  engine  and  interface  was  a  significant  game-­‐changer.  In  the  years  following  the   introduction  of  the  Endeca  interface  at  NCSU,  the  Triangle  Research  Libraries  Network  invested  in   further  development  of  features  that  enhanced  the  utility  of  the  Endeca  software  itself.   Programmed  enhancements  to  the  interface  provided  additional  services  and  functionality.  In   some  cases,  these  enhancements  were  aimed  at  improving  discovery.  In  others,  they  allowed   researchers  to  make  new  and  better  use  of  the  data  that  they  found  or  made  it  easier  to  obtain  the   documents  that  they  discovered.   Faceting  and  Limiting  Retrieval  Results   Perhaps  the  most  immediately  striking  innovation  in  the  Endeca  interface  was  the  introduction  of   facets.  The  use  of  faceted  browsing  allowed  users  to  parse  the  bibliographic  record  in  new  ways   (and  more  ways)  than  had  preceding  catalogs.  There  were  several  fundamentally  important  ways   faceting  enhanced  search  and  discovery.   The  first  of  these  was  the  formal  recognition  that  keyword  searching  was  the  user’s  default  means   of  interacting  with  the  catalog’s  data.  NCSU’s  initial  implementation  allowed  for  searches  using   several  indexes,  including  authors,  titles,  and  subject  headings,  and  this  functionality  remains  in   place  to  the  present  day.  However,  by  default,  searches  returned  records  containing  the  search   terms  “anywhere”  in  the  record.  This  behavior  was  more  in  line  with  user  expectations  in  an   information  ecosystem  dominated  by  Google’s  single  search  box.   The  second  was  the  significantly  different  manner  in  which  multiple  limits  could  be  placed  on  an   initial  result  set  from  such  a  keyword  search.  The  concept  of  limiting  was  not  a  new  one:  certain   facets  worked  in  a  manner  consistent  with  traditional  limits  in  prior  search  interfaces,  allowing   users  to  screen  results  by  language,  or  date  of  publication,  for  example.       ENGINE  OF  INNOVATION:  BUILDING  THE  HIGH-­‐PERFORMANCE  CATALOG  |  OWEN  AND  MICHALAK       doi:  10.6017/ital.v34i2.5702   10   It  was  the  ease  and  transparency  with  which  multiple  limits  could  be  applied  through  faceting  that   was  revolutionary.  A  user  who  entered  the  keyword  “java”  in  the  search  box  was  quickly  able  to   discriminate  between  the  programming  language  and  the  Indonesian  island.  This  could  be   achieved  in  multiple  ways:  by  choosing  between  subjects  (for  example,  “application  software”  vs.   “history”)  or  clearly  labeled  LC  classification  categories  (“Q  –  Science”  vs.  “D  –  History”).  These   limits,  or  facets,  could  be  toggled  on  and  off,  independently  and  iteratively.   The  third  and  highly  significant  difference  resulted  from  how  Library  of  Congress  Subject   Headings  (LCSH)  were  parsed  and  indexed  in  the  system.  By  making  LCSH  subdivisions   independent  elements  of  the  subject-­‐heading  index  in  a  keyword  search,  the  Endeca   implementation  unlocked  a  trove  of  metadata  that  had  been  painstakingly  curated  by  catalogers   for  nearly  a  century.  The  user  no  longer  needed  to  be  familiar  with  the  formal  structure  of  subject   headings;  if  the  keywords  appeared  anywhere  in  the  string,  the  subdivisions  in  which  they  were   contained  could  be  surfaced  and  used  as  facets  to  sharpen  the  focus  of  the  search.  This  was   revolutionary.   Utilizing  the  Power  of  New  Indexing  Structures   The  liberation  of  bibliographic  data  from  the  structure  of  MARC  record  indexes  presaged  yet   another  far-­‐reaching  alteration  in  the  content  of  library  catalogs.  To  this  day,  most  commercial   integrated  library  systems  depend  on  MARC  as  the  fundamental  record  structure.  In  NCSU’s   implementation,  the  multiple  indexes  built  from  that  metadata  created  a  new  framework  for   information.     This  change  made  possible  the  integration  of  non-­‐MARC  data  with  MARC  data,  allowing,  for   example,  Dublin  Core  (DC)  records  to  be  incorporated  into  the  universe  of  metadata  to  be  indexed,   searched,  and  retrieved.  There  was  no  need  to  crosswalk  DC  to  MARC:  it  sufficed  to  simply  assign   the  DC  elements  to  the  appropriate  Endeca  indexes.  With  this  capacity  to  integrate  rich  collections   of  locally  described  digital  resources,  the  scope  of  the  traditional  catalog  was  enlarged.   Expanding  Scopes  and  Banishing  Silos   At  UNC-­‐Chapel  Hill,  we  began  this  process  of  augmentation  with  selected  collections  of  digital   objects.  These  collections  were  housed  in  a  CONTENTdm  repository  we  had  been  building  for   several  years  at  the  time  of  the  Library’s  introduction  of  the  Endeca  interface.  Image  files,  which   had  not  been  accessible  through  traditional  catalogs,  were  among  the  first  to  be  added.  For   example,  we  had  been  given  a  large  collection  of  illustrated  postcards  featuring  scenes  of  North   Carolina  cities  and  towns.  These  postcards  had  been  digitized  and  metadata  describing  the  image   and  the  town  had  been  recorded.  Other  collections  of  digitized  historical  photographs  were  also   selected  for  inclusion  in  the  catalog.  These  historical  resources  proved  to  be  a  boon  to  faculty   teaching  local  history  courses  and,  interestingly,  to  students  working  on  digital  projects  for  their   classes.  As  class  assignments  came  to  include  activities  like  creating  maps  enhanced  by  the     INFORMATION  TECHNOLOGY  AND  LIBRARIES  |  JUNE  2015     11   addition  of  digital  photographs  or  digitized  newspaper  clippings,  the  easy  discovery  of  these   formerly  hidden  collections  enriched  students’  learning  experience.   Other  special  collection  materials  had  been  represented  in  the  traditional  catalog  in  somewhat   limited  fashion.  The  most  common  examples  were  manuscripts  collections.  The  processing  of   these  collections  had  always  resulted  in  the  creation  of  finding  aids,  produced  since  the  1930s   using  index  cards  and  typewriters.  During  the  last  years  of  the  twentieth  century,  archivists  began   migrating  many  of  these  finding  aids  to  the  web  using  the  EAD  format,  presenting  them  as  simple   HTML  pages.  These  finding  aids  were  accessible  through  the  catalog  by  means  of  generalized   MARC  records  that  described  the  collections  at  a  superficial  level.  However,  once  we  attained  the   ability  to  integrate  the  contents  of  the  finding  aids  themselves  into  the  indexes  underlying  the  new   interface,  this  much  richer  trove  of  keyword-­‐searchable  data  vastly  increased  the  discoverability   and  use  of  these  collections.   During  this  period,  the  Library  also  undertook  systematic  digitization  of  many  of  these  manuscript   collections.  Whenever  staff  received  a  request  for  duplication  of  an  item  from  a  manuscript   collection  (formerly  photocopies,  but  by  then  primarily  digital  copies),  we  digitized  the  entire   folder  in  which  that  item  was  housed.  We  developed  standards  for  naming  these  digital  surrogates   that  associated  the  individual  image  with  the  finding  aid.  It  then  became  a  simple  matter,  involving   the  addition  of  a  short  JavaScript  string  to  the  head  of  the  online  finding  aid,  to  dynamically  link   the  digital  objects  to  the  finding  aid  itself.     Other  library  collections  likewise  benefited  from  the  new  indexing  structures.  Some  uncataloged   materials  traditionally  had  minimal  bibliographic  control  provided  by  inventories  that  were  built   at  the  time  of  accession  in  desktop  database  applications;  funding  constraints  meant  that  full   cataloging  of  these  materials  (often  rare  books)  remained  elusive.  The  ability  to  take  the  data  that   we  had  and  blend  it  into  the  catalog  enhanced  the  discovery  of  these  collections  as  well.   We  also  have  an  extensive  collection  of  video  resources,  including  commercial  and  educational   films.  The  conventions  for  cataloging  these  materials,  held  over  from  the  days  of  catalog  cards,   often  did  not  match  user  expectations  for  search  and  discovery.  There  were  limits  to  the  number   of  added  entries  that  catalogers  would  make  for  directors,  actors,  and  others  associated  with  a   film.  Many  records  lacked  the  kind  of  genre  descriptors  that  undergraduates  were  likely  to  use   when  seeking  a  film  for  an  evening’s  entertainment.  To  compensate  for  these  limitations,  staff  who   managed  the  collection  had  again  developed  local  database  applications  that  allowed  for  the   inclusion  of  more  extensive  metadata  and  for  categories  such  as  country  of  origin  or  folksonomic   genres  that  patrons  frequently  indicated  were  desirable  access  points.  Once  again,  the  new   indexing  structures  allowed  us  to  incorporate  this  rich  set  of  metadata  into  what  looked  like  the   traditional  catalog.   Each  of  the  instances  described  above  represents  what  we  commonly  call  the  destruction  of  silos.   Information  about  library  collections  that  had  been  scattered  in  numerous  locations—and  not  all   of  them  online—was  integrated  into  a  single  point  of  discovery.  It  was  our  hope  and  intention  that     ENGINE  OF  INNOVATION:  BUILDING  THE  HIGH-­‐PERFORMANCE  CATALOG  |  OWEN  AND  MICHALAK       doi:  10.6017/ital.v34i2.5702   12   such  integration  would  drive  more  users  to  the  catalog  as  a  discovery  tool  for  the  library’s  diverse   collections  and  not  simply  for  the  traditional  monographic  and  serials  collections  that  had  been   served  by  MARC  cataloging.  Usage  logs  indicate  that  the  average  number  of  searches  conducted  in   the  catalog  rose  from  approximately  13,000  per  day  in  2009  to  around  19,000  per  day  in  2013.  It   is  impossible  to  tell  with  any  certainty  whether  there  was  heavier  use  of  the  catalog  simply   because  increasingly  varied  resources  came  to  be  represented  in  it,  but  we  firmly  believe  that  the   experience  of  users  who  search  for  material  in  our  catalog  has  become  much  richer  as  a  result  of   these  changes  to  its  structure  and  content.   Cooperation  Encouraging  Creativity   Another  way  we  were  able  to  harness  the  power  of  Endeca’s  indexing  scheme  involved  the  shared   loading  of  bibliographic  records  for  electronic  resources  to  which  multiple  TRLN  libraries   provided  access.  TRLN’s  Endeca  indexes  are  built  from  the  records  of  each  member.  Each   institution  has  a  “pipeline”  that  feeds  metadata  into  the  combined  TRLN  index.  Duplicate  records   are  rolled  up  into  a  single  display  via  OCLC  control  numbers  whenever  possible,  and  the   bibliographic  record  is  annotated  with  holdings  statements  for  the  appropriate  libraries.   We  quickly  realized  that  where  any  of  the  four  institutions  shared  electronic  access  to  materials,  it   was  redundant  to  load  copies  of  each  record  into  the  local  databases  of  each  institution.15  Instead,   one  institution  could  take  responsibility  for  a  set  of  records  representing  shared  resources.   Examples  of  such  material  include  electronic  government  documents  with  records  provided  by   the  MARCIVE  Documents  Without  Shelves  program,  large  sets  like  Early  English  Books  Online,  and   PBS  videos  streamed  by  the  statewide  services  of  NC  LIVE.   In  practice,  one  institution  takes  responsibility  for  loading,  editing,  and  performing  authority   control  on  a  given  set  of  records.  (For  example,  UNC,  as  the  regional  depository,  manages  the   Documents  Without  Shelves  record  set.)  These  records  are  loaded  with  a  special  flag  indicating   that  they  are  part  of  the  shared  records  program.  This  flag  generates  a  holdings  statement  that   reflects  the  availability  of  the  electronic  item  at  each  institution.  The  individual  holdings   statements  contain  the  institution-­‐specific  proxy  server  information  to  enable  and  expedite  access.   In  addition  to  this  distributed  model  of  record  loading  and  maintenance,  we  were  able  to  leverage   OAI-­‐PMH  feeds  to  add  selected  resources  to  the  SearchTRLN  database.  All  four  institutions  have   access  to  the  data  made  available  by  the  Inter-­‐university  Consortium  for  Political  and  Social   Research  (ICPSR).  As  we  do  not  license  these  resources  or  maintain  them  locally,  and  as  records   provided  by  ICPSR  can  change  over  time,  we  developed  a  mechanism  to  harvest  the  metadata  and   push  it  through  a  pipeline  directly  into  the  SearchTRLN  indexes.  None  of  the  member  libraries’   local  databases  house  this  metadata,  but  the  records  are  made  available  to  all  nonetheless.   While  we  were  engaged  in  implementing  these  enhancements,  additional  sources  of  potential   enrichment  of  the  catalog  were  appearing.  In  particular,  vendors  began  providing  indexing   services  for  the  vast  quantities  of  electronic  resources  contained  in  aggregator  databases.     INFORMATION  TECHNOLOGY  AND  LIBRARIES  |  JUNE  2015     13   Additionally,  they  made  it  possible  for  patrons  to  move  seamlessly  from  the  catalog  to  those   electronic  resources  via  OpenURL  technologies.  Indeed,  services  like  ProQuest’s  Summon  or   EBSCO’s  Discovery  Service  might  be  taken  as  another  step  toward  challenging  the  catalog’s   primacy  as  a  discovery  tool  as  they  offered  the  prospect  of  making  local  catalog  records  just  a   fraction  of  a  much  larger  universe  of  bibliographic  information  available  in  a  single,  keyword-­‐ searchable  database.   It  remains  to  be  seen,  therefore,  whether  continuing  to  load  many  kinds  of  MARC  records  into  the   local  database  is  an  effective  aid  to  discovery  even  with  the  multiple  delimiting  capabilities  that   Endeca  provides.  What  is  certain,  however,  is  that  our  approach  to  indexing  resources  of  any  kind   has  undergone  a  radical  transformation  over  the  past  few  years—a  transformation  that  goes   beyond  the  introduction  of  any  of  the  particular  changes  we  have  discussed  so  far.   Promoting  a  Culture  of  Innovation   One  important  way  Endeca  has  changed  our  libraries  is  that  a  culture  of  constant  innovation  has   become  the  norm,  rather  than  the  exception,  for  our  catalog  interface  and  content.  Once  we  were   no  longer  subject  to  the  customary  cycle  of  submitting  enhancement  requests  to  an  integrated   library  system  vendor,  hoping  that  fellow  customers  shared  similar  desires,  and  waiting  for  a   response  and,  if  we  were  lucky,  implementation,  we  were  able  to  take  control  of  our  aspirations.   We  had  the  future  of  the  interface  to  our  collections  in  our  own  hands,  and  within  a  few  years  of   the  introduction  of  Endeca  by  NCSU,  we  were  routinely  adding  new  features  to  enhance  its   functionality.   One  of  the  first  of  these  enhancements  was  the  addition  of  a  “type-­‐ahead”  or  “auto-­‐suggest”   option.16  Inspired  by  Google’s  autocomplete  feature,  this  service  suggests  phrases  that  might   match  the  keywords  a  patron  is  typing  into  the  search  box.  Ben  Pennell,  one  of  the  chief   programmers  working  on  Endeca  enhancement  at  UNC-­‐Chapel  Hill,  built  a  Solr  index  from  the  ILS   author,  title,  and  subject  indexes  and  from  a  log  of  recent  searches.  As  a  patron  typed,  a  drop-­‐ down  box  appeared  below  the  search  box.  The  drop-­‐down  contained  matching  terms  extracted   from  the  Solr  index  in  a  matter  of  seconds  or  less.  For  example,  typing  the  letters  “bein”  into  the   box  produced  a  list  including  “Being  John  Malkovich,”  “nature—effects  of  human  beings  on,”   “human  beings,”  and  “Bein,  Alex,  1903–1988.”  The  italicized  letters  in  these  examples  are   highlighted  in  a  different  color  in  the  drop-­‐down  display.  In  the  case  of  terms  drawn  directly  from   an  index,  the  index  name  appears,  also  highlighted,  on  the  right  side  of  the  box.  For  example,  the   second  and  third  terms  in  the  examples  above  are  tagged  with  the  term  “subject.”  The  last  example   is  an  “author.”   In  allowing  for  the  textual  mining  of  LCSH,  the  initial  implementation  of  faceting  in  the  Endeca   catalog  surfaced  those  headings  for  the  patron  by  uniting  keyword  and  controlled  vocabularies  in   an  unprecedented  manner.  There  was  a  remarkable  and  almost  immediate  increase  in  the  number   of  authority  index  searches  entered  into  the  system.  At  the  end  of  the  fall  semester  prior  to  the   implementation  of  the  auto-­‐suggest  feature,  an  average  of  around  1,400  subject  searches  were     ENGINE  OF  INNOVATION:  BUILDING  THE  HIGH-­‐PERFORMANCE  CATALOG  |  OWEN  AND  MICHALAK       doi:  10.6017/ital.v34i2.5702   14   done  in  a  week.  Approximately  one  month  into  the  spring  semester,  that  average  had  risen  to   around  4,000  subject  searches  per  week.  Use  of  the  author  and  title  indexes  also  rose,  although   not  quite  as  dramatically.  In  the  perpetual  tug-­‐of-­‐war  between  precision  and  recall,  the  balance   had  decidedly  shifted.   Another  service  that  we  provide,  which  is  especially  popular  with  students,  is  the  ability  to   produce  citations  formatted  in  one  of  several  commonly  used  bibliographic  styles,  including  APA,   MLA,  and  Chicago  (both  author-­‐date  and  note-­‐and-­‐bibliography  formats).  This  functionality,  first   introduced  by  NCSU  and  then  jointly  developed  with  UNC  over  the  years  that  followed,  works  in   two  ways.  If  a  patron  finds  a  monographic  title  in  the  catalog,  simply  clicking  on  a  link  labeled  “Cite”   produces  a  properly  formatted  citation  that  can  then  be  copied  and  pasted  into  a  document.  The   underlying  technology  also  powers  a  “Citation  Builder”  function  by  which  a  patron  can  enter  basic   bibliographic  information  for  a  book,  a  chapter  or  essay,  a  newspaper  or  journal  article,  or  a   website  into  a  form,  click  the  “submit”  button,  and  receive  a  citation  in  the  desired  format.     An  additional  example  of  innovation  that  falls  somewhat  outside  the  scope  of  the  changes   discussed  above  was  the  development  of  a  system  that  allowed  for  the  mapping  of  simplified   Chinese  characters  to  their  traditional  counterparts.  Searching  in  non-­‐Roman  character  sets  has   always  offered  a  host  of  challenges  to  library  catalog  users.  The  TRLN  Libraries  have  embraced  the   potential  of  Endeca  to  reduce  some  of  these  challenges,  particularly  for  Chinese,  through  the   development  of  better  keyword  searching  strategies  and  the  automatic  translation  of  simplified  to   traditional  characters.   Since  we  had  complete  control  over  the  Endeca  interface,  it  proved  relatively  simple  to  integrate   document  delivery  services  directly  into  the  functionality  of  the  catalog.  Rather  than  simply   emailing  a  bibliographic  citation  or  a  call  number  to  themselves,  patrons  could  request  the   delivery  of  library  materials  directly  to  their  campus  addresses.  Once  we  had  implemented  this   feature,  we  quickly  moved  to  amplify  its  power.  Many  catalogs  offer  a  “shopping  cart”  service  that   allows  patrons  to  compile  lists  of  titles.  One  variation  on  this  concept  that  we  believe  is  unique  to   our  library  is  the  ability  for  a  professor  to  compile  such  a  list  of  materials  held  by  the  libraries  on   campus  and  submit  that  list  directly  to  the  reserve  reading  department,  where  the  books  are   pulled  from  the  shelves  and  placed  on  course-­‐reserve  lists  without  the  professor  needing  to  visit   any  particular  library  branch.  These  new  features,  in  combination  with  other  service   enhancements  such  as  the  delivery  of  physical  documents  to  campus  addresses  from  our  on-­‐ campus  libraries  and  our  remote  storage  facility,  have  increased  the  usefulness  of  the  catalog  as   well  as  our  users’  satisfaction  with  the  Library.  We  believe  that  these  changes  have  contributed  to   the  ongoing  vitality  of  the  catalog  and  to  its  continued  importance  to  our  community.   In  December  2012,  the  Libraries  adopted  ProQuest’s  Summon  to  provide  enhanced  access  to   article  literature  and  electronic  resources  more  generally.  At  the  start  of  the  following  fall   semester,  the  Libraries  instituted  another  major  change  to  our  discovery  and  delivery  services   through  a  combined  single-­‐search  box  on  our  home  page.  This  has  fundamentally  altered  how     INFORMATION  TECHNOLOGY  AND  LIBRARIES  |  JUNE  2015     15   patrons  interact  with  our  catalog  and  its  associated  resources.  First,  because  we  are  now   searching  both  the  catalog  and  the  Summon  index,  the  type-­‐ahead  feature  that  we  had  deployed  to   suggest  index  terms  from  our  local  database  to  users  as  they  entered  search  strings  no  longer   functions  as  an  authority  index  search.  We  have  returned  to  querying  both  databases  through  a   simple  keyword  search.     Second,  in  our  implementation  of  the  single  search  interface  we  have  chosen  to  present  the  results   from  our  local  database  and  the  retrievals  from  Summon  in  two,  side-­‐by-­‐side  columns.  This  has   the  advantage  of  bringing  article  literature  and  other  resources  indexed  by  Summon  directly  to   the  patron’s  attention.  As  a  result,  more  patrons  interact  directly  with  articles,  as  well  as  with   books  in  major  digital  repositories  like  Google  Books  and  HathiTrust.  This  change  has   undoubtedly  led  patrons  to  make  less  in-­‐depth  use  of  the  local  catalog  database,  although  it   preserves  much  of  the  added  functionality  in  terms  of  discovering  our  own  digital  collections  as   well  as  those  resources  whose  cataloging  we  share  with  our  TRLN  partners.  We  believe  that  the   ease  of  access  to  the  resources  indexed  by  Summon  complements  the  enhancements  we  have   made  to  our  local  catalog.   CONCLUSION  AND  FURTHER  DIRECTIONS   One  might  argue  that  the  integration  of  electronic  resources  into  the  “catalog”  actually  shifts  the   paradigm  more  significantly  than  any  previous  enhancements.  As  the  literature  review  indicates,   much  of  the  conversation  about  enriching  library  catalogs  has  centered  on  improving  the  means   by  which  search  and  discovery  are  conducted.  The  reasonably  direct  linking  to  full  text  that  is  now   possible  has  once  again  radically  shifted  that  conversation,  for  the  catalog  has  come  to  be  seen  not   simply  as  a  discovery  platform  based  on  metadata  but  as  an  integrated  system  for  delivering  the   essential  information  resources  for  which  users  are  searching.   Once  the  catalog  is  understood  to  be  a  locus  for  delivering  content  in  addition  to  discovering  it,  the   local  information  ecosystem  can  be  fundamentally  altered.  At  UNC-­‐Chapel  Hill  we  have  engaged  in   a  process  whereby  the  catalog,  central  to  the  library’s  web  presence  (given  the  prominence  of  the   single  search  box  on  the  home  page),  has  become  a  hub  from  which  many  other  services  are   delivered.  The  most  obvious  of  these,  perhaps,  is  a  system  for  the  delivery  of  physical  documents   that  is  analogous  to  the  ability  to  retrieve  the  full  text  of  electronic  documents.  If  an  information   source  is  discovered  that  exists  in  the  library  only  in  physical  form,  enhancements  to  the  display  of   the  catalog  record  facilitate  the  receipt  by  the  user  of  the  print  book  or  a  scanned  copy  of  an  article   from  a  bound  journal  in  the  stacks.     In  2013,  Ithaka  S+R  conducted  a  local  UNC  Faculty  Survey.  The  survey  posed  three  questions   related  to  the  catalog.  In  response  to  the  question,  “Typically  when  you  are  conducting  academic   research,  which  of  these  four  starting  points  do  you  use  to  begin  locating  information  for  your   research?,”  41  percent  chose  “a  specific  electronic  research  resource/computer  database.”  Nearly   one-­‐third  (30  percent)  chose  “your  online  library  catalog.”17     ENGINE  OF  INNOVATION:  BUILDING  THE  HIGH-­‐PERFORMANCE  CATALOG  |  OWEN  AND  MICHALAK       doi:  10.6017/ital.v34i2.5702   16   When  asked,  “When  you  try  to  locate  a  specific  piece  of  secondary  scholarly  literature  that  you   already  know  about  but  do  not  have  in  hand,  how  do  you  most  often  begin  your  process?,”  41   percent  chose  the  library’s  website  or  online  catalog,  and  40  percent  chose  “search  on  a  specific   scholarly  database  or  search  engine.”  In  response  to  the  question,  “How  important  is  it  that  the   library  .  .  .  serves  as  a  starting  point  or  ‘gateway’  for  locating  information  for  my  research?,”  78   percent  answered  extremely  important.     On  several  questions,  Ithaka  provided  the  scores  for  an  aggregation  of  UNC’s  peer  libraries.  For   the  first  question  (the  starting  point  for  locating  information),  18  percent  of  national  peers  chose   the  online  catalog  compared  to  30  percent  at  UNC.  On  the  importance  of  the  library  as  gateway,  61   percent  of  national  peers  answered  very  important  compared  to  the  78  percent  at  UNC.   In  2014,  the  UNC  Libraries  were  among  a  handful  of  academic  research  libraries  that  implemented   a  new  Ithaka  student  survey.  Though  we  don’t  have  national  benchmarks,  we  can  compare  our   own  student  and  faculty  responses.  Among  graduate  students,  31  percent  chose  the  online  catalog   as  the  starting  point  for  their  research,  similar  to  the  faculty.18  Of  the  undergraduate  students,  33   percent  chose  the  Library’s  website,  which  provides  access  to  the  catalog  through  a  single  search   box.19   A  finding  that  approximately  a  third  of  students  began  their  search  on  the  UNC  Library  website   was  gratifying.  OCLC’s  Perceptions  of  Libraries  2010  reported  survey  results  regarding  where   people  start  their  information  searches.  In  2005,  1  percent  said  they  started  on  a  library  website;   in  2010,  not  a  single  respondent  indicated  doing  so.20     The  gross  disparity  between  the  OCLC  reports  and  the  Ithaka  surveys  of  our  faculty  and  students   requires  some  explanation.  The  Libraries  at  the  University  of  North  Carolina  at  Chapel  Hill  are   proud  of  a  long  tradition  of  ardent  and  vocal  support  from  the  faculty,  and  we  are  not  surprised  to   learn  that  students  share  their  loyalty.  For  us,  the  recently  completed  Ithaka  surveys  point  out   directions  for  further  investigation  into  our  patrons’  use  of  our  catalog  and  why  they  feel  it  is  so   critical  to  their  research.   Anecdotal  reports  indicate  that  one  of  the  most  highly  valued  services  that  the  Libraries  provide  is   delivery  of  physical  materials  to  campus  addresses.  Some  faculty  admit  with  a  certain  degree  of   diffidence  that  our  services  have  made  it  almost  unnecessary  to  set  foot  in  our  buildings;  that  is  a   trend  that  has  also  been  echoed  in  conversations  with  our  peers.  Yet  the  online  presence  of  the   Library  and  its  collections  continues  to  be  of  significant  importance—perhaps  precisely  because  it   offers  an  effective  gateway  to  a  wide  range  of  materials  and  services.   We  believe  that  the  radical  redesign  of  the  online  public  access  catalog  initiated  by  North  Carolina   State  University  in  2006  marked  a  sea  change  in  interface  design  and  discovery  services  for  that   venerable  library  service.  Without  a  doubt,  continued  innovation  has  enhanced  discovery.   However,  we  have  come  to  realize  that  discovery  is  only  one  function  that  the  online  catalog  can   and  should  serve  today.  Equally  if  not  more  important  is  the  delivery  of  information  to  the     INFORMATION  TECHNOLOGY  AND  LIBRARIES  |  JUNE  2015     17   patron’s  home  or  office.  The  integration  of  discovery  and  delivery  is  what  sets  the  “next-­‐gen”   catalog  apart  from  its  predecessors,  and  we  must  strive  to  keep  that  orientation  in  mind,  not  only   as  we  continue  to  enhance  the  catalog  and  its  services,  but  as  we  ponder  the  role  of  the  library  as   place  in  the  coming  years.  Far  from  being  in  decline,  the  online  catalog  continues  to  be  an  “engine   of  innovation”  (to  borrow  a  phrase  from  Holden  Thorp,  former  chancellor  of  UNC-­‐Chapel  Hill)  and   a  source  of  new  challenges  for  our  libraries  and  our  profession.   REFERENCES     1.     Cathy  De  Rosa  et  al.,  Perceptions  of  Libraries  and  Information  Resources:  A  Report  to  the  OCLC   Membership  (Dublin,  OH:  OCLC  Online  Computer  Library  Center,  2005),  1–17,   https://www.oclc.org/en-­‐US/reports/2005perceptions.html.   2.     Karen  Calhoun,  The  Changing  Nature  of  the  Catalog  and  Its  Integration  with  Other  Discovery   Tools,  Final  Report,  Prepared  for  the  Library  of  Congress  (Ithaca,  NY:  K.  Calhoun,  2006),  5,   http://www.loc.gov/catdir/calhoun-­‐report-­‐final.pdf.   3.     Roger  C.  Schonfeld  and  Kevin  M.  Guthrie,  “The  Changing  Information  Services  Needs  of   Faculty,”  EDUCAUSE  Review  42,  no.  4  (July/August  2007):  8,   http://www.educause.edu/ero/article/changing-­‐information-­‐services-­‐needs-­‐faculty.   4.     Ross  Housewright  and  Roger  Schonfeld,  Ithaka’s  2006  Studies  of  Key  Stakeholders  in  the  Digital   Transformation  in  Higher  Education  (New  York:  Ithaka  S+R,  2008),  6,   http://www.sr.ithaka.org/sites/default/files/reports/Ithakas_2006_Studies_Stakeholders_Di gital_Transformation_Higher_Education.pdf.   5.     Xi  Niu  and  Bradley  M.  Hemminger,  “Beyond  Text  Querying  and  Ranking  List:  How  People  are   Searching  through  Faceted  Catalogs  in  Two  Library  Environments,”  Proceedings  of  the   American  Society  for  Information  Science  &  Technology  47,  no.  1  (2010):  1–9,   http://dx.doi.org/10.1002/meet.14504701294;  and  Cory  Lown,  Tito  Sierra,  and  Josh  Boyer,   “How  Users  Search  the  Library  from  a  Single  Search  Box,”  College  &  Research  Libraries  74,  no.   3  (2013):  227–41,  http://crl.acrl.org/content/74/3/227.full.pdf.     6.     Charles  R.  Hildreth,  “Beyond  Boolean;  Designing  the  Next  Generation  of  Online  Catalogs,”   Library  Trends  (Spring  1987):  647–67,  http://hdl.handle.net/2142/7500.   7.     Kristen  Antelman,  Emily  Lynema,  and  Andrew  K.  Pace,  “Toward  a  Twenty-­‐First  Century   Library  Catalog,”  Information  Technology  and  Libraries  25,  no.  3  (2006):  129,   http://dx.doi.org/10.6017/ital.v25i3.3342.   8.     Karen  Coyle,  “The  Library  Catalog:  Some  Possible  Futures,”  Journal  of  Academic  Librarianship   33,  no.  3  (2007):  415–16,  http://dx.doi.org/10.1016/j.acalib.2007.03.001.   9.     Karen  Markey,  “The  Online  Library  Catalog:  Paradise  Lost  and  Paradise  Regained?”  D-­‐Lib   Magazine  13,  no.  1/2  (2007):  2,  http://dx.doi.org/10.1045/January2007-­‐markey.     ENGINE  OF  INNOVATION:  BUILDING  THE  HIGH-­‐PERFORMANCE  CATALOG  |  OWEN  AND  MICHALAK       doi:  10.6017/ital.v34i2.5702   18     10.    Marshall  Breeding,  “Next-­‐Gen  Library  Catalogs,”  Library  Technology  Reports  (July/August   2007):  10–13.   11.    Jia  Mi  and  Cathy  Weng,  “Revitalizing  the  Library  OPAC:  Interface,  Searching,  and  Display   Challenges,”  Information  Technology  and  Libraries  27,  no.  1  (2008):  17–18,   http://dx.doi.org/10.6017/ital.v27i1.3259.   12.    Michael  J.  Bennett,  “OPAC  Design  Enhancements  and  Their  Effects  on  Circulation  and   Resource  Sharing  within  the  Library  Consortium  Environment,”  Information  Technology  and   Libraries  26,  no.  1  (2007):  36–46,  http://dx.doi.org/10.6017/ital.v26i1.3287.   13.    Eric  Lease  Morgan,  “Use  and  Understand;  the  Inclusion  of  Services  against  Texts  in  Library   Catalogs  and  Discovery  Systems,”  Library  Hi  Tech  (2012):  35–59,   http://dx.doi.org/10.1108/07378831211213201.   14.    Lorcan  Dempsey,  “Thirteen  Ways  of  Looking  at  Libraries,  Discovery,  and  the  Catalog:  Scale,   Workflow,  Attention,”  Educause  Review  Online  (December  10,  2012),   http://www.educause.edu/ero/article/thirteen-­‐ways-­‐looking-­‐libraries-­‐discovery-­‐and-­‐ catalog-­‐scale-­‐workflow-­‐attention.   15.    Charles  Pennell,  Natalie  Sommerville,  and  Derek  A.  Rodriguez,  “Shared  Resources,  Shared   Records:  Letting  Go  of  Local  Metadata  Hosting  within  a  Consortium  Environment,”  Library   Resources  &  Technical  Services  57,  no.  4  (2013):  227–38,   http://journals.ala.org/lrts/article/view/5586.   16.    Benjamin  Pennell  and  Jill  Sexton,  “Implementing  a  Real-­‐Time  Suggestion  Service  in  a  Library   Discovery  Layer,”  Code4Lib  Journal  10  (2010),  http://journal.code4lib.org/articles/3022.     17.    Ithaka  S+R,  UNC  Chapel  Hill  Faculty  Survey:  Report  of  Findings  (unpublished  report  to  the   University  of  North  Carolina  at  Chapel  Hill,  2013),  questions  20,  21,  33.   18.    Ithaka  S+R,  UNC  Chapel  Hill  Graduate  Student  Survey:  Report  of  Findings  (unpublished  report   to  the  University  of  North  Carolina  at  Chapel  Hill,  2014),  47.   19.    Ithaka  S+R,  UNC  Chapel  Hill  Undergraduate  Student  Survey:  Report  of  Findings  (unpublished   report  to  the  University  of  North  Carolina  at  Chapel  Hill,  2014),  39.   20.    Cathy  De  Rosa  et  al.,  Perceptions  of  Libraries,  2010:  Context  and  Community:  A  Report  to  the   OCLC  Membership  (Dublin,  OH:  OCLC  Online  Computer  Library  Center,  2011),  32,   http://oclc.org/content/dam/oclc/reports/2010perceptions/2010perceptions_all.pdf.