White paper on OpenMinTed Community Requirements December ...openminted.eu/wp-content/uploads/2016/12/... · White paper on OpenMinTed Community Requirements December 1, 2016 Author(s):

WhitepaperonOpenMinTedCommunityRequirementsDecember1,2016

Author(s):

Agroknow AkriviKatifori,PanagiotisZervas,NikosManouselisGESIS MandyNeumann,PeterMutschke

INRA Sophie Aubin,Mouhamadou Ba, Philippe Bessières, Robert Bossy,EstelleChaix,LouiseDeléger,PatriciaGeretto,ClaireNédellec

ARC NataliaManolaOU NancyPontica,LucasAnastasiouFrontiers FrederickFenterEMBL-EBI ChristophSteinbeck,C.J.RuppCNIO MartinKrallingerEPFL RenaudRichardetUNIMAN MattShardlow

The objective of this document is to summarize the Community Requirements Analysisreport delivered in D4.2, which recorded and presented the different needs of theparticipatingusercommunities,identifyingcommonalitiesanddifferences.Thedeliverablepresents a set of the most representative stakeholder personas and concludes withconcrete requirements both for the OpenMinTeD platform and the foreseen use caseapplications.

H2020-EINFRA-2014-2015/H2020-EINFRA-2014-2Topic:EINFRA-1-2014Managing,preservingandcomputingwithbigresearchdataResearch&InnovationactionGrantAgreement654021

WhitepaperonOpenMinTedCommunityRequirements

•••

Public Page1of27

TableofContents

1. INTRODUCTION................................................................................................................................................5

2. METHODOLOGY................................................................................................................................................6

3. OVERVIEWOFTHECOMMUNITIES.................................................................................................................7

3.1 SCHOLARLYCOMMUNICATION..............................................................................................................................73.2 LIFESCIENCES.....................................................................................................................................................83.3 AGRICULTURE/BIODIVERSITY.............................................................................................................................103.3.1 THEAGRISSUB-COMMUNITY...........................................................................................................................103.3.2 THEFOODSAFETYSUB-COMMUNITY.................................................................................................................103.3.3 MICROBIOLOGYBIODIVERSITYRESEARCHERSUB-COMMUNITY.............................................................................113.3.4 CROPPLANTRESEARCHERSUB-COMMUNITY......................................................................................................113.4 SOCIALSCIENCES...............................................................................................................................................12

4. MAINSTAKEHOLDERSANDTHEIRNEEDS....................................................................................................14

4.1 TEXT-MININGRESEARCHERS–TDMEXPERTS.......................................................................................................144.1.1 CURRENTPRACTICEANDCHALLENGES...............................................................................................................144.1.2 TDMRELATEDNEEDS......................................................................................................................................154.2 RESEARCHERS....................................................................................................................................................174.2.1 CURRENTPRACTICEANDCHALLENGES...............................................................................................................184.2.2 TDMRELATEDNEEDS......................................................................................................................................184.3 CONTENTPROVIDERS.........................................................................................................................................214.3.1 CURRENTPRACTICEANDCHALLENGES...............................................................................................................214.3.2 TDMRELATEDNEEDS......................................................................................................................................224.4 AGGREGATORS.................................................................................................................................................244.4.1 CURRENTPRACTICEANDCHALLENGES..............................................................................................................244.4.2 TDMRELATEDNEEDS......................................................................................................................................25

5. CONCLUSIONS.................................................................................................................................................27


•••

Public Page2of27

TableofFiguresFigure1Scholarlyproductioniscyclicalinnature. ____________________________________________________________ 7Figure2MainstakeholdersrelatedtotheOpenMinTeDobjectives______________________________________________ 14

IndexofTables

Table1MariosAntoniou–RepresentativeoftheTextminingresearcherpersona..................................................................16Table2MarienVanderBeek,TheSocialScientist,representativeoftheresearcherstakeholdertype...................................20Table3MarieFoss,Datacurator-RepresentstheContentproviderstakeholdertype............................................................22Table4JuanCampos,theTechnicalManager,representativeoftheaggregarorandapplicationdeveloperstakeholdertypes

..............................................................................................................................................................................................25


•••

Public Page3of27

DisclaimerThis document contains description of the OpenMinTeD project findings, work and products.CertainpartsofitmightbeunderpartnerIntellectualPropertyRight(IPR)rulesso,priortousingitscontentpleasecontacttheconsortiumheadforapproval.

In case you believe that this document harms in any way IPR held by you as a person or as arepresentativeofanentity,pleasedonotifyusimmediately.

The authors of this document have taken any available measure in order for its content to beaccurate, consistent and lawful. However, neither the project consortium as a whole nor theindividual partners that implicitly or explicitly participated in the creation and publication of thisdocumentholdanysortofresponsibilitythatmightoccurasaresultofusingitscontent.

ThispublicationhasbeenproducedwiththeassistanceoftheEuropeanUnion.ThecontentofthispublicationisthesoleresponsibilityoftheOpenMinTeDconsortiumandcaninnowaybetakentoreflecttheviewsoftheEuropeanUnion.

The European Union is established in accordance with theTreatyonEuropeanUnion(Maastricht).Therearecurrently28Member States of the Union. It is based on the EuropeanCommunitiesandthememberstatescooperation inthefieldsofCommonForeignandSecurityPolicyandJusticeandHomeAffairs. The fivemain institutions of the European Union arethe European Parliament, the Council of Ministers, theEuropean Commission, the Court of Justice and the Court ofAuditors.(http://europa.eu.int/)

OpenMinTeDisaprojectfundedbytheEuropeanUnion(GrantAgreementNo654021).


•••

Public Page4of27

PublishableSummaryThis paper presents a summary of the results recorded during the requirements elicitation phase ofOpenMinTeD in deliverable D4.2 Community requirements analysis report. The deliverable is the seconddeliverableofWorkPackage4titled“CommunityDrivenRequirementsandEvaluation”andaimstoprovideanoverviewoftherequirementsrelevanttoTextandDataMining(TDM)fromresearchcommunitiesthathavebeenidentifiedaspotentialendusersoftheOpenMinTeDprojectservices.ThefirststepintheprocessoftheOpenMinTeDprojectdevelopingitsTDM-poweredservices,aswellasitsservicestoendusersande-infrastructure,istheidentificationoftheactualneedsofthecommunitiesthatarecurrentlyinneedoftheseservices.Asdescribedinthefollowingsections,therequirementshavebeenelicitedthroughtargetedonlinesurveys, interviews, focus groups and workshops that engaged different stakeholder representatives,includingtextminingresearchersaswellasvariousenduserstakeholdertypesandcommunities,basedonthemethodologydescribedinD4.1Requirementsmethodologydeliverable.


•••

Public Page5of27

1. IntroductionThe aim of the OpenMinTeD project is to enable the development of an infrastructure that fosters andfacilitatestheuseoftextanddatamining(TDM)technologies inthefieldofscientificpublicationsbutnotlimitedto it,bytwomainusercategories:applicationdomainusersandtext-miningexperts.OpenMinTeDaimstotakeadvantageofexistingtoolsandtextminingplatforms,facilitatingaccesstothemthroughtheappropriate registries, and enabling or enhancing their interoperability through an interoperability layerbasedonexistingstandards.OpenMinTeDsupportsawarenessof thebenefitsand trainingof textminingusers and developers alike and demonstrates themerits of the approach through a number of use casesidentified by scholars and experts from different scientific areas. It brings together different types ofstakeholders, including content providers and scientific communities, text mining and infrastructurebuilders,legalexperts,dataandcomputingcenters,industrialplayersandSMEs.

OpenMinTeD involves researchers fromfour scientific communities ranging fromscholarlycommunication(OpenAIRE, UK/CORE, LIBER, Frontiers), to Life sciences, focusing on biochemistry (EMBL-EBI) andneuroinformatics (Human Brain Project), Agriculture (INRA, Agroknow/FAO) and Social sciences (GESIS)domains. In thecontextofWP04CommunityDrivenRequirementsandEvaluation theprojectaims to (i)gatherrequirementsandcharttherespectivefieldsastoTDMusageandpracticesaswellastools,resourcesandstandardsused, (ii)defineprototypeapplications thatserve thecorrespondingscientificcommunitiesvia the OpenMinTeD infrastructure, and (iii) evaluate these applications in relation to the infrastructure.Deliverable D4.2 Community Requirements Analysis Report focuses on the first point and includes theanalysisofeachofthefourusercommunitiesthatwillassistinprioritizingtheimportanceoftheidentifiedneedsanddefiningthemainprojectstakeholderpersonas.ItwillalsoproduceaconcreterequirementslisttofeedthefunctionalspecificationsprocessaswellasprovideasetofmainstakeholderpersonasandtheirneedsinordertoguidethedesignanddevelopmentoftheOpenMinTeDplatformandapplicationusecases.

Theremainderofthedocumentisstructuredasfollows:Section2providesabriefoverviewoftheprocesses

and activities carried out for requirements elicitation. Section 4 provides an overview of the relevantcommunitiesinvolvedinOpenMinTeDandtheirparticularneedsinrelationtoTDM.Andsection4highlightstheneedsofthemainidentifiedstakeholdersandtheirrepresentativepersonas.


•••

Public Page6of27

2. RequirementselicitationmethodologyAsmoredetailsontherequirementselicitationmethodologycanbefoundindeliverableD4.1RequirementsMethodology, this section summarizes the evaluation activities that were realized in the context of thismethodology.Themethodologyproposedageneralevaluationprocesstobeadaptedtothespecificitiesofeachcommunitysoastorecordtheneedsofeachofthecommunities,aswellasageneralapproachacrosscommunities to validate the collected requirements of each of the main stakeholder groups. Morespecifically,therequirementselicitationprocesswascompletedinthreemainphases:

Phase1includedapreliminarysurveytoidentifyafirstsetofneeds,implicitorexplicit,asexpressedbytheendusersofTDMservices, and to recordapreliminaryoutlookof theusers towards issues likeaccess tocontent, licensing, storage, etc., in relation to TDM. These surveyswere created based on a basic surveyoutlineandadaptedbyeachCommunity leadertotheneedsofeachcommunity.Theyweredistributedason-linesurveystopotentialstakeholdersofTDMservices.

Phase2includedasetofactivitieswiththeobjectiveto:

1. MapeachCommunityandidentifymainstakeholders.ThisactivitywasinitiatedduringtheUserWorkshopthattookplaceinAthensinFebruary2016,consolidatedduringtheperiodimmediatelyafterthemeetingandservedasthebasisoftheCommunityreports.

2. Enrichrecordedneedsthroughfocusgroupsandinterviewswithrepresentativesoftheidentifiedstakeholderswithinandoutsidetheprojectconsortium.ThisactivitytookplaceattheCommunitylevelbytheconsortiumpartnersofeachCommunity.

3. Compilethe“communityreports”essentiallyrecordingtheoutlookofeachCommunitytowardsTDMissuesthroughtheperspectivesofthemainprojectpartnersineachone.

4. Recordtextminingresearcherneeds.ThemainstakeholdergroupofOpenMinTeDwastargetedthroughfocusgroupsessionsandinterviews.

Phase3 includedthevalidationofrequirementsthroughsurveysandinterviewsandthroughacontinuousvalidation process with the project partners and the creation of the main stakeholder personas. Aquestionnairepermainstakeholdertypewascreatedanddistributedasanon-linesurveyorusedtoguidestructured interviews. In parallel, aworking list of requirementswas available in order for all consortiumpartnerstoreviewandcommentonandtoguideearlyonthefunctionalspecificationsprocess(DeliverableD4.3).

Throughtherequirementselicitationprocessasetofrepresentativepersonashasbeenidentified.Apersonais a representation of the goals and behavior of a hypothesized group ofusers, personified through afictional,representativeofthisgroup.Personasareastrongtooltosupportadeeperunderstandingoftheusersthroughoutthewholedesign,developmentandtestingcyclesofasoftwareproduct.


•••

Public Page7of27

3. OverviewofthecommunitiesThis section presents the four communities that have been examined within OpenMinTeD to recordchallengesandneedsintermsofTextandDatamining.

3.1 ScholarlycommunicationScholarly communication progresses on research and the sharing of information that results from thatresearch. The cycle (Figure 1) starts with a question or an idea that is pursued and discussed leading toresearch.Thenextstepmaytaketheformofinformalcommunicationsuchastalkingtocolleaguessharingthesameinterestsandsharingemails.Thesemayresulttoapresentation,aposter,apanelpresentationataprofessional meeting. At the end of this process, information sharing becomes more formalized asresearcherspublishinajournalarticle.

Figure1Scholarlyproductioniscyclicalinnature.

In Scholarly Communication, bottlenecks arise when information is required at every stage. Traditionalstructures for literaturearchiving, searchinganddiscoveryarearchaic in the faceof 2Marticlespublishedeveryyear.Reviewing,indexing,alertsbecomeexponentiallymoretediousandcomplicated.LicensingandIntellectualProperty rightsand restrictions,whileprotecting theoriginal authors, impede the informationsharingandprogressionintothenextcycle.

Text andDataMining innovations shouldmake these cyclesworkmoreeffectively. Text andDataMiningtoolsshouldprovideallstakeholderswithtoolsthatworkintwodirections:


•••

Public Page8of27

• Provideadvancedsearchandretrievalmechanismsforitems(data/articles/etc.)thatdirectlytouchupontheirinterestsandexpertise

• Allowthemtoleveragelargedatasetsofarticlestodrawgeneralconclusions(meta-studies,epidemiologicalstudies,etc.)

Atthemoment,mostcontentprovidersandaggregatorsareinformedaboutthebenefitsofTDMforamoreeffective content provision and explore different ways to integrate TDM tools and services in theirworkflows.

TheScholarlycommunicationscommunityiscomposedofthefollowingmainstakeholders:

Policymakers:Thesedefinetheresearchlandscapebyintroducingpoliciesforfundedresearchandincludepublicandprivatefundingbodiesthatsponsorresearchprojects.

Contentproviders:Theseinclude:

a) Publishers: there are mainly two categories of publishers, the proprietary/traditional/subscriptionbased/toll access publishers, whosemain income derives from a subscription basedmodel, and the pureopenaccesspublisherswhooperateundervariousbusinessmodels.

b)Repositories:thesecanbemainlydividedfurtherintotwocategories,theinstitutionalrepositories,whichcollect the scientific outputsofone institution (i.e. academic, research, etc.) and the subject repositories,whichcollectscientificoutputsinspecificfield(s).

c)Libraries:Thereareavarietyofdifferent libraries,suchasacademic, research,special (corporate, legal,industry,etc.),public,online.AllthesecanserveasacontentproviderforTDMpurposes.

d) Scholarly communication databases: These can be bibliographic and citation databases, or othersubscriptionbaseddatabasesthatofferaccesstopeer-reviewedscientificcontent.

Aggregators: Aggregators or harvesters, like CORE or OpenAIRE collect in one searching point (peer-reviewedandnon-peer-reviewed)scientificinformationfromavarietyofdataproviders.AggregatorsserveanimportantroleinTDMfortworeasons.Firstly,themajorityoftheharvestedcontentcanbeusedforTDMpurposes.Whenwerefertoanaggregatorofopenaccesscontent,likeCORE,thenthistaskismadeeasier,due to the open access status of the aggregated content and their liberal licensing conditions. Secondly,aggregatorscanprovidecontenttootherservices,suchastheOpenAIREproject,whichaimsatthewidestdisseminationoftheEuropeanResearch.

ResearcherNetworks:Theresearchnetworkingsites,whichhostintheirdatabasescollectionsofscientificoutputs, are becoming lately increasingly popular. The useof theseoutputs from text anddataminers ischallengingthough,sincethesesitesdonotcomplywithinteroperabilitystandards,mostofthetimeshaveproprietary systems, they do not follow metadata and other standards and do not always make theircontentavailableforTDM.

3.2 LifeSciencesThe biggest challenges facing society today are biological: security of food supply, water integrity,controllingdiseaseoutbreaksinhumansandplants,andcaringforanageingpopulation,tonamejustafew.


•••

Public Page9of27

The life sciences encompass the study of all aspects of living things, and have developed within diversespecialised communities over the past century. With the emergence of data-driven science, thesecommunities are coming closer together, developing shared approaches and becoming moreinterdisciplinary.

Thelifesciencesdomainisinastateof'datadeluge'withthousandsofnewpublicationseachweek.Eveninasub-fieldaresearchermayhavetoreadandprocessdozensofrelevantpublicationseachdayinordertokeepuptodatewiththefield.Clearly,this isanintractableproblembyhumanmeans.Asasolution,manyresearchersareturningtoTDMforinformationretrievalandextractiontohelpprocessthelargevolumesofdata.Fortunately,thetechniquesareadvancedandcapableofsatisfyingmanyofthecommunity'sneeds.

The Life sciences community is particularly heterogeneous with diverse needs intersecting between sub-communities.Themetabolomicsspecialistandthefunctionalneuroscientistwillhavesomecoreinterestsincommon - however their specialist needs will differ. Whilst the core areas of the life sciences are wellresourcedforTDM,thespecialismsonthefringemaysufferdoublyfromalackofsufficientresourcesandTDMknowledge. Exceptions exist in someareas, for example theBLUIMApackageprovidesUIMA-basedinteroperableTDMtoolsforneuroscience.

Curation and annotation are two typical tasks in this domain which both require highly specialisedknowledgeofthespecificsub-field.Whereassomeareasmayusecrowdsourcingforthesetasks,thisisverydifficult in the life sciences as the crowd will not have the correct background knowledge to make aninformeddecision.Thismakesthesetasksveryexpensiveaswemustemployhighlyeducatedpeopletodoverypainstakingtasks.Automationofcuration,andmaybealsoannotation,wouldalleviatethecosttotheresearcherandreducetheamountoftimeacuratorspendslookingthroughirrelevantdocumentation.

LifesciencesincludethefollowingstakeholderswithinterestinTDM:

Database curators carrying out literature curation, includingmodel organism databases (MOD) like TAIR,MGI,RGD,WormBase,FlyBase,MaizeGDB;Functionalgenomeannotationdatabases(e.g.GOA),proteomicsdatabases(BioGRID,IntAct,MINT),comparativetoxicogenomics(CTD).

Experimental biomedical and basic science researchers: (1) to improve the interpretation and design ofexperimental research by improving the access to previously published information on the studiedbioentities; (2) using literature mining and knowledge discovery software for the generation of newhypothesesthatwillbeexperimentallyvalidated.

Clinicians: improve the information access for evidence based clinical practice using text miningtechnologiesandbiomedicalsemanticsearchengines

Chemists: systematic access to chemical information (structure associated chemical entities) described intheliteratureandpatents.

Biocurators and Bioinformaticians: Text-mining assisted curation results are useful as Gold Standardvalidationsetsforpredictivebioinformaticsresults.

Pharma Industry: Drug discovery and target selection, identifying adverse drug effects, competitiveintelligenceandknowledgemanagement


•••

Public Page10of27

Publishers:semanticannotationofonlinepublications,structureddigitalabstracts

Scientificpapersauthors:authorderivedannotations(assistedcompletionofstructureddigitalabstracts)

Patients: improved search engines, especially important for the detection of cases of similar rare diseasecasesandpersonalizedmedicine(automaticdetectionofmutationsdescriptionspublishedintheliterature)

Computer scientists: useful training data provided by BioCreative to improve the performance of cuttingedgestatisticalmachinelearningalgorithmsandfeatureselection/exploration

3.3 Agriculture/BiodiversityTheAgricultural/Biodiversitycommunityisquitebroadanddiversesincetherearemanydisciplinescoveredunder the label “Agricultural Science”. In general, agricultural science is a broadmultidisciplinary field ofbiology that encompasses the parts of exact, natural, economic and social sciences that are used in thepracticeandunderstandingofagriculture.Themostimportantdisciplinesareagriculturalandfoodsciences,forestry, environment and natural resources, ecology and nutrition among others. Therefore, theAgriculture/BiodiversityCommunityincludesstakeholdersfromtheabovementioneddisciplines,disciplinesthatusuallytendtointeractandtheyconstitutetheagriculturalecosystem.Allthesedisciplinesfacemanychallenges when it comes to extracting and combine meaningful information from disperse sourcesespeciallyconcerninghumanandanimalhealth,diseaseoutbreaks,environmental issuesetc.OpenMinTeDapproachedthiscommunitythroughtheperspectivesofdifferentsub-communities.

3.3.1 TheAGRISsub-community

The Agris ecosystem comprises all stakeholders involved in the agricultural research and practices. ItincludesallAgriculturalScienceandTechnologyinformationfromresearchbeingconductedinvariousareasoftheagriculturalscience,likehorticulture,viticulture,apiculture,farming,agriculturaleconomicsetc.

A major stakeholder group is that of researchers who are carrying out research in major domains ofagriculture, such as sustainable and organic agriculture, horticulture, viticulture, plant breeding and plantpathology,plantphysiology,domainswhoarecloselyrelatedtothedomainsoffoodscience,suchasfoodsecurity, foodsafetyandnutrition.These researchersmaybeworking inpublicorprivate institutionsandcompanies,universities,researchinstitutesandtheirresearchdataoutcomesareofgreatimportancetothewhole community. Any support given to the researchers (in the formof applications, services, tools etc.)thatwill help themdo their jobmoreefficiently inorder for them toprovidemeaningful conclusions andsolutions.

Theresearchdataoutcomesareinter-connectedwiththeotherstakeholdersofthecommunitywhoareinthe position to take advantage of these outcomes and serve other target groups and users of thecommunity,suchascompaniesetc.inthemostefficientway.

3.3.2 TheFoodSafetysub-community

Foodsafetyisconsideredasaglobalpublicgoodandacomplexproblematthesametime.Globalizationofthe food supply has resulted in food safety risks being widely extended beyond domestic borders.


•••

Public Page11of27

Foodbornediseaseoutbreaksarecommoninboththedevelopedanddevelopingworldsandhaveseriousimplicationsforpublichealthandtrade.

TheFoodecosystemorFoodcasecomprisesallstakeholdersinvolvedinthefoodresearchandpractices.Itincludes Food Science and Technology information from research being conducted in the areas of foodsecurity, food safety, food alerts etc. Researchers in the food sub-community who are working in thedomainsoffoodsecurity, foodsafetyandnutrition,either inresearch institutions,universitiesorpublicorprivate organizations, are looking for ways to connect the different models of risk assessment that aredeveloped at themoment. Also, they are trying to collect all data andmodels andmore importantly, toextract the various pathogens fromwithin the publications for example. All this information thatwill beextractedcouldspeedupthefoodoutbreakscauseidentificationandcouldsolveatthesametimethedataexchangeproblemwithlocalauthorities,companiesandotherorganizations.

3.3.3 MicrobiologyBiodiversityResearcherSub-community

All stakeholders involved in microorganism studies, either researchers, industrials or service providers,advocateaformalandunifiedrepresentationofmicroorganismbiotopeinformation(habitats,propertiesofthe habitat, interaction of the species with its environment (phenotypes and molecules)) from varioussources,e.g.literatureanddatabases.Thisneedispresentbothinappliedmicrobiology(e.g.foodindustry,health science or waste treatment) and in fundamental research (e.g. metagenomics studies, microbialbiogeographyandphyloecology).

One of the main needs of the microbial biodiversity community is the completion of the knowledgedescribed indatabaseswithknowledgefromthe literatureandotherdatabasesonthissubject,aswellastheircomparison for furtheranalysis.While this isasignificantcommunityneed, there isno resource thatcentralizestheknowledgeonmicroorganismhabitats.Mostoftheinformationisexpressedinfreetext,andisnoteasilyusedbecause it isnotstandardized.The information isdescribedinscientificpapers,freetextfieldsofdatabases,internationalmicroorganismculturecollectionsandbiodiversitysurveys.

Information content analysis and standardization calls for information extraction tools that canautomaticallyanalyzedescriptionsofmicroorganismbiotopessothatbiotopedescriptionsoriginatingfromdifferentexperimentscanbecomparedatalargescale.Hereanalysismeansnotonlytheextractionoftherelevantspansoftext,butalsothenormalizationorcategorizationwithreferenceresources(e.g.taxonomyoforganisms,ontologyofhabitat,ontologyofphenotypes,ontologyofphysico-chemicalproperties,etc.).

3.3.4 CropPlantResearcherSub-community

TheCropPlant community belongs to the Plant Biology community that is broad anddiverse and rangesfromfundamentalresearchonplantbiologytoseedcompanies.Improvementofplantbreedinginparticularisakeytofoodcrisesandworldhungerinthecontextofclimatechange.Moregenerally,improvementofplantspeciesofagronomical interest in thenear futurehasbecomean internationalstakebecauseof theincreasing demand for feeding a growing world population and to mitigate the reduction of industrialresources(oilespecially).


•••

Public Page12of27

The end-user objectives of the Plant use case is thus the design of formal biological models for generegulation networks in order to better understand the components and their interactions and to identifybiologicalparametersof interest. Theplant communityexpressedgreat interest in informationextractionamongotherTDMtasks.Themainmotivationseemstoextendtheanalysispossibilitiesalreadyofferedbydatasets and corresponding tools and platforms. Information extracted from texts is considered anadditional, complementary type of data. In that context, annotation resources need to be designedconsideringthealreadywidespreadreferencevocabularieswithsomeadaptationrequiredtofitTDMtasks.

Agriculture/BiodiversityincludesthefollowingstakeholderswithinterestinTDM:

• Foodsafetyofficersandagencies,waterqualitymanagers• Informationmanagers• Researchers• Breeders• Bioinformaticsapplicationproviders• Veterinarians,epidemiologists

3.4 SocialSciencesThe Social Sciences Community is rather diverse, asmany disciplines are covered under the label “SocialSciences”.Ingeneral,socialsciencesaresciencesthatstudythehumansocietyfromdifferentperspectivessuchassocietalstructuresanddynamicsandtheprocessesandrelationshipsbetweenindividualandsociety.Important disciplines are sociology and political sciences, but also economics, communication studies,psychology,education,anthropologyandothers.

A significant sub-community within Social Sciences is empirical social sciences. The term “empiricalresearch”refers to theresearchperformed usingempirical evidence collected through direct andindirectobservationorexperience.Onthebasisofatheoreticalframeworkthescientistcreatesanempiricalconcept formeasuring his/her research question in an appropriateway. The empirical evidence (the datarecordofone'sdirectobservationsorexperiences)canbeanalysedquantitativelyorqualitatively.Throughquantifying the evidence or making sense of it in qualitative form, a researcher can answer empiricalquestions,whichshouldbeclearlydefinedandanswerablewiththeevidencecollected(usuallycalleddata)andusetheseanswerstoformpublicationswithadirectrelationtothedata.

Today,TDMseemstobeinsufficientlyestablishedinthesocialsciences,whichmaybeexplainedbythefactthatthefocusineducation/teachingliesonmethodologicaleducation,andthatqualitycontrolplaysamajorrole in this field of study. So, social scientists don’t want to rely completely on complex text miningworkflowstheydon’tunderstandasitisnotrealistictoexpectthemtohavedeeperknowledgeincomputerscienceornaturallanguageprocessing.Theywouldratherwanttousesomesimpletextminingtoolstohelpthem detect patterns in big amounts of textual data, link them to other relevant items, get a newperspectiveonthetexts,maybegeneratenewhypothesesfrompatternsetc.Butthisisalwaysfollowedbyclosereadingofthetextstoperformamanualqualitativeanalysis.

As the complexity of software tools related to text mining is steadily growing (from simple counting ofwords in thebeginningup todetectionofargumentativestructuresnowadays), thepublicationoutputof


•••

Public Page13of27

text mining related research in the social sciences also grows. So it seems that this method of blendedreading(ormixedmethods)gainspopularityhere.Thusthereispotentialtowinevenmoresocialscientistsaspotential usersof an integrated textminingplatform.Theessential stepswhere the researchers couldrelyoncomputerassistancearethecollectionoftextstodefinecorporatoworkwith,andfrequencyandco-occurrence analyses (because they are intuitive and easily interpretable). More complex steps pose aprobleminthesocialsciencesintermsoftheirdemandfortransparencyandreproducibility.

TheSocialsciencescommunityincludesthefollowingstakeholderswithinterestinTDM:

• Socialscientistsseekingforinformationrelevanttotheirresearch• Contentandserviceproviders• Socialscienceresearcherswhoneedtoanalysetextualdatafortheirresearch


•••

Public Page14of27

4. MainstakeholdersandtheirneedsFigure2presentsanoverviewof themain stakeholdergroups in relationwithTDMand theOpenMinTeDobjectives. The remainder of this section focuses on the TDM related challenges and needs of eachstakeholdergroup.

Figure2MainstakeholdersrelatedtotheOpenMinTeDobjectives

4.1 Text-miningresearchers–TDMExpertsThis group includes TDM experts who focus their scientific research on automated ways to derive high-quality informationfromtext.Thesemay includeresearchersofdifferent levels,moreexperienced,senioronesorevenPhDcandidates.

4.1.1 Currentpracticeandchallenges

Text mining researchers employ a variety of textual content in the context of evaluating their researchproducts, algorithms, tools and services. Publications, social media records and news are the mostprominent ones. They also employ content metadata, to either facilitate the selection of most relevantdocumentsorimprovetheirtextminingresults.

Thesurveyresultsindicatethatalthoughabout50%ofthetextminingresearchersarefamiliarwithlicensing-relatedtermstheother50%isonlyfamiliarwiththeterm“license”withoutanyfurtherdetails.

Textmining researchers consider several TDM related challenges as very important. These are related toretrievalofusefulcontent,toolsandservicesfortheirresearch,evaluationandbenchmarkingoftheirownresearchoutcomes,aswellasmakingthemavailabletotheircommunity.Issuesoflicensingandretrievalofopensourcematerialseemtobeprominentchallenges,alongwithlocatingandretrievingrelevantcorpora.Leveragingvarious,occasionallyconflicting,languageresourcesonthesamedomainalsooftenseemstobeanissue.Morespecifically:

TDMresearchstateoftheartandpublicawareness.

In text mining research the notion of "off the shelf" product is still not fully established. Users of TDMsoftware need tomanually re-examine, curate and improve results of textminingprocesses. The tools in


•••

Public Page15of27

manycasesareerrorprone,noisyorincompleteandrevisionsarenecessary.Furthermore,therearealotof"small"toolsthatareresearchproductsandaremarginallyusefultobothotherresearchersandtheIndustry.SMEs and the Industry in general cannot re-use them in a profitableway and it is generally easier to re-develop tools than locate and re-use them. On the other hand, the public has increasingly higherexpectationsasaresultofemergingcommercialTDMapplicationsandend-to-endhighquality;robustandeffective,resultsareneeded.ThispointiscrucialforraisingpublicawarenesstoTDMandtheOpenMinTeDplatformcouldbeameanstothateffect.

Useofstandardsandframeworks.

WorkingonlywithexistingstandardslikeUIMAandGateandwebservicesleavesoutthemajorityofTDMresearchwork.IndustryframeworksarenotsufficientlyinvestigatedandtakenintoaccountincurrentTDMresearchpractice.

Componentpackagingandinteroperability.

Research centers in general do not have the policy or mentality to publish code developed by theirresearchers,ascommercializationandproductdevelopment isavery lowprioritytothem.Interoperabilitycannotbeensuredatthemomentasthetextminingscientificcommunitydoesnothavetheexperienceto"package"components, incontrastwiththeIndustry.MostTDMresearchersremainskepticaltowardsthefeasibilityofafullyinteroperablesolutionforaworkfloweditor.

Incentivestowardsinteroperability.

Researchers at the moment lack a strong incentive to commit to interoperability frameworks, such asvisibility of the researcher’swork, ease of use by others andbeing able to use other tools like their ownavailablethroughtheplatform.OMTDcouldbeapotentialmeanstodisseminatetheseincentives.

Contentandlanguageresourcesregistryandlicensing

Content iscrucialas it isnecessarytoperformlargescaleexperiments.However, licensing isan importantissue:Cantheresultsafterminingonthiscontentbeused?Whatwouldbethespecialattributionsneeded?Licensing issues tend tobe ignoredwithin the researchdomainbutarecrucial for the Industry.Thesameappliesforlanguageresources:Againitisveryimportant,especiallyforSMEswhichneedtousethemalsoforcommercialpurposes.

4.1.2 TDMrelatedneeds

TDM researchers considered several incentives for adopting the OMTD platform in their work practice,includingthevisibilityoftheircreatedtoolsandservicesandthedisseminationoftheirwork,alongwiththeability to easily combine their work with others’ and build composite workflows, also for evaluationpurposes,areverystrongincentives.Morespecifically,theenvisionedplatformisexpectedtosupporttheirworkinseveralways:

OMTDplatformasadisseminationtool,forscientificandcommercialpurposes.


•••

Public Page16of27

OMTDcanbeusedastrongdisseminationtoolofthetoolsandservicescreatedbytextminingresearchers.A platform offering access to a registry of software related to TDM can act as a common point for theresearcherstogainvisibility.

OMTDplatformcansupportTDMresearch.

TextminingresearchersagreedthatacommonregistryofTDMtoolsandservicesofferedbyOMTDwouldbevaluableasa tool topreviewanduseother researchers’software. It isoftenthecasethat researchersandTDMservicedevelopersneedto:

• AccesscomponentsusefulfortheirTDMworkflows• ReviewthestateoftheartonspecificaspectsofTDMtoseeiftheirresearchisredundant• Comparecomponents,toprovethattheirsisbetterortochoosethebestsolution

ManagingTDMworkflows.

There isa strongneed for the researchers tobeable tomanageTDMworkflows throughaplatformthatensures reproducibility and long term availability of results. Visibility of the whole processing of theworkflowsisconsideredveryimportant,withthepossibilitytolookintotheintermediateresultstoidentifyand remedy possible issues. The workflow editor offered should provide comprehensive workflowmanagementfunctionalitieswithemphasisoneaseofuseandagraphicaluserinterfacethatwillguideevenlessexperiencedusersinneedofTDMservices.

AccesstotextualcontentandTDMresources.

Researchers would like to be able, through the same platform, access textual content and linguisticresources(referencecorpora,lexica,ontologies,etc.)tosupporttheirresearch.Thereisaneedtosupportthepotentialevolutionofresources,throughautomaticandmanualcuration.Thereisthusanimplicitneedformanaging these resources, including issues of versioning, confidence, providence, etc. Comparison ofsuchannotationiscrucial,forevaluationpurposesandtoresolvepossibleconflicts.

Hardwareresourcesandstorage.

Researcherswould like to have access to resourceswhere they could run theirworkflows and store theproducedprocesseddataandresources.

Thispersonahasbeencreatedtorepresentthetextminingresearchers’needsandobjectives.

Table1MariosAntoniou–RepresentativeoftheTextminingresearcherpersona

Quote: “My research would be much easier if Icouldhavedirectaccess to the stateof theart in

TheText-miningresearcher

Name:MariosAntoniou

Age:28

Location:Athens,Greece

Technical comfort: experienced softwaredeveloper

Profession:TextminingPhDresearcher


•••

Public Page17of27

tools,content,resources.”

Backstory

MariosisacomputerscientistwhohasrecentlystartedhisPhDintheDepartmentofInformaticsandTelecommunicationsoftheUniversityofAthens.Hisgeneraldomainofresearch isTDMand,morespecifically,entityextraction frompublications.He is investigatingthestateof theartwhileat thesame timepreparing experiments to test algorithmshehasbeendeveloping in collaborationwithotherresearchersinhisteam.

Motivations

Mariosneedstobeabletohaveacomprehensiveoverviewofthestateoftheartinhisfield,soastodefinemoreclearlyhisresearchdomainandselectthemostcuttingedgeareastofocus.Heisofteninneedofcertaintextminingcomponentstocombinethemwithhisowntorunspecificworkflowsfortestingpurposes.

Frustrations

Marios findshimselfspendinga lotof time just to researchtheexistingstateof theart.Heknowsthere are several relevant tools and services to his work, through the existing publications hediscovers,howeverishasbeenprovenverydifficulttolocatethesoftware.Eveninthecaseswhenhe is able to find the implementations of certain approaches, it is a significant challenge tomakethemworkandevaluatetheminthecontextoftheworkflowsheistesting.

Locatingandretrievingtestcorporaandlanguageresourcesisagainanimportantchallengeforhimas he is faced with licensing issues and he is frequently being blocked when trying to downloaddocumentcollectionsfromspecificpublishers.

Idealexperience

Marioswould like an easy, directway to locate and retrieve resources related to his research. Hewouldliketobeable,throughasingleaccesspoint,toexploreexistingtoolsandservices,combinethem inworkflowsandevaluate them,comparing themwithhisown tools.Similarly,he feels thathavingdirectaccesstolanguageresourcesandcorporatouseasinputtohisworkflowswouldbeasignificantbenefittohiswork.

A single registry of TDM tools and services combined with workflow editing and managementfunctionalitywill support him in all aspects of hiswork, including defining his research objectives,examiningthestateofthearttocompareitwithhisownwork,runbenchmarksandevaluationsofhistools,withrelevantdocumentcorporaandlanguageresources,andmakehisworkavailableandvisibletotherestoftheTDMcommunity.

4.2 ResearchersResearchersareactivelyinvolvedin(i.e.conducting)scientificresearchpartofwhichisthesearch,retrievalandstudyoftextualinformation,includingscientificpublications.Theyaretheendusersofthetextminingproducts accessible through the OpenMinTeD platform, the domain researchers who will employ TDM-powered services to retrieve content relevant to their scientific or other interests. In this group we canincludeotherstakeholderswhohavesimilarsearchandretrievalneeds,butnotnecessarilyforpurescientificpurposes.Thesecouldincludepolicymakers,fundingbodies,etc.


•••

Public Page18of27


Researchersemployavarietyof textualcontent,accordingtotheirspecificdomainof interestandneeds,withpublications,socialmediarecordsandnewsbeingthemostprominentones.

Mostoftheresponders(63%)werefamiliarwiththeterm“license”buthadnoclear ideaofdetails inthisdomain,althoughtheysometimesfacelicensingissueswhenlookingforcontent.

ResearchersconsideredseveralTDMrelatedchallengesasimportanttotheirwork:

Searchandretrievalofcontent

Researchersstill face issueswhensearchingfor relevantcontent,both inprecisionandrecall.Precisionoftheresults ispresentedasamain issue.Usersoftengetunusable resultsduetoambiguities in thesearchterms.

As one of the responders explained, “Finding suitable information at all is a challenge. Often the keywordtermsuseddonotproduceanyresultsandthus,onehastocheckallpossiblesynonymstofindanythingatall,especiallyregardingtopicsonwhichverylittleresearchhasbeendonesofar.”

Insomecases,theusersneedtorestricttheirsearch inparticularsectionsofthetextualcontent,andthelackofthisfunctionalitycanresultinlackofprecisioninthesearchresults.

Processingthesearchresults

Havingretrievedtextualcontentthatispotentiallyusefulfortheirresearchobjectives,researchersneedtoanalyse the results. Thequick identificationof the topicof thedocument is againa challenge, inorder toassessitsrelevance,aswellas,correspondingly,theidentificationofsimilardocuments.A

Themajorityoftherespondersfeelastrongneedforaddedvalueservicesandfunctionalitytosupportthemin searching and processing of the search results. This additional processing functionalitywould not onlyhelpthemexcludeirrelevantresultsbutalsomakesenseoftherelevantonesbyquicklydetectingpatternsinthecontentitself.


Researchershighlightedtheneedforeffectivesearchandretrieval intermsofprecisionandrecallofopenaccessfulltextdocuments,forvisualizationstopresenttextminingresultsandforqueryrefinementbasedonrelevantresultsalreadyidentified.Specifically,researchersfeltthattheywouldbenefitsignificantlyfromthefollowingTDMrelatedservices:

Multi-stepsearchresultsrefinement.Userswouldlikeadditionalsearchfunctionalityamongtheresultsofthe initialquery,withdifferentcriteria.Theywould likea“multi-step”searchprocess tobeable todefinecomplexcriteriabeyondBooleanoperatorsthatwouldmakeitpossibletoexcludedirectlyirrelevantresults

Usingcontrolledvocabularies.Theresearchersmentionedtheuseofcontrolledvocabulariesasanecessityin different contexts. Firstly, they felt that standardizedmetadata vocabularies are crucial for search andretrieval as they can be used to locate documents whichmay not include the search term (for example


•••

Public Page19of27

“health behaviour” but rather a related sub-term (“alcohol consumption”). Vocabularies can also guideauthorstowardsstandardterminologiesandmoreeffectivemanualclassificationofpublications.Theuseofvocabularies can also support effective inter-disciplinary searchers. GESIS provides search termrecommenderbasedoncontrolledvocabulary,butdisambiguationofsearchtermsisstillanopenissue.

Automatic extraction of named entities and terms. Users would find very helpful to be able to extractautomatically from the search results specific information.Objects tobe identified in texts should includenotonlyNamedEntities(e.g.taxaorgenes)butalsoterms(e.g.anatomypartsorphenotypictraits)whichcanpresentvariationsuchassyntacticvariation(e.g.lengthofpaniclevs.paniclelength).

Theywouldalsoliketoproducestatisticsonthecorpusaboutthenumberofoccurrencesofspecificfacts,forexamplenumberofdemonstrationsmentionedinacorpusofnewspapersarticles

Identificationofentityrelations.Animportantcaseofinformationextractionneedexpressedbytheuserswas the identification and presentation of relations between different information objects, for example,relations of citations between publications and research data, topic relationships between publications /authorsandrelationsbetweenentitieswithinthesamepublication.

Resolution of ambiguities. The researchers considered very helpful to their research the resolution ofconceptualambiguities("Washington”thecityand“Washington”thepresident),but,moreimportantlytheresolutionoftopicambiguities.Forexample,inthecaseofsearcheson“youthandviolence”theywouldliketo be able to differentiate between results that discuss "violence BY youth”, and those that discuss"violenceAGAINSTyouth"

Open access. Researchers feel it is very useful to be able to view clearly in results hit list whether adocumentisopenaccessornotandwhatistheexactlicensingschemaofspecificdatasetstheyneedtouse.Mostof the researchers felt thatopenaccess iscrucial;however, theydidnot report that licensing issueshave been a serious impediment to their work. As one of the researchers commented, “If I need theparticulardocumentorcorpus,Iwillfindawaytosolvethisissue”.

Language. The issue of languagewasmentioned as important. The researchers would like to be able toautomatically search indifferent languagesat thesametimeusing theequivalent translated terms. In thecaseofresearchinolderdocuments,theissueofarchaiclanguageintheterminologyusedwasmentioned,whichneedstobetakenintoaccountintextminingtoolsandservices.

Linkingpublicationsanddatasets.Thisissueisconsideredcrucialbyresearcherswhoworkwithquantitativedatasets (survey results). They would like from the publication a reference and access to the relateddatasets.And,equally importantwouldbe tobeable todiscover thepartof thedata relevant to specificvariablesmentionedinthepublications

Reliabilityandtrust.Ingeneral,althoughtheresearchersexpressedaneedfortextminingfunctionalitytobecomeapartoftheirresearchworkflow,theyequallyexpressedconcernsonthematurityandqualityofsuchservices,especiallyincomplexusecasesastopicambiguityandsentimentanalysis.Theyfeltthattheyneedanobjectiveand“correct”measureoftheperformanceoftheautomatictextminingprocessontheirparticularinformationneed,explicitlypresentasanindicationofferedwiththeresultlist.WithoutawaytotrusttheperformanceoftheTDMservices,theywillnotfeelreassuredthattheydidnotmissanysignificant


•••

Public Page20of27

partofthedocumentsordataduetoerrors inTDMprocess.And inthiscasetheywouldfeeltheneedtomanuallyverifytheresults.Asoneoftheresearcherscommented,“IamscepticalwhethertheresultsoftheautomatedprocessingoftextsaresorobustthatIcanpresentthemtomyscientificcommunity.”

A domain dedicated portal. Users want federated access to textual content from various literaturedatabases.Thisportalwouldofferadaptedsearch facilities including facets corresponding to finegrainedobjects of interest for the researcher, e.g. genes, phenotypic traits, markers, etc. The portal should beconstantlyupdatedtowardsitssources.

Sharedvocabulariesandreference lists.Theyshouldbeusedtoannotatetexts inordertocreatebridgesbetweendatafromdatabasesanddataextractedfromtexts.Uniqueandpersistent identifiersforobjectsdescribedinannotationresourcesmustbeincludedinannotationsetsandTDMoutputs.Tofacilitatetheiruse in other tools, TDM outputs should be represented in open standard formats. Whenever possible,existing expert resources developed for other purposes should be considered for use in TDMworkflows,avoidingduplicateeffortsandirrelevantresults.

TDMbasedapproachesforvocabularydesignandadaptation.TDMusersofthedomainaskthattheyaremade available to the community in order to leverage the production of knowledge. Easy access toextractiontoolsandresourcesisnecessaryaswellastraininganddocumentation.TDMoutputcurationandvisualizationtoolsincludingvocabularyeditingonesareofmajorinteresttothisend.

GenericTDMworkflowsadaptabletospecificneeds.Genericworkflowsshouldbebuiltandsharedsothattheycanbereusedandadaptedtospecificneeds,e.g.samequestionbutdifferentspeciesorplantpart.

ThepersonainTable2isrepresentativeoftheresearcherstakeholdertype.

Table2MarienVanderBeek,TheSocialScientist,representativeoftheresearcherstakeholdertype

Quote:“Ascientificanalysisofthepastcanhelpusbuildabetterfuture.”

TheSocialScientist

Name:MarienVanderBeek

Age:28

Location:Eindhoven,Netherlands

Technicalcomfort:mainstreamsoftwareuser

Profession:HistoryofTechnologyPhDStudent

Backstory

MarienisaPhDstudentintheTU/eTechnischeUniversiteitinEindhoven.Sheispartoftheresearchteam focusing on history of technology and, in particular, the effects of the European railwayinfrastructuretowardsEuropeanization.

Motivations

Marienisoftenworkingwithpublicationsonherresearchinterests,aswellastechnicalreportsand,veryoften,surveysonthepreferencesanduptakeofdifferenttransportationapproaches,overthelastyearsbythepublic.

Asshe is interested in theEuropeanperspectiveof transportation,sheneedstoreviewmaterial in


•••

Public Page21of27

differentlanguages.SheisfluentinEnglishandGerman,apartfromDutch,butattimessheneedstoreview relevant documents in different languages and she employs the help of colleagues ortranslators.

Frustrations

Forheritiscrucialwhensearchingforcontenttogetasmanyrelevantresultsaspossibleand,atthesametime,nottogettoomanyirrelevantresultsthatshehastowastetimetosortthrough.

Whenworkingwithsurveys,Marienneedstoidentifyquicklytherelevantsubpartsofdatasetsthatdealwithquestionssheisinterestedin.Currently,shehastorelyonthemetadataofthesurveyforthistask,whichinthebestcasecontainsalistofdescriptorsonthesurveylevel.

OneofMarien’sbiggestproblemsistheamountoftextualdatashehastodealwith.Asshedoesn’thavethetimetoreadeverysingledocument,e.g.fromacorpusof50yearsofnewspaperarticles,shewouldneedthehelpofatooltoidentifythemostinterestingtextstoselectforclosereading.

Idealexperience

When searching for content, Marien wants to have her information need matched as closely aspossible.Ifsheentersambiguoussearchterms,shewantsthesystemtodetectthisandpresentherwiththepossiblealternativesshecouldchoosefrom.

Whenworkingwithsurveydata,Marienwouldappreciateifpublicationmetadatawasequippedwithinformation about and direct links to survey variables that were examined in the publication.Browsingthedataset,shewantstoseeinwhatotherpublicationsthevariableswerereferredtosothatshecangoandreadwhatanalysisotherresearchersperformedonthem.Ideally,anintegratedsearch system would even be able to present her variables from different surveys that matchkeywordsorconceptssheislookingfor.

When working with large corpora of texts, Marien wishes for a tool that helps her identifydocuments that are worth a closer investigation. The identification may be based on wordfrequencies, topics, interestingword co-occurrences etc. AsMarien is not at all familiarwith TDMalgorithmsandimplementations,sheneedstobepresentedwithameasureofcorrectness,precisionof the results through a tool that allows her to perform TDM on her corpus of documents in anintuitiveandtransparentway.

She would like to visually explore the results in the form of plots and graphs to identify certaininteresting structures she would like to inspect closely, maybe performing close reading ofinteresting(partsof)textsidentified.

4.3 ContentprovidersContent providers include publishers, entities (persons or organizations) that produce and distributespublications such as journals and books in printed or digital form. It also includes librarians, informationscientists, knowledge and information managers, content and repository managers and curators, thecontentspecialists,whichareresponsibleforaninstitution'scollections(digitalandanalogue).



•••

Public Page22of27

TheresponderstotheOpenMinTeDsurveyprovideaccessmostlytoscientificpublications.XMLseemstobethemostprominentformat(59%),whereas33%oftheprovidersreporttobeusingDublinCoreasastandardand22%nostandard.56%oftherespondersofferthedocumentlocationwithinitsmetadatarecord.

89% of the responders are familiarwith licensing issues and terms and consider a particular challenge toselecttheappropriatelicensingschemefortheircontenttoincludebothmodificationsandTDMprocessing.The most preferred licensing policy seemed to be Creative commons (56%). On the question for theexistence of provision of themodification of content, all responders replied negatively, except two,whoexplained that “in the archival contract there is a clause that states that content may be modified by thearchive(necessaryforarchivingreasons)”.

Inascaleof1to5,onhowfamiliartheyarewiththeapplicationoftextminingtechniquestocontent,theresponders’average responsewas2.83 (std 1.34).Whenasked if theywouldallowtheapplicationof textminingontheircontentmostoftheresponders(63%)werepositive,whereas26%wouldprobablyallowit,underspecificconditions.

ContentprovidersviewTDMasapotentialsolutiontoasetofchallengestowardstheirultimateobjectivewhichistoprovidetotheirusersaddedvalueservices,includingunifiedaccesstoresourcesofvarioustypesandsourcesandadvancedsearchfunctionality,specializedonspecificpartsofthedocumentandtakingintoaccountdomain-specificvocabularies.


To promote their objective for advanced and effective search functionalities on their content, providersneedtoenrichtheirmetadatathroughappropriateTDMservices:

Automatedmethodforenrichingexistingmetadatarecords

UsingtheRegistryofferedbyOpenMinTeDwillallowthemtoselect the tool that ismoreappropriate formetadataextractionandusingitthroughtheprojectinfrastructuretoextractcontentmetadata.

Buildingdomainspecificterminologyforfurthertextannotation/indexing/search

Content providers need to be able tomanage terminology and vocabularies related to their content andenrichthemwithappropriatesemi-automaticmethods.

AnimportantpointisthatthedifferenceinlevelofexpertiseoftheusersshouldbetakenintoaccountintheOMTDplatformfortheWorkfloweditor.Reusabilityoftoolsandresourcesisalsoakeyofsuccesstoreducethedevelopmentcostsandtechnicalbarriersfornon-experts.

Trust in TDM results is an issue thatoftenemerged indiscussionswithdata curators. ConfidencemetricsshouldbeincludedinTDMresultsand/oreasilyaccessibleontheplatform.CurationtoolsarealsonecessarytoreviseTDMoutputsbeforetheir integrationinotherdomainapplications.Whencorrectingoutputs,thepossibilityofrevisingannotationresourcesshouldalsobeofferedthuscreatingavirtuouscircle.

Table3presentsthepersonacreatedfortheContentproviderstakeholder.

Table3MarieFoss,Datacurator-RepresentstheContentproviderstakeholdertype


•••

Public Page23of27

Quote: “If we want researchers to access ourcontent we need to invest on enriching itsmetadata.”

TheDataCurator

Name:MarieFoss

Age:48

Location:Lyon

Technical comfort: experienced software user,basicprogrammingskills

Profession:Datacurator

Backstory

Marie works for several years now for an open access publisher hosting scientific journals andconference proceedings. Her role in the organization is to guide the content curation processestowards metadata enrichment to make the content more searchable and accessible through thesearchandretrievalfunctionalityofferedtoresearchersinthepublisherwebsite.

Motivations

Followingtherecentdevelopments inthescientificcommunicationdomainandtheavailabilityofahuge volume of publications on-line,with an increasing part of them available as open access fulltext, Marie feels that content providers should invest in new and advanced ways of metadataenrichmenttosupporttheneedsofresearchersformorepreciseandcompleteresultsets intheir,often complex and demanding, information needs. Part of the content that her organizationmanagesisscientificpublicationsfromthelifesciencesandagriculturedomains,fieldswhereshehasnoticedthatthereiscomplexandrichterminologywhichisinconstantflux.Shewouldliketobeableto take advantage of the particularities and updates in each domain in terms of vocabularies andterminologytooffermoreadvancedsearchandretrievaltotheirusers.

Frustrations

Marie,althoughnotacomputerexpertoriginally, shehas triedover theyears tokeepupwiththeadvancesinthefieldsofinformationsearchandretrievalingeneralandTDMinparticular.Althoughsheknows that the small ITdepartmentofherorganizationhasnot the resources todevelopandintegrateTDMservicesintheirworkflows,sheconstantlystrivestofindoff-the-shelfTDMsolutionsthat could provide input to their workflows in the form of content terminology taxonomies orontologiesorevenTDMsoftwaretoapplyonthecontent.Locatingtheappropriatesolutionsforhercase,evaluatingthemandapplyingthemfortheircontentisforherapainstakingprocesswhichhasnot been particularly fruitful so far. Even for her as a manager of a technical team includingapplication developers, it has been a significant challenge to locate, install, adapt and re-use TDMsoftware.

Idealexperience

Mariewouldideallywishforasinglepointofaccesswhereshewouldbeabletoviewinformationonthe latest advances in the field of TDM. She would like to locate easily new solutions, not onlysoftware,butalsothelateststandardontologiesinspecificfields.Shewouldthenliketobeabletotest these new approaches on her content, preferably through external services that would notrequirehertoinstallandimportsoftwarecomponentsintoherorganizationworkflows,andthen,iftheresultswerefoundsatisfactory,tointegratethemintothemetadataschemeoftheircontent.

AplatformbringingtogetherinformationonTDMwithrelatedsoftwareandotherresources,along


•••

Public Page24of27

withaneasywaytoapplyandevaluatetoolsandservicesbeforeintegratingtheminherworkflowswouldbeidealforher.

4.4 AggregatorsTheaggregatorgroupofstakeholdersincludesorganizationsandcompanieswhichdevelopandsupporte-infrastructures that provide federated access to a multitude of content providers’ material. These oftenemploy software developerswho require the integration of TDM components and services to their own,whichmayformpartoftheservicesprovidedbyaggregatorsorcontentproviders.


Aggregators, likeOpenAIRE, COREandAGRIS, aggregateopen access researchoutputs from repositoriesandjournalsworldwideandmakesthemavailablethroughasetofservices.Theirobjectiveistoprocessandenrichtheaggregatedcontentthroughadvancedsearchandretrievalfunctionality.

After an initial content provider and dataset registration phase, the aggregator harvests the contentmetadata andperformsquality control and enrichment, followedby an indexingphase tomake the dataavailablethroughitssearchandretrievalfunctionality.

AllinterviewedrepresentativesoftheaggregatorswithintheprojectconfirmedtheimportanceofTDMforthe semanticenrichmentphase.They reportedusing tools ranging from topic/subject identification in theabstract, based on specific domain ontologies (the case of the AGROVOC ontologies in AGRIS) to morecomplexenrichmentprocesseslikeproject,fundingorcitationextractioninthecaseofOpenAIRE.

Thesemanticenrichment isperformedaspartof theaggregationworkflowofaggregator,either throughoff-the-shelftoolsorwithcustomcomponentsdevelopedinpremisesbytheaggregatordevelopmentteamtotailortheTDMapproachtotheirspecificobjectives.

Interoperability at the level of metadata harvesting was not reported as an issue, as usually contentprovidersareaskedtocomplywithspecificharvestingspecifications,insomecasesevenstandardslikeOAI-PMHorDublinCore,thatcanenabletheaggregatortoperformpossibletransformationsandcontrolstotheharvestedmetadatainastraightforwardmanner.

ThereweretwomainchallengesreportedinrelationtoTDM:

Licensing

Atthemomentthereisnotanexplicitlegalframeworkorguidelinesforthelicensingprocessinrelationwiththeuseofneitherthemetadatanotthefulltextharvestedbycontentprovidersfortheusefortextmining.Some aggregators handle this issue through the service agreement with the content provider at theregistration level. Others make no explicit mention for TDM during the dataset registration phase butconsiderimplicittheagreementofthecontentprovidersfortheuseofmetadatafortextminingpurposes.AsTDMgraduallybecomesastandardpracticeasthesemanticenrichmentphaseofaggregatorharvestingprocesses,thereisastrongneedforacomprehensive,explicitlicensingframeworktoguidetheprocessoflicensingforTDMduringtheaggregationofmetadataandcontent.


•••

Public Page25of27

AccessingpublicationfulltextforTDM

A related issue is full text access for aggregated publications, either for TDM purposes or to provide itdirectlytotheendusersoftheaggregatorservices. Insomecases,a linktothefulltext isprovidedinthemetadataofthepublication.Ifthislinkisnotexplicitlypresentinthemetadataitcanbemorecomplicatedtoextractandprovide.Additionally,itisoftenthecasewhentheaggregatorneedstoaccessthefulltextforTDMpurposes, though, forexample,webcrawlingthat theyareblockedby thecontentproviders. In thiscaseanappropriatelicensingframeworkisneededtoensurethatthereisasmoothcollaborationofcontentprovidersandaggregators.


Aggregatorsexpressedcommonneedswiththoseofthetextminingresearchers,inrelationwith:

• ARegistryoftextminingtoolsandservicesthatwillsupportthemtoeasily locateandre-usetext-miningsoftware.

• AworkfloweditorwheretheycancreateandmanageworkflowsforspecificTDMpurposes• Thepossibilitytoruntheworkflowscombinedwithlanguageresourcesandenrichesthemetadata

offeredbytheirservices.• ThelatestinformationonlicensingschemesrelatedtotheprovisionofcontentingeneralandTDM

issuesinparticular

ThepersonainTable4hasbeencreatedtorepresenttheaggregatorandapplicationdeveloperneedsandobjectives.

Table4JuanCampos,theTechnicalManager,representativeoftheaggregatorandapplicationdeveloperstakeholdertypes

Quote: “Making content fromdifferent providerssearchable and accessible is by nomeans an easytask.”

TheTechnicalmanager

Name:JuanCampos

Age:43

Location:Madrid,Spain

Technical comfort: experienced softwaredeveloper

Profession:Technicalmanager

Backstory

Juanworks for an organizationwhich offers federated access to open access publications from amultitudeofsources,privateandpublicpublishersaswellasinstitutionalrepositories.Hemanagesatechnical team consisting of application developers and content curators with the objective toconstantly improve their information system search and retrieval functionality and ensure thesmoothandeffectiveinclusionofnewcontentproviderstotheiraggregationservices.


•••

Public Page26of27

Motivations

As the aggregated content is constantly growing, Juan wishes to make sure that researchersaccessinghisorganization’sportalwillbeabletoeffectivelyretrievethecontenttheyneed.Itseemsthat the quantity of the content available can only become an asset for the users as long as it isavailablethroughaneasytouseandprecisesearch interfacewithaddedvalueaccessandretrievalfunctionality.

Juan tries to stay informed of the new advances in TDM and feels that such solutions wouldsignificantlyboostthequalityoftheirsearchandretrievalservices.Richmetadataseemstobethekeytomakingthecontentdiscoverableandaccessiblebytheendusers.Forthepastfewyearshisorganization has invested in including TDM in their metadata management workflows, testingdifferentapproachesandcomponents.

Frustrations

Juanandhisteamhavetoconsolidateawidevarietyofcontentmetadata. Providers,accordingtotheirpoliciesandneeds,registertheircontentwithmetadataindifferentformats,followingdifferentstandardsandindifferentlevelsofdetail.AlthoughJuan’steamprovidesguidelinesandhasinplacespecificpolicies for the registrationofa contentprovider, thequalityof importedmetadata isnotalwaysuptothedesiredstandardsexpectedbyhisorganizationsohisteamstrivestoenrichitwiththeirinternalTDMprocesses.

However,thisisnotatrivialtask.Ononehand,uncertaintyonlicensingforTDMimpedesthemassprocessing of the content needed for metadata enrichment, as providers’ reaction to this vary.Additionally,JuanisnotsurewhatisthebestwaytogetinformedandbeuptodatewiththestateoftheartonTDManddiscovernewprocessesforpossibleinclusionintheirworkflows.

Idealexperience

JuanwouldliketohaveaccessinaconsolidatedwaytoexistingTDMapproaches,withadirectwayto evaluate their effectiveness and suitability for his needs, compare them with otherimplementations and even download and include them in his organizations workflows withoutsignificanteffort.

Apart fromdiscoveringTDMsoftware, Juanneeds tobe informed regularlyon latest advancesondomain-specificvocabularies,ontologiesandterminologythatcanbeusedtoannotatethecontentandmakeitaccessibletothedomainexpertintheterminologyofthedomain.


•••

Public Page27of27

5. ConclusionsThe requirement elicitation activities in the context of the OpenMinTeD project confirm that there is acomplexmosaicofstakeholdersthathaveadirector indirect interest inTDMsolutions inrelationtotheirworkingpractices.Guidedbytheneedsof theresearchers,whicharetheendusersof theTDMsolutions,with constantly more demanding and complex information needs, content providers, aggregators,applicationdevelopersfindthemselvesinthepositiontoseekoutTDMsolutionsthatcanbeincorporatedtotheirofferedservices. Inthissensetheyare inneedoftheresearchproductsoftext-miningresearchers,astakeholder group who approaches OpenMinTeD with more direct needs, in relation to registering,organizing,evaluatingand, thus,makingmoreaccessibleandavailable theseTDMresearchproducts.Thisdocumentedattemptedtosummarizethesedifferentperspectivesandpresentanoverviewofthecurrentstatus, challenges and needs in relation to TDM, which eventually lead to concrete requirements forOpenMinTeD.

ThisdocumentpresentedthesummaryoftheresultsofthisprocesswhichresultedindeliverableD4.2,withthe ultimate objective to (a) offer a concrete functional requirements list to feed the functionalspecificationstobeproducedinthecontextofD4.3Functionalspecificationsand(b)describethroughthedefinitionofstakeholderpersonastheneedsofthemainstakeholdertypes,tobeusedasuserarchetypesthatwillguideallsubsequentdesignanddevelopmentactivitieswithintheproject.

Documents

White paper on OpenMinTed Community Requirements December ...openminted.eu/wp-content/uploads/2016/12/... · White paper on OpenMinTed Community Requirements December 1, 2016 Author(s):