Investigation As a Member of Research Discourse* © Vasily Bunakov Scientific Computing Department, Science and Technology Facilities Council, Harwell OX11 0QX, United Kingdom vasily.bunakov@stfc.ac.uk Abstract and take a look at the simultaneous circulation of them in our own research domain of experimental science Investigations are specific intellectual entities utilizing large research facilities: neutron sources, that circulate in large research facilities with synchrotrons, and powerful lasers shared by multiple shared access by multiple research teams; researchers (visiting scientists). investigations have some common features with research papers (publications) and can be 2 Facilities research lifecycle and data included in citation networks. We consider different approaches to modelling the relations modelling between research papers and investigations and 2.1 Facilities science landscape discuss opportunities for matching these two members of common research discourse. The Research facilities can be thought of as well- analysis undertaken can be of interest for equipped hubs where research teams or individual research centres that consider information researchers come to perform their experiments on their services based on data and publications own samples. The research facility core is typically contextualization. represented by a unique scientific instrument: a particle accelerator, a neutron source, a powerful laser, a 1 Introduction telescope, or a supercomputer that allows detailed simulation of natural phenomena, or by a few such The journal articles, e-prints, reports and other instruments that offer researchers different research similar artefacts that irrespective of their physical techniques. The examples include European manifestation can be seen as derived from their paper- Synchrotron Radiation Facility (www.esrf.eu), neutron based “document” ancestors are the well-established source in The Institut Laue-Langevin (www.ill.eu), means of research communication and a popular aide Siberian Synchrotron and Terahertz Radiation Centre for tracking the state and the trends of research (http://ssrc.inp.nsk.su/CKP/eng/) or the future Extreme discourse. The “papers” have clear identity, allow Light Infrastructure (www.eli-beams.eu). review (of different kinds) and participate in citation networks; this supports performing the aforementioned Research conducted in facilities bears characteristics functions of the quality research communication and of “big science” such as a long-term capital investment, measurable research tracking; this also makes “papers” permanent support staff, scalable computing valuable intellectual entities worth capturing in library infrastructure; and “bench science” with individual catalogues, and worth sharing via advanced information scientists and small research teams that may have services. specific and short-time research goals. The user community of European facilities counts tens of We suggest that other type of intellectual entities, thousands scientists who pursue different applications: investigations, have essential features similar to the crystallography reveals the structures of proteins document-like entities hence are the natural candidates important for the development of new drugs; neutron to supplement “papers” as valuable members of scattering identifies stresses within engineering research discourse. We consider the types of relation components such as turbine blades, and tomography can between the document-like and investigation entities, image microscopic details of biological tissues ([1]). Proceedings of the 16th All-Russian Conference "Digital A business model for user research on large facilities Libraries: Advanced Methods and Technologies, Digital that emerged a few decades ago has been influenced by Collections" ― RCDL-2014, Dubna, Russia, October 13–16, the advances in instrumentation and data analysis that 2014. are now more automated and more user friendly than in early days of facilities a few decades ago. This has led, * This work is related to the projects of PaNdata collaboration among other effects, to a lesser significance of the www.pan-data.eu supported by the EU 7th Framework Programme for Research and Technological Development. instrumentation “gurus” ([2]), and to the emergence of The author would like to thank his colleagues in PaNdata for specific services for research and industry that allow their input for this paper although the views expressed are the users sending their samples for remote investigation views of the author and not necessarily of the collaboration. according to one of the service plans ([3]). 160 Yet the facilities business model has proved to be  Series of Experiments (that altogether effective and is a foundation for a specific research constitute Investigation with the proclaimed goals): lifecycle, and for specific information modelling and the user will bring samples, and sometimes an information services in support of it. additional equipment to the facility, calibrate the experimental environment and actually take 2.2 Generic research lifecycle measurements Despite the variety of facilities instruments and  Data Archiving: facilities offer high-throughput experimental techniques, the following distinct stages data collection and archiving services; archiving of raw are typical across facilities and thus represent a generic data collected in the facility data storage is often a facilities lifecycle: policy requirement  Research Proposal: the facilities are often  Data Analysis: it can be done through multi- oversubscribed so the researcher (investigator) should layer computing environment where some tools are justify the value of her research and the suitability of a offered by facilities, and others applied by scientist particular experimental technique individually  Approval Process: multilateral assessment by  Results Publication: journal articles and alike; the facility, including risk assessment (as the experiment facilities often require the visitor scientist to report back may involve hazardous materials or techniques) on any publications derived from the experiments.  Experiment Scheduling: allocation of the time This generic lifecycle is illustrated by Figure 1. slot within a facility operating cycle, and registration of all visitor scientists Figure 1. Facilities research lifecycle 2.3 Data modelling effort so far “cores” to the future investigation entities – on the national research portals, e.g. ISIS neutron and muon Facilities collect raw experimental data in a variety source (www.isis.stfc.ac.uk) intends to publish the of formats yet there is a movement towards unification descriptions of all approved proposals on the UK best represented by NeXuS standard and community common gateway to publicly funded research around it (www.nexusformat.org). There are of course (http://gtr.rcuk.ac.uk/). The internal representation data checks and data replication services, as well as format for these entities is going to be CERIF (see under some recent attempts to form and curate archival www.eurocris.org) that is widely used in the European packages according to OAIS reference model ([4]). grant information systems. The aforementioned generic lifecycle gave birth to the rich CSMD metadata model ([5], [6]) which is 3 Research data in research discourse implemented, with some modifications, in the popular ICAT software platform ([7]). 3.1 The modes and purposes of sharing research data Some facilities started assigning persistent The earlier mentioned NeXuS format, Research identifiers to datasets ([8]) and there is a recent effort of Objects and persistent identifiers for data present three having persistent identifiers for other aspects of different modes of sharing research data. facilities research such as instruments or experimental NeXuS file includes both data and data context techniques ([9]). (metadata) and thus offers research result as a The promotion of the research idea through the “package” that can be interpreted by other researchers – facilities lifecycle has inspired the concept of Research or the same research team in future – with the help of Objects for facilities science ([10]) that acquire more format-compatible software. It is a responsibility of the and more detail whilst the investigation proceeds from “package” creator to embed all essential information in its conceptual stage through the experiment to the there; the boundaries of information context are very research paper and associated artefacts. well defined (it is literally one data file). An interesting recent development is the intention of Research Objects suggest the enrichment of some facilities to start publishing the descriptions of the information according to a specific model while the approved research proposals (grants) – that are the intellectual entity moves through the research lifecycle; 161 this implies that there is a “creator” to the model and the The Figure 2 illustrates provenance relations “curator” of intellectual entity on each phase of between investigations and research papers that are a lifecycle; the boundaries of intellectual entity are more foundation for appropriate “citations”. flexible (it may be an aggregation of various components) but are still well-defined. The supply of nothing more but persistent identifiers for data, perhaps associated with some moderate contextual description (metadata), implies the paradigm of “open world” where intellectual entities can be deliberately constructed by various agents, hence there are no clear (predefined) boundaries to the entities, and virtually everyone can be considered a data “curator”. Sharing data or information, however, is not the end in itself and can be considered as a means to empower Figure 2. Research provenance chain research discourse, to supply some intellectual entities Similarities between investigation and research into it. So quite often, when people speak of “research paper as intellectual entities are summarized in the data” they actually mean intellectual entities where data Table 1. may be just a component, or something associated with a “quantum” of research discourse. Table 1. Common features of investigations and This can be illustrated by observations over DataCite research papers. (www.datacite.org) – a platform that proclaimed goal is supplying data with dereferencaeble persistent Feature / aspect Publication Investigation identifiers (well-formed DOIs). The data centres who (research paper) actually use DataCite in fact tend to assign DOIs not to Is an intellectual Yes Yes datasets but to “quantums” of research discourse, e.g. to entity doctoral theses (that may of course contain some data Is a subject of peer Yes Yes but is not the data per se). In case of facilities science, review (via proposal we observe that DataCite DOIs are in fact approval) dereferenceable to the landing Web pages that contain Can cite all Yes Yes descriptions of investigations which are, as we significant explained it earlier, the series of experiments performed intellectual entities with a certain research goal on the assigned instrument of research discourse within a dedicated timeslot. Citation chains exist Yes Yes So when a researcher cites “data” via DataCite DOI, (steps of discourse observed she in fact quite often cites an intellectual entity – which can be a paper or something else, e.g. event (such as an Universal identifiers Yes Yes earthquake) in geophysics, or investigation in the case available of facilities science. 1 This attitude towards “data” DOIs assignment is only natural as what researchers tend to Looking into what intellectual entities can refer to cite may not be “data” per se but certain identifiable what other intellectual entities (with the inclusion of elements of research discourse. datasets and software – which may or may not bear a clear identity) suggests the asymmetry in the direction 3.2 The place of investigation and the place of data in of references so that e.g. a research paper can cite a facilities research discourse dataset but not vice versa: Investigation as an intellectual entity bears some Table 2. Cross-references of intellectual entities features that are common with traditional research paper. Indeed, an investigation proposal is peer- References Paper Investi- Dataset Software reviewed; investigation can be cited from papers by the (“from” row gation “to” column) well-formed DOI and from other investigations, too, as when a researcher submits proposal, she refers to the Paper Yes Yes Yes Yes relevant past publications and past investigations. 2 Investigation Yes Yes Yes Yes Dataset No Yes Yes Yes (e.g. simulation) 1 Examples of dereferenceable “data” DOIs that in fact resolve in investigation or research paper descriptions: Software Yes (e.g. No Yes Yes to paper http://dx.doi.org/10.5286/ISIS.E.24066298 about http://dx.doi.org/10.5167/UZH-27029 algorithm) 2 Looking into the ICAT database for ISIS facility indicates the existence of investigation “chains” when the next investigation refers to the previous one, with as many as four investigations in a row undertaken in the last 10 years. 162 In fact, research discourse in facilities science splits investigations lifecycle) and those in the institutional into the two distinctive layers that can be called “research library catalogues. per se” and “data management”; this is illustrated by To estimate the viability of automated techniques, Figure 3. we tried to match the bibliographic records for the papers that were the “input” to the investigations performed on ARGUS muon spectrometer. 3 We managed to visually identify the small number of the well-formed bibliographic records in the institutional repository that for sure match the corresponding poorly- formed ARGUS bibliographic records. We then applied different modifications to the ARGUS records in combination with measuring the Levenstein distance Figure 3. Directions of typical references and two ([13]) between them and those in the library catalogue. layers of research discourse The first experiments suggest that bibliographic records from two systems: ePubs which is the The two layers only loosely interact with each other institutional papers repository and ISIS ICAT which is and the bottom one can be considered a service layer in the data catalogue supported by ISIS neutron and muon support of the top one, despite recent attempts to facility, can be successfully matched if we measure promote a view that information departments can play a Levenstein distance between modified bibliographic role of data curation units, hence expanding their remit records. A particular pretty simple technique could be from the mere support of information technology to the extraction and normalization of the numeric catering for richer tastes of researchers interested in components from the bibliographic record (volume, semantic representation of information and in its pages and year), measuring distances between such sensible reuse ([11]). normalized extracts – in effect, between two strings with only numbers in them – then playing with the 3.3 Problems, challenges and opportunities threshold (the particular Levenstein distance) that The above analysis contributes to modelling of allows to distinguish between matches and non- research discourse in facilities science with the matches. This technique was tried out via bespoke Java suggestion that data and software should play a modest software module and is illustrated by Table 3. (supportive) role compared to research papers and well- The technique tuning, including the measurements defined investigations. Different information models of precision and recall, should be done with the larger that can be applied to the same facilities research numbers of bibliographic records; there is about a discourse. One of them is the model based on Research thousand records in ICAT data catalogue that have Objects ([10]) that suggest the “enrichment” of the core bibliographic components – candidates for matching Investigation entity while it moves down the facilities them with bibliographic records in ePubs papers research lifecycle illustrated by Figure 1 – turning into a repository. Yet it has to be understood that mere rich aggregation of data, data context (metadata), and matching bibliographic records is just the first step in software. Another view is seeing research discourse as what we aspire to: a reasonably automated technique for “grid” composed of provenance chains similar to that in linking investigations to research papers in situations Figure 2; the Research Activity model ([12]) offers a where there are no bibliographic records catalogued for basic semantic means to support this view. investigations, only investigations textual descriptions Irrespective of what of the two models we adhere to, and other metadata. they are likely to use the same techniques, e.g. for There are more than ten thousand papers in ePubs matching research papers with investigations. repository that are marked up by the librarians as having One problem here is that, despite it is a requirement of relation to ISIS neutron and muon facility with no facilities to submit the “input” to the investigation proposal indication which investigation (series of experiments) or and then the “output” of it in terms of research papers that instrumental work they actually relate to. For the majority led to the idea of the experiment, or have been resulted of these papers, there are no corresponding bibliographic from it – there is no good curation of these bibliographic records in the facility investigations catalogue; hence records, or a clear requirement for their format. On the other techniques are required to match the papers to other hand, when the institutional library eventually and investigations. We consider decomposition of, on one independently collects the facility output in the form of hand, the bibliographic records from ePubs institutional research papers, they do it in a systematic way with good repository and, on the other hand, the investigation coverage and according to the best cataloguing practice but descriptions from the ISIS investigations database into there is no record of the investigation that the paper has the corresponding elements, then looking into distances been resulted from as there is no requirement to capture it between elements with the further aggregation of them in the bibliographic record, also the investigation is often into sensible metrics. The analysis of bibliographic mentioned only implicitly in the paper. So if we want more records and investigation descriptions suggests the context for the research papers and for the investigations, following elements as the candidates for mutual mapping: there is a task of matching bibliographic records coming from facilities User Office (the unit that looks after 3 http://www.isis.stfc.ac.uk/instruments/argus/argus6461.html 163 Table3. Matching bibliographic records in ICAT data Another opportunity for the validation of the catalogue and ePubs papers repository (ARGUS case) investigations-to-publications matching technique will be looking into descriptions of research proposals Levenstein (grants) in the research information portals. For ISIS distance Levenstein Levenstein between facility, it will be Gateway to Research portal distance distance “numeric” (http://gtr.rcuk.ac.uk/) that is about to start collecting ICAT ePubs Reference reference between full between parts with the investigation proposals in a systematic manner so that bibliographic “numeric” year sometime after the investigations are over, they will be references parts normalized and the last supplemented by the submission of research papers page removed resulted from them. It will be possible then to use the Pratt et al, Phys. Phys Rev newer investigations accompanied by papers resulted Rev. Lett. 96, Lett 96 from them (as submitted by the researchers themselves) 17 0 0 247203 (2006) 247203 for the calibration of the automated matching technique (2006) Lancaster et al, Phys Rev that can be applied to the large corpus of past Phys. Rev B73, B 73 investigations and research papers. 24 1 1 020410(R) 020410 Validated via two independent sources of (2005) (2006) Blundell and J Phys bibliography: ePubs institutional repository and Pratt, J. Phys.: Condens (forthcoming) records in the Gateway to Research Condens. Matter Matter 16 portal, the automated matching technique may become 16, R771 (2004) R771- 30 3 0 a useful tool for research contextualization and for R828 (2004) enrichment of the existing records in publications and data catalogues. M.T.F.Telling J Phys and Condens Apart from matching research papers with S.H.Kilcoyne, Matter 19 investigations, an interesting theme for further research Electron transfer 2 026221 could be looking into the cases of “indirect citations” in dextran, J. (2007) 81 6 6 when (see Figure 2) one research paper does not directly Phys.: Condens. Matter 19 No 2 cite another one but there is an identifiable connection (17 January from one to another through the intermediary 2007) investigation; or the similar consideration from the J Tomkinson and Phys investigations network perspective where one M.T.F Telling, Chem investigation does not explicitly refer to another but Ammonium ions Chem in alkali metal Phys 8 they are in fact connected through the intermediary halide crystals: 4434- 113 12 5 research paper(s). Discovering these sorts of “indirect Tunnelling and 4440 citations” may contribute to the development of spin relaxation, (2006) alternative metrics for measuring research output, in PCCP 2006 8 38 4434 addition to traditional paper citation metrics. 4 Conclusion The mentioned massive of records in the ISIS ICAT data catalogue (about a thousand of them) – for which the Our analysis indicates that Investigation in facilities association with ePubs papers catalogue can be established science is an intellectual entity that has a clear identity, via the earlier outlined bibliographic records matching is involved in structured information exchange and technique – can be used for the validation of automated bears some essential features similar to traditional matching between investigation metadata records and research papers. There are various opportunities for the (more than ten thousand) bibliographic records for all ISIS information modelling and for the formation of links instruments. Then validation by the researchers themselves between investigations and other intellectual entities, will be required, as well as some technical means in namely research papers that can be either an input to the support of that validation – such as online polls. investigation, or an outcome of it. This study can be considered an analysis and a roadmap that precede the scalable experiments on the information contextualization in the domain of facilities science. It is also a call for information practitioners to share their views on the research information contextualization and on the role of various intellectual entities in their research domains, as the popular notion of “data” and its widely accepted importance may sometimes misrepresent the actual content of research discourse where other domain-specific intellectual entities could be more appropriate for sensible Figure 4. Mapping research papers bibliography to information management and for measuring research investigations metadata output. 164 References in DataCite Summer Meeting, Digital Research Data in Practice (DataCite2012), Copenhagen, [1] Vasily Bunakov, Brian Matthews and Catherine Denmark. http://purl.org/net/epubs/work/62852 Jones. Towards the Interoperable Data [9] PaNKOS: Proton and Neutron Knowledge Environment for Facilities Science. A chapter in Organisation System. www.purl.org/pankos “Collaborative Knowledge in Scientific Research [10] B. Matthews et al. Investigations as research Networks” (AKATM book series). In press by IGI objects within facilities science. In 1st Workshop Global. on Linking and Contextualizing Publications and [2] J. Mesot. A need to rethink the business model of Datasets, Malta, September 26th, 2013. user labs? Neutron News, 2012, 23 (4), 2–3. http://purl.org/net/epubs/work/11912059 [3] S.J. Coles and P.A. Gale. Changing and [11] Vasily Bunakov and Brian Matthews. Data Challenging Times for Service Crystallography. Curation Framework for Facilities Science. In 2nd Chemical Science, 2012, 3 (3), 683–689. International Conference on Data Technologies [4] Reference Model for an Open Archival and Applications, Reykjavik, Iceland, 29-31 Jul Information System. CCSDS 650.0-M-2 (Magenta 2013, (2013): 211–216. Book) Issue 2, June 2012. http://public.ccsds.org/ http://purl.org/net/epubs/work/10938269 publications/archive/650x0m2.pdf [12] Vasily Bunakov. Core semantic model for generic [5] B. Matthews et al. Using a Core Scientific research activity. In 15th All-Russian Conference Metadata Model in Large-Scale Facilities. "Digital Libraries: Advanced Methods and 5th International Digital Curation Conference, Technologies, Digital Collections", Yaroslavl, London, UK. Russia, 14-17 Oct 2013, CEUR Workshop [6] B. Matthews et al. Model of the data continuum in Proceedings (ISSN 1613-0073) 1108 (2013): Photon and Neutron Facilities. PaNdata ODI, 79–84. http://ceur-ws.org/Vol-1108/paper10.pdf Deliverable D6.1. http://pan-data.eu/sites/ [13] Владимир Левенштейн. Двоичные коды с pan-data.eu/files/PaNdataODI-D6.1.pdf исправлением выпадений, вставок и [7] D. Flannery et al. ICAT: Integrating Data замещений символов. Доклады Академии Наук Infrastructure for Facilities Based Science. In e- СCCP (in Russian) 163 (4): 845–8, 1965. Science: Fifth IEEE International Conference on (Appeared in English as: Vladimir Levenshtein. e-Science. Binary codes capable of correcting deletions, [8] Michael Wilson. Meeting a scientific facility insertions, and reversals. Soviet Physics Doklady provider's duty to maximise the value of data. Talk 10 (8): 707–710, 1966.) 165