The report presents the tasks on graphical resources management thoroughly describing applied ontologies of GrafOnto research graphics collection used for solving problems of spectroscopy. The problems of ontology modularity and automatic classes` generation are being discussed. Examples of solving reduction problem as well as applied ontologies metrics are presented. In the middle of 2000s the emergence of digital scientific libraries with publications as well as Semantic Web approach oriented on semantic description of information resources induced the work on decomposition of resources into smaller parts that require the creation of semantic annotations oriented on the description of domains and various data representations used in them. Various forms of data representation are always used in scientific publications (text, tables, graphics, symbols (for example, formulas), etc …). On the other side researcher got the facilities for storing and presenting large amounts of information, although published data and information was needed for the control of this information quality. Virtual data centers in various domains appeared in the second half of the 2000-th. These data centers usually contained the published data represented in publications in tabular form. In the end of 2000s publications on scientific graphical resources' systematization started to appear Ref. [1-4]. An example of an approach to creating a collection of graphical resources in High Energy Physics is presented in Ref. [5]. The report presents the results of the final stage of scientific plots' systematization in three disciplines of spectroscopy. At the first stage we formed GrafOnto collection of graphical resources [6-10] describing the results of studies on the problems of a water molecule spectral lines' continuum and on spectral properties of weakly bounded complexes and absorption cross-sections used for the photochemical reactions rates' calculation. At the second stage the typification of plots and figures as well as the first version of GrafOnto resources ontology was done (see Ref. [11-13]). In order to upload new datasets into GrafOnto system and support them one has to solve the tasks on managing graphical resources. These are such tasks as specification of informational resources' structure for spectroscopy problems and analysis of re-
sources’ validity, control of data completeness and trust estimation. The decision support system which used in management of the collection GrafOnto is based on ontologies describing the primitive and composite plots and figures. Description of these ontologies is the aim of this report. 2
The collection is based on a digital library, containing more than a thousand articles. These articles are dedicated to spectroscopy research such as spectral lines’ continuum, weakly bound complexes’ properties and spectral functions in near and far ultraviolet range. A distinctive feature of the above problems of spectroscopy is that the major part of published data is represented in a form of plots, figures and images.
In order to create a collection, graphical objects should be manually extracted and converted into a digital form. Software used to upload, storage, view, search and integrate graphical resources into collection is original. At present, the collection contains about 3000 primitive plots included into 625 composite plots and 104 composite figures as well as about 4000 primitive plots ready for the upload. The uploaded plots describe properties of 19 molecules, 25 complexes and 50 mixtures. Almost a half of primitive plots characterize properties of a water molecule. Collections’ plots are related to dozens of physical quantities (functions) and a dozen of physical quantities (arguments). Table 1 illustrates spectral lines’ collections and a number of primitive plots related to these functions for substance groups. It is worth noting that, at present, only a part of the plots from the publication chosen by experts is uploaded into the collection. Other plots will be processed automatically after the software for machine processing of graphical resources is developed. The collection of plots that has already been created will be used as a data set for training a neural network aimed at automatic recognition of scientific graphics. The principal tasks of graphical resources management are to control resources structure and data quality. An ontology knowledge base accumulating all computergenerated information on collection components is used for making decisions during the management.
Resources structure contains plots of various types, their description, substances, functions and their arguments, physical quantities’ units, units table as well as coordinate systems and level of detail of their description, etc. Control of plots and figures validity is based on the analysis of calculated values of paired relations between cited plots and original plots related to them. Such a relation is characterized by a reference to publication, figure number and an identifier of a curve. Note that, at present, the collection of cited plots contains 693 primitive and 248 composite plots. The ontology describing the present state of the collection resources is presented below. 4
Taxonomy of some of the most important artifacts of research publishing [5] includes concepts: figure (composite figure, plot (exclusion area plot, GenericFunctionPlot, histogram), diagram, picture. In our work we defined additional concepts characterized by the methods of acquiring physical quantities (FTP, Cell, etc …) as well as their types (Theoretical, Experimental, Fitting, Asymptotic), slang names of physical quantities, etc. and declared them as subclasses o GenericFunctionPlot class. These definitions are oriented on physical quantities used as plots’ axes.
We defined the following hierarchy for forming ontologies in spectroscopy domain. Basic ontology of spectroscopy graphical resources contains three parts and each part is related to one of the three problems of spectroscopy. These problems are the following: problems of continuum absorption, weakly bound complexes as well as the specific task of spectral functions related to photochemical reactions in the atmosphere. 4.1
Basic ontology contains some classes and properties, which are used in applied ontologies of domain problems. In our case, these problems are weakly related to each other and are represented by the following independent modules: graphical resources of continuum absorption, weakly bound complexes and absorption profiles, defining rate of photochemical reaction. Each of these modules is split into three parts: the first part characterizes coordinate systems used in GrafOnto collection, the second one characterizes physical quantities, while the third one characterizes the substances, the properties of which are presented in the collection. 4.2
In ontologies classes define many resources presented in our work in a form of plots and figures from GrafOnto collection as well as in a form of description of their properties. All the classes are explicitly defined in OWL 2 syntax with the use of Manchester syntax for their definition.
In the framework of the chosen model the main entities in spectroscopy are substances (Substance class) – molecules as well as complexes and mixtures, and methods of acquiring physical quantities’ values (Method). Graphical representation is related to graphical system entity (GraphicalSystem). The components of a graphical system are, for example, the coordinate axes of plots representing physical entities. In GrafOnto each published plot or figure is related to the description of its properties (Description, ResearchPlotDescription classes). One of such properties is a bibliographic reference to a publication (Reference). The Problem class contains three individuals (Continuum, Complex and CrossSection) each identifying a problem related to a graphical resource.
Domain problems are closely related to the tasks for their solution. GrafOnto collection contains graphical resources related to the problems mentioned in the introductory abstract of this paragraph. Classes of spectroscopy problems’ ontologies contain numerous resources and their description. PhysicalQuantity class consists of two non-adjacent subclasses named SystemPhysicalQuantityDepended and SystemPhysicalQuantityIndepended. The first class contains physical quantities the dependency of which on other physical quantities is presented in plots and figures, while the second one contains physical quantities the dependency of which is presented in plots and figures.
In order to understand the names of ontology classes we have to describe the etymology first. A name may consist of several words. These words correspond to the names of individuals in the corresponding classes MethodType, SystemPhysicalQuantityDepended and Substance. Fig. 1 presents examples of schemes for creating subclasses names in A classes (Physical quantity and related substances) and B classes (Substance and related physical quantities) presented in simplified syntax.
Fig. 1. Word order in the names of A and B groups’ classes
For example, a class named Description_Experimental_Absorption _Coefficient__cm2mol_1atm_1_ contains all the descriptions of measured absorption coefficients with cm2molecule-1atm-1 dimension for a series of substances being a subclass of Physical quantity and related substances. The third group of classes related to subclasses of GraphicalSystem class is not presented in this work. 4.3
Comments for all the properties used in natural language are presented in OWL 2 ontologies code. Here we present a simplified classification of some properties related to physical quantities and descriptions of plots and figures. Description of properties related to Description and CoordinateSystem classes as well as to Temperature and Pressure quantitative characteristics is omitted.
Table 2 lists ontology properties defining their domains and ranges. The last column of the table shows abbreviations of properties used in the scheme of individual presented in Fig. 2.
Qualitative properties characterized physical quantities are hasOriginType, hasSourceType and hasMethodType. The values of hasOriginType property indicate the origin of dataset related to the plot: it should be original and should be obtained by digitizing the curve of a primitive plot. The values of hasSourceType property can describe primary data, i.e. the data obtained by the authors of the publication as well as the previously published curves (i.e. cited) and commonly known curves (i.e. expert). The values of hasMethodType property characterize qualitative acquisition of datasets of primitive plot: Theoretical is a calculation using physical or mathematical model, Experimental is measurement, Fitting is a continuous curve creation using the method of fitting to experimental values.
The relations between plots and figures are defined by 6 properties (has{OPPD, CPPD, OCPD, CCPD, MCPD}, hasPrototype) . First five mereological properties describe composition of composite plots (OPPD, CPPD) and figures (OCPD, CCPD, MCPD). The value of hasPrototype property used in the description of cited primitive plot is the corresponding original plot. This property defines the descriptions that contain datasets with closely related values. 4.4
Being equivalents of figures and plots from published graphical resources on the above problems images generated in GrafOnto system are related to the description of their metadata making the most significant part of ontology individuals included in Abox. Typification of figures and plots given in Ref. [14] is defined by the property values. Abbreviation of corresponding values is used in the names of such individuals (for example, OCP – Original Composite Plot). Fig. 2 illustrates the structure of one of such plot types, i.e. original primitive plot. Ovals stand for ontology individuals, rectangles stand for literals and directed arcs stand for objective (OP) and determined (datatype – DTP) properties. Cited primitive plot have a similar structure with an addition of observations with hasPrototype, hasChild and hasParent properties. Special cases of individuals characterizing properties of coordinate system and its axes are shown in the lower part of Fig. 2. A series of individuals are related to the classes defined by enumeration of its individuals. Primitive Plot Description (PPD) Domain Description PrimitivePlotDescription PrimitivePlotDescription PrimitivePlotDescription PrimitivePlotDescription PrimitivePlotDescription СitedPrimitivePlotDescription CoordinateSystem CoordinateSystem CoordinateSystem Y-axis Y-axis Y-axis Y-axis X-axis or Y-axis X-axis X-axis CitedPrimitivePlotDescription
Property hasReference hasSubstance hasSourceType hasOriginType hasCurveType hasCS hasCitedReference hasCSType hasX-axis hasY-axis hasMethod hasMethodType hasPY-axis hasSY-axis hasAxisScale hasPX-axis hasSX-axis hasPrototype PrimitivePlotDescription hasTemperature PrimitivePlotDescription hasPressure PrimitivePlotDescription hasSystemFigureNumber ResearchFigureDescription hasOriginalImageOfPlot ResearchFigureDescription hasOriginalPlotInformation FigureDescription hasFigureCaption FigureDescription isPartOfFigureNumber ResearchFigureDescription hasNumberOf Points ResearchFigureDescription hasPlotCaption Original Composite Plot Description (OCPD) OriginalCompositePlotDescrip- hasOPPD tion Cited Composite Plot Description (CCPD) CitedCompositePlotDescription hasCPPD Composite Figure Description (CFD) CompositeFigureDescription hasOCPD CompositeFigureDescription hasCCPD CompositeFigureDescription hasMCPD
Range Reference Substance {Primary, Expert, Cited} {Digitized, Original} {Line, Point} CoordinateSystem Reference {2D-Decartes} X-axis Y-axis Method {Theory, Experiment, Fitting} PubPhysQuanDepended SysPhysQuanDepended {Linear, Logarithmic} PubPhysQuanIndep SysPhysQuanIndep OriginalPrimitivePlotDescription float float integer URI URI string integer integer string
Abbr OP1 OP2 OP3 OP4 OP5 OP6 OP7 OP8 OP9 OP10 OP11 OP12 OP13 OP14 OP15 OP16 OP17 OP18 DT1 DT2 DT3 DT4 DT5 DT6 DT7 DT9 DT12 Op19 Op20 Op21 Op22
Op23 OriginalPrimitivePlotDescription PrimitivePlotDescription OriginalCompositePlotDescription CitedCompositePlotDescription
MultipaperCompositePlotDescription
Ontologies metrics are used for comparing ontologies of different parts of a domain or of different domains, characterizing quantitative and qualitative peculiarities of ontological description. In OWL ontologies the number of object properties characterizes the number of paired relations between individuals. Some of these individuals may have quantitative estimation. The estimated relations are described by certain (datatype) properties.
Table 3 contains metrics for applied ontologies of three spectroscopy problems as well as the unification of these ontologies (Σ Ontology). As for GrafOnto resources collection the equality of numbers characterizing the number of properties, their domains and ranges means that they are characterized by identical properties. However, the difference in classes` numbers indicates the use of a greater number of spectral functions in Continuum problem in comparison with Complex and Cross Section problems. As individuals of one and the same group of types are used applied ontologies we may conclude that ontology on Continuum problem describe the highest number of primitive plots.
Metrics comparison of Σ Ontology with ontologies of tabular information resources Ref. [14] reveals that in our work on graphical resources ontology we managed to significantly increase the number of classes in one year. It clearly indicates that Σ Ontology contains the highest number of obvious answers on typical user requests. 5
The report presents applied ontologies of scientific plots and figures used for managing
graphical resources in three problems of spectroscopy. Ontologies describe a collection
of plots and figures published in the period from 191
The future of GrafOnto collection is related to automatic recognition of plots and figures used in spectroscopy as well as to the generation of applied ontologies characterizing validity analysis and confidence estimation of its resources. 10. Lavrentiev, N.A., Rodimova, O.B., and Fazliev, A.Z.: Systematization of graphically plotted published spectral functions of weakly bound water complexes. Proc. SPIE 10035 (2016). doi: 10.1117/12.2249159 11. Lavrentiev, N.A., Privezentsev, A.I., and Fazliev, A.Z.: Tabular and Graphic Resources in
Quantitative Spectroscopy. In: L. Kalinichenko et al. (eds.) DAMDID/RCDL 2018, CCIS 12. Lavrentiev, N.A., Privezentsev, A.I., and Fazliev, A.Z.: Systematization of Tabular and Graphical Resources in Quantitative Spectroscopy. CEUR Workshop Proceedings, Selected Papers of the XX International Conference on Data Analytics and Management in Data Intensive Domains. Edited by Leonid Kalinichenko, Yannis Manolopoulos, Sergey Stupnikov, Nikolay Skvortsov, Vladimir Sukhomlin 2277, 25–32 (2018). 13. Lavrentiev, N.A., Privezentsev, A.I., and Fazliev, A.Z.: Applied Ontology of Molecule Spectroscopy Scientific Plots. Proc. of Conference "Knowledge, Ontologies, Theories", DigitPro 2, 36–40 (2017). 14. Odintsova, T.A., Tretyakov, M.Yu., Pirali, O., and Roy, P.: Water vapor continuum in the range of rotational spectrum of H2O molecule: New experimental data and their comparative analysis. Journal of Quantitative Spectroscopy and Radiative Transfer 187, 116–123 (2017). doi: 10.1016/j.jqsrt.2016.09.00