LIBRES: Library and Information Science Research
Electronic Journal ISSN 1058-6768
2000 Volume 10 Issue 2; September 30
Bi-annual LIBRES 10N2


Modular Bibliometric Information System with Proprietary Software (MOBIS-ProSoft): a versatile approach to bibliometric research tools.

 

Gilberto R, Sotolongo-Aguilar *, Carlos A. Suárez-Balseiro **,  Maria V. Guzmán-Sánchez *

 

* The Finlay Institute; POBox 16017, Cod. 11600 La Habana, CUBA. E-mail: finlayci@infomed.sld.cu

** Faculty of Communication, University of Havana Calle G, No.506, Vedado, La Habana 10600, La Habana, CUBA. E-mail: csbgv@bib.uc3m.es

 

Abstract

 

This paper presents a platform outline for bibliometric research. Conceived as a modular system it is based on proprietary software. This proposal intends to show a low complexity framework software which is reasonably widely available including artificial neural networking software. This approach works smoothly with small and medium size corpora and may be very useful for both research and educational purposes.

 

Introduction

One of the challenges today for information professionals is to guide the way through huge volumes of information generated by different means. The birth and development of new disciplines such as “data mining” and “knowledge discovery”, shows the increasing importance of quantitative and qualitative analysis of huge corpora data (Dhar and Stein, 1997; Swanson and Smalheiser, 1997).

 

Bearing this in mind, bibliometric research becomes one of the fundamental tools used by information professionals in their quest of indicators; allowing them “critical appraisals” of scientific research, as well as interaction among researchers, institutions and knowledge areas.

 

The above reasons have conditioned increasing efforts for the systematization and standardization of methods and tools used in bibliometric research. Classical studies have supported the importance of clearly defining the problems in the field, emphasising the application of statistics as a key factor in discovering new knowledge (Egghe and Rousseau, 1990). For Glanzel (1996), bibliometrics is a complex discipline which, although it  may be classified as a social science, is closely conditioned by pure and technological sciences. Therefore any methodological characterization requires well-documented data processing methods, a clear description of the sources and exact definition of indicators and, on the other hand, there is a need for an effective selection and integration of the applied technologies. Ravichandra Rao (1996) asserts that no bibliometric technique alone can be applied to all research, but instead different procedures should be used for different problems. Grivel, Polanco and Kaplan (1997) emphasize what they call  “informatic infrastructure” where bibliometrics could develop all its potential. According to these authors, bibliometrics should have a methodology characterized by not only an adequate mathematical representation but also an effective “informatic architecture”.

 

Therefore bibliometric information systems are the workbench of bibliometric research. Being an  important part of this field of endeavor, they require a flexible design in order to obtain accurate and customized indicators and should integrate new features resulting from the latest developments.

 

Many colleagues have found their way into bibliometrics by building in-house applications. At the end of the 80’s Terrence Brooks prepared a set of computer programs written in Turbo Pascal called the Bibliometrics Toolbox in order to measure the bibliometric aspects of a literature (Brooks, 1987; McLain, 1990). For Van Raan (1996) in Leiden the chosen name was “The Machine”; in CRRM they use a software suite, with DATAVIEW (Rostaing et al., 1996) as flagship, and the CUIB-METRIC system is proposed by specialists at the UNAM in Mexico (Portal and Thompson, 1994). There is the application of TOAK -Technology Opportunity Analysis Knowbot - at the Technology Policy & Assessment Center, at Georgia Institute of Technology, in Atlanta, USA (Porter and Detampel, 1995) and HENOCH (Grivel et al., 1997), and NEURODOC (Polanco et al., 1998) which are used at INIST in France. Bibexcel (Bibmap before Excel), developed by Professor Olle Persson, from Inforsk, Umeå University in Sweden, and BibTechMon (Kopcsa and Schiebel, 1998) are other available approaches. The work of Katz and Hicks (1997), Small (1998), and White and McCain (1998) takes the same direction. Finally we have to mention the experiences of Chen (1995), Lin (1995, 1997), and Orwing, Chen and Nunamaker (1997) in the application of artificial neural networks for bibliometric purposes based on the Kohonen’s self-organizing map (SOM), which  is an orderly mapping of a high-dimensional, eventually structured distribution of data onto a regular low-dimensional grid. The Kohonen´s SOM is probably the best know network model geared towards unsupervised training and essentially consists of a regular grid of processing units or “neurons” associated with a model of some multidimensional observation to represent all the available observations with optimal accuracy, using a restricted set of models ordered on the grid so that similar models are close to each other and dissimilar models far from each other  (Kohonen, et al., 1999; Kohonen, 1998).

 

However the problem arises of  when generalization should be done. In-house applications are rarely well documented and their use by others becomes difficult. The result is that only the members of the team are able to replicate the use of such  applications. The standardization fails and it is not only a handicap for practical research;  it becomes a barrier for teaching purposes because many educational institutions are not be able to obtain and implement in-house applications to support bibliometric educational programs.

 

This problem may be overcome in part using a set of proprietary software, which is well-documented, widely available and more accessible than in-house applications. On the other hand, the validation of techniques is obvious and many teams of developers are continuously improving the performance of such software.

 

This paper describes a methodology based on the utilization of a set of proprietary software working together in order to perform bibliometric analysis of a literature. This methodology is explained as an open and flexible bibliometric information system in compliance with a simple modular design and connectivity for desktop work. It is useful for practical work as well as for education and training purposes. We envisage our task as seeking the integration of different available software with the objectives of consolidating an informatic infrastructure for our bibliometric research and developing standard methods that fit with this purpose.

Methodology

Our approach consists of five modules based on proprietary software integrating the system. The modules perform the following functions:

 

(1)   Bibliographic Searches

(2)   File Conversion & Handling

(3)   Bibliographic Reference Management

(4)   Indicators including (experimental) Artificial Neural networking

(5)   Bibliometric Analysis

 

Bibliographic Searches are conducted online or on CD-ROM.  Resulting files are downloaded and converted by module (2) File Conversion & Handling. Resulting files are the input to module (3) Bibliographic Reference Management, where the standardization of the database is performed. Different fields under study or a combination of them are exported and saved as text files. Afterwards, those files are processed in module  (4) Indicators. In this module several statistical analysis may be carried out  to obtain the inputs for the bibliometric analysis in the module 5.  Different scenarios could be implemented, varying  the elements inside each module.

 

In our experience, for small and medium size corpora, the following packages appear to work very  well:

 

1.      Bibliographic Searches. Dependent on the topic of research e.g. in biomedicine SPIRS, WINSPIRS, both from Silver Platter, PubMed and Internet GratefulMed or The Query E-mail Retrieval System from NLM.

 

2.      File Conversion & Handling. Resulting files are downloaded and treated by BiblioLink™ converting them according to a selected configuration that depends on host and fields to be studied. BiblioLinkä convert the files to Prociteä format.

 

3.      Bibliographic Reference Manager. Procite™ (Research Information Systems Inc.), works very well for the managing of bibliographic references allowing standardization of data. Furthermore BiblioLinkä and Prociteä are totally integrated in their latest versions.

 

4.      Indicators (obtained in this module). The functions of this module are performed by different statistical packages, e.g. Excel™ (Microsoft Corp.) and its complement xlStat™ (Stat@Com Inc.) and STATISTICA® (StatSoft Inc.). The former gives the possibility of profiting from all the built-in features of this program including graph and functions features. We have recently introduced an experimental submodule for Artificial Neural Networking. We used Viscovery ® SoMine from Eudaptics Software Gmbh for this purpose, allowing us to work without models and statistical assumptions by using the powerful Self-Organizing Maps (SOM) Technology. It leads to a very good representation of high-dimensional data by maintaining similarities implicit in the data.

 

5.      Bibliometric analysis. Finally in this module the analysis of indicators is performed according to the aims of the particular task undertaken.

 


Results & Discussion

 

The above mentioned scenario operates according to the following procedure: bibliographic searches are conducted online or on CD-ROM. Resulting files are downloaded and treated by BiblioLink™ converting them according to a selected configuration that depends on host and fields to be studied. The resulting converted file is already in Procite™ format having the possibility to switch directly to the bibliographic reference management features of Procite™. Here standardization of the database is conducted. Many different treatments could take place including the building of authority lists with the contents of different fields including an authority list of all words in any field or in the whole database. The different fields under study or a combination of them are exported and saved as text files. Afterwards, Excel™ imports those files. Frequency analysis may then be performed aided by the Pivot Table feature of Excel™ complemented by built-in Analysis Functions available in the Tools Menu. With Excel™, using xlStat™ it is also  possible to build the matrices that produce the input for cluster analysis, factor analysis, PCA, and multidimensional scaling  and undertake these analysis. Alternatively, those matrices may be exported as Excel™ sheets and imported into STATISTICA® (StatSoft Inc.) and finally processed. More recently we have been experimenting with Viscovery SoMine. Beyond its visual exploration capabilities, Viscovery® also supports in-depth statistical analysis of data. The combination of the non-linear data representation of the SOM approach with classical statistical methods - such as regressions or principal component analysis (PCA) - results in the improvement of the final model in terms of precision  and efficiency.

 

This system platform guarantees a comprehensive traceability of all data from the first data downloaded, to the last chart obtained. At the same time, consistent results are attained by means of the reproducibility at all the steps performed. Bibliographic data in the database could also be used for building-up formatted bibliographies.

 

This approach has been applied in different studies (Macías-Chapula et al., 1999; Guzmán-Sánchez et al., 1998; Sanz-Casado, et al., 1998) mainly focused on the biomedicine field and more specifically in the area of vaccines research. However, other domains such as library and information science or economics have been explored (Sanz-Casado et al., 1999; Sotolongo-Aguilar, 1999)[1]. Bibliometric output data of the system could perform, among others, the following activity and relations measurements:

 

1.      Counts of papers by the following fields or a combination of them:

·        Authors

·        Sources

·        Keywords (e.g. MESH)

·        Years

·        Substances

·        Documents types

·        Languages

·        Affiliation

·        Country of publication

·        Authors/papers

·        Substances/papers

·        Keywords/papers

·        Document types/papers

 

2.      Co-occurrence matrices for multivariate analysis of the following fields or a combination of them:

·        Authors

·        Keywords

·        Substances

·        Document types

·        Self-Organized-Maps for spatial representation of linear or multidimensional data

 

For bibliographic searches in biomedicine we have been using The Query E-mail Retrieval System from NLM. This is a very nice retrieval engine by e-mail and works very well. In the case of the bibliographic reference management software, we have extensively used Procite™ beginning with version 2.02 (MS DOS) up to the latest available 5.0 (for Windows). The advantage of the latter is that it integrates its companion file-conversion software BiblioLink™. It also works very smoothly. Other reference management software has been tested e.g. EndNote™ and Reference Manager™ including the latest versions. There are  nearly 40 reference management software packages on the market eligible for these tasks. Statistical packages are another important component. Undoubtedly EXCEL® is widely used and complies very well with many bibliometric tasks. A very good complement to EXCEL® , as already mentioned, is xlStat™ with many useful features for building matrices, cluster analysis, factor analysis and PCA, and multidimensional scaling. Finally the above mentioned Viscovery® SoMine (Eudaptics Software Gmbh 1999) seems to be a very powerful application; it is based on the concept of Self-Organizing Maps (SOM) which is a particularly robust form of unsupervised neural networks. Teuvo Kohonen first introduced the SOM method which can be viewed as a non-parametric regression technique that converts multi-dimensional data spaces into lower dimensional abstractions. Much as a regression plane is an abstraction of the original data, Viscovery® SoMine generates a representation of the data distribution, with the difference that this representation is non-linear. Inside Viscovery, a two-dimensional hexagonal grid realizes the SOM. Starting from a set of numerical, multivariate data records, the "nodes" on the grid gradually adapt themselves to the intrinsic shape of the data distribution. Since the order on the grid reflects the neighborhood within the data, attributes and features of the data distribution can be read off from the emerging "landscape" on the grid. The resulting "map" contains the representation of the original data distribution. In a second step, this data representation is systematically converted to visual information in order to enable the application of a number of evaluation techniques. By means of a new network scaling method the learning process has been improved significantly. For illustrations see the Appendix.

Conclusions

The benefits resulting from the developments outlined in this paper could be threefold. Besides integrating public domain software in a flexible modular design, comprehensive automated processing and data representation stages of research could be achieved in contrast to the cumbersome tasks that  are performed by other means. This platform is supported on software which is widely used  and regularly updated and upgraded; in contrast with adhoc software that becomes outdated very rapidly.

 

Last but not least, teaching bibliometric research seems to benefit from this approach, bearing in mind its use of proprietary software available world wide, regularly updated and supported by well established developing teams. Improved bibliometric research practices must be supported by theoretical and practical education in which technologies have a key role. However, if the methods and tools used are not widely accessible and it is necessary to depend on an in-house application, the experience could be frustrating.

 

Finally we have to emphasise the fact that MOBIS-ProSoft is not new software, it is not a new in-house developed application; it is a methodological approach using a set of different proprietary applications  working in modules. This methodology has been shown to be a working platform that could be up-graded, is flexible and has utility performance. Moreover it is a practical alternative for educators to improve their teaching programs. Improvements to MOBIS-ProSoft are foreseen. Participation in the testing of this methodology is welcome, as well as new ideas for incorporating modules or improving the existing system.

 


Appendix

 

This appendix contains figures illustrating some features of  the MOBIS-ProSoft (modular bibliometrics information system with proprietary software) approach.

 

Figure 1. Matrix-building module showing (partially): raw data (upper-half); frequency of ocurrence (lower half – left); coocurrence matrix (lower half – right)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 


One of the most interesting features of using  the MOBIS-ProSoft approach, is the matrix-builder feature implemented by running biblio.xla module under xlStat™. A selected multiword field of the database, for instance the Subject field or the Author field, is exported as comma delimited text and after it is imported into Excel™. Then the biblio.xla module can be run.  The resulting Excel™ sheet of the current workbook looks like the upper half of figure 1. Column A is manually added for assuring that all elements will be included in the matrix-building process. The results of running biblio.xla appears in the lower half of figure 1; at the left the frequency table is built up; after skipping one column the symmetric coocurrence matrix appears at the right. The square matrix can have  a maximum size of 252 x 252. In the illustration is a data set from ISA (1966-1998) descriptors related to Library & Information Science Research Methods in Latin America & the Caribbean.

 


Figure 2. Dendogram resulting from clustering using the Ward method including clusters observation / clusters size

 

 

This illustration shows the results of  clustering similarities e.g. Pearson Product Moment rp, using the Ward method. In this case country clustering according to the Activity Index from ISA (1966-1998)  Library & Information Science in Latin America & the Caribbean. The illustration includes the Clusters Observation / Clusters Size for the best partition suggested by running the results i.e. three groups of size 5, 5 and 9, with the corresponding country names by group.

 


Figure 3.  Map based on  principal component analysis with Varimax transformation

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 


This illustration shows the map corresponding to the same data set used in figure 2. In this case the display is a PCA map with Varimax transformation. The groups founded in the clustering procedure shown in figure 2 are highlighted.

 


Figure 4. Map based on  multidimentional scaling

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 


 Figure 4  shows  an MDS map which displays another view of the data set used in figures 2 and 3. In this case dissimilarities were calculated as 1- rp.

 


Figure 5.  Self Organized Map (SOM) and windows of map-layers & values based on an Artificial Neural Network (ANN) trained with  Kohonen unsupervised algorithm.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 


The most intriguing feature recently incorporated on an experimental basis to the MOBIS-ProSoft approach is the SOM-ANN. Again, as in figure 1, a data set from ISA (1966-1998) descriptors related to Library & Information Science Research Methods in Latin America & the Caribbean is used, in this case time series of descriptor occurrence for the years 1983-1984,1986, 1990-1998. Viscovery® SoMine was used for training the network based on 133 data records from same number of descriptors characterized by 12 components  (dimensions or features). Forty four cycles in normal exact mode of training were needed  for generating a map size of 100:61 with 1905 nodes that represent the trained network .  Figure 5 shows a screenshot with some results. In the upper-left part of the screenshot is a  map showing the clusters formed, at the upper-right the values of some of the year component of the cluster corresponding to bibliometrics that is located in the lower-left corner of the cluster map. The rest are all the map-layers corresponding to all the dimensions (one for each displayed from left-right  top-bottom. A different tone of gray (different colors in the original) shows different “landscapes” views. The first year of the time series (left map in the first row of maps-layers) displays isolated spots of activity, while as time goes by (second row of maps-layers) the activity increases and non-linear correlation could be observed. Resulting data from the trained network could later be evaluated.

 


References

 

Brooks, T (1987). The Bibliometrics Toolbox, version 2.8.  North City Bibliometrics. Available at: ftp.u.washington.edu/public/tabrooks/toolbox/

 

Chen, H. (1995). Machine Learning for Information Retrieval: neural networks, symbolic learning, and genetic algorithms. Journal of the American Society of Information Science 46(3), 194-216

 

Dhar, V. and Stein, R. (1997). Seven Methods for Transforming Corporate Data into Business Intelligence.  Prentice Hall.

 

Egghe, L. and Rousseau, R. (1990). Introduction to Informetrics. Quantitative Methods in Library Documentation and Information Science. Netherdlands: Elsevier Sciences Publisher.

 

Eudaptics Software Gmbh. (1999). Viscovery® for CRM-applications (Viscovery White Paper). Available from: http://www.eudatic.com/

 

Glanzel, W. (1996). The need for standards in bibliometric research and technology.  Scientometrics 35(2), 167-176.

 

Grivel, L., Polanco, X. and Kaplan, A. (1997). A computer system for big scientometrics at the age of the worldwide web.  Scientometrics, 40(3), 493-506.

 

Guzmán-Sanchez, M.V., Sánz-Casado, E. and Sotolongo-Aguilar, G. (1998). Bibliometric study on vaccines (1990-1995) in Iberian-American countries. Scientometrics 43(2), 189-205 .

 

Katz, J.S. and Hicks, D. (1997). Desktop Scientometrics. Scientometrics 38(1),141-153.

 

Kohonen, T., Kaski, S., Lagus, K., Salojärvi, J., Honkela, J., Paatero, V., and Saarela, A. (1999). Self-Organization of a massive text document collection. In: Oja, E. and Kaski, S. editors. Kohonen Maps. Amsterdam, Elsevier pp.171-182.

 

Kohonen, T. (1998). Self-organization of very large document collection: State of the Art. In: Niklasson, L., Boden, M. and Ziemke, T., editors. Proceedings of ICANN98, 8th International Conference on Artificial Neural Networks, vol. 1, Springer, London. pp. 65-74.

 

Kopcsa, A. and Schiebel, E. (1998). Science and technology Mapping: A New Iteratio Model for Representing Multidimensional Relationships. Journal of the American Society of Information Science, 49(1), 7-17.

 

Lin, X. (1995). Searching and Browsing on Map Displays. Proceedings of ASIS 95, Chicago, 13-18.

 

Lin, X. (1997). Map Display for Information Retrieval. Journal of the American Society of Information Science 48(1), 40-54.

 

Macias-Chapula, C.A., Sotolongo-Aguilar, G.R., Madge, B. and Solorio-Lagunas, J. (1999). Subject content analysis of AIDS literature as produced in or about Latin America and the Caribbean. Scientometrics 46(3), 563-574

 

McLain, J. P. (1990). Bibliometrics Toolbox.. Journal of the American Society for Information Science 41(1), 70-71.

 

Orwing, R., Chen, H. and Nunamaker, J. (1997). A Graphical, Self-Organizing Approach to classifying electronic meeting output. Journal of the American Society of Information Science 48(2), 157-170.

 

Polanco, X., Francois, C. and Keim J. P. (1998). Artificial neural network technology for the classification and cartography of scientific and technical information. Scientometrics  41(1-2), 69-82.

 

Portal, S. G. and Thompson, A. C. (1994). CUIB-METRIC: an integral system for metric analysis of bibliographic information. Investigacion Bibliotecológica, 8 (16), 27-31.

 

Porter, A. L. and Detampel, M. J. (1995). Technology Opportunities Analysis. Technological Forecasting and Social Change 49, 239-255.

 

Ravichandra Rao, I.K. (1996). Methodological and conceptual questions of bibliometric standards.  Scientometrics, 35(2), 265-270.

 

Rostaing, H., Dou, H., Hassanaly, P. and Paoli, C. (1996). Dataview: bibliometric software for analysis of downloaded data. Available from: http://crrm.univ-mrs.fr

 

Sanz-Casado, E., García-Zorita, C., García-Romero, A., and Modrego-Rico, A. (1999). Research by spanish economists. Characteristics in terms of the scope of publications. Proceedings of the 7th  International Conference on Scientometrics and Informetrics, University of Colima, Colima, Mexico, July 5-9, pp. 593-595.

 

Sanz-Casado, E., Suárez-Balseiro, C.A., and García-Zorita, C. (1998). Estudio de la producción científica española en biomedicina durante el período 1991-1996. Actas de las Jornadas de Documentación en Ciencias de la Salud, Zaragoza, marzo 1998. (Copies available from the authors).

 

Small, H. (1998).  A general framework for creating large-scale maps of science in two or three dimensions: the SCIVIZ system. Scientometrics 41(1-2), 125-133.

 

Sotolongo-Aguilar, G. (1999). Library and Information Science Research Methods in Latin America and the Caribbean (1966-1998). Report 23, Scientific University Council (SUC), University of Havana, Cuba. (Available from the authors).

 

Swanson, D. R. and Smalheiser, N. R. (1997). An interactive system for finding complementary literatures: a stimulus to scientific discovery. Artificial Intelligence 91, 183-203. Available from: http://kiwi.uchicago.edu/webwork/AIabtext.html

 

Van Raan, A.F.I. (1996). Scientometrics: state-of-the-art. Scientometrics 38(1), 205-218.

 

White, H. D. and McCain, K. W. (1998). Visualizing a Discipline: An Author Co-Citation Analysis of Information Science. Journal of the American Society of Information Science 49(4), 327-355.

 

 


 

This document may be circulated freely
with the following statement included in its entirety:

Copyright 2000

This article was originally published in
LIBRES: Library and Information Science
Electronic Journal
(ISSN 1058-6768) September 30, 2000
Volume 10 Issue 2.
For any commercial use, or publication
(including electronic journals), you must obtain
the permission of the authors.

Gilberto R, Sotolongo-Aguilar
The Finlay Institute; POBox 16017, Cod. 11600 La Habana, CUBA. E-mail:
finlayci@infomed.sld.cu

Carlos A. Suárez-Balseiro
Faculty of Communication, University of Havana Calle G, No.506, Vedado, La Habana 10600, La Habana, CUBA.
E-mail: csbgv@bib.uc3m.es

Maria V. Guzmán-Sánchez
The Finlay Institute; POBox 16017, Cod. 11600 La Habana, CUBA. E-mail:
finlayci@infomed.sld.cu


To subscribe to LIBRES send e-mail message to
listproc@info.curtin.edu.au
with the text:
subscribe libres [your first name] [your last name]
________________________________________

Return to Libres 10n2 Contents
Return to Libres Home Page

 

 



[1] All the figures used in the appendix are related to the report prepared by Gilberto Sotolongo-Aguilar for the Scientific University Council of the University of Havana in 1999. This study focused on library and information science research methods in Latin America and the Caribbean from 1966 to 1998.

CRICOS provider code: 00301J