Aspects of Modern Electronic Publishing Systems

s. After the publications are found, the system performs a similaritysearch of submitted articles and recent publications of the referees. We thinkthat we are able to assure the quality of publication into a long term withoutany additional human interaction! 9.5.2 Search for Publications of Referees In the following some possible solutions to the problem of finding the publica-tions of our referees are discussed. Some some of the freely available serviceswere explored. To enable similarity search the goal is to find at least theabstract of papers written by a particular referee. Several available search-engines index a lot of material stored all over the world.The most powerful search-engine at the time of writing is the Google searchengine with more than 2 ∗ 10-indexed web pages. An application programinterface (API) for Google using SOAP (Simple Object Access Protocol) isavailable for different programming languages. The problem of formulating theright search query still exists and there are some problems in extracting theright entities from the result. Therefore Google is unfortunately not suitablefor our task in searching the publications of referees. Many referees do supply us with the address of a personal homepage where alist of publications is very often provided. Often authors are allowed to makeat least the abstract available at their home page after signing a copyrighttransfer statement with the publisher of the paper. Older papers (usuallypapers older than one year) are often available in fulltext on the homepage.Nevertheless, homepages differ in many respects (structure as well as format)and an automatic extraction of the publications of an author is not easilypossible. There are still some referees without a homepage or without a list of 138CHAPTER 9. SUPPORT IN THE REVIEW PROCESS publication page linked from the homepage. It is conspicuous that informationextraction from homepages is difficult and unmanageable in a long perspective. Unstructured data is difficult to put into the right context therefore we werelooking for structured, freely available material about publications. For com-puter science related material a service located at University of Trier ([Ley,2002]) is very appropriate. The data is stored in an XML formatted file andis accessible via a high sophisticated search interface. Hyperlinks to onlineavailable electronic resources are provided. It is easy to search using first andlast name of an author and get back a list of known publications of this authorincluding some meta data (year, type, title etc.). We think that this service isa good start point for further explorations of publications of an author. Another well-known service for the search of publications is NEC ResearchInstitute’s ResearchIndex ([NEC Research Institute, 2002]). Additionally toinformation offered by the system from University of Trier, ResearchIndexoffers a citation index and caches also the publication itself. Since we aregoing to explore this system also in another project, we decided to use thissystem as a source of publication-gathering from our referees as well. With these two systems we are able to gather at least some of the publicationsfrom our referees and are able to extract either the fulltext or the abstractof the publication. We are going to write adapters to make this informationavailable to our system. Additionally we are going to encourage our refereesto supply the system with abstracts of their recently published papers. At the moment we are developing tools for gathering information about ourreferees concerning publications and are also developing a web-application toenable upload of abstracts and additional information about publications ofour referees. We use web-data-extraction tools (a very nice overview of someof these tools is given in [Laender et al., 2002]) as well as simple text extrac9.5. SEARCH FOR THE RIGHT REFEREE139 tion tools to extract information out of web resources. Each publication isrepresented in the Hyperwave Information Server (HIS) as an entity consistingof a collection holding all the necessary meta data (title, year, keywords etc.)and the abstract of the publication. This collection structure makes it possibleto add future developments like fulltext of the publication or citations of theentity. We are aware of many problems (e.g. author’s name representationdiffer in format: one system supply the full information i.e. first-, middle-and lastname, other systems just supply the first letter of the firstname andthe lastname...) already solved in the mentioned systems. The implementa-tion of the prototype and the upcoming feedback and evaluation procedurewill demonstrate if the described approach for gathering information aboutpublications is applicable. 9.5.3 Similarity Search Once the publications of the referees are stored in our database, searching forsimilar articles is just “a mouse click away” and very easy to handle in theused publishing server system. We use a highly sophisticated knowledge man-agement systems (Hyperwave Information Server, HIS) where features such as“find similar articles” or “find a category for an article” are already integrated.Two systems are available at the moment: [Verity, 2002] and [Autonomy, 2002].These systems support the programmer with slightly different parameters andimplementations. Nevertheless, getting a similar document is very easy toimplement. Since similarity search may also be scoped (i.e. the search scope may berestricted to a specific part of the server) there is an enormous potential forthis function. It is possible to limit the search scope to all articles publishedlast year. If no appropriate referees are found then the search scope may be 140CHAPTER 9. SUPPORT IN THE REVIEW PROCESS extended to a larger time-scope. If there is a lack of potential referees for atype of topic then the editorial board must be extended etc. Search results may be sorted in different ways to guarantee quality in thereview of articles. To give some examples: if the referee is the only author ofthe article and the article was written just a few months ago, then this refereeis very suitable to review this new article. If, on the other hand, the referee isone of several co-authors and the article was published a long time ago, thenthe referee may be not the appropriate person to review the selected topic. Asall reviewing activities of the referees are logged in the system, the activity ofthe referee may also be considered when selecting the referee for reviewing anarticle. Articles published in journals may be endowed with a higher prioritythen those published in conference proceedings. Further parameters includinglength of the article, additional ratings of other users of the article, relevanceto the community represented by a citation index parameter etc. may be takeninto account. It has been shown that there are many parameters to adapt to find the rightreferee for a submitted paper. After an evaluation period of the system feed-back from the referees will be evaluated to improve the performance of thesystem. 9.6 Further Applications Many other applications are possible with the shown approach of informationgathering and searching for similar documents. In every situation where ex-perts about a topic must be found, this approach may be a good choice toimprove the selection of experts. 9.6. FURTHER APPLICATIONS141 Conference management systems (e.g. ConfMan, [Halvorsen et al., 1999], usedat EurAsia 2002) support conference organizers in the reviewing process ofsubmissions. Submitted papers are uploaded to the server system and metadata is added by the authors via web form. Additional information about asubmission (like keywords) are added by the authors themselves. Members ofthe program committee sign-up for a review and return the results via email orweb-form. These systems may be extended by the described ideas: automatickeyword extraction from the submitted papers will guarantee appropriate key-words, the most appropriate referee of a submission may be found by simplydo a similarity search between the content of the submission and the abstractsor fulltexts of articles written by the referees. The number of referees per sub-mission may be reduced and therefore time is saved without losing quality ofreviews. Refereeing of grant proposals may be improved by automatically selecting theappropriate referee for the proposal. Whenever an expert or expert group of anumber of given persons must be found the described approach may improvethe current situation. This idea can be extended to introduce a support sys-tem for writing scientific papers ([Krottmaier, 2002a]). Experts about a topicare found automatically and the system may generate an automatic referencesection for publications in preparation. It is and will not be possible to exploreand gather information about all experts. Nevertheless, it is planned that atleast a selection of experts who published at least one paper in a well-known,reputable journal are considered to be included in the data base. 142CHAPTER 9. SUPPORT IN THE REVIEW PROCESS 9.7 Conclusion and Future Work Recent publications of authors express their current research interests in muchgreater details than information provided by the authors like unstructuredformatted lists of keywords and/or ACM categories. Interests change overtime and many experts do not update information given on their homepage. It has been shown that there are several applications where experts must befound. An application is the assignment of a number of referees to submittedpapers of a journal. Constant quality of publications and a minimum amountof time for the referees and editor(s)-in-chief are the main advantages of thedescribed system. Several problems when finding publications of an authorare already solved in other systems ([Lawrence et al., 1999] or [Ley, 2002])therefore the system must not deal with these problems. Future work is addressed to implement some of the ideas presented in thispaper. Gathering of information about a referee will be implemented usingthe two mentioned systems. Contact to the operator of one system is alreadyestablished. Other technologies of gathering information (like intelligent agenttechnology) are currently explored and compared to the described approaches.If publications are found, information must be extracted and the appropriatedata must be inserted into the database system. Usability studies and feedbackwill guarantee a fast and easy to use application.