Werner Robert Svoboda Project leader or project partecipami in numerous mattonai (Austria and FRG) and international (EECf Council of Europe) projects. Ai ihe moment guest researcher ai ihe « Gesellschafi fùr Mathematik und Datenverarbeitung » (Institut fùr Datenverarbeitung im Rechi, Bonn),
This paper is a slightly cfaanged version of the letture presentaci at the Fifth Symposium on Legai Data Processing (Vienna, 7-8 May 1979), Especially important sources of this paper are the « Technical Study in Legai Information Retrieval » (Com-mission of the Eoropean Communities) and the « Study on Common Standards for Query Languages in Computerlsed Legai Retrieval System » (Council of Europe), .
Àt present there Is a num-ber of systems whose basic structures fot..the most part date back to the early 70ies? thus representing the then available and feasible knowledge concerning users and System technology.
Revolutionary innovations are not to be expected for the next five years (such as the utilisation of artificial intelligence ori a large scale, decisive new findings in user analysis or computer linguistics), at the moment it Is rather of primary importaoce to utilise the (detailed) findings originating from the gradually developing research in this field - to adapt available systems (in a narrower sense) to man and in this context, in particular, to the legai user by user adequate and user friendly system organisation, but also Page 36 to enable the man using the System to recognise the davantages of a System utilisation andito use the System optimally.
In other words, in the reality of available systems the point Is to adapt user behaviouir and interactión standards of systems to each other. le thls context it is of primary importante that nelther the adapiation of - the system to the user alone- nor the adaptation of the user to the System will lead to succes? but that based on both approaches - with emphasis on the first - we shall have to try to achleve optimal restiits,
This .paper is _ to indicate the most important approaches to such an adaptation - coming from two sides - and the 'course which these adaptation processès should take. il adaptation of the system to the legàl user by user Friendly System Organisation
@1. The Retrieval Language
Some Gemerai Requirements for a Retrieval Language
The possibilities offered by a given software package (STAIRS, MISTRÀL, GOLEM? STATUS, FIND etc.) compared with the possibilities actually-used generally represent ony a fraction of them. Gettale basic facilities and details of the retrieval System are almost universally designa ted by the-users of ' various systems as particularly important or desirable (see below).
Since legai information systems are in many cases planned for use by the lawyers who actealiy need the information (direct end-user operation and not mid-used opera tion), there will exist a relatively large group of legai users who only occasionally make use of the system. For such occasionai users there should be available, alongside the generai user language (ali retrieval facilities), a highly simplified user language, This language could thee serve novice users as well.
To build up such a simplified language two ways seem to be feasible:
- to simplify the existing retrieval language by means of a preprocessor (i, e, to chaoge aed combine originai commands, to improve erro-r messages etc, as for example DATEV did it with STAIRS) (short term approach);
- to create a new retrieval langeage with main emphasis- on naturai langutge communicatio-n (i. e. in using the results of the artificial-intelligen-ce-research, esperìally query-answering systems) (long term approach).
Some terminals provide a certain amount of keys which may be reco-gnized by the application software as specific commands. The user - instead of -typing a còmm$ind with ali the necessary formalism - then has only to push the key. This facility is very comfortable for the user. The mostPage 37 comfortable way of usieg t System would be to use it ' by means- of a terminal specially designed tot this System (as It is done in XEXIS).
Fot the European scene this solution seems to be not feasible: In their ncrmal working environment, the acquisition of a terminal only for being connected to a legai information System wo«uld, for the large majority of eed-usersj not be justifiable, neither now nor in the near future (espe-cially considering the costs - whether or not the user -would have to pay them himself). With this in mind, direct use of an information System by such end-user s would only be practised if the terminal could also be used for other purposes (legai information being a by-product) and there-fore the terminal can't be dedicated only to use the information System. Therefore it is very important to provide mnemotechnically well built com-mands for the user.
The thinking and formulating habits pecoliar to lawyers lead to diiiculties in applying the formai and mathetnatical components of the retrieval language, The retrieval langttage of a legai information System should take this into consideration by, for example, inteoducing appropriate symbols for the operators, making parenthetical expressions avoidable and by adjusting the priorities of the various operators to lawyers1 way of thinking.
Starting and Finishing a Dialogue . -
Àfter the user has established the connectio-n between his terminal and the computing centre, e. g. by phone dialling, the System should ask the user for the password immediately. Starting of the dialogue by the user himself, e, g. by entering LOGON, is superfluous.
Procedure of admission is important above ali for correct accounting of services effected by the System; there has to be the possibility of settling. accounts in regard to the terminal and also in regard to the users which necessitates user specific passwords. When entering the password there should be the possibility of avoiding its display on- the screen or of making it illegible.
The usually available protective measures of blocking the terminal after three unsuccessful attempts of entering appears useful. If this happens, only the System controller is able to unblock the terminal.
During work; at the ' terminal, a sivitch from otte database to another (or several others) sho-uld be possible tt any time by entering a-command: the queries then should be preserved as potential search elements for a new query.Page 38
The user should be -able to stop the dialogue at every stage by a specific ccmmand. The folle wing particular forms of terminating the dialogue should be considered:
- termlnafion by switching off the terminal (by mistake);
- precauticms in case of System break downs or line disturbances ;
- termination due to inactivity of the user during a certain period of time;
- interruption of a dialogue and later continuation within the current availability or a subsequent availability of the System.
Law is the principal area of documentation in which the aetual texts of documents are frequently included in the information System. Fulltext Storage (i. e, the Storage of the originai text of the. documents or of abstraets with considerable length) in the legai field has effeets on the need of opera tors.
Fulltext search results in special search difficulties (which, however, are related to' the peculiarities of the respeetive languages). These difficulties,. include the grammatica! variations in which words can appear in a text and compound words. Most of these difficulties can be soived with the truncation, especially the right hand truncation and therefore this feature is of particular importance. Compared with right hand truncation, left hand truncation and masking is of considerable less importance.
The numiber of words to be included in the query via truncation must be sufficiently large (in practice up to 200 have been shown to be to few but this is only for dates, which could be treated specially).
Only the Boolean operators and and or are very important; not is seldom applied by users - it can have a « dangerous » and « unexpected » effect. Extensions of these operators are not-used.
When condueting fulltext retrieval the following extension of the basic boolean operators sho-uld be taken in account: In the case that several document units forni one physical document, a logicai and on the level of the physical document (i. e. the group of document units which form the physical document) could be of importance,
Positional (relazionai) operators gain in signifìcance in the light of fulltext Storage. According-to user opinion it is important to have one positional operator and it seems that it would be sufficient to hmé the possibility to define that two or mo-re search terms will appear in one sentence or within a certaie (maximum) distance. The argument of getting irrelevantPage 39 documents unless being able to define the exact position of words seems to be an artificial one: The chance of retrieving irrelevant documents by lack of a variety of positional operators is, combined with the usuai relevance rate of 50-75% of retrieved documents, not considerable.
Arithmetic operators (gfeater than, less than e te.) are of importance, taking into account the importance of dates and other struetured informatica in legai documents.
The following priorities of combination operators seem to fit best to lawyers' thinking:
1) ADJ (if existing)
3) (other) positional operators
5) AND NOT
For practical use the combining priority of or in relation to and is particularly important, Experiences with users other than naturai science experts revealed that the priority of or before and corresponds more to their intuitive understanding than vice versa. It is characteristic that the systems LEXIS, NOVA STATUS and DÀTEV with their user...