Resulting, factual and literature data bases are comple-
menting not substituting each other.
There are some efforts to develop software systems which:
can manage literature as well as factual data bases. But
until now there are only a faw document retrieval systems
which have some features of fact retrieval Systems. As a
rule two software systems will have to be implemented.
Just one example for the reasons shall be given:
A document retrieval system usually manages character
strings, only like INTERNATIONAL SOCIETY FOR PHOTOGRAMME =
TRY AND REMOTE SENSING. Typical functions are to find an
entry in an alphabetically ordered list or to identify a
substring, e.g. PHOTOGRAMM. A fact retrieval system must
be able to handle numbers. But in an alphabetically or-
dered list the character string "200" representing the
value 200 precedes "21" and "21" is different from "21.0".
Therefore fact retrieval systems are designed to manage
numerical data formats which are ordered by value. Alpha-
numerical fields are usually provided, too but free for-
matted character strings with the corresponding functions
are not,
2. Some features of a system in operation
A software package for the documentation of measured valu-
es was developed by the Betriebsforschungsinstitut in Diis-
seldorf, Federal Republic of Germany. At the Betriebsfor-
schungsinstitut the software is. used for as different in-
formation systems as data bases on the properties of iron
and steel and a data base on plants of the iron and steel
industry. A German enterprise and EURATOM, the joint rese-
arch centre for nuclear research of the Commission of the
European Communities, purchased the software for implemen-.
ting data bases of their own.
The main functions of the system are:
. Selection of values out of the data base or files;
- descriptive statistics;
- regression and correlation analysis.
Where appropriate the results are given in graphical form
on an interactive cathode ray tube terminal or an a plot-
ter.
Figure ? represents an example of descriptive statistics.
The values are exhibited in a normal probability chart
where the cumulative frequency function becomes a straight
line. In addition the 95 % confidence limits are given.
Thus from point A for example it can be concluded that less
than 10 7? of the values will be lower than 14.3 kpm/cm?
with a probability of 97.5 %. This is an important result
e.g. for a quality control system.
Dahte 3
