Search techniques numerical databases

CrossFire is the enhanced inhouse version of the Beilstein database. The database consists of about 30 million reports, of which five million are reactions. Hence this database is a reaction database. But apart from that, it covers structural, factual and bibliographic information. This means that CrossFire can also be classified as a structural and most of all as a numeric database. In a recently published article by Lawson, an example is shown, illustrating the use of substance-, reaction-, and document-based search techniques on the same problem. The special power of hyperlinks between reactions, substances, and documents is shown. 

A column index can be created in order to speed access to the data in the column. For example, molecular weight might be an important column used in most searches of a mass spectroscopy database. If every row of the table has to be examined to determine if the value is between, say, 100 and 200, this would be slower than searching only those rows known to contain these values. One technique used to index columns of numerical data is to presort the values into a relatively small number of bins. The index records which rows belong to which bin. When searching for values between 100 and 200, only rows that belong to the appropriate bin need to be examined. 

Each plastic in this database is first characterized by descriptive data such as its trade name, manufacturer, product group, form of supply, or additives. Then follows complex technical information on each material, with details of fields of application, recommended processing techniques, and special features. The central element of this material database is the numerical values it gives on a wide range of mechanical, thermal, electrical, optical, and other properties. All these items can be searched for individually or in the combination of properties that was the subject of the enquiry. 

For those techniques that give data that are so characteristic of a particular molecular structure that they can be termed a fingerprint , there is the attractive possibility of searching through databases of reference compounds so that a positive identification can be made quiekly. Numerous publicly-accessible databases are available for mass spectrometry, and NMR and vibrational spectroscopy. There are also utilities for predicting the appearance of particular types of spectra for a given input model. We list a number of them in the on-line supplementary section for Chapter 2 on databases, and note that a number of commercial databases also exist. 

References to molecular structure and energies as derived from spectroscopy, electron diffraction and X-ray diffraction and theoretical calculation are included. They are indexed by compound, but the numerical data are not included. These references form the basis for the work on statistical-mechanical calculations of ideal-gas functions. References to publications on theoretical calculations and correlations, molecular dynamics, equations of state, compilations, error analysis and experimental techniques are also included and can be searched by topic area. It is possible to extract from the database recent publications containing experimental data selected on the basis of the properties measured. 

---