An Algorithm for Translating Chemical Names to Molecular Formulas, (Thesis)

An algorithm for translating directly from chemical names to molecular formulas is described. The validity of the algorithm was tested both manually and by computer. Molecular formulas of several hundred randomly selected chemicals were calculated successfully, verifying the linguistic analyses and the logic of the computer program. The algorithm for manual human translation consists of eight simple operations. The procedure enables non-chemists to compute molecular formulas quickly without drawing structural diagrams. Th e machine translation routine is rapid and requires a program of less than 1000 instructions. If the experimental dictionary were expanded to include low frequency morphemes, formulas for all chemical names could be handled. The problem ofchemical nomenclature is discussed in terms of the information requirements of chemists, The approach of the linguist to the problem of nomenclature is contrasted with that of the chemist, It is shown that there is only one language of chemical nomenclature though there exist many systems of nomenclature, The difficulties in syntactically analyzing Chemical Abstracts (C. A.) nomenclature results from C. A/s ambiguous use of morphemes such as imino, not the use of so-Cal led trivial nomenclature. The more systematic 1.U. P. A.C. nomenclature includes idiomatic expressions but eliminates all homonymous expressions. of the most frequently occurring segments. Approximately forty morphemes such as 10, e, y! and allomorphs such as thi and sulf were isolated, A list of their 200 actual co-occurrences were compiled. These studies are particularly valuable in identifying idiomatic expressions such as diaz, the meaning of which cannot be computed from the referential meanings of di and az. Morpheme classes are illustrated by the bonding morphemes (an, en, yn, iun, etc.) and the homologous alkyl morphemes meth, eth, prop, but, etc. The syntactic analyses include the demonstration of transformational properties in chemical nomenclature as e.g. in primary amines (R-N) where aminoRune= Rylamine. To complete the grammar onewould have to expand the inventory of morphemes, morpheme classes, and the list of transformations, Chemical name recognition is not simply a word-for-word translation procedure a Rather the syntactic analysis required is comparable to the procedure employed by Harris, IIiz, et al (Transformations and Discourse Analysis Projects, Univ. of Pennsylvania) for normal English discourse. The structural linguistic data is supported by a summary of 1. U. P. A.C. rules for generating chemical names. In order to relate this study to the general problem of chemical information retrieval, the historical development of chemical nomenclature is traced from the 1892 Geneva Conference to the present. The relationship between nomenclature, notation, indexing and searching (retrieval ) systems is discus sed, In particular, the need for linguistic studies to solve the intellectual facet of the “retrieval” problem is discussed in contrast with the manipulative aspects which are more readily amenable to machine handling. The problem of synonymy in chemical nomenclature must be resolved if computable syntactic analyses of chemical texts are to be used for mechanized indexing. Th e completion of the detailed grammar of chemical nomenclature would not only permit the calculation of molecular formulas but also the generation of structural diagrams, systematic names, line notations, and other information required in machine searching systems, With suitable modifications the procedures could easily be applied to foreign nomenclature. The field of chemico-linguistics is of interest to the organic chemist as it can improve methods for teaching nomenclature. Similarly, for the linguist chemical nomenclature is a fertile field of study, One can control the experimental conditions more easily than in normal discourse, However, conclusions can be drawn which may have more general application.