Computer aided extraction of chemical reactions from natural language text documents represents a problem of high complexity. Thus, chemicals may be mentioned as starting materials, catalysts, supports, solvents, side products or products � all representing different roles to be captured. Further, additional information includes temperature, yield or other quantitative or qualitative data that increases the value of the information on a particular chemical reaction.We will present a semantic text mining system that uses chemical ontologies on chemical compounds, chemical classes and substituents in conjunction with a chemical reaction relationship model to extract information on chemical reactions from scientific publications. The system first identifies all named entities such as chemical named entities, classifies chemical names as specific compounds, chemical compounds classes, substituent lists or general reaction related named entities. In addition, relevant units, numeric or qualitative value and terms such as for example �excellent yield of >90%� are identified.In a second step, terms are combined into complex, nested named entities to facilitate a syntax based rule assignment of specific roles the identified chemical named entities. Reaction specific syntax rules are implemented to understand and classify chemical reactions � from metabolic reactions up to the range of well-known named chemical reactions. As a result, the presented text mining system allows to scan millions of text documents in few days and extract chemical reaction information in CML or other chemistry aware data file formats for filling chemical reaction databases. As a case study we will present the automated generation of a metabolic database for phytochemicals.

Extraction of chemical reactions from full text documents: From n-tuples of value attribute pairs toward the automated construction of reaction databases

248th National Meeting (2014)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.