An Information Extraction Method for Sedimentology Literature with Semantic Rules

Because of the high demand for text information, the semantic network is under tremendous pressure to improve semantic accuracy. The domain data set expands the types of entities in order to ensure that entity analysis is performed directly between contexts, thereby saving the time consumption of information search and improving the quality control of key data. However, literature data in the field of geological sedimentology relies mainly on manual annotation, which consumes considerable time. It is still a big problem in reducing human error and dynamically expanding entity classification. For solving these problems, this paper proposes a batch document information extraction method based on sentence part of speech rules (ESM). Technically speaking, nltk (Natural Language Toolkit) is used to identify specific sentence components by adding sedimentology specific semantic rules and Bi-LSTM + CRF (bidirectional Long Short-Term Memory network and conditional random field). Therefore, this paper makes an experimental evaluation to prove the efficiency of ESM.