Article Preview
Top1. Introduction
Named entities are the atomic elements in text belonging to predefined categories such as, names of people, places, expression of time, quantities, money values, percentages, etc. Named Entity Recognition (NER) is the task of identifying such named entities. Named-entity recognition is often broken down into two distinct problems: (i) Identification of named entities (ii) Classification of the named entities to the type of entity they refer to (e.g. organization, person, money value and others). Temporal expressions and some numerical expressions (i.e., money, percentages, etc.) may also be considered as named entities in the context of the NER task. Therefore, the definition of the term “named entity” is not strict and often has to be explained in the context in which it is used (Li Zhang, Yue Pan, & Tong Zhang, 2004; Ritter, A., Clark, S., Mausam, & Etzioni, O., 2011; Daniel Sanchez-Cisneros & Fernando Aparicio Gali, 2013).
The categories chosen for a particular NER system may depend on the requirements of the system. For Eg: For a biomedical NER system, the names of proteins, genes, drugs and diseases etc. are identified. Similarly, if product classification is important, it may be necessary to identify the color, size, shape and brand names as named entities. Named Entity Recognition is one of the most important tasks which plays a vital role in several natural language processing areas including machine translation, question-answering, information extraction, information retrieval and recommender systems. There is a significant amount of work in the field of Named Entity Recognition (NER), where the goal is to locate and classify parts of free text into a set of predefined categories. Nowadays the number of text documents are vast and they contain many redundant information. Balaji et al., (Balaji, Geetha, & Ranjani Parthasarathi, 2014) focused on removing such redundant information and they identified the essential components from various documents.
Chemical entities are everywhere throughout the chemical literature and biomedical literature. The development for text-mining systems that can efficiently identify chemical entities are required. Due to the unavailability of sufficient corpora and data resources, the researchers have focused on the development for gene and protein Named Entity Recognition systems. Finding the mentions of chemical compounds in text is of interest for several reasons. An annotation of the entities enables a search engine to return documents containing elements of this entity class, to find relations, to map the entities to corresponding structures or substructures which enables a chemist to search for similar structures or substructures and combine the knowledge in the text with databases or to integrate other tools handling chemical information. But due to the various nomenclatures, this task is highly complex and should be supported by computational tools. Recently, a framework is designed to analyze different user profiles and interests while query processing including relevance analysis (Ramesh, Ganapathy, Bhuvaneshwari et al., 2015).
This paper presents the development of a chemical Named Entity Recognition system based on Pattern Matching approach. This approach extracts the chemical terms from unannotated texts using regular expressions developed from the orthographic and morphological feature sets. Tf-Idf (Term frequency- Inverse document frequency) and PMI (Pointwise Mutual Information) association measures have been used to filter out the non-chemical entities from the extracted entities. The main aim of this paper is to investigate the effects of the various patterns for the Named Entity Recognition task. These patterns are represented using prefix, suffix, special symbols, digits etc. The performance of the proposed system is studied with different pattern combinations in chemical documents. The process flow is described in Figure 1.
Figure 1. Flow diagram for pattern matching chemical NER and functional group classification