Feasibility of Using the Position as Feature for Idea Identification from Text

Abstract

Abstracts of research papers are meant to provide a brief condensed overview of respective research topics. This includes a glimpse of the new idea that the paper proposes. The aim of the research presented here is to investigate the feasibility of the effect of text position in the idea identification. The abstracts are structured in the form of introduction, body, and conclusion. It is hypothesized that research ideas tend to be phrased in conclusion section of paper abstracts. 25 abstracts of the scientific papers were used to automatically identify the position of ideas within abstract sections. The results support the notion that the conclusion of the abstracts significantly represents the ideas.

Optimization of Idea Mining Model based on Text Position Weight

“…A modified idea mining model to capture the term position within text phrases is utilized [3], in which a term position measure is incorporated into the idea mining model, proposed in [29], to enhance the performance. This modified model assigned as a probability weight to be incorporated into the less considered sub-measure which enables calculating the importance of the position for the candidate ideas.…”

“…Authors in [3] performed an experimental investigation by implementing 10000 combinations and produce a new set of random weights based on rules assigned for discovering the best ranking. The findings of these experiments have shown that the highest MAP obtained when integrating position measure highly indicated ideas in conclusion sections more than other sections.…”

“…As an approximation of the entire content of the paper or pattern, the abstract provides a more manageable text for computation of idea mining model. A further investigation to the feasibility of using text position as an additional feature for idea identification is conducted through a preliminary study [3]. This investigation attempted to discover the position of the piece of text that have a higher likelihood to contain the idea.…”