Rugpong Grachangpun. Keyword recommendation for academic publications using flexible N-gram. Master's Degree(Management Information Systems). King Mongkut's University of Technology North Bangkok. Central Library. : King Mongkut's University of Technology North Bangkok, 2011.
Keyword recommendation for academic publications using flexible N-gram
Abstract:
Keywords in academic publications are important to explain the trend of a
paper. This thesis aims to develop and evaluate a method of keyword
recommendation for English academic publications using Flexible N-gram in the
domain of Information Retrieval. The generated keywords are flexible in terms of
length and number of words in a phrase. The advantage of a flexible length means the
increase chance of informative gain. Techniques from two major fields, Statistics and
Natural Language Processing domains, are incorporated into this thesis. The
techniques from Statistics such as TF-IDF, and Correlation Coefficient in order to
measure the degree of importance and relationship degree. The technique from
Natural Language Processing such as Part-of-Speech Tagging which is used to filter
the generated phrases derived from the statistical procedures to obtain the phrases that
comply to natural language structures. The algorithm is evaluated by comparing the
top 8, 10 15, 20 generated results with the original keywords that were assigned by
the author. In conclusion, best performance is found when the top 5 words are
presented to the user with 33.59%, 28.46%, 40.97% for F-measure, Precision and
Recall respectively.