Text Mining Techniques for Analyzing Unstructured Manufacturing Data
Yazdizadeh Shotorbani, Peyman
Manufacturing companies are increasingly enhancing their web presence as a strategy for improving their visibility in the global market. The exponential growth of manufacturing websites has resulted in a drastic increase in the size and variety of unstructured manufacturing information available online. This poses both challenges and opportunities. The challenge is related to efficient information search and retrieval when dealing with a large volume of heterogeneous information. Traditional search methods, such as keyword search, can no longer meet the information retrieval and organization needs of the manufacturing cyberspace. At the same time, the textual data available online contains a wealth of technical knowledge that, if mined properly, may result in discovery of new patterns and trends that were otherwise unknown. There is an acute need for more advanced computational tools and techniques that can help search, organize, and summarize large achieves of text pertaining to technological capabilities of manufacturing suppliers. In this research, three text mining techniques, namely, Classification, Clustering, and Topic Modeling are applied to analyzing manufacturing data. R programming package is used for implementation of the aforementioned techniques. The novelty of the proposed classification technique is in adopting concept- based method rather than term-based method that results in higher semantic relevance of the results. Also, clustering and topic modeling are serialized to improve the likelihood of discovering useful knowledge patterns.
Text mining, Topic modeling, Classification, Clustering, Manufacturing, Machine learning, Supplier discovery
Yazdizadeh S. P. (2016). <i>Text mining techniques for analyzing unstructured manufacturing data</i> (Unpublished thesis). Texas State University, San Marcos, Texas.