Abstract:
|
The availability of publicly accessible product review websites is producing a large and rapidly growing volume of customer feedback data that can be used to improve customer experience and marketing performance. Using this data is not without challenges. It is complex and poorly structured. Much of the potentially important information is in text generated by posters. It is in this language data that the details of customers' opinions about the "aspects," or characteristics, of their experiences are expressed. We are conducting a research project involving the development and testing of methods for extracting aspect-specific sentiments. We are using over 1.6 million reviews of 12,000 hotel properties. We summarize our application of statistical natural language processing and sentiment lexicon methods, with particular focus on conditional random field models for part of speech tagging and named entity recognition.
|