Introduction to Text Mining Part 2

A continuation of the last article introducing text mining. In the previous article, we discussed the first two benefits of text mining which included: Finding missing information by looking at word patterns rather than meaning. Quicker analysis than other qualitative research techniques.

Both of these are common reasons that people enjoy using text mining techniques. Another benefit includes: Quantifying Focus Areas Finally, text mining gives you something numerical to look at, so that you can analyze the importance of various words accordingly. It doesn't necessarily tell you whether the view of those words is primarily positive or negative, meaning that if 50% of respondents praise your company bathrooms, and 50% disapprove of them, you are not going to know that it is split and simply see one large number.

But it would indicate numerically that bathrooms are (or are not) important to your customer.

That said, text mining is not without its weaknesses. Its primary weakness is being unable to tell you the meaning or degree of importance assigned to the word. In the bathroom example above, it would be difficult to know whether feelings about the bathrooms are good or bad, and simply seeing the word show up could lead to erroneous conclusions.

Text mining can also be affected by problems or questions in your survey. For example, if you have most of your questions centered around one product, you may receive more responses related to that product when another product is more valuable to the customer. Also, survey confusion can lead to recurring words that are indicative of how the person feels about your survey, not how they feel about your company.

If they don't understand a question, for example, a word may recur that isn't necessarily related to what you hoped to find. Finally, these days very few people seem to be able to spell correctly when it comes to typing online, and while there are a number of text mining programs that account for this, it's still a problem that could affect your data collection.

Overall, text mining is probably not a standalone analysis tool. It doesn't really get to the meat of the data, and focuses instead on quantifying something that human beings may struggle to quantify. But it is no doubt useful, especially as a form of secondary analysis, and may be considered if you have a survey that is heavy in qualitative questions.