Intention Extraction

Before we do anything, we need to pin down exactly what it means to extract intention from unstructured text. This is will be a high-level overview of intentions. I'll use Lexalytics, the company I work for, to illustrate intention extraction. For a more in-depth explanation of our method of extracting intention, read through “Intentions: What Will They Do?." It drills down a few levels deeper and makes for an interesting read. Also, our web demo functions as another educational touchstone.

Intentions are an indication of future behavior of a speaker. We have 4 intentions that come out of the box: buy, sell, quit, and recommend. Unlike other companies that provide intentions, we actually provide the tuple of intender, intention, and object of the intention.

Intentions

Let's dive right in. Historically, text mining has been a very retroactive business: analyzing text to see what’s already been expressed in the past. Intention analysis technology sends this paradigm on a 180-degree course change; it's effectively predicting the future.

Well, okay, not quite predicting. But with intention extraction, we determine the expressed intent of customers and reviewers. Will a given person buy, quit, sell, or recommend your product? Let's look at an example:

I don’t like it. It keeps crashing on me (I think it’s a hardware issue). I’ll be returning it tomorrow.

 Your run-of-the-mill text analytics system will extract basic information, such as sentiment (negative) and the object (“laptop”), and classify the content into a category like “errors”.

However,  intention extraction takes the analysis one step further: it unlocks the capability of revealing that the customer bought the laptop, but now intends to return or “quit” it.

Uh-oh, you just discovered that due to a crashing issue, you’re about to lose a customer. Better start resolving the problem! Better yet, dig into the root of the problem to prevent future issues cropping up.

With intentions extraction, you’ll:

• Improve lead flow
• Decrease customer churn
• Understand reviews better
• Provide highly-targeted advertising to customers who are expressing an immediate need for your product
• And much, much more.

Just like other text analysis features, Intentions are fully customizable at Lexalytics: we allow our users to configure existing intentions and define their own intention-laced keywords, specific to their market and business. Just tell us what verbs indicate intent for you (“desire," “want," “wish," etc.) and we’ll pull out the intender and the intended for these new intent types.

Intention extraction is just another reason text mining is moving from a luxury to a necessity. Modern analytics engines like Lexalytics' offer a rich mix of configurability and filtering provides high-quality results with all the information that’s necessary to determine exactly what a customer wants and what they intend to do. Without listening to these voices your company will not be able to adapt. And if a company cannot adapt, it will die.