Context

Below you will find a high-level explanation of the importance of context in text mining, and a brief discussion of how Lexalytics determines context. For a more in-depth explanation of our context determination functions, read through “Dealing with Context in Text Mining”. Check out our web demo to see Lexalytics in action, or get in touch to schedule a live demo with our team of data ninjas.

Determining Context

It’s all well and good to know who’s being discussed and in what tone, but entity extraction and sentiment analysis are just two pieces of the text analytics puzzle. In fact, entity sentiment taken alone can be dangerously misleading.

A mention of windows, for example, needs context: are consumers complaining about the broken glass in their newly-installed dining room windows, or are they commenting on bugs in a recent version of Microsoft Windows? They could even be talking about seeing someone’s soul through their eyes. Without understanding the context behind the word, you’re left grasping at straws.

Lexalytics offers two primary methods of context determination:

Themes

Themes are short phrases, usually an adjective-noun phrase. Themes are determined through a complex process called lexical chaining, which we won’t go into here: suffice it to say that the most contextually relevant noun phrases are sorted into entities and themes, the latter of which do a great job conveying the general subjects and context of a document of text.

Facets

Facets, introduced with the Salience 5 release, represent a totally new way of extracting meaning from text. Theme extraction relies on noun phrases to work, and not all text contains a noun phrase to work with. Think about this sentence: The bed was hard. There’s no qualifying theme here, but the sentence contains important information for a hospitality professional to be aware of. Facets are based on subject-verb-object phrases, and handle those tricky cases where even Lexalytics’ powerful theme processing isn’t suited for the job.

Entity extraction shows who is being discussed, and sentiment analysis reveals the tone of those discussions. With themes and facets, Lexalytics provides the context our users need to understand why consumers feel the way they do.