Attensity Understands Emoticons

March 13, 2015

Part of big data is being able to make sense of unstructured data, including the pieces that natural language processing software cannot understand like emoticons. Emoticons are an Internet phenomenal where people use grammatical symbols, numbers, and letters to represent feelings and ideas.

They cannot be spoken, so if an organization wants to analysis all of its data they need to be able to interpret emoticons. PC World tells us that Attensity has already created a way to understand emoticons without turning to a teenage girl to translate: “For Attensity’s BI Parsing Tool, Emoticons Are No Problem.”

Attensity’s Semantic Annotation natural-language processing tool was designed to handle large data loads. It can monitor and extract insights from unstructured data, including data from social media platforms and internal information like customer surveys and calls.

“Rather than relying on traditional keyword-based approaches to assessing sentiment and deriving meaning, Attensity’s Java-based product takes a more flexible natural-language approach. By combining and analyzing the linguistic structure of words and the relationship between a sentence’s subject, action and object, it’s designed to decipher and surface the sentiment and themes underlying many kinds of common language—even when there are variations in grammatical or linguistic expression, emoticons, synonyms and polysemies.”

This means Attensity can generate data straight from sentences rendered entirely in emoticons and acronyms.

Another practical use for Attensity’s Semantic Annotation would be to create a translation app for parents trying to decipher their teenager’s text messages.

Whitney Grace, March 13, 2015
Sponsored by ArnoldIT.com, developer of Augmentext

Comments

One Response to “Attensity Understands Emoticons”

  1. Foundation for Defense of Democracies on March 13th, 2015 12:30 pm

    Foundation for Defense of Democracies

    Attensity Understands Emoticons : Stephen E. Arnold @ Beyond Search

  • Archives

  • Recent Posts

  • Meta