How does Birdie deal with Ambiguity?
When dealing with user-generated textual data, this is a challenge that is tackled by pretty much everyone in the Natural Language Processing field: how to deal with ambiguity?
A word can have a multitude of meanings that depend entirely on its context and usage. For instance, “store”: could mean the act of storing something or it could be referring to an actual physical brick-and-mortar store.
They didn’t have that device in the store so I had to order it online.
They didn’t store the yogurts in the freezer properly.
Birdie is also affected by this: the opinions we collect from all over the web can have all sorts of flavors and possible contexts. To fix this, we’ve implemented an incremental feedback-based model that automatically sorts the right meaning for each keyword in our database!
Here’s the layman’s explanation for the model: based on your feedback, we create clusters of all the sentences that match what you want to see. If you click “snippet context is right” on a sentence in which “store” refers to physical locations, we bring all the sentences with that context together and deprioritize the other meanings. If you click “snippet context is wrong” on a sentence in which “store” refers to the act of storing something, we cull all sentences that have the same context.
We update this model every day with feedbacks and properly sort the opinion snippets depending on your needs. Our taxonomy squad can help you get a customized keyword taxonomy ready in no time! Send us a message and we can show you a live demo of our wide array of insight creation software.
