Factual Consumer Insights: what an accurate sentiment model means.

7th May 2020

When it comes to data, the average Marketing Manager is faced not with the issue of quantity, but clarity, and informed decisions permissible with current insights. A huge part of this is an accurate sentiment model. This allows a brand to have a true picture of public opinion, or brand reputation, and allows them to leverage customer experience.

NLP Sentiment Consumer Insights

Take the example above. Even the more intelligent models would rule the post as positive or neutral in sentiment i.e First clause +, second clause (giving credit to those able to take into account the negator ‘wasn’t’). 

So what if you’re a rice company and interested in gaining clear market insights? Highly sophisticated sentiment models, with focus in the food industry, would see the irrelevance of ‘the table’ within a rice conversation.

Also, key in this example would be taking into account the adverb ‘super’, which would modify the overall feeling toward the rice. This formula, applied to millions, upon millions, of data points, allows the data scientist to provide accurate information to creative teams.

Even emojis can be a headache. For example, they can vary due to macro trends that spread into the vernacular. This has been evident with the creation of relevant, COVID-19 related emojis. They also vary depending in dialect, sector, sub-sector, age, and more.

New COVID-19 Emojis

Similar intricacies, and nuances, must be reckoned with when looking at spam (excuse the pun), which we all know can clutter our social media threads. Whether they be low value product advertising or fake accounts/bots, this data can highly sway the overall picture.

Sword Stone Failing Sentiment Models

Within the Middle East, building suitable tools, which can glean insights from Arabic dialects has always been a challenge. Like the sword in the stone, many have tried, but most fail  (although, broader academic research continues).

That said, as part of Digital Ape’s proprietary tool -Sila– we approach it with relevant staff experience within the sector/sub-sector. Models are then programmed which form the true picture of the modern-day, linguistically diverse, consumers in the Middle East.

Overall, a functional sentiment model can only be built by highly specialized staff and tools constructed that match their skills. Specificity is key as brands should seek out their insights, and creative thereafter, from agencies with relevant experience.

More about Sila

DA has built a custom data warehouse service called Sila, powered by data science we turn that raw social data into actionable insights.

To find out more, please click here and use the contact form to get in touch.

This article first appeared on a LinkedIn post by Digital Ape’s James Revely.