Sign up for our newsletter and get the latest big data news and analysis.

Interview: Unlocking Audience Understanding with NLP AI

I recently caught up with Andrea Vattani, Co-Founder & Chief Scientist, Spiketrap, to explore the path forward for brands looking to unlock true understanding of their audiences and their communications in the context of the world. Only two months into 2021 and nearing our second year of the pandemic, consumer behavior has changed rapidly and will continue to evolve as we adapt to new ways of life. So, how can brands keep a pulse on their shifting audiences this year and beyond?

insideBIGDATA: How has consumer behavior changed over the course of the pandemic — and how can companies keep a pulse on their shifting audiences in 2021? 

Andrea Vattani: As people have physically distanced, they’ve become more digitally connected. Conversations that once took place in-person or on a limited number of channels now sprawl online with increasing velocity. Seemingly small stories can trend quickly, and audience reactions on one platform can reverberate across others, leaving teams struggling to keep up with ever-shifting audience sentiment.

To that end, companies must also revisit how they tune in. Reliance on 1:1 insight for every consumer is neither realistic with privacy considerations nor scalable in a way that can provide meaningful insight. Instead, it is more important than ever to seek extrapolated insights by leaning into machine learning.

insideBIGDATA: How does natural language processing (NLP) AI work and help companies unlock true audience understanding?  

Andrea Vattani: Without NLP AI, organizations must sift through content manually. While boolean searches can help teams attain some semblance of the scale of a conversation, they are riddled by internal biases, fatigue and a never-ending cycle of balancing false positives and false negatives. Indeed, capturing the entirety of a conversation — and just that conversation — is practically impossible with keyword-dependent solutions alone.

NLP AI can help organizations extract the signal from the noise — identifying and attributing seemingly disparate pieces of online discourse into a coherent conversation or narrative. By providing a contextualized structure for otherwise unstructured data at scale, NLP AI provides a mechanism for making sense of high velocity audience conversation in real time. When these conversational threads are then processed through the lens of sentiment or safety, organizations can begin to understand the tenor and health of any given topic that their audience is talking about.

insideBIGDATA: What do companies looking to utilize NLP AI need to know about the technology to successfully understand conversation across digital platforms?

Andrea Vattani: Understanding how your NLP AI works is essential. Identity attribution, conversation identification, contextual sentiment analysis and the sophistication of brand safety monitoring are all factors that can vary wildly from one solution to another.

Conversations exist outside of keywords, and how people talk is constantly evolving. Pay close attention to how a given solution attributes content to entities or conversations. Is it able to disambiguate without constant human intervention? For IP holders with complex franchises or brands with somewhat generic or ambiguous terminology, this is a highly important consideration.

For instance, it is not enough for an AI to associate Peter Parker with the Spider-Man franchise. Rather, it must be able to ascertain whether the text thread is in reference to a particular movie, television series, video game, comic or toy — whether or not the text thread explicitly provides this information — and then attribute this particular mention to a related narrative about that entity.

Beyond identity recognition, can a given NLP AI identify common threads across different platforms and data sets? Does it successfully delineate between the sentiment of the original content and that of its branching reaction threads? Can it identify early when a conversation might be turning toxic?

insideBIGDATA: You say 2021 will be the year of AI-powered understanding at scale. Why is this? 

Andrea Vattani: Over the past year, considerable investments into text generation such as GPT-3 have proven that it is possible to mimic how humans write. Of course, this addresses only one aspect of communication — understanding is the next step. Already, numerous groups like Spiketrap are working to find meaning within the mayhem of ever-evolving digital discourse, unlocking deeper language and conversation comprehension.

As we move forward, NLP-based AI will focus around true understanding of communication in the context of the world. Expect NLP AI to evolve from merely processing for the sake of mimicking to processing for the sake of true understanding.

insideBIGDATA: What does the future of audience understanding bolstered by NLP AI look like — and what do companies need to know to be ahead of this trend? 

Andrea Vattani: As more organizations turn to NLP AI to better understand their audiences, they will also be freeing their teams to ask more meaningful questions and iterate more rapidly on solutions. It allows teams to move beyond the question of what people are saying to why. By facilitating real-time accessibility to answers for deeper questions, NLP AI will help organizations make better and more informed decisions.

Getting ahead in this regard simply requires digging deeper. Challenge existing listening and research methodologies and instead push for solutions that can provide meaning — not just metrics.

About the Interviewee

Andrea Vattani is Co-Founder and Chief Scientist at Spiketrap, where he leads the science team, overseeing and guiding the advancement of the conversation analytics platform’s Clair AI. Prior to Spiketrap, Andrea was a Senior Lead Engineer at Amazon Goodreads, where he led the effort around multiple machine learning applications. He received his PhD in Computer Science from the University of California, San Diego, where his research spanned the domains of machine learning, big-data analysis, social networks and game theory. Andrea is also a graduate of StartX, Stanford’s startup accelerator for top entrepreneurs, as well as a member of various scientific program committees, including NeurIPS, ICML, AAAI and ICLR. When Andrea isn’t tinkering with data, he can be found playing a competitive game of tennis, leisurely biking along the beautiful Bay Area or making delicious pizza with his wife.

Sign up for the free insideBIGDATA newsletter.

Join us on Twitter: @InsideBigData1 – https://twitter.com/InsideBigData1

Comments

  1. Thank you so much for sharing this important post. I learned a lot from the post about understanding listeners unlocked with this NLP AI

Leave a Comment

*

Resource Links: