Published October 31, 2022
Large language models - Another consumer intelligence first
How large language models drive actionable insights in real time
When a leading fashion company recently underwent a major brand makeover, it launched several international campaigns. To measure the impact of these, 1,000s of surveys were collected from customers across the globe.
However, the brand didn’t consider how these surveys were to be analyzed on such a large scale.
The head of insights initially struggled to analyze these surveys in a timely manner. Even with a team of seven analysts, it took three weeks to provide management with an insightful status update on the performance of the campaign. This didn’t provide the team adequate time to adapt its marketing.
The team estimated that the three weeks cost the company around $60K in resources.
When the brand worked with Talkwalker…
Running the same analytics on the Talkwalker platform, the brand can now quickly understand the customer response in real time and at no additional cost.
This helped it identify where the campaign resonated, and where it fell flat. The brand could quickly adapt their brand messaging to tackle the negative points the surveys raised, and redistribute marketing budgets to boost efforts in regions that were less engaged.
The head of social at a leading car manufacturer needed to categorize millions of social mentions relating to the launch of its new model, so that the team could understand how consumers were rating its latest models by design, price, comfort, environmental impact, performance, etc.
To do this, the team segmented the data in several ways, using complicated boolean rules and custom AI models available on generic social listening platforms. It took the team several days to come up with effective and accurate segmentation that would provide accurate data.
When the brand worked with Talkwalker…
Running the same segmentation on Talkwalker, the brand was able to segment data, providing short descriptions of the parameters it required.
A few days into the campaign launch, a crisis emerged when a large activist group targeted the brand with a disinformation campaign. The structured segmentation immediately identified this as a threat to the brand’s reputation.
As part of the brand’s crisis management plan, it could create an additional category around these mentions, monitor the risk effectively, and counteract it with targeted PR.
Post-launch, it added further segmentation to the project, to monitor the buyer journey. Tracking customer behavior across different parts of the journey (awareness, intent to buy, purchase, etc.) the brand could better understand how and when to engage consumers to boost sales.
Talkwalker - first to bring the power of LLMs
Now’s the time to bring the power of large language models to your consumer intelligence.
Talkwalker's Blue Silk™ is the AI secret sauce behind the magic of Talkwalker. The integration of large language models presents a significant competitive edge that is difficult to replicate.
We have been first-to-market with many technological advances that create value for our customers and this is the latest step in the enhancement of the already-significant capabilities of Talkwalker's Blue Silk™.
A history of technological innovation and thought leadership
Social media runs at a near-impossible pace, and to keep up, brands need a future-ready platform. That’s why Talkwalker is constantly investing in new and innovative AI to disrupt the industry and provide clients with faster and easier ways to identify insights, as part of our Blue Silk™ AI.
All these innovations are based on proprietary technologies - Talkwalker’s awesome team of data scientists created the technology, and have the expertise to constantly develop it.
Applying large language models is another industry first in a long line of firsts.
When Talkwalker introduced image recognition in 2016, it was seen as a breakthrough, enabling brands to find their logo in images across social media. Now, image recognition is seen as an industry standard, critical for crisis management and brand monitoring. But, that’s only because Talkwalker did it first.
And now, it’s the turn of large language models - the future that consumer intelligence must have, ready for today.
What is a large language model?
Large language models (LLMs) are language models that are significantly larger than usual.
Maybe not. Let's go a little bit deeper…
Language models are neural networks that have been trained on a large body of language data in order to predict the next word in a sentence. In this way the model can learn the relationships between words and phrases in a language and generate new text, translate between languages, and perform other natural language processing tasks, such as question answering, classification, or summarization.
Simply put, the AI predicts what you want, based on a significant amount of language data. You may have come across a language model in the form of autocorrect or text generation in Google docs.
The shortcomings of first-generation language models was that each model had to be trained for specific tasks. Which meant significant hours of manual effort in training the custom AI models to perform the task at hand.
Enter large language models, the natural next step. Large language models are AI models with significantly larger neural networks, pre-trained on vast sets of language data - 1,000x larger neural networks. This means they’re capable of performing a wide variety of tasks with little or no need for further training. These artificial neural networks are designed to mimic the complexity of the human brain, therefore, the larger the network the greater the brain power.
Depending on the size and significant training of the networks, you can get few-shot and zero-shot learning.
- Those that require a little training are called few-shot learning, meaning the AI needs a few shots to get it right.
- While those that require no training are called zero-shot learning, as the AI gets it right off the bat. Talkwalker Blue Silk™ AI applies large language models with zero-shot learning.
What can Talkwalker’s large language models do?
Initially, we’ve launched not one, but two new applications for the Blue Silk™ large language models, Blue Silk™ Insight and 1-Click AI Classifier.
Blue Silk™ Insight
Reviews, customer satisfaction surveys, and customer support emails offer a detailed look at what customers think of your products or brand. They can quickly highlight product defects, customer service issues, or delivery problems, and be used for product innovation. If 1,000 customers tell you that they wished your new smartphone had more memory, guess what you should include in your next release.
Yet these types of data can also be very detailed. And numerous. If your company generates a lot of reviews, CSAT surveys, and customer support emails every day, it’s time-consuming to keep on top of them.
Blue Silk™ Insight uses the power of LLMs to do the hard work of reading through 1,000s of CSAT surveys, reviews, and customer support emails, and gives you the essence of all that feedback from a nuanced understanding of the human language. All the vital information you gathered, displayed in a handful of sentences. Giving you actionable insights in real time, and enabling you to take action faster than ever.
All of this is done by Talkwalker’s Blue Silk™ AI, instantaneously, without the need for customized training. The massive amount of data contained in all the consumer feedback doesn’t need to be manually labeled or annotated.
Blue Silk™ Insight - All your reviews are simplified into easy-to-understand insights.
1-Click AI Classifier
When handling large amounts of social data, it helps to segment and categorize your data to make it more manageable. This could be to split your data by products, or it could be used to help remove false positives if your brand name is also a generic noun (Apple, Orange, Visa, Amazon, etc.).
In the past, this could be done manually, or by creating complicated boolean rules or AI classifiers. But, if you have millions of mentions, that’s no longer possible. Using Blue Silk™ LLMs, 1-Click AI Classifier will automatically segment and categorize your data in a way that is both scalable and with low error. Again, with zero training needed.
This marks a new era of using AI in consumer intelligence.
Improved time-to-insight on social data classification.
How do large language models work for marketers?
So what does all this mean in practical terms?
The introduction of LLMs to our Blue Silk™ AI will enable marketers to turn large amounts of data into fast, easy-to-action insights. Instantly.
Benefits of Blue Silk™ Insight
When it comes to Blue Silk™ Insight, it’s about translating all your reviews, customer support emails, and CSAT surveys into actionable insights in real time. You can:
- Turn hours of work into seconds, enabling insights professionals to focus their time on more comprehensive research.
- Validate business hypotheses in real time with automatic AI-generated insights.
- Create near-instant engagement strategies in response to your data, to enhance the customer experience across your enterprise.
- Run data on-demand. Expecting an influx of data from a new product launch? It can be incorporated into your analysis in real time.
- Run Blue Silk™ Insight on different subsets of the data. For example, thousands of customer reviews on a smartphone brand can be split by various model types.
Blue Silk Insight™ simplifies both review and survey data.
Benefits of 1-Click AI Classifier
1-Click AI Classifier is the new essential for handling big, big data. With it, marketers can:
- Quickly and accurately categorize millions of relevant social conversations, blogs, and articles without the need to spend hours setting up complicated rules.
- Instantly filter for relevant online customer data.
What are the alternatives to large language models?
Now, there isn’t much alternative to adding AI analysis to your consumer intelligence. Brand mentions can run into the millions, making manual analysis impractical.
AI saves time and effort and enables teams to focus their efforts on boosting customer satisfaction, market share, and revenue through data-driven insights.
Large language models take AI analysis further.
As the only consumer intelligence platform AI-powered by large language models, Talkwalker enables our customers to work faster, with even more accurate data.
Large language models are the future of state-of-the-art consumer intelligence. If you want the fastest way to actionable insights in real time, you need Talkwalker’s Blue Silk™ powered by large language models.
This is just the beginning at Talkwalker. As ever, we continue to invest in our Blue Silk™ AI, to provide the ultimate consumer intelligence platform, and give users more ingenious, time-saving artificial intelligence support.
Expect to see innovations in the near future.
In the meantime, you can experience how Blue Silk™ will help your brand, with a free demo. Request yours by clicking below.
Subscribe to our newsletter for social resources
Join 70,000+ professionals and become a better social media marketer. Get social media resources and tips in your inbox weekly.