Spanish Sentiment and Toxic Speech Models boosts near-human-benchmark accuracy rates representing significant advances in continuous listening capabilities.
Aware, the AI Data Platform delivering contextual intelligence to power Experience Management, today announced the release of new Spanish sentiment and toxic speech models to its ever-expanding ecosystem of natural language processing (NLP) and computer vision (CV) models. With Aware’s cutting-edge capabilities to create and deploy accurate models quickly and cost-effectively, multilingual organizations now gain continuous listening capabilities needed to deliver a better employee experience.
Aware CEO, Jeff Schumann, underscores the importance of understanding the authentic voice of employees for Experience Management. “Aware’s capability for rapid and cost-effective model deployment empowers multilingual organizations to adopt continuous listening strategies today. In turn this approach, coupled with Aware’s comprehensive suite of applications, extracts more value from collaboration data, ensures inclusivity, and boosts engagement and productivity.”
Today, many executives rely on infrequent surveys and unreliable anecdotes to understand the needs of their employees, customers, and business. Such methods can result in delayed, incomplete, and ineffective data, with language, cultural and location barriers often leading to inequitable employee experiences. Even survey giant Qualtrics acknowledges the emerging need for continuous listening. “The world is changing much quicker than once a year conversation where you get the results two, three months later,” Qualtrics’ chief customer officer Donnchadh Casey said. These rapid changes are more apparent than ever in today’s global workforce, where teams in Barcelona may collaborate with those in New York, or on the frontline in California, where 55% of the workforce now identifies as Hispanic.
Aware is uniquely equipped to quickly respond to this market’s emerging needs, thanks to over six years of the targeted development of its Aware IQ platform. This purpose-built AI Data Platform provides real-time, contextualized intelligence into the voice of the employee by collecting data, structuring, and enriching digital workplace conversations at scale. Leveraging twelve different customizable NLP and computer vision models, the Aware IQ platform analyzes workplace conversations across Slack, Microsoft Teams, WebEx by Cisco, Zoom, WorkJam and dozens of additional applications.
Many companies who seek to adopt AI in continuous listening strategies default to publicly available large language models (LLMs). These LLMs provide broad set of capabilities but lack specificity and carry a risk of unknown biases particularly when being trained predominately on English language content scraped from the Internet. Aware, on the other hand, understands that cultural context and conversational nuance are crucial to responsible AI and truly representative models. Aware’s models are trained on a proprietary, ever-growing dataset of tens of millions of digital workplace conversations, including those conducted in other languages. This targeted approach, combined with Aware’s ability to quickly and cost effectively develop and deploy new models, enables companies everywhere to take advantage of continuous listening today.
“Maintaining precise AI models is a complex task. Each model is meticulously fine-tuned to grasp linguistic and cultural nuances, while avoiding bias—a challenge due to the evolving nature of languages and cultures, said Jason Morgan, VP of Data Science at Aware. “Using techniques like Reinforcement Learning from Human Feedback (RLHF), which use human guidance to train AI, is crucial to keep the models effective as languages and cultures change. The process demands substantial computational resources, expertise, and time, to deliver accurate rates and the contextual intelligence to make confident decisions. And the results speak for themselves with near-human-benchmark accuracy for our Spanish models.”
These actionable insights combine with the power of Aware’s applications to deliver organizations more utility from collaboration data, and ensure every voice is heard—leading to increased engagement, improved productivity, and a more inclusive workplace environment:
- Surface Sentiment. The new Spanish Sentiment model boasts an impressive 86% accuracy—surpassing market leaders by over 30% and enabling a nuanced understanding of messages, whether positive, negative, or neutral.
- Measure Conversation Health. The Spanish toxic speech model offers data on the health of conversations, with accuracy near the human-benchmark, helping organizations identify and address potentially harmful communication patterns.
- Extract Actionable Themes. Uncover themes from Spanish messages within Trending and Emerging Theme dashboards and stay informed about hot topics across the organization.
- Enhanced Collaboration Search. Employ Aware’s fast, federated search functionality to identify instances of toxic speech originating in Spanish within collaboration tools.