A Natural Language Processing AI/ML Platform for Businesses of All Sizes

Google Cloud’s advanced AI and ML will expand access to Cohere’s NLP platform for businesses of all sizes.

Just about every digital interaction creates data that is useful to companies hoping to glean additional information about their platforms and customer usage. Web pages, invoices, records and business documents are just the tip of the text-heavy data iceberg generated during online transactions. Companies can use natural language processing (NLP) technologies such as voice assistants, chatbots and advanced data analytics to gain insights from this data. However, access to NLP traditionally required an extensive financial investment from companies and a team of engineers and data scientists.

To make NLP technologies accessible to companies of all sizes, Cohere provides industry-leading NLP services that can understand, process and generate language that mimics normal human communication. Recently, Cohere announced that the AI infrastructure of Google Cloud would power its platform through a new multiyear partnership.

Cohere’s products will be developed and deployed on Google’s supercomputers—Cloud Tensor Processing Units (TPUs)—optimized for large-scale AI and machine learning (ML). Together, the partnership will enable more companies to access Cohere’s advanced large language models (LLMs) and NLP technologies.

“Leading companies around the world are using AI to fundamentally transform their business processes and deliver more helpful customer experiences,” said Thomas Kurian, CEO of Google Cloud. “Our work with Cohere will make it easier and more cost-effective for any organization to realize the possibilities of AI with powerful NLP services powered by Google’s custom-designed Tensor Processing Units (TPUs).”

Unlocking the Power of Big Data with NLP

NLP technology is at the intersection of linguistics and artificial intelligence, helping computers interpret and communicate with human languages. This involves computers processing and communicating in “natural language”—mediating exchanges using the nuances of the everyday language we use to communicate with one another. Most people have interacted with NLP technology in AI-driven chatbots and virtual assistants or even voice-operated GPS systems. NLP is used in any program that tries to communicate using the conversational language we would use to chat with a friend.

Natural language processing (NLP) takes text, or auditory inputs, of languages and converts them into data that computers can understand. (Image courtesy of Cohere.)

Natural language processing (NLP) takes text, or auditory inputs, of languages and converts them into data that computers can understand. (Image courtesy of Cohere.)

Because human language is naturally ambiguous and nuanced, it can be difficult for a computer to determine the intention and meaning behind voice or text data. For example, grammar, sarcasm and varying sentence structure may come easy to humans fluent in a particular language, but they can be difficult to interpret and mine for information in a digital space. The goal of NLP is to structure the highly unstructured voice and text data generated through human interactions with technology.

NLP encompasses many tasks that can break down complex inputs to correctly digest and interpret human text and voice data. Speech recognition is used to convert data from speech into text. This can be difficult due to the variation in human voices and grammatical accuracy—and the speech-to-text translation needs to consider everything from tone, to accents and intonation.

Once information is translated from speech to text, the data needs to undergo grammatical tagging to identify the different components of speech, such as verbs and nouns. Word sense disambiguation can then use semantic analysis to determine the meaning of a word in a given context. For example, the word bass can refer to a fish or a musical instrument depending on the context. Coreference resolution can also improve ambiguity in a text by linking a person or object with a specific pronoun (i.e., “she” and “Susan”). Finally, sentiment analysis can help NLP software tract subjective information for voice and text data to interpret emotions or sarcasm.

Another useful NLP application is named entity recognition (NER), which identifies words representing important pieces of data. For example, a company may want to identify a customer’s location (i.e., “Canada”) or their name (i.e., “Susan”). This is often crucial for advanced data analytics.

NLP provides an immensely powerful toolset for delivering comprehensive customer experience (CX) services on digital platforms. Advanced chatbots and virtual assistants are already mediating many CX interactions on websites, and NLP and AI help these systems better understand and interpret a customer’s needs.

Beyond CX applications, NLP is also increasingly used in business operations to help increase employee productivity and simplify business processes.

Cohere’s NLP Platform Helps Companies Make the Most of Their Data

Powered by Google Cloud’s technology, Cohere helps companies build products and services using its NLP models. Custom-built applications using the company’s models can then be used to understand and process language by classifying, summarizing and ultimately generating text.

With their single API offering, engineers can modify billions of parameters for applications in NLP. The company’s models can read numerous web pages to teach themselves to understand everything from individual words to the overall meaning of a given document. Moreover, the API can deliver NLP without expensive supercomputing infrastructure as it is now driven by Google Cloud’s AI and ML capabilities.

Cohere can make more intelligent platforms with a better understanding of human language using its large pretrained Transformer language models (large language models). Cohere trains massive language models and then makes them accessible through its API for easy addition to any third-party system. Likewise, models can be further customized by training with a company’s proprietary data.

Cohere breaks down its services into three main applications: composition, comprehension and comparison. The API can write copy, summarize text, complete prompts or fill in the blanks to compose human-like text. It can also compute the likelihood of text and group semantically similar sentences into a given topic. For example, it can determine the many iterations of the question “What are your store hours?” such as “When are you open?” and “What are your hours tomorrow?” Finally, the API can compare the similarity between different text passages and make predictions based on the likelihood of different text outcomes occurring. For example, the API can compare the likely end of a sentence. The platform can determine that the most likely ending to a sentence such as “in the future, computers will be able to better understand…” could be either “humans” or “language.”

NLP can determine if a question is similar to a FAQ with a known answer to streamline customer queries. (Image courtesy of Cohere.)

NLP can determine if a question is similar to a FAQ with a known answer to streamline customer queries. (Image courtesy of Cohere.)

Plus, Cohere’s platform can compare a customer question to common FAQs to simplify and streamline the process of answering customer queries. With many ways to ask the same question, NLP can help determine if a question is similar to a known answer.

“By partnering with Google Cloud and using Cloud TPUs, Google’s best-in-class machine learning infrastructure, we are bringing cutting-edge NLP technology to every developer and every organization through a powerful platform,” said Cohere’s cofounder and CEO, Aidan Gomez, who coauthored the breakthrough paper “Attention is All You Need,” which introduced the Transformer architecture that now defines NLP. “Until now, high-quality NLP models have been the sole domain of large companies. Through this partnership we’re giving developers access to one of the most important technologies to emerge from the modern AI revolution.”

NLP Architecture Continues Evolving to Meet the Modern Needs of Companies

Using NLP, a company can gain important insights into its platform usage and business operations. For example, NLP can help a company glean information about how customers use its website and identify common issues occurring during consumer interactions. A company can also better understand bottleneck stages in customer service, business operations, or other platform functions by analyzing its proprietary data. In the future, companies of all sizes will be able to access these powerful data analytics tools without the need for data science teams. As NLP architecture continues to evolve, our interactions with digital tools and platforms will be able to improve alongside the increased communication capabilities of computers.