April 25, 2022
Say Hello to NLU Freedom
OneReach.ai takes a different approach to NLU - learn why we have an award winning philosophy.
The award-winning OneReach.ai NLU philosophy is radically different—a future-proof strategy that puts the power of the entire marketplace at your fingertips
In the fast-paced conversational AI marketplace, it seems like there’s a new NLU/NLP leader crowned every week. General language models like GPT-3 outpace their predecessors and cause massive disruptions only to be outpaced by new technologies that disrupt in their own ways—like Macaw, a model demonstrated by the nonprofit Allen Institute for AI that emerged from their research into creating AI that can perform at human levels on standardized tests. Macaw is as good or better than GPT-3 at answering questions but is one-tenth the size (https://techcrunch.com/). Seismic shifts like these come in unexpected shapes at unexpected times, and locking into any one engine amounts to NLU purgatory—where your NLU capabilities are limited by a single vendor’s roadmap.
The OneReach.ai philosophy around NLU technology mirrors our broader philosophy about conversational AI: success is all about speed and flexibility—you need to be able to leverage any solution in the marketplace at a moment’s notice. With that in mind, we designed a platform that lets customers reap the rewards of optionality, plugging in new engines as soon as they become available. Our approach is unique, but it’s catching on—CogX recently named OneReach.ai Best Innovation in NLP in their 2022 awards honoring innovators, visionaries, and change-makers who are “helping us get the next ten years right.”
All of this is powered by OneReach.ai’s proprietary NLU Engine, which amalgamates multiple systems and selects the best performing one—as determined at runtime to satisfy specific requests.
You don’t need to choose between Microsoft, AWS, Google, IBM, or any other NLU vendor to serve as your permanent choice. Use them all, and shift to the best one on a case-by-case basis. Using multiple engines in the same conversation increases success rates dramatically. When choosing OneReach.ai, you have the ability to leverage the entire market for NLU capabilities—future-proofing your organization against a rapidly evolving NLU marketplace, and enabling use of simultaneous engines within the same conversation.
How it works
OneReach.ai allows you to amalgamate multiple systems from multiple providers in real-time to deliver better results than any single solution can provide by itself. Using a GANs network, we hone our models based on the best performing model out of many engines. We might train the same utterances with Google, AWS, Azure, IBM, and a solution-specific vendor like Deepgram. Then we rank the results from each vendor on an utterance-by-utterance basis and create a proprietary model using the best results from each.
This novel method of model efficiency improvement ensures that, at any given moment, OneReach.ai customers always have the best possible models the market can provide. Even better, when these external players improve their NLU capabilities, our customer’s models automatically improve.
Other providers typically require customers to create an account within the third-party service, log in, and build their NLU models. Then, those vendors simply call that service via API at runtime. We support these APIs, but we also offer a consolidated native training interface within OneReach.ai that automatically creates a proprietary custom model based on any combination of third-party vendors you choose.
Figure 1: Choose NLU engines to create custom models.
Figure 2: Confidence scores produced from multiple NLU engines by intent.
The rewards of speed and flexibility
OneReach.ai offers customers the ability to utilize infinite NLU providers in the same conversation, in any sequence. This is especially useful when one vendor has entity models that another doesn’t. For example, you can use LUIS & Rasa for intent recognition and Dialogflow for entity recognition. The ability to use any NLU engine at every step in a conversation gives you extreme flexibility, mixing and matching vendors to utilize the entire market’s capabilities.
This granular control over which engines you use extends down to each individual response. If you find that an engine outperforms others on geography, use it in contexts specific to location. Customers have the ability to use common engines such as: Microsoft LUIS, IBM Watson NLU, Amazon Lex, Google Dialogflow, RASA, Microsoft QnA Maker, Salesforce, and SpaCy. OneReach.ai’s open architecture also enables seamless integration with any custom NLU engine, and customers can connect to these systems through our native amalgamation engine or through direct connections. In sum, the speed and flexibility built into our platform allows our customers to avoid any trace of vendor lock-in. This allows for faster creation, testing, iteration, and evolution of your automated solutions.
Total entity control
Customers can use OneReach.ai’s proprietary, prebuilt entity library or build their own custom entities on top of OneReach.ai’s NLU engine. Our native engine is designed to support the following capabilities:
- Machine Learning Entities: Custom entities built with unstructured conversational data.
- List Entities: Custom entities built from a simple list of terms and bolstered by a custom feature model.
- Regex Entities: Custom entities built from regex (or regular expressions). We have a library of prebuilt regexp patterns for common use cases like passwords.
- Pattern Entities: Custom entities built from common language patterns.
- Prebuilt Entities: Utilize our native entities or external prebuilt entities from any vendor (Lex, LUIS, Dialogflow, etc). You can use any combination of vendor entities in a single conversation, affording extreme flexibility and avoiding vendor lock-in.
- Hierarchical Relationships: We can create sophisticated role-based entities with our annotation GUI—and can even create hierarchical entity structures with parent/child relationships that can extend to infinite levels of depth.
- Flow Entities: Custom entities built with OneReach’s flow builder can take into account formal logic and context, and can even combine multiple prebuilt, machine learned, regex, list, and pattern entities. These flow-built entities have unlimited potential for customization as they have all the functionality of OneReach.ai’s 600+ function-deep library and are accessible for granular access at the source code.
- Entity Marketplace: As additional entity models are created, we include them in our marketplace, so clients can easily leverage them for their solutions, further future proofing your NLU capabilities.
Figure 3: Utilize prebuilt entities or create custom models.
Put our philosophy to work for you
We have a growing list of customers who have seen how our open and flexible NLU strategy can be a massive force accelerator as organizations work to leverage conversational AI. Whether your organization is already on that list, or if you’re ready to take your automation efforts to radically new heights, OneReach.ai can embolden your efforts and future-proof your ecosystem.
Stay up to date
December 18, 2023