Pryon’s AI engine automatically performs advanced natural language processing, ushering in a new era of innovation for the enterprise.
Pryon’s Central Portal is where you create and maintain the collections of data needed for your applications.
Pryon’s connectors, integrations, and modern architecture make deployment as fast and painless as possible.
Pryon’s AI Engine automatically performs all of the steps that typically require teams of data scientists and months of time. It understands data that is not specially prepped, doesn’t rely on the construction of taxonomies, ontologies or other classification schemas (creates them on the fly as a matter of fact) and is accurate out-of-the-box without training. It creates a model of your content, processing it for immediate use in applications.
Natural Language Processing & Understanding
Natural Language Processing is Artificial Intelligence that enables computers to analyze and understand human language. Pryon’s AI is trained to perform read, organize and search in ways that previously required humans. This powerful capability is used in every interaction, both to understand a request and to retrieve the accurate response. The success of any NLP project is directly correlated to the sophistication of the underlying natural language technologies used.
Automatic Model Generation
To make your content ready for use in chatbots, search, automations, etc. – it must be broken into specific pieces so a user can get the exact answer, result or snippet needed. This can be done manually as when a specialist breaks information into intents and entities. Pryon creates a dynamic model of your content for automatically identifying and attaching rich metadata to each piece of information. When you need to add, change or remove content this model is regenerated with a click.
Automatic Tuning Through Unsupervised Learning & No Code Tuning
Pryon offers superior accuracy out-of-the-box. Once Pryon creates a model of your content it also automatically tunes that model for your content by generating synthetic data and applying unsupervised learning. You can also easily tune your model by providing any lists of terms, acronyms, product parts, etc. that are specific to your company or industry. Just create a document defining your terms and drop it in your collection.
Automatic Ingestion of All Common Business Documents And Web Pages
Enterprise data is very challenging for basic transformer models to ingest. Web pages, annual reports, marketing presentations, legal documents and contact center knowledge base articles all have graphic elements and other clues about meaning and significance that need to be recognized and normalized to create an accurate model. Pryon’s inventions make it possible to fuse messy enterprise data into one model. Supported file types include Microsoft Word, Microsoft PowerPoint, Google Docs, Google Slides, text files, PDFs, URLs, and secure web pages.
Our portal is where you create and maintain the collections of data needed for your applications. It’s highly intuitive and requires no special skills. You can create a collection in minutes following simple steps. You do not need to migrate anything.
Once you push “Ingest” for a collection, Pryon generates a model of your content. You can test the model immediately to see how well it responds to text or voice requests in natural language.
Want to see Pryon in action? This video shows how easy it is to add a collection and how quickly a content model is generated.
When Pryon provides information in response to a query, it presents signals that help end users understand where information is coming from and how much they can trust it. For example, if a subject matter expert has explicitly submitted pairs of questions and answers, these are automatically provided as “verified” answers because an expert has reviewed them.
Once a collection is in use, Pryon tracks every request made, every response provided and usage analytics including explicit feedback provided by end users. This provides visibility into what questions end users are asking and other insights.
Pryon supports several types of system and user feedback, automatically taking actions to improve the quality of the information in a collection. These include:
Positive Feedback – When users provide positive feedback about the accuracy of an answer Pryon provides an opportunity for the collection admin to officially make that a verified answer
Negative Feedback – When a user provides negative feedback, the collection admin is alerted and provided an opportunity to change the answer without leaving Pryon or to change the underlying information and reingest the collection
Out of scope alerts – When a question is asked for which there is no good answer in a collection, Pryon alerts the collection admin and provides an instant ability to create an answer to the question without leaving Pryon.
Once content is formed into a collection, transformed by AI and tested for accuracy – it’s ready to use in applications. Our connectors, integrations, and modern architecture are here to make deployment as fast and painless as possible. Use our developer tools to configure the presentation that works best for your project.
"First Mile" Connectors
Out-of-the-box connectors are available for Box, Google Drive, SharePoint online, ZenDesk, and Internet/Intranet sites. Connections to other knowledge sources and enterprise applications are added regularly. There is no need to move data.
“Last Mile” Connectors
Integrate Pryon into any existing application with our documented REST API. Pryon also provides pre-built end user interfaces designed for the web, mobile applications
The best way to configure the presentation of information retrieved by Pryon depends on who the users are, the underlying data, the application they are using to access information and how the data will be used. Pryon is extremely flexible and easy to configure to meet the needs of your business and end users.
You can configure:
- Number of results returned
- Length of answer returned (e.g., shorter for voice)
- Amount of surrounding (contextual) information
offered if scrolling is supported
- Trust factors presented with results (e.g., confidence
scores, providence of data, etc.)
- Formatting of link to source document and other
We provide live demos to companies building applications such as Automations, Chatbots, Customer Service, Search, and Voice Assistants. We can even show you Pryon working on your data. Let’s get started!