Datagramar Business Language Processing has been built from the ground-up to solve technological challenges inherent to the understanding of conversations with heavy business interpretations.
The combination of a data-oriented deterministic algorithm and business specific grammar rules makes BPL a unique NLP engine that requires no training. Adding business synonyms is done in real time without any need to retrain any model. It saves considerable amount of time for operating the engine in live deployments.
All questions asked by end-users augment a non-regression dataset used to continuously measure the accuracy of the engine and identify potential regressions or lack of understanding. It guarantees the virtual assistant performs at the best of its capabilities based on real and live user intents.
Understand the business domain of the dataset and the context of the end user. BLP is capable of discerning subtleties behind words, acronyms, expressions, as well as business intents or concepts. It can also make meaningful inferences necessary to constraint the data accessibility and visibility:
To meet scalability and security requirements, the BLP engine supports multiple deployment models, including in-house, dockerised or full SaaS architecture.
To prevent non-authorised access to data insights, BLP enforces the necessary data access controls according to the user permission scope and company security rules
To maximise the integrability of Datagramar, the BLP engine is agnostic to the data sources thanks to its abstraction layer. An administration framework allows to configure anything related to business interpretability (data relations and classifications).
Currently Datagramar is available in English, Spanish and French. 10 other languages will be added as part of the roadmap.
During the initial setup, the data structure of your sources is enriched with the business information necessary for the correct natural language interpretations. This is done specifically for your data. Datagramar always manages to resolve a question. In case the end-user is expecting another interpretation, he can precise his sentence to improve the response.
This is done during the initial setup. Each data source object (or table) is mapped with a business interpreter that adds knowledge about your data, including object type (dimension, metric, date), quantities (volume, mass,...), currency, units and advanced data characteristics.
Users interacting with the virtual assistant expect it to understand who they are, in which department they work and what data is available to them based on the company rules. Datagramar transforms possessive expressions (like 'my department', 'my KPI') into a data request that carries data scope and permissions details.
Product overview
Solutions
Contact
