Axél is built on proprietary algorithms, that makes it one of the most innovative solutions on the market, thanks to specific and distinctive features.
TRAINING BY CONCEPTS
Training made by massive examples, as common in any standard solution, is based on thousands of questions and answers, full of concepts and periphrasis, with the aim to get a perfect match between user’s questions and the input examples.
Axél, instead, creates concepts and needs only one training, which joins all the possible actions to the correspondent answer.
Training made by massive examples is always related to an error tolerance criteria, based on the match between the text entered by the user and the questions encoded as examples.The outcome is the result of a probability calculation, which is returned with a 75% degree of confidence.
Axél creates the concept related to its logical inference only once. Just one single training is needed and it is error tolerance free. Its confidence level is higher than 99%.
Standard solutions need a long time to be developed, often more than six months. The training time is proportional to the complexity of the context.
Axél dramatically reduces the time requested for modeling the knowledge base to only some weeks.
Axél is naturally able to get data-flows from different sources and can combine together concepts and simple data, as belonging to the same graph-structured neural network. This feature boosts to more advanced applications, even in real time, which would call for data from external systems, such as RSS feeds, IOT applications or CRM, BI, DAM software.
Standard solutions do not support automatic learning and need continuous training to improve their performance.
Axél supports automatic learning, increases its knowledge base and improves its performance through data analysis and user’s experience.
PROFILING AND RECOMMENDATION
Although is was specifically created for NLP, thanks to its unique flexibility, Axél also manages to profile users, by analyzing their requests. As a result, it will react to the same input in different ways and will sort the customer into different categories (regular, medium, occasional client) according to their emotional and economic behavior.
TTS AND STT SYSTEMS WITH NO PRE-CABLED ONTOLOGIES
Thanks to its flexibility, dynamic modeling system and concepts interpretation, Axél’s engine can integrate TTS (Text-to-Speech) and STT (Speech-to-Text) systems, in order to allow voice interaction (i.e to manage phone calls, with no need of pre-cabled ontologies, which have deep impact on software architecture, as well as being significantly expensive.
All the standard solutions today available on the market support multilingual translation, although algorithms are mostly created for English and then adjusted to other languages. They work with the same rules of online translation, namely they are often unable to match the algorithms of other languages.
Axél is a “native Italian speaker”, and Italian is in fact its most accurate language. However, as it is based on concepts and not on single phrases and examples, Axél can understand sentences from any language, with no need of literal translation.
NO BINDING PATTERNS OF SPEECH
Standard solutions do not allow open conversation and mostly rely on multiple answers systems, faking an AI system. Being trained by concepts, Axél can answer to any question which is related to the same concept, and is able to recognize different verbs and terms, without using pre-modeled patterns of speech. Axélero’s solution is definitely a platform for open conversation.
CLOUD AND ON PREMISE
All competitors design their platform for the exclusive use of their customers and cannot retrieve that material for big players needs and projects. Axél relies on a flexible IT architecture, which can be partly or entirely installed on the customer’s softwares, according to their needs.