Voicetel Communications is a high-tech company that operates in the Polish market of services aimed at supporting call centres. It specialises in the implementation of voice technologies in the automatic customer service sector. The Partner Solutions are applied primarily in businesses that provide call centre services, as well as in companies offering general services, operating in the manufacturing sector and trade which have their own telephone customer service departments.
The Voicetel Communications technologies – the Esense Dialogue Interface Platform, and the Brilliance™ Dialogue Interface, which is still in the pilot phase – help to make communication between businesses and clients easier. They also allow to do away with oversized departments that employ phone consultants and replace them with an advanced and fully automated virtual assistant.
The business fields of Voicetel Communications S. A. include:
- the implementation of pilot activities concerning the BrillianceTM Dialogue Interface (a ready automatic customer service solution) in businesses that have contact centre departments,
- the lease of Esense Platform to technological enterprises that create IT systems.
The current research and business activity of Voicetel Communications has resulted in the creation of a dialogue platform which is still being developed.
It was initially based on Google’s speech recognition and VTML language, and it allowed to create non-advanced dialogue systems of automatic telephone customer service, i.e. systems that allow its users to make fairly unimpeded utterances; these systems, however, do not allow for dynamic changes of a dialogue, and they lack advanced forms of contact personalisation with the user.
On this Platform, the Partner has created a series of bots that handle, among others, such processes as making medical appointments, voice-based FAQ for insurance agents, voice-based handling of enquiries concerning transport orders, customer satisfaction assessment, or calculating third-party liability insurance premiums.
Throughout the creation of these dialogue systems, Voicetel was doing its best to make sure that the solution elements that were being devised would boast top reusability and modularity levels, thanks to which each new robot would allow for the extension of the grammar and vocabulary base, and fragments of dialogues that could be reused.
Based on the numerous limitations and imperfections of Google’s speech recognition technology that had been spotted, as well as the high labour-output ratio and the different types of risk arising from the willingness to apply the Google technology in Voicetel solutions, a decision was made to change the speech recognition technology to Nuance technology.
Given the completely different nature of the new speech recognition engine, and the changed approach to the outsourcing of technological infrastructure, work began on a new range of integration of component services, including Nuance ASR, Ivona TTS, Asterisk, and VoiceXML-format interpreter, expanded with several functionalities required by the specific nature of guidelines devised by Voicetel Communications.
The efforts launched in 2014 have allowed to build a local trial infrastructure that makes it possible to develop bots in a new format. The dialogue systems that were brought to life on the new platform (e.g. booking of plane tickets, hotel rooms and cinema seats, soft debt collection, lending VOD films and cars) had higher levels of ergonomics thanks to a major improvement of speech recognition quality.
Throughout its current activity, Voicetel has commissioned a variety of research and development activities aimed at developing the dialogue interface for three entities. Speechguru sp. z o. o. was the major contractor in this area in 2015. A subsidiary was doing the ordered works through external resources (subcontractors – individuals employed based on specific task contracts, and external entities).
Voicetel will keep its current model of developing services based on external subcontractors, because it is beneficial for the company from the financial point of view – there is no need to employ qualified staff on a permanent basis, from the point of view of quality – ordering the execution of the tasks by selected entities allows to effectively assign the key projects to top-quality experienced IT experts, and it allows to manage the risk related to the timeliness and quality of the assigned tasks to a much fuller extent than in the case of employing staff in a direct manner. The company’s Board stay focused on supervising, commissioning and coordinating research and development activities.
One of the resources of Voicetel Communications is also the knowledge it has gained whilst looking for information on the current dialogue systems and classification algorithms applied for their construction, as well as commissioning research that would look at the mode and quality of the “human being – robot” communication.
At the request of the company, a series of studies on users was conducted, based on the created dialogue systems. The study was based on meticulously selected groups of users, and the very tests of system utility and the quality of their operation were preceded by detailed analyses of the expectations and needs of their potential users. This knowledge has allowed to provide the right direction for further research and development activities geared at enhancing the company’s technology.
Throughout the activities carried out so far, Voicetel has also gained experience with the Nuance SSM (Statistical Semantic Models) classifier, learning about its drawbacks and limitations, which has pushed the company to make plans for the nearest future to bring forth its own classifier that would be free of any of the barriers found in the previous solution.
The new classifier allows for an incomparably more accurate interpretation of the intentions of a Brilliance™ Dialogue Interface user thanks to the application of the innovative approach based on a multi-level speech classification.
Overcoming the barrier of accurate classification of users’ intentions has allowed the company to develop other features of Brilliance™ Dialogue Interface, and to realise the need to make them fully available in its product, i.e. non-linearity, contextuality, and a high degree of dialogue personalisation.
The implementations of Voicetel technology carried out so far have allowed above all to gain experience related to better integration with a client’s infrastructure, and the stability of the platform itself in real conditions. They have also given priority to the forthcoming activities focused on BrillianceML description language, which will allow to carry out further implementations more swiftly and economically.