Development of a multiple-domain Outbound Metabot

Información del proyecto

POBUCA Bot

Identificador del acuerdo de subvención: 877256

DOI

10.3030/877256

Proyecto cerrado

Fecha de la firma de la CE 14 Agosto 2019

Fecha de inicio 1 Septiembre 2019

Fecha de finalización 29 Febrero 2020

Financiado con arreglo a

• INDUSTRIAL LEADERSHIP - Innovation In SMEs

Coste total

€ 71 429,00

Aportación de la UE

€ 50 000,00

Coordinado por

SIEBEN INNOVATIVE INFORMATION TECHNOLOGY SOLUTIONS LTD
Greece

Periodic Reporting for period 1 - POBUCA Bot (Development of a multiple-domain Outbound Metabot)

Período documentado: 2019-09-01 hasta 2020-02-29

Resumen del contexto y de los objetivos generales del proyecto

Disrupt the Corporate Software environment
Through POBUCA Bot, Sieben attempts to disrupt the Corporate Software environment by introducing Artificial Intelligence (AI) in business communication. Further, POBUCA Bot, targets directly the forthcoming 5G-IoT-related vertical industries, where low-latency communications will need to be assisted by ultra-reliable man-machine communication and user experience (UX), augmenting human capabilities.
In effect, POBUCA Bot Project is an ICT-oriented effort based on three pillars:
- Natural Language Processing (NLP): It is an inter-disciplinary domain within Computer Science focusing on how computers can process and analyze natural language data.
- Deep Learning (DL): It is part of the broader concept of machine learning. It includes algorithms that allow cascading information processing, i.e. processing on layers, whereby each successive layer uses the output from the previous layer as input, thus resembling the “learning” process of humans.
- The 5th generation of wireless communications (5G), due by 2020. 5G will be the successor of current 4G (LTE/WiMax) systems, offering higher data rates, reduced latency, increased reliability and massive device connectivity.


Objectives
POBUCA Bot basic concept is to provide global markets with an Outbound Virtual Assistant with multi-domain capabilities on a quasi “plug-and-play” mode.
Targeting the Brands & Retailers markets: Initially, the company will develop an outbound VA, which will be an add-on to the company’s current business software solutions, to initiate effort in a business sector that the company already possesses significant expertise and clientele. Therefore, the first POBUCA Bot editions will be directed towards the Brands and Retailers markets, offering outbound VA capabilities on domains such as Sales & Marketing, Customer Support, Sales Teams Support, and Line of Business Support.
Entering the hospitality sector: Next, efforts will be targeted into the hospitality sector. In this attempt, main objective will be to provide travelers with an Outbound Virtual Concierge (OVC), as part of the tourist establishment’s (hotel, travel agency, attraction, etc.) overall offering. POBUCA Bot OVC is intended to be distributed as a B2B2C solution, hence it is foreseen to be technologically agnostic, i.e. not connected to specific tools and platforms. Additional domains: Accommodation Support, Shopping & Dining Support, Transportation Support, and Emergencies Support.

The future: Ultimately, POBUCA Bot will be offered as a flexible cross-disciplinary Outbound metabot to global 5G-IoT solution providers’ markets.

Trabajo realizado desde el comienzo del proyecto hasta el final del período abarcado por el informe y los principales resultados hasta la fecha

"Improving POBUCA Bot concept and functionality has been a nodal component of POBUCA Bot Project. Development of an innovative solution aiming at disruption of global markets is a challenge in itself, so Sieben needs to take stable and cautious steps towards her objectives. To this end, during the past months, work on POBUCA Bot prototyping focused on the following solution components:

-Analysis and processing of natural language, in multiple languages.
Technical team has focused heavily to identify techniques to support multiple languages and Bot response capabilities to interlocutor’s language. To this end, a fully functional chat bot (beta version) has been successfully implemented which understands user intent in English and Greek and performs fuzzy searches in contacts database to find related data. The Greek- speaking and the English-speaking version of the chat bot is currently in use by selected pilot users globally who provide Sieben with important user requirements and feedback.

- Improving TTS and STT algorithms.
Sieben developers have been working on ""text to speech” and “speech to text” algorithms and their integration with telephony services. TTS and STT algorithms can be considered as the spinal cord of any Bot solution, since they define the ability of the Bot: firstly, to translate natural language into computer-processable commands, and then to translate the process output into natural language, as seemingly as possible.

- Adding emotional intelligence (EQ) to Bot functionality.
To-date, global ICT community – including Sieben – has mainly focused on adding Cognitive Intelligence (IQ) to chat bots. Indeed, most of the effort spent within POBUCA Bot project, was directed towards adding cognitive intelligence to POBUCA Bot. At the same time however, Sieben tech team acknowledged that key POBUCA Bot component towards a multi-domain outbound bot is EQ. Within the past months, Sieben Tech team has worked mainly on the following POBUCA Bot attributes: (a) on understanding interlocutor’s state, to minimise inconvenience and (b) on developing a natural-sounding agent with minimum lag between answers in order to seamlessly interact with a human over the phone. However, the above are considered rather elementary steps towards adding EQ to POBUCA Bot and the team has acknowledged the challenges ahead.

- Datasets.
Data availability is considered the single most important parameter for AI engines training. Hence, during the project, significant effort was directed towards securing important anonymised data. In this effort, Sieben’s data centric approach in product development has made available a significant amount of properly anonymised data, which can be used for training of new POBUCA bot models.

- Other advancements. Within Project 877256, technical team made significant advancements towards a fully functional POBUCA Bot prototype, which can be detected on the following main thematic areas:
[i] Speech to text using deep learning and connectionist temporal classification (CTC),
[ii] Speaker diarisation using speaker embeddings and spectral clustering,
[iii] Sentiment analysis using transformers,
[iv] Language modeling for Greek, English and various word embeddings"

Avances que van más allá del estado de la técnica e impacto potencial esperado (incluida la repercusión socioeconómica y las implicaciones sociales más amplias del proyecto hasta la fecha)

Democratizing AI

The 5G-ΙοΤ ecosystem is an unparalleled effort to create new communication standards, norms and end-user experiences. It consists of two main pillars, namely TELCOs (telecommunication services providers) and OEMs (equipment manufacturers), while a global ecosystem incorporating thousands of software houses, solutions providers, start-ups, etc., are working on potential end-user solutions making use of the 5G-IoT technologies’ capabilities.

The ability of POBUCA Bot to train itself and adapt its functionality to the business needs of individual industries makes it unique in its field and pioneer in the democratization of technology, as its presence makes the process of direct selling feasible and accessible to small and medium-sized enterprises.

Advancement of POBUCA Bot technology into a “plug-and-play” type utility (i.e. a technology requiring minimum effort to be incorporated into third parties’ solutions), is expected to be positively adopted by the above solution providers, since they will be able to incorporate Outbound VA capabilities in their own solutions at minimum development cost.

POBUCA Bot technology incorporates a significant societal element, in particular concerning inclusion of individuals with special needs, such as with visual impairments or hearing disabilities. Also, POBUCA Bot is expected to help overcome language barriers, in particular for travelers abroad, businesspersons expanding their activities in foreign countries, or even professionals operating in a multilingual environment.

POBUCA Bot