=Paper=
{{Paper
|id=Vol-2747/paper12
|storemode=property
|title=Free software virtual assistants for designing pervasive gaming experiences to promote active aging: State of the art
|pdfUrl=https://ceur-ws.org/Vol-2747/paper12.pdf
|volume=Vol-2747
|authors=Eduardo Nacimiento-García,Carina S. González-González,Francisco L. Gutiérrez-Vela
}}
==Free software virtual assistants for designing pervasive gaming experiences to promote active aging: State of the art==
https://ceur-ws.org/Vol-2747/paper12.pdf
Free software virtual assistants for designing
pervasive gaming experiences to promote active
aging: State of the art
Nacimiento-Garcı́a, Eduardo1[0000−0002−4075−0944] , González-González. Carina
S.1[0000−0001−5939−9544] , and Gutiérrez-Vela, Francisco L.2[0000−0001−6629−7597]
1
Universidad de La Laguna, La Laguna, Spain
{enacimie,cjgonza}@ull.es
2
Universidad de Granada, Granada, Spain
fgutierr@ugr.es
Abstract. The design of pervasive gaming experiences to promote ac-
tive aging through virtual voice assistants makes us raise two funda-
mental requirements, such as respect for user privacy and the possibility
of developing proactive applications. These two requirements make the
use of the main commercial assistants, unfeasible. In general, these de-
vices are always listening and processing conversations without the need
for explicit consent each time. These virtual assistants usually restrict
some functionalities such as proactivity in Applications. Thus, in this
paper, we present different free and open-source solutions to fulfill these
requirements in implementing pervasive games.
Using free software-based tools, we can develop virtual assistants that
meet our requirements and implement our pervasive gaming experiences
to promote active aging. After analyzing the main free tools available,
we highlight Mozilla DeepSpeech and Rasa, on the one hand, and on
the other hand, the tools offered by the Stanford Open Virtual Assistant
Lab (OVAL). It is important to note that these tools are modular to be
mixed depending on our tastes and needs.
Keywords: pervasive game · active aging · virtual assistant · voice as-
sistant · deep speech · rasa · free software.
1 Introduction
Nowadays, homes with a multitude of interconnected digital devices are increas-
ingly common and these, along with mobiles, are already part of our daily lives.
In the same way, transversally, technology is linked to the daily lives of many
people. We also have that the percentage of the world population over 60 years
old does not stop increasing[1], and it is estimated that between 2000 and 2050,
this percentage will more than double, reaching 11% to 22%[2].
It is essential to highlight the importance of active aging for people. For this,
we must consider the specificities of this sector of the population and investigate
Copyright c 2020 for this paper by its authors. Use permitted under Creative Commons
License Attribution 4.0 International (CC BY 4.0).
1
2 Nacimiento-Garcı́a, Eduardo et al.
how to promote this active aging, which must include, for example, from per-
forming physical activities to avoiding the problems derived from loneliness[2],
because many times these types of people suffer from it.
Games, and especially pervasive games, together with the proliferation of
interconnected digital devices, offer us a clear opportunity to deal with the issue
of active aging from this perspective.
There are currently various virtual voice assistants on the market[3] and free
tools to create virtual voice assistants specially adapted to our requirements[4].
We propose the use of virtual voice assistants together with pervasive games to
promote active aging. There are two fundamental requirements when choosing
which platform or technology to use. The first requirement is the possibility of
the assistant’s proactivity, and the second is respect for the user’s privacy.
When it comes to using virtual assistants, we run into problems from the
functionalities perspective and the privacy point of view. In the first case, we
find the limitation of proactivity by the main existing virtual assistants, but this
is a functionality that we consider very important. In the second case, we find
how companies use and process the data of users who use their devices, which
causes a severe loss of privacy. Thus, we will focus mainly on the analysis of free
software tools that allow us to build virtual assistants that adapt to our needs
and respect privacy.
This study aims to find tools that allow us to design experiences based on per-
vasive games through virtual assistants and that comply with the requirements
that we have raised regarding proactivity and privacy.
This paper is organized as follows: first, we describe the main characteristics
and needs in the design of pervasive gaming experiences to promote active aging;
second, we introduce the use of virtual assistants considering the impossibility
of adapting non-free virtual assistants to all our requirements, from among the
free ones, we will see the most appropriate, and finally, the conclusions comment
on the main findings of this work.
2 Use of pervasive gaming experiences to promote active
aging
The percentage of people over 60 continues increasing, and there seems to be a
clear trend for this percentage to continue to grow. Due to the large number of
transformations produced in societies with the incorporation of ICT in daily life
and especially in the field of digital entertainment, it is necessary to redesign
and adapt these transformations to the needs and preferences of older people in
order to incorporate this segment of the population into the digital culture and
at the same time promote active aging through the use of these technologies.
It is necessary to adapt to these circumstances in all areas, including virtual
assistants and games, among other things, to try to achieve active aging of the
population. It is essential to analyze this type of technology, primarily through
using the voice and avoiding graphic interfaces that could hinder older adults[5].
2
Free Software virtual assistants for designing pervasive gaming experiences ... 3
Along with aging, people are prone to present some problems derived from
age, such as cognitive deterioration, loneliness, depression, etc. The use of virtual
assistants can serve to try to avoid or improve the situation of older people
regarding these problems[6]. Voice-controlled assistants have great potential in
monitoring older people’s health conditions, mainly due to the variety of sensors
available that can be used. For example, sensors can be used to detect movement
or pressure on a bed and depth sensors, among others, that can help us control
the parameters of health and activity that these people carry out, respiratory or
heart rates, behavior patterns, etc[7].
We want to place particular emphasis on proactivity due to the target au-
dience for the resulting product. The devices must be capable of generating
interactions with older people without waiting to be activated previously. This
feature is essential for the correct use of virtual assistants’ possibilities because
we may be dealing with people unfamiliar with the use of technology, but also
with people who suffer depression as a consequence of loneliness, among other
things.
Technological evolution has allowed games to play an essential role in mobile
devices today[8], and increasingly their redesign to be used with virtual assis-
tants will become more common. A challenge arises when adapting games to
the characteristics that these devices offer us and discover the new possibilities
that arise. The field of pervasive games is wide and diverse in the approaches
and technologies used. These games are used to create new experiences that
can combine elements of real and virtual games[9]. With the advancement of
pervasive games, we can achieve until now incredible or unthinkable forms of
entertainment[10].
It is crucial to design experiences based on older adults’ pervasive games
since evidence indicates that older people who use games regularly tend to have
a more positive attitude towards aging than other people at that age. From the
socio-emotional point of view, applications must adapt to the requirements and
preferences of the elderly be considered and developed[11].
The development of pervasive games focused mainly on older adults with
virtual assistant technologies with voice and preserving users’ privacy can serve
this field’s innovations. These innovations could be extrapolated to other areas of
society with requirements or conditions other than the hegemonic ones in games
and virtual assistants, such as childhood or people with functional or intellectual
diversity.
3 Virtual assistants
One of Artificial Intelligence (AI) ’s main objectives has been to achieve dialogues
between people and computers. In recent years, the branch of AI focused on con-
versation systems has grown the most. Apart from voice recognition, other topics
such as gesture recognition, image recognition, or videos are investigated[3].
In this section, we will see the main problems related to using the most used
virtual assistants, especially considering the stated objectives related to the use
3
4 Nacimiento-Garcı́a, Eduardo et al.
of these in the development of experiences based on pervasive games to promote
active aging. Later we will focus on analyzing the main alternatives in the world
of free software to get a virtual assistant that meets our requirements.
3.1 Top virtual assistants on the market
Many large companies implement their virtual assistant systems, such as Ama-
zon Alexa, Google Home, Microsoft’s Cortana, Apple’s Siri, or Facebook’s M.
A critical issue to keep in mind regarding the most popular commercial voice
assistants is that some of its functionalities are limited. So, features such as
proactivity are not available[12], which could seriously limit this line of research’s
possible applications.
Serious security problems have been detected due to the way these virtual
voice assistants work. In part due to the lack of a reliable authentication method,
since it is possible that identity can be easily supplanted and, for example,
perform purchases not authorized by the person who owns the assistant[13],[14].
Given the mistrust of these devices’ users’ vulnerabilities, a clear tendency has
been detected to use them in private environments, mainly and avoid using them
in public spaces[15].
Around the main virtual assistants, special attention must be paid to the
privacy and data management that corporations do, because their commercial
interests make them seek to maximize profits and take advantage of the value of
user information. We have that while the people who use it express their concern
that their rights are violated and want to control their personal information[16].
This concern regarding privacy and data use without user permissions seems
to stand out, especially in Amazon Alexa, based on the number of articles that
refer to it. However, this is probably mainly because Amazon controls 70% of the
virtual voice assistant market. However, probably none of the different commer-
cial systems is free from this concern. The use made of the collected data derived
from a non-voluntary interaction with the devices is especially delicate. Analysis
of the traffic produced by Alexa shows that this is part of regular operation[17].
The fact that these virtual assistants are continuously “listening” represents a
serious privacy problem in which many people are either unaware of its possible
implications or finally end up resigning themselves in exchange for using the
technological advances that this type of platform offers[18].
3.2 Virtual assistants with Free Software
The main virtual assistants do not meet our demands for proactivity and privacy.
After searching for alternatives, it is necessary to focus our efforts on finding free
alternatives that allow us to implement virtual assistants who are proactive and
respect the privacy of people who use them. We also need these tools to offer us
the possibility of adapting it as much as possible to the people’s demands and
tastes for whom they are intended. In this case, mainly for older adults, without
forgetting the other sectors of the population.
4
Free Software virtual assistants for designing pervasive gaming experiences ... 5
After analyzing the main free software tools available, it is observed that
there are some tools with the capacity to be used to implement a virtual assistant
with our demands. It is important to note that we usually do not have a single
software that does all the work necessary to implement our virtual assistant.
However, there are usually programs that do part of the work, and then each of
these tools interconnects with each other, thus giving rise to the final product.
We must emphasize that this modular approach often allows us to interconnect
these “modules” with each other to adapt it even more to our needs. Even when
the same project offers us several of these “modules,” we can often use some
of them and interconnect them with others without using all the same project
elements.
Therefore, below we pay special attention to free tools that can allow us to
build systems that respect the user’s privacy and personal environment.
Also, we have that these systems can be adapted to the needs of our re-
search and requirements, contrary to what can happen when the possibilities of
adapting a proprietary system are usually limited.
• Deep Speech: The Baidu research team proposed a voice recognition system
that uses deep learning and simplifies voice systems[19]. Other essential features
of Deep Speech are that they do not use a phoneme dictionary, and that “you do
not need hand-designed components to model background noise, reverberation,
or speaker variation, but rather learn it directly”[19].
There is an implementation of Deep Speech made by Mozilla and released
under the MPL (Mozilla Public License)[20]. It uses a neural network to convert
the captured voice’s spectrogram into the transcribed text; it is beneficial for us
to use on the Internet of Things (IoT) devices. It can be used with or without
a continuous internet connection. It includes pre-trained data sets. However, we
can also add new sets[21].
Mozilla Deep Speech uses Google’s Tensorflow to facilitate the speech’s im-
plementation to the text engine[22]. Such a project requires a large amount of
heterogeneous data to function correctly. There is the Common Voice project,
also from Mozilla, to collect this data using crowdsourcing through more than
50,000 people. The information has been collected in 38 languages, making it
the largest body of free audio for voice recognition[23].
In non-ideal and rather noisy environments, the Mozilla implementation has
a much lower error rate than other commercial systems from Google, Apple, or
Microsoft. It also has a lower error rate in noise-free and combined environments,
although the difference, especially with the Google system, is not wide[4].
We can find success stories where Mozilla tools can be integrated with other
free tools such as Rasa. A virtual assistant can be built in a local environment,
careful with privacy since it avoids sending packets over the internet. Mozilla
DeepSpeech and Mozilla TTS are used to convert speech to text and vice versa.
In this case, Rasa would be in charge of understanding natural language and
managing dialogue[24].
5
6 Nacimiento-Garcı́a, Eduardo et al.
• OVAL: Also notable is the “Stanford Open Virtual Assistant Lab (OVAL),”
which aims to create a free virtual assistant ecosystem that respects privacy[25].
This project is divided into other projects, such as Thingpedia, LUInet, Genie,
Almond, a communication protocol for a virtual assistant, and Brassau.
Almond is a free virtual assistant through crowdsourcing, and that preserves
privacy; Almond is programmable and is designed to offer online and internet
of things services[26]. Thingpedia offers us a public knowledge base with an
open API and natural language interfaces[26]. Brassau has a graphical virtual
assistant that converts natural language commands into a graphical interface[27].
We also have Genie, a natural language semantic parser generator for virtual
assistant commands[28]. Soundr allows us to use an array of microphones like
those included in most intelligent voice systems and infer the speaker’s spatial
locations and the head[29].
• Rasa: When we analyze Deep Speech, we talked about Rasa as part of a joint-
use proposal between both tools. Rasa is a machine learning framework that
allows us to automate texts and voice conversations. Rasa is free software under
Apache 2 license. With Rasa, we can understand messages from conversations,
but it also offers us the possibility of connecting to a wide variety of messaging
channels and APIs[30].
Basically, Rasa’s operation consists of receiving a message and converting it
into a dictionary to be processed. Rasa has two main components, Rasa Core
and Rasa NLU. Rasa Core uses dialogue management based on machine learning,
and with this framework, we can create, for example, chatbots. Rasa NLU offers
us a module of classification of intentions and extraction of entities[31]. Rasa
NLU handles natural language processing.
• Others tools: To finish this section of virtual assistants, we will see other
free tools that we could also use to construct a virtual assistant.
Tacotron 2 is a neural network architecture for speech synthesis from the
text, developed by Google, which allows us to synthesize waveforms from spec-
trograms[32]. Tacotron allows an end-to-end voice system, which achieves very
similar to that of a real human voice[33].
Studies have been carried out that mix the use of Deep Speech and Tacotron
in the same project. Tacotron is used to recognize speech and convert it to text
automatically, and Deep Speech is used to convert that text to speech[34].
Another system, designed by Facebook and under MIT license, is wav2letter++,
a light, and straightforward tool to build speech recognition systems, which uses
the ArrayFire library and is written in C++[35].
As an alternative to Rasa NLU, we also have Snips NLU, a machine learning-
based voice platform connecting with embedded systems[36]. Snips are designed
to work without an internet connection, so it allows you to perform processing
without using the cloud, although you can also connect to the internet if it is
wanted[12]. Snips allow us to guarantee the privacy of users[37].
6
Free Software virtual assistants for designing pervasive gaming experiences ... 7
4 Defense of privacy
The privacy of virtual assistant users is an important point to keep in mind. We
have that the main commercial services do not comply with this requirement
because confidential information is sent through the internet to be later processed
in the cloud without the person’s explicit consent who uses the device consciously
or unconsciously. It is important to establish protocols that assure us that this
type of “cheating” practice is not carried out.
The need arises to process the voice input locally, even in a “more primitive”
way, until a request or an affirmative response to the assistant’s proactive action
is detected. After obtaining the user’s permission, it would be possible to analyze
the content obtained below and during that conversation. These protocols must
allow us to enjoy the benefits of virtual assistants without violating privacy.
Many of the privacy problems that we have using the main virtual voice
assistants are not unique in this type of device. However, problems derived from
the Service as a Software Substitute (SaaSS)[38] are added to the voice that may
be continuously being recorded and processed.
The problem related to privacy is not only solved with the drafting of terms
and conditions of use. Your data will be collected and processed at all times. Also,
many people will accept these abusive conditions even if they do not agree[15],
only to enjoy these devices’ technical advantages, so there should even be legal
directives that regulate this type of practice by companies.
Table 1. Relationship between the tools and the required objectives.
Tool Proactivity Privacy Easy to use
Google Home Low Low High
Amazon Alexa Low Low High
Apple Siri Low Low High
Mozilla & Rasa High High Medium
OVAL High High Low
5 Conclusions
In this paper, we analyzed virtual assistants’ use to design pervasive gaming
experiences to promote active aging, and together we propose two essential re-
quirements that virtual assistants must meet. These requirements are that the
user’s privacy is respected and that our application can be proactive because it
is mainly intended for older people.
The main virtual assistants in the market are not respectful of user pri-
vacy, and also because these assistants usually have limited functionality, such
as proactivity. For this reason, we decided to investigate whether there are free
tools that allow us to develop virtual assistants that meet our requirements and
allow us to create pervasive gaming experiences to promote active aging.
7
8 Nacimiento-Garcı́a, Eduardo et al.
After analyzing the main free tools available, we have mainly chosen Mozilla
and Rasa’s tools, on the one hand, and on the other hand, the tools offered by
the Stanford Open Virtual Assitant Lab (OVAL). It is important to remember
that these tools are modular to be mixed depending on our tastes and needs.
Simplifying, we can divide the analyzed tools based on their proactivity,
respect for privacy, and ease of use in our experiences based on pervasive games.
In Table 1, we show the classification of the different tools based on the analysis
carried out.
Acknowledgement
Supported by “Predoctoral training program for research staff in the Canary
Islands of the Ministry of Economy, Knowledge, and Employment co-financed
by the European Social Fund (ESF), with a co-financing rate of 85% ” and the
project “Design of pervasive gaming experiences using virtual assistants to pro-
mote active aging in older people within the digital home environment (RTI2018-
096986-B-C32)”.
References
1. Lutz, W., Sanderson, W., Scherbov, S.: The coming acceleration of
global population ageing. Nature 451(7179) (February 2008) 716–719.
https://doi.org/10.1038/nature06516
2. Arslantas, H., et al.: Loneliness in Elderly People, Associated Factors
and Its Correlation with Quality of Life: A Field Study from Western
Turkey. Iranian Journal of Public Health 44(1) (January 2015) 43–50
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4450013/
3. Këpuska, V., Bohouta, G.: Next-generation of virtual personal assistants (Microsoft
Cortana, Apple Siri, Amazon Alexa and Google Home). In: 2018 IEEE 8th Annual
Computing and Communication Workshop and Conference (CCWC). (January
2018) 99–103. https://doi.org/10.1109/CCWC.2018.8301638
4. Ordiales, H.: Comparativa de métodos de conversión de voz a texto Open Source.
(2019) 6
5. De los Santos Cicutto, Santiago: Experiencia de uso de asistentes de voz sin GUI
en personas mayores (2017)
6. Sunghoon, K., Parasuraman, G.M., Jaunbuccus, S.: Elderly Care Assistant: A
Discreet Monitoring Tool. In Fleming, P., Lacquet, B.M., Sanei, S., Deb, K.,
Jakobsson, A., eds.: Smart and Sustainable Engineering for Next Generation Ap-
plications. Lecture Notes in Electrical Engineering, Cham, Springer International
Publishing (2019) 287–301. https://doi.org/10.1007/978-3-030-18240-3 27
7. Shalini, S., et al.: Development and Comparison of Customized Voice-Assistant
Systems for Independent Living Older Adults. In Zhou, J., Salvendy, G., eds.:
Human Aspects of IT for the Aged Population. Social Media, Games and Assistive
Environments. Lecture Notes in Computer Science, Cham, Springer International
Publishing (2019) 464–479. https://doi.org/10.1007/978-3-030-22015-0 36
8
Free Software virtual assistants for designing pervasive gaming experiences ... 9
8. Arango-López, J., Gallardo, J., Gutiérrez, F.L., Cerezo, E., Amengual, E., Valera,
R.: Pervasive games: giving a meaning based on the player experience. In: Pro-
ceedings of the XVIII International Conference on Human Computer Interaction.
Interacción ’17, Cancun, Mexico, Association for Computing Machinery (Septem-
ber 2017) 1–4. https://doi.org/10.1145/3123818.3123832
9. Magerkurth, C., Cheok, A.D., Mandryk, R.L., Nilsen, T.: Pervasive games: bringing
computer entertainment back to the real world. Computers in Entertainment 3(3)
(July 2005) 4. https://doi.org/10.1145/1077246.1077257
10. Hinske, S., Lampe, M., Magerkurth, C., Röcker, C.: Classifying Pervasive Games:
On Pervasive Computing and Mixed Reality. 21
11. Allairea, Jason C., et al.: Successful aging through digital games: Socioe-
motional differences between older adult gamers and Non-gamers. (2013).
https://doi.org/10.1016/j.chb.2013.01.014
12. Jesús-Azabal, M., et al.: Voice Assistant to Remind Pharmacologic Treatment in
Elders. (February 2020) 123–133. https://doi.org/10.1007/978-3-030-41494-8 12
13. Lei, X., et al.: The Insecurity of Home Digital Voice Assistants – Amazon Alexa
as a Case Study. (November 2019) http://arxiv.org/abs/1712.03327
14. Zhang, N., et al.: Understanding and Mitigating the Security Risks of Voice-
Controlled Third-Party Skills on Amazon Alexa and Google Home. (June 2018)
http://arxiv.org/abs/1805.01525
15. Easwara Moorthy, A., Vu, K.P.L.: Voice Activated Personal Assistant: Accept-
ability of Use in the Public Space. In Yamamoto, S., ed.: Human Interface and
the Management of Information. Information and Knowledge in Applications and
Services. Lecture Notes in Computer Science, Cham, Springer International Pub-
lishing (2014) 324–334. https://doi.org/10.1007/978-3-319-07863-2 32
16. Norberg, P.A., Horne, D.R., Horne, D.A.: The Privacy Paradox: Personal Infor-
mation Disclosure Intentions versus Behaviors. Journal of Consumer Affairs 41(1)
(2007) 100–126. https://doi.org/10.1111/j.1745-6606.2006.00070.x
17. Ford, M., Palmer, W.: Alexa, are you listening to me? An analysis of Alexa voice
service network traffic. Personal and Ubiquitous Computing 23(1) (February 2019)
67–79. https://doi.org/10.1007/s00779-018-1174-x
18. Lau, J., Zimmerman, B., Schaub, F.: Alexa, Are You Listening? Privacy Percep-
tions, Concerns and Privacy-seeking Behaviors with Smart Speakers. Proceedings
of the ACM on Human-Computer Interaction 2(CSCW) (November 2018) 102:1–
102:31. https://doi.org/10.1145/3274371
19. Hannun, A., et al.: Deep Speech: Scaling up end-to-end speech recognition. (De-
cember 2014) http://arxiv.org/abs/1412.5567
20. Mozilla: mozilla/DeepSpeech (May 2020) original-date: 2016-06-02T15:04:53Z.
21. Firmansyah, M.H., Paul, A., Bhattacharya, D., Urfa, G.M.: A.I.
based Embedded Speech to Text Using Deepspeech. (February 2020)
http://arxiv.org/abs/2002.12830
22. : Welcome to DeepSpeech’s documentation! — DeepSpeech 0.7.1 documentation
23. Ardila, R., et al.: Common Voice: A Massively-Multilingual Speech Corpus. (De-
cember 2019) https://arxiv.org/abs/1912.06670v2
24. Petraityte, J.: Build an AI voice assistant with Rasa Open Source and Mozilla
tools | Rasa (August 2019)
25. OVAL: Stanford Open Virtual Assistant Lab (2020)
26. Campagna, G., Ramesh, R., Xu, S., Fischer, M., Lam, M.S.: Almond: The Ar-
chitecture of an Open, Crowdsourced, Privacy-Preserving, Programmable Virtual
Assistant. In: Proceedings of the 26th International Conference on World Wide
9
10 Nacimiento-Garcı́a, Eduardo et al.
Web, Perth Australia, International World Wide Web Conferences Steering Com-
mittee (April 2017) 341–350. https://doi.org/10.1145/3038912.3052562
27. Fischer, M., Campagna, G., Xu, S., Lam, M.S.: Brassau: automatic gener-
ation of graphical user interfaces for virtual assistants. In: Proceedings of
the 20th International Conference on Human-Computer Interaction with Mo-
bile Devices and Services, Barcelona Spain, ACM (September 2018) 1–12.
https://doi.org/10.1145/3229434.3229481
28. Campagna, G., et al.: Genie: a generator of natural language semantic parsers for
virtual assistant commands. In: Proceedings of the 40th ACM SIGPLAN Confer-
ence on Programming Language Design and Implementation - PLDI 2019, Phoenix,
AZ, USA, ACM Press (2019) 394–410. https://doi.org/10.1145/3314221.3314594
29. Yang, J.J., et al.: Soundr: Head Position and Orientation Prediction Using a
Microphone Array. In: Proceedings of the 2020 CHI Conference on Human
Factors in Computing Systems, Honolulu HI USA, ACM (April 2020) 1–12.
https://doi.org/10.1145/3313831.3376427
30. : Build contextual chatbots and AI assistants with Rasa
31. Singh, A., Ramasubramanian, K., Shivam, S.: Introduction to Microsoft Bot,
RASA, and Google Dialogflow. In: Building an Enterprise Chatbot: Work with
Protected Enterprise Data Using Open Source Frameworks. Apress, Berkeley, CA
(2019) 281–302. https://doi.org/10.1007/978-1-4842-5034-1 7
32. Shen, J., et al.: Natural TTS Synthesis by Conditioning WaveNet on Mel Spectro-
gram Predictions. (February 2018) http://arxiv.org/abs/1712.05884
33. Wang, Y., et al.: Tacotron: Towards End-to-End Speech Synthesis. (April 2017)
http://arxiv.org/abs/1703.10135
34. Chandran, S., Giri, S.: Voice Converter Using DeepSpeech and Tacotron. (2019)
6
35. Pratap, V., et al.: wav2letter++: The Fastest Open-source Speech Recog-
nition System. ICASSP 2019 - 2019 IEEE International Conference on
Acoustics, Speech and Signal Processing (ICASSP) (May 2019) 6460–6464.
https://doi.org/10.1109/ICASSP.2019.8683535 arXiv: 1812.07625.
36. Staudigl, F.: Design and Implementation of an End-to-End Speech Assistant. (June
2019) https://dspace.cvut.cz/handle/10467/83444
37. Coucke, A., et al.: Snips Voice Platform: an embedded Spoken Language Un-
derstanding system for private-by-design voice interfaces. (December 2018)
http://arxiv.org/abs/1805.10190
38. Richard Stallman: Who does that server really serve? (2010)
10