AutoUI 2021 is going virtual, for the second year in a row because of the current situation. This year, AutoUI 2021 is supported by L3Pilot Driving Automation, WIVW, Google, ACM, and SIGCHI.

We will be joining AutoUI 2021 this year, and we will be organizing WP6 — Designing Biofeedback of Driver State and Emotion in Automated Vehicles.

Empathic Labs is teaming up with the HumanTech institute, the University of St.Gallen, and Cerence. Multidisciplinary backgrounds joined to create and present this workshop.

Vehicles are still an important mode of transportation. Autonomous vehicles have been on the rise with a global sales estimation of 33 million cars in 2040 according to IHS Markit. Yet the humans are part of the driving task and will always be part of it while going through a critical situation. We believe in the importance of drivers' self-awareness and self-knowledge. For this reason, having an overview and detailed view of a driver's emotions and affective states should be a concerning research and development topic. By displaying the driver's state we hope to influence their driving.

We will tackle this topic by brainstorming with other researchers and practitioners about the biofeedback response modalities and visualizations. Since nowadays cars are understanding the user and being able to communicate with them, an empathic medium consisting of companionship, trust, and confidence should be created. At this point, autonomous driving can reach a higher stake releasing the driver from their main task and allowing other activities.

Multiple research and development topics will be discussed during this workshop. We will be working on reaching precise objectives while sharing and collaborating with all the participants about the context of this workshop. Creativity and brainstorming sessions will be used to model, design, and visualize the concepts and ideas.

The final objective can be summarized as:

Conceptualization, design, and visualization of biofeedback in autonomous car scenarios

Workshop Dates:

• Session 1: Thursday 09.09–15h10–16h10 CET
• Session 2: Friday 10.09–12h50–13h50 CET

More details about the workshop are found here

The full list of workshops is found here.

You can find the full program of AutoUI 2021 here.

If you would like to have access to the materials of this workshop please don’t hesitate to contact us at guys@empathiclabs.ch

Workshop: Designing Biofeedback in Automated Vehicles was originally published in Empathic Labs on Medium, where people are continuing the conversation by highlighting and responding to this story.

When building chatbots with Natural Language Understanding (NLU), we need a lot of training samples to train intent detection systems. One way is to collect data from users by extracting conversations or using crowdsourcing platforms like Amazon Mechanical Turk, but this is tedious work or costly processes (in addition to the fact that writing sentences by hand is not a fun job). The other way is to develop a language model that is capable of paraphrasing sentences. A paraphrase is a restatement of a text using other words while keeping the same semantic. In this way, it will be possible to create a fraction of the sentences manually and then generate tens of new sentences based on them by paraphrasing them. Adding emotions to these generated paraphrases allows to increase the diversity of the paraphrase and also lays the foundation for the creation of an emotional dialogue architecture.

Transformer-based language models are becoming extremely powerful, so we will see how we can leverage their power to generate emotional paraphrases.

Concept

We propose an architecture composed of three modules: 1) corruption of the emotion present in the input sentence, 2) creation of a sentence that is enhanced emotionally for a specific emotion using a fine-tuned GPT-2 model, and 3) generation of a paraphrase of the emotionally augmented sentence.

1 — Data Corrupter

The Data Corrupter detects and removes emotional words from the sentence. This is done by comparing every word of the sentence against 3 emotional lexicons: DepecheMood, NRC, and SentiWordNet. We use this process on a massive emotional dataset made by unifying 7 emotional datasets under Paul Ekman “6 basic emotions” model: anger, disgust, fear, joy, sadness, and surprise. The 7 datasets we used are GoEmotions, Emotion-Stimulus, CrowdFlower, ISEAR, SMILE, SemEval-2018 Task 1, and TEC. All these datasets use a different emotional model so we need to map these emotions under Ekman’s model using Plutchik’s “Wheels of emotions”, and Parrot’s tree structure model.

Using the Data Corrupter, we can create 6 datasets, one for each emotion of Ekman’s model, of pairs of emotional sentences and corrupted sentences. If we do not find emotional words in a sentence, we should ignore it.

2 — Emotion Enhancer

Using these datasets, we can fine-tune 6 GPT-2 models, one for each emotion. The goal of each model is to reconstruct the emotional sentence using the corrupted sentence, hence learning how to construct a sentence conveying a specific emotion. We call these models Emotion Enhancer.

3 — Paraphrase Generator

Our last module, the Paraphrase Generator, is simply composed of a publicly available fine-tuned GPT-2 model for the paraphrase generation task. We use a GPT-2 model provided by RASA.

Evaluation

To evaluate our system, we create 3 testing sets from popular paraphrases datasets: MSCOCO, QQP, and PARANMT. We conducted a qualitative study with human judges, as well as a quantitative evaluation.

Automatic Evaluation

We use the well-known metrics BLEU, METEOR, and TER to evaluate the paraphrasing aspect of our system. These metrics work well for the paraphrases evaluation and correlate well with human judgments (source: Madnani et al. 2012 and Wubben et al. 2010). To evaluate the emotional aspect of our system, we use an emotion classifier and simply compute the percentage of time the emotional classifier predicts that our paraphrase has the expected emotion. We compare our paraphrasing results with state-of-the-art model GAP from Yang et al. (2019).

Our system is inferior in pure paraphrasing metrics. The Paraphrase Generator module achieves decent scores, but our full emotional architecture performs pretty badly. There are no significant differences between the models for the various emotions.

For the emotion evaluation, the emotion score is higher for the paraphrases generated by our system for almost all the emotions for every testing set. Note that the surprisingly high values for the surprise emotion in the QQP dataset are accountable to a bias. Indeed, the classifier apparently thinks that a text ending with a question mark suggests a surprise, and there are many of them in QQP. Overall, the results obtained suggest that our joy, anger, sadness, and fear models work bests.

Human Evaluation

A human evaluation is required to assess the performance of the system qualitatively. Twenty sentences (6 from MSCOCO, 7 from QQP, and 7 from PARANMT) are randomly chosen and processed by our system that generates 6 sentences for each input, one for each emotion. The 20 reference paraphrases of these 20 base sentences were also added. In total, the human evaluation testing set is composed of 140 pairs of base sentences and emotional/reference sentences. Human annotators are asked to rate each pair on a scale of 1 to 5, where 1 is worst and 5 is best, the content preservation, readability, and diversity of the sentence compared to the base sentence. They are also tasked with classifying the sentence based on the primary emotion it conveys: anger, disgust, fear, joy, sadness, surprise, or no emotion. Each pair is evaluated by 20 human judges hired on Amazon Mechanical Turk.

The ground truth performs better for content preservation and readability for almost every emotion. Regarding diversity, different emotions can perform better than the ground truth sentences, even if the average is almost the same.

For the emotion evaluation, we compared how the human judges classified the emotion of the references with the emotion of the various models. For example, the number of samples classified by the judges as containing anger doubled between the references and the sentences from the anger model, thus an increase of 100%. We see that our disgust, fear, and sadness models work better than the others as they mostly impact the number of votes for their particular emotion. Asking human judges to classify a text based on seven emotions is challenging, as we reach a Fleiss κ of 0.07, meaning a slight inter-annotators agreement only.

Conclusion

We propose a new approach to perform emotional paraphrases by leveraging pre-trained GPT-2 models. We propose an architecture made of three modules. The first removes initial emotion markers from the original sentence, the second augments the emotion of it using a process to create parallel emotional/corrupted sentences, and the third generates a paraphrase of this augmented sentence. We evaluated our system on popular paraphrase datasets and noticed that both in our automatic and human evaluations our models fail to perform as well as state-of-the-art models in paraphrasing-related metrics. On the other hand, our system is able to generate emotion with some degree of success. Both evaluations agree that the fear and sadness models perform well. Only the quantitative evaluation shows that joy and anger are efficient, whereas, on the contrary, the judges found that disgust was working well.

Such technology can greatly facilitate the automatic creation of training phrases for natural language understanding (NLU) systems, but it can also be integrated into an emotional dialogue architecture to create emotional conversations.

Thanks for reading!

If you want to know more about NLU and how it is used for chatbot creation, read the following article:

What is NLU? How to make chatbots understand what you want?

Emotional Paraphrasing was originally published in Empathic Labs on Medium, where people are continuing the conversation by highlighting and responding to this story.

As part of our Master degree, we were asked to reflect on health-care chatbots and how we could improve them.

The health-care chatbots nowadays and in a near future

This kind of chatbot can be classified into two categories: physical and mental health.

In the physical health category, we have for instance bots that can give a basic diagnosis based on symptoms the user describes and decide if this user needs to go to a doctor or not. Similar bots were designed to help people act the right way in suspicion of Covid-19.

Mental health is also a trending domain as it takes more and more importance in this world we live in. It was shown that, for some people, it was easier to confess to a bot rather than a human that could judge them. This is also true for people with person-to-person communication issues like people with autism. Having a conversation with human can be a challenge for this population and the possibility to chat with a chatbot instead can greatly help them feel better.

The healthcare chatbots domain has grown a lot in the recent years and is called to developed even more in the future, as the demand tends to grow faster than the offer. It will probably be soon reality to follow a psychotherapy with a chatbot rather than a human for instance, or have your symptoms analysed by a chatbot to help you decide if you should consult a doctor or not.

Our chatbot

As we listed existing solutions, we realised something: healthcare chatbots are vastly about being sick and healing from this sickness. We therefore decided to tackle another aspect that seems to be little to not represented in the chatbots we could find: preventing rather than caring.

Based on this analysis, we designed “Mon Pote Santé” (My Health Buddy), a companion to help you deal with your daily hurts and bobbins (please appreciate the pun between “Pote” and “Bot”).

Mon Pote Santé has two sides: the caring and the preventing one.

On the caring side, it can give important phone numbers like the emergencies or poison control center for instance. The preventing side is about pollen allergies and sun sensitivity of the user. With the user’s location it is possible to inform them when there is a risk for them regarding pollen or UV.

Thus, you can ask “Do I need to put sun screening before going out?” or “Is there a risk regarding my allergies today?”, Mon Pote Santé knows you and can give you sound advice and help you avoid unnecessary harm.

We used the platform Deeplink to develop Mon Pote Santé and an API called Ambee to retrieve the data about UV and pollen as well as the geolocation API from the Swiss Confederation to translate addresses into coordinates.

Future work on Mon Pote Santé

As it is, Mon Pote Santé is just a proof-of-concept with the bare minimum functionalities, but we believe it has a real potential. We can expand its functionalities further by adding proactive scenarios, like automatic alerts in case of high pollen level or medicine reminders for people taking a daily treatment for instance. Another scenario that would be worth including is the monitoring of the air pollution and preventing the user from being exposed too severely to this pollution.

There are tons of similar scenarios that could be added (advice on injuries, how to properly disinfect a wound, symptoms tracking, etc.). Maybe Mon Pote Santé development will continue someday, time will tell.

Conclusion

Chatbots are a growing part of our world. We see them appearing on the website we browse, in the messaging apps we use, and soon healthcare chatbots will probably be part of our life too.

Health prevention is an important topic and it makes sense to include it in the health chatbots. Maybe in the future we will see more chatbots to include it in their functionalities, working at keeping us safe and sound in the first place rather than helping us to get better.

Thank you for reading this article, we hope you enjoyed it! We’d love to hear about your opinions and questions in the comments. If you found this article interesting please have a look to the Empathic Labs page for more stories.

• Patrick the packet, a chatbot to enhance your experience with package deliveries
• FoodCoach-bot, the chatbot that helps you keep track of your food intake

From caring to preventing was originally published in Empathic Labs on Medium, where people are continuing the conversation by highlighting and responding to this story.

During our Master degree, we were given the task to enhance customer service themed chatbots and to research on what could be some new innovative features.

After some initial research we stepped on package tracking which has a lot of potential in our eyes.

There are not a lot of chatbot solutions to manage the delivery of your package. The only one that we found were simple assistants that redirected to the webpage when ask for specific topics.

It mostly feels unfinished and unnecessary for customer that have experience with more complete chatbots.

What can we do better?

Our idea started with feature ideas that can replace the customer portal and make it unnecessary to use it altogether.

We are looking at basic features like:

• Package location
• Estimated arrival
• Deposit options

Using the chatbot should enable customers to act in a more natural way and with less burdens. Continuing on this path, we explored ideas to make the interaction more memorable and fun.

That is were the final concept of Patrick emerged. We want customer to experience a new way of package tracking by actually asking their package themselve!

Our concept personifies the deliverable and tries to make it as easy as chatting with a friend to manage your packages.

What does it looks like?

The following screenshots explain the main features that are implemented.

Asking and saving the package id

Patrick understands the requirement of the package id to perform most of the supported tasks.

Once received and saved, Patrick can use it to request information about himself.

Asking about the package location

Patrick will request information with the package id about it’s current location and will share it with the customer with a map pointer.

Defining deposit options

Patrick understands when a customer defines a specific deposit instructions and saves it in the system.

It will be added as information when asked for the arrival estimation as well.

How does it work?

Patrick is for the moment a chatbot only accessible on Telegram. The following diagram shows an overview on the internal architecture of our chatbot:

Patrick is implemented in Python and uses the python-telegram-bot wrapper to communicate with Telegram. A simple function that echoes the same message back to the user will look like the following

def start(update, context):
   [...]

def text(update, context):
    update.message.reply_text(update.message.text)

def main():
    TOKEN = "TELEGRAM TOKEN HERE"

    updater = Updater(TOKEN)
    updater.dispatcher.add_handler(MessageHandler(Filters.text, text))

    updater.start_polling()
    updater.idle()

Rasa NLU offers a REST API to parse messages which needs to be activate on the server launch:

$ rasa run --enable-api

Rasa NLU includes an OpenApi file on their website which can be used to generate the python client code with OpenAPI Generator.

A trained Rasa NLU server can then be requested from the Python implementation:

import openapi_client
from openapi_client import InlineObject1
from openapi_client.rest import ApiException

configuration = openapi_client.Configuration()


def parse_message(message):
    with openapi_client.ApiClient(configuration):
        api_instance = openapi_client.ModelApi()

        try:
            api_response = api_instance.parse_model_message(InlineObject1(text=message))
            return api_response
        except ApiException as e:
            print(f'Could not connect to Rasa')

Conclusion

We presented our work on chatbots during our Master degree in multiple steps going from necessity exploration to concept ideas and finishing with explanation on the implementation.

The chatbot prototype was developped by Michel Sahli and Kevin Pantillon.

Thanks for reading and discover more interesting stories of Empathic Labs

Empathic Labs

Patrick the packet, a chatbot to enhance your experience with package deliveries was originally published in Empathic Labs on Medium, where people are continuing the conversation by highlighting and responding to this story.

Hi, I’m Paul. During the 6th semester of my Bachelor’s degree in computer science at the High School of Engineering and Architecture in Fribourg, I had to carry out a project about gamification, in collaboration with the HumanTech Institute.

About the project

This project is the follow-up of the Master’s thesis of another student, Marco Mattei. The goal of his project was to develop a virtual reality (VR) app to help people suffering from specific phobias confront their fear. This was brilliantly reached with the creation of 2 separate 3D environments in which arachnophobes (who fear spiders) & resp. acrophobes (who fear heights) can move freely using an HTC Vive, while the therapist manages the world’s elements (spider appearance/behaviour or elevator altitude).

The goal of my project was to enhance Mr. Mattei’s work by adding gamification elements to it. This would allow patients to use the application autonomously while also inciting them to interact with their fear more & more. Note that the focus was solely on the arachnophobia part, the acrophobia treatment scene has not been gamified.

But, what is gamification?

A gamified application is not to be confused with a game. Gamification consists in adding game-design elements and principles to products whose 1st purpose is not related to a game. Thus, it shouldn’t take the user away from the use for which the application was designed (in this case, treating a phobia). Gamification simply helps to improve user engagement by keeping a user-centered approach.

How to apply gamification?

The early stage of this project was all about research and design. It was essential to first analyze similar products in order to collect information about what is efficient and what isn’t. Then, I had to identify the best practices to use when adding gamification to an application.

Making the application’s use autonomous

To make sure the app can be used without any assistance, I had to design a simple workflow and an intuitive user interface (UI). Multiple levels have been created with the aim to match milestones of a real exposure therapy. Each level is defined by a simple spider interaction objective (e.g. “Touch the spider with a tool”) and comes with 6 different difficulties (cartoonish or realistic spider, immobile or moving, alone or surrounded by other spiders).

The level selection is made through a UI displayed on a screen in the VR world, and the user has a pointer to click on the buttons. Filters are available to simplify the selection.

Encouraging user engagement

Common game design principles were applied in order to make patients want to use the application more. During each level, the user gain points based on the difficulty of the task and the duration of the interaction with the spider. This gives the patient a feedback on the progression of its ability to face its fear. Then, a medal is assigned based on the score. This makes the patient feel that its progress is being recognized, while also encouraging it to play the levels again and again to improve its score (with the aim of winning gold medals). In the same perspective, global achievements (who aren’t necessarily related with spider interactions) have been defined. For example, an achievement is won after having completed 25 levels, another after having won at least 1 gold medal.

Result

Now, what?

Even though the objective of this project was reached, some improvements could be made. First of all, user customization is lacking in this application. Indeed, this aspect is often mentioned in gamification’s best practices as it helps making the user feel familiar with the product.

Secondly, the diversity of the levels could be improved. The various interaction types and difficulties are nice, but after some time using the app, the tasks can become boring. This is mostly due to the fact that the spider is always located at the exact same place. Adding some randomness to the level “generation” would allow to avoid this problem.

Because of a lack of time at the end of the project, I couldn’t make proper user tests. The only feedback I could get was from my roommates and my girlfriend. They were however all very positive, as none of them had difficulties using the app (without any assistance) even though they’re not specially friends with IT and video games in general. It may be very interesting and constructive to test my work on real arachnophobes, and on a long-term basis to see how effective it is.

Thank you for reading my article! I hope it was interesting enough to awake your interest in gamification of mental health applications and in general :)

Gamifying a VR phobia therapy app was originally published in Empathic Labs on Medium, where people are continuing the conversation by highlighting and responding to this story.

Hi, my name is Jérôme Vial. I am a student at the School of Engineering and Architecture of Fribourg in Switzerland. My study plan include two semester projects and for the second one I chose to work on a sentiment analysis application in collaboration with the HumanTech Institute.

A quick summary about the project

This project is part of the Coléarnis project. Coléarnis is a platform created in collaboration with the School of Engineering of Neuchâtel. Coléarnis allows enterprise to create video in order to pass the knowledge from experienced employees to the other employees. Employees can then react to the video through comments and that’s where my project take place.

What are the project goals ?

This sentiment extraction application has many uses, but the main goal is to create a pipeline to extract the sentiment of video comments written in french. Sentiment analysis in french and on short texts are the two main constraints for this project.

During the project I reached theses two goals, it means I figured out how to extract sentiment on video comments written in french. To reach the goals I did research on existing technologies and I tested each of them. I figured out that using Google CNL and PyFeel would be the best options as Google CNL has strong results and PyFeel is developed especially for the french language.

Which technology and why ?

A big part of the project is research. I needed to find existing technologies that are working on short texts and with french texts. I found many technologies to perform sentiment analysis:

• Google CNL
• PyFeel
• Amazon Comprehend
• Azure Cognitive
• IBM Natural Language Understanding

All theses technologies are working with french texts, but after I performed some tests and compared them with each other I chose to use Google CNL and PyFeel. The main reasons are the fact that Google CNL has the best performances overall and the fact that PyFeel is a local library, which means no network data transfer.

Emoji’s importance in sentiment analysis

Since the project is to extract sentiment of video comments, I figured out that many comments contain emojis. In fact, those little yellow faces born in 1999 became the standard in text writing and in our way of expressing ourself. In 2019, 92% of the people use emojis! Emojis are really important in comments, they allow the user to express his feeling quickly and without having to write a lot of text. That’s why in the pipeline I added the library called Emosent to analyse emojis as well.

The pipeline, how does it work ?

The project has the shape of a pipeline, it means each part has a specific task. To begin, the first part is the pre-processing. It’s the part where the data are collected, cleaned and put in shape in order to facilitate the sentiment extraction.

The second part is the processing, it’s the part where the three technologies (Google CNL, PyFeel and Emosent) work together to extract the sentiment. I chose to extract the sentiment using Google CNL and PyFeel together to enforce the result.

After the extraction is done there is the last part, the analyzing part where I check the results and prepare the output. The output is mainly constituted of statistics like “80% of the users had a good feeling during this video.”.

What about the future ?

This project is an open door to many other projects. Even if the main goal is to extract sentiment of video comments, we can easily transform it into something else.

We could use this project and tune it to be able to create an app where the user drops a dataset, then the pipeline extracts the sentiments and finally the user manually check if the extraction is correct or not. Possibilities like this allows us to include human-in-the-loop process and also to upgrade our extractor to perform better.

I also recommend to create a real dataset that contains comments from the project platform. For this project I extracted comments on YouTube and I used existing dataset, but they poorly represent the comments that would be on the Coléarnis platform.

A self-reflection about my work

At the end of the schedule, I am satisfied because I reached all of my milestones in time and I reached the main goal with providing an application to extract sentiments of video comments written in french. I learned a lot during this project and it created an interest for this domain.

The main project is far to be done, but the purpose of this project was to do some research and test in order to help the development of the Coléarnis platform.

Thank you for reading my article and I hope I created interest in sentiment analysis !

Sentiment analysis of video comments was originally published in Empathic Labs on Medium, where people are continuing the conversation by highlighting and responding to this story.

Studying social behavior scientifically can be a difficult task: Real interactions between people typically lack standardization, but standardized experiments are often so artificial that their representativeness towards everyday-life situations remains questionable. New evidence suggests that the use of Virtual Reality (VR) can play a role in overcoming this dilemma.

Marius Rubo works as a postdoc researcher at the Clinical Psychology and Psychotherapy workgroup at Fribourg University (head: Prof. Simone Munsch). He studies how VR is perceived and how it can be used to improve the assessment of psychological phenomena such as social attention.

Studying social behavior scientifically: It’s complicated

The social behavior of humans is so multifaceted that its investigation spans over a variety of scientific fields such as sociology, social psychology, clinical psychology or behavioral economics. Even in an everyday encounter, people orchestrate their behavior in ways which are more complex than we can currently model: interpersonal distances are regulated, body postures communicate attitudes or mutual expectations, gaze contact is sought in some moments and avoided in others. A variety of professions — perhaps most intensely psychotherapists — are trained in systematically evaluating this rich flow of information and employing it to solve interpersonal problems. At the same time, the scientific analysis of social behavior does not live up to many practitioners’ expectations in providing tangible data regarding concrete problems (e.g.: What can and what can’t we infer when we observe that someone avoids gaze contact with us?). Scientists in the field, on the other hand, are confronted with an abundance of influencing variables which are difficult to control for statistically (e.g. individuals’ personality traits and relationship history, the social hierarchy expressed in a situation) and a methodological dilemma: When observing real human interactions in field research, a variety of potentially important variables may remain uncontrolled; by contrast, when employing standardized laboratory tasks (e.g. when participants view videos of social situations while their gaze is being tracked), it remains unclear how behavior in the laboratory relates to real social encounters.

Using VR to combine the best of laboratory and field research

Several researchers have speculated that the use of VR may allow to create a bridge between these two methodological approaches by immersing participants in situations which are perceived as more lifelike compared to the viewing of videos, but which can still be controlled at high levels of detail by experimenters. While generally still lacking broader empirical evidence, this idea is now backed by a recent study of ours (Rubo & Gamer, 2021). In this study, participants viewed an avatar either in VR or on a computer screen as it repeatedly approached them and sometimes smiled at them. Similarly to what one would expect in a real social encounter, participants in the VR group relatively robustly reciprocated the avatar’s gaze, while this behavior was clearly weaker in participants viewing the same scene on a computer monitor. The study therefore provides clear evidence that VR can constitute a methodological advancement over more traditional laboratory setups when investigating social behavior. At the same time, an abundance of open research questions persists (e.g. In which important ways does VR not resemble real social encounters?; What do individual characteristics in social behavior in such a VR scenario reveal about a person?) and will hopefully be addressed in the future.

Rubo, M., & Gamer, M. (2021). Stronger reactivity to social gaze in virtual reality compared to a classical laboratory environment. British Journal of Psychology, 112(1), 301–314. https://bpspsychub.onlinelibrary.wiley.com/doi/full/10.1111/bjop.12453

Virtual Reality may Enrich the Science of Social Behavior was originally published in Empathic Labs on Medium, where people are continuing the conversation by highlighting and responding to this story.

The Swiss Text Analytics Conference, June 14–15–16, 2021

Switzerland is a small country, but counts 12 universities and 8 universities of applied sciences. This implies a lof of researchers and a ton of research done.The fields of Natural Language Processing (NLP) and Artificial Intelligence (AI) are in vogue, in the research world of course, but also in companies in Switzerland and elsewhere. Large companies have been taking the lead for several years, but for some time now these technologies have been infiltrating even small and medium-sized companies (SMEs).

Therefore, it makes sense to have a conference dedicated to these topics, in order to highlight the research done in Switzerland, but also to encourage the transfer of these technologies in the industry.

Introducing SwissText

This conference is called SwissText and the 6th edition is happening soon, more exactly from 14 to 16 June 2021. For the second time, the conference will be held online.

SwissText is an annual conference that brings together text analytics experts from industry and academia. It is organised by the Swiss Association for Natural Language Processing (SwissNLP) in collaboration with the University of Applied Sciences and Arts Northwestern Switzerland (FHNW) as well as the Zurich University of Applied Sciences (ZHAW) and the data innovation alliance. SwissText 2021 is part of the Networking Event Series — Natural Language Processing funded by InnoSuisse, the Swiss Innovation Agency. It is supported by more than 10 universities and scientific associations. — Source: https://www.swisstext.org

The conference is composed of 4 main tracks, namely:

• The Applied Track: Focus on Swiss technology and its application in the industry. Where experts from academia and industry meet (17 papers).
• The Scientific Track: More technical research, done by international researchers (7 papers).
• The Demo Track: A way to showcase projects and prototypes (5 demos).
• The Highlights Track: A handpicked selection of high quality papers presented in top NLP conferences in the past year.

And there is more than the official tracks, because the conference features workshops and shared tasks. There are 3 shared tasks and 3 workshops:

• Task 1: Text Normalization for Swiss German
• Task 2: Sentence End and Punctuation Prediction in NLG Text
• Task 3: Swiss German Speech to Standard German Text
• Workshop 1: The European Language Grid
• Workshop 2: NLP efforts against COVID-19 in Switzerland
• Workshop 3: NLP in Finance

As you can see, the topics are oriented on the challenges brought by the linguistic diversity in Switzerland, on important current events as well as on societal issues, and the important and impactful advances in the NLP world.

The preliminary version of the program can be found here: https://www.swisstext.org/program

Each year, the conference highlights high quality researchers to give a keynote presentation. The 2021 edition features 3 experts:

• Sebastian Welter — AI Lead Architect at Accenture
• Lucia Specia — Professor of NLP at Imperial College London
• Lluís Màrquez — Principal Applied Scientist at Amazon Research in Barcelona

For more info about them and the conference in general, please check the official website: https://swisstext.org

As a side note, I’d say that the HumanTech Institute of the HES-SO Fribourg is proud to be a partner of the conference for the second year.

We really appreciate the way the conference is going. It is very professionally organized and the content is always interesting. Therefore, we really look forward to this year edition of the conference!

The registration is now open and the tickets are very affordable since it’s an online conference. Early-bird tickets are open until May 5, then the regular price apply.

Get your ticket now at https://www.swisstext.org/registration!

If you plan to attend, I’ll be happy to discuss with you about any NLP topic. Have a nice day, and see you there the 14–15–16 of June!
Jacky Casas

PS: Just click on the “follow” button to be notified when Empathic Labs publishes a new article, thanks! Or follow on Twitter if you prefer @empathiclabs 🐦

SwissText 2021 — NLP conference in Switzerland was originally published in Empathic Labs on Medium, where people are continuing the conversation by highlighting and responding to this story.

Digital psychotherapy (DPT lab) is where computer scientists and clinical psychologists team up to combine theories and expertise.

The lab develops and evaluates new technologies in order to improve the effects as well as the accessibility of evidence-based psychotherapeutic interventions.

Psychotherapy

People can struggle with challenges like coping with unfortunate events, quitting smoking, having a better lifestyle, improving relationships or dealing with a mental disorder. The challenges end up weakening the individual and worsening the overall content and satisfaction. Psychotherapy helps to raise awareness towards thoughts and behavioral patterns and to adjust them by ensuring conditions allowing trust, openness, neutrality and support.

Computer science

A computer scientist’s role is the creation, design, and implementation of software and hardware systems for the advancement of technology and innovation.

The collaboration

Computer scientists interested in psychology at the HumanTech institute joined forces with clinical psychotherapists at the University of Fribourg to create Digital Psychotherapy Lab, a group who’s mission is to develop and evaluate innovative therapeutic tools using novel technologies, enabling therapists to provide affordable and flexible state-of-the-art care to a larger number of affected persons. The group’s developments are rooted in fundamental research, comprehensively examining the benefits and difficulties in the use of each new therapeutic approach.

Areas of expertise

Virtual Reality

Virtual Reality, where users can immerse themselves into reactive artificial environments, have been employed in the treatment of specific phobias for more than two decades, but treatment programs targeting several other mental disorders are only now coming within reach. A key benefit of employing this technology is to allow patients to repeatedly experience problematic situations along the treatment process, relying on well-established therapeutical principles while expanding the possibilities for corrective experiences.

Some projects:

• VR for fear of spiders and fear of heights therapy

The fear of spiders (arachnophobia) and fear of heights (acrophobia) belong to the group of specific phobias. This term designates an irrational fear felt by the individual. It’s a highly frequent type of phobia that can cause severe troubles to affected people, as they will seek to avoid at all costs any contact with the concerned situation.

A way to overcome these phobias is called exposure therapy. This technique consists in letting the patient confront his fear, progressively so that fear decreases as the exposure increases.

This method may offer many challenging requirements for the therapist (e.g.: possessing a set of spiders, taking the patient to high places, etc.). A virtual reality app allows its user to be confronted with a spider whose appearance and size can vary, or to take an elevator in the middle of a city. The therapist always has the control of the spider (size, movements, appearance…) or of the altitude and sees the patient’s point of view in real time.

• VR Biofeedback for Eating Disorders — Training gastric interoception
• VR Cyberball

Chatbots

A chatbot is an artificial intelligence agent that can have a conversation with a user using messaging applications, websites, or through a phone call.

Some projects:

• Virtual assistant for teenagers with binge-eating disorders

Mobile apps

Digital media such as a mobile phone application have the potential to complement and improve existing psychotherapeutic services and to close existing gaps in healthcare services. The current COVID-19 pandemic also highlights the need to expand existing digital care services.

Some projects:

• App-based aftercare services for inpatients with psychological illnesses
• Daily questionnaire app

Websites

The advantage of online interventions (website, app) is that they can be used anonymously, independent of time and location on a smartphone, tablet or laptop. They are also less expensive. In this way, access to treatment is also made possible for socio-culturally disadvantaged or isolated patients.

Some projects:

• iBEAT

Many more projects and collaborations to come!

DPT Lab will be collaborating with Empathic Labs and using their blog to publish future projects.

Feel free to contact any of the lab’s members to discuss, collaborate or suggest ideas by using the following url: https://digitalpsychotherapylab.ch/team/

Fouad

Introducing DPT Lab was originally published in Empathic Labs on Medium, where people are continuing the conversation by highlighting and responding to this story.

by Marius Rubo, Andrea Wyssen, Felicitas Forrer & Simone Munsch

This project is conducted in collaboration between the Clinical Psychology and Psychotherapy workgroup at Fribourg University (head: Prof. Simone Munsch) and the HumanTech Institute (head: Prof. Elena Mugellini). Its purpose is to make virtual reality treatments for specific phobias easily available to therapists and patients.

Specific Phobias: Frequent and debilitating, but often respond well to treatment

How do you feel right at the top of a tower with the view on the abyss? For many people, the mere imagination of such a situation, let alone the real experience, provokes intensive anxiety. With a point prevalence of around 3% height phobia is one of the most frequent specific phobias. Less severe manifestations of this mental disorder are even much more frequent and affect up to one quarter of individuals in community samples.

Specific phobias tend to generalize, which means that people who suffer from intense anxiety while standing on a high tower may later also experience anxiety when driving over a bridge for example. If such situations are avoided, phobias may therefore strongly affect people’s daily lives (for instance if someone drives a longer route to work every day in order to avoid a bridge).

Fortunately, phobias can be treated effectively within a few therapy sessions, where individuals, after a preparation phase, learn to expose themselves to their phobic anxieties in real situations, but it is well known that treatment in virtual reality (VR) is actually similarly effective. After exposure training, therapists offer support to transfer behavior change from the laboratory to daily life in order to decrease the individual’s safety behaviors and avoidance.

Novel approaches to therapy strive for easier dissemination

In recent years, the traditional face-to-face psychotherapy is more and more often combined with other techniques such as internet-based or technique-supported interventions (so-called “blended treatments”). In treating specific phobias, exposure in VR are especially advantageous. Exposure trainings in VR can be conducted repeatedly, independent of time and location, in varying intensity and difficulties. In addition, they allow the exposure to situations which are difficult to plan (e.g., height exposure on a mountain peak). Moreover, the therapist can fully control the situation. Once therapists are familiar with the application of the VR setting, these interventions are less time consuming. Nevertheless, dissemination of such VR tools in clinical practice is still unsatisfactory. Therefore, the present project aims at developing and distributing an easily applicable VR intervention to treat fear of height.

In the initial phase of the project, we develop software and test its feasibility, acceptance, and efficacy in a sample of patients at the outpatient psychotherapy unit at the University of Fribourg (Psychotherapeutische Praxisstelle). Later, we will provide the novel therapeutic tool to interested members (psychotherapists) of the Swiss Society for Cognitive-Behavioral Therapy. The complete treatment program includes four standardized face-to-face, email- or internet-based sessions with a guideline for preparation, performance and postprocessing of the exposure sessions. It contains the software for the VR exposure intervention addressing height phobia (with different levels of intensity and difficulty) and a tutorial for its application. The outcome of the VR-supported treatment will be evaluated.

We thank the Swiss Society for Cognitive-Behavioral Therapy for the financial support of our project and Marco Mattei for his work on the project.

“Face-to-face” with your fears was originally published in Empathic Labs on Medium, where people are continuing the conversation by highlighting and responding to this story.