SolcEllen
Solar Panels Information Assistant
Chatbot
https://www.rvn.se/sv/Utveckling/Miljo/energikontoret-vasternorrland/solceller/
Region Västernorrland & Västernorrland, Sweden
region.vasternorrland@rvn.se
(0046) 0611-800 00
Service Skills
Solar Panels Installation
Solar Panels Efficency
Costs
Languages
English
Swedish
Technical Resources
Check our GitHub project
Find the code and instructions to run a chatbot
Data
During the chatbot development process lots of data was collected and created. We are making it available for others to train their AI assistants.
Check it here.
Other useful resources we created
Lorem ipsum dolor sit amet, consectetur adipisicing elit.
Benefits of the Service
More efficient customer service
The objective of this service is to provide comprehensive, understandable information about solar energy and solar panel installations: information that is often spread through different online sources and not easy to find. Its aim is also to facilitate and encourage the contact with local energy advisors in the region who can provide individual, cost-free advice – the service that many inhabitants in the region are not aware of. By providing information and support, the service would encourage people to invest in solar panels and contribute to the development of solar energy in the region. Hopefully, the chatbot will inspire and encourage investing in renewable energy in the region.
Better use of internal resources
By using a Chatbot to provide information about solar energy and solar panel installations and as an advisory service information and advice becomes more available for the main user group. Both by making information and advice easier to find, but also by making existing local energy advisors more available when sources of information and advice diversify. A chatbot can therefore result in better use of resources when the advisors are assisted in their task to provide advice to the community on how to be more energy efficient.
Address Sustainability and Climate Change Issues
By making information about solar energy and solar panel installations more comprehensive and understandable through the service, and by assisting local energy advisers in their work this service can be of help to facilitate the expansion of solar energy and an additional push for clean electrification.
Targeted Groups
Marketing and Promotion
The service has been promoted on Region Västernorrlands social media, such as Facebook, Linkedin, Instagram, and on Region Västernorrlands webpage. The service has also been promoted during Demo sessions.
Development and Implementation Process
Development Process
Development of the chatbot
1. Developing scenarios – initially 16 scenarios with different questions and answers were developed by regional energy advisors’ group. Scenarios covered such topics as: installation, costs, direction of solar panels, their sustainability etc. In the following phases scenarios were further developed and new ones were added.
2. Preparing a survey for testing – an evaluation survey was developed online (Microsoft Forms, in Swedish) to check how the user interacts with the chatbot, what is their general feedback and to correct errors. It focuses on exploring the existing scenarios (not creating new scenarios) and checks how (interactions flow looks when following certain scenario.
3. Internal and external testing, including focus groups and survey feedback collection.
4. Developing two language versions – the initial chatbot (16 scenarios) was developed in Swedish. After first testing phase and first round of improvements the text was sent for professional translation. After that the Swedish version was developed that was used for further testing. Following changes and updates were made consecutively in both languages. The user can switch between the languages during the conversation. Testing of the chatbot
1. I internal testing – December 2020
This testing was done among people working at RVN Regional Development Dept. 6 people from RVN took part in testing and 2 energy advisors – in total 8 people. Participants were given questions to interact with the chatbot and were asked to later fill in an evaluation survey. The feedback was evaluated and chatbot improved accordingly.
2. II internal testing (January 2021) was done with the same group and the mistakes identified in the phase I were reviewed.
3. III external testing – January 2021 Chatbot went through the external testing. Focus group was organized (7 participants) and chatbot was sent to test among wider public with the help of the online survey (13 participants). Focus group was organized with energy and climate advisors in Västernorrland, who are main stakeholders in the pilot and have most knowledge about the topic. During the online meeting, the different scenarios were reviewed, and multiple comments were given.
4. IV internal testing – February 2021 – the updates made after the external testing were reviewed and checked internally and further comments for improvement were made.
5. V external testing – March 2021
4 March a focus group with energy advisors and NUIG was organized (in English). During the meeting the energy advisors commented on the updated version of the chatbot and left further comments. The second stage of this phase was a focus group with the students from the Mid-Sweden University. This focus group was organized on 4 May and 4 students and 1 university lecturer took part. The focus group was held in English and the partners from NUIG participated as well. Additionally, survey feedback was collected from several partners identified by the project coordinator in the region.
Technical Requirements
The chatbot requires a server and platform such as a website to be hosted on. User will need Wi-Fi or mobile data to access the chatbot.
Technology Used
The chatbot is a form of artificial intelligence (AI) used in messaging apps. This tool helps add convenience for customers—they are automated programs that interact with customers like a human would.
Challenges and Recommendations
Most common challenges in developing the chatbot:
Chatbot not understanding questions: the most common problem during the testing was that the chatbot did not understand questions in the way they were formulated. This was a challenge especially in Swedish as the algorithms in Swedish are not so developed to teach chatbot to understand different versions of a question. It happened that even small differences in the text of the question (e.g. a different word) resulted in chatbot not understanding it. The challenge was overcome by adding as many alternatives as possible and update the questions database with the results of the testing (when new questions appeared). The NUIG partners implemented technical solutions for this challenged to and managed to improve chatbot’s ability to understand questions. During the focus group 4 may it was proposed that the chatbot could suggest some alternatives (Did you mean …?) – this solution was further tested in chatbot’s development.
Vague or difficult answers: the testing showed that some answers were written in a technical language that was difficult to understand by the general public. The challenge was overcome by consulting the text with the solar energy expert and the communication expert (we are still working on it).
Problem with algorithms in Swedish (not existing knowledge base for this language): this was a technical problem that occurred in the Swedish version. NUIG was trying to adapt some technical solutions for the Swedish context in order to overcome this challenge.
Lack of people to consult the chatbot (interest groups, stakeholders’ groups): it was difficult to involve new groups in testing. Persons interested in solar panels are rather limited group of users (homeowners with suitable budget) so it was difficult to reach this specific group. Project coordinator tried to reach different groups of users (homeowners’ associations, etc.), but faced very little interest or total lack of response. As a solution some online testing was done with the smaller group of users and testing with key stakeholders was done in few rounds in order to check the chatbot properly. The 3 focus groups were organized in total: 2 with the local energy advisors and 1 with the Mid Sweden University. A recommendation is to build a group of potential users / stakeholders very early in the project development to build commitment and ownership, so that those groups feel more obliged to participate in the testing on the later stages.
Lessons Learnt
Useful tips in developing a chatbot:
– Test it with many, different user group. Conduct few rounds of testing and keep track of all the updates.
– Document each round of testing and track all the changes that are made in the chatbot (write down and document each change or comments that come up under testing).
– Build as big database of questions as possible. Especially if English is not your main language – spend some time on proposing as many ways of putting a question as possible. Doing it at the early phase will spare you time on improvements during the testing phase. When possible try to have at least 5 versions, even more.
– Check the text with the communication expert – it will let you adjust the language to the general public.
– People like to have pre-selected questions to choose from – it is a good idea to have a list with sample questions in a chatbot, but also give an opportunity to formulate their own questions.
– Build a network of potential users / stakeholders at the very early stage of the project. Try to include different users group and build ownership and interest over the project so that the groups are willing to test the service and participate in its development.
– Visual aspects are as important as the content – think of colors, avatar, introductory sentences etc.
– Include a sentence at the beginning about handling personal data – mind users not to write personal data.