Robot Videos

Furhat talks about hypertension, exercise and diabetes (2025, 4 minutes)

A Furhat robot talks about hypertension, exercise and diabetes using a knowledge graph in a Neo4j database and an OpenAI LLM. The knowledge graph was constructed by LLM using LLMGraphTransformer (LangChain/Neo4j). The dialogue system uses vector and graph retrieval (GraphRAG) from PDF documents to ground the LLM.

Furhat talks about restaurants using ChatGPT (2023, 7 minutes)

A Furhat robot talks about restaurants in Tokyo Waterfront using ChatGPT. In comparison with responses by CityTalk based on knowledge graphs in databases, responses generated by ChatGPT are more friendly and natural, but the provenance of the data is unknown and ChatGPT responses sometimes include false information. 

Bilingual responses in Japanese and English (2023, 3 minutes)

A Furhat robot gives bilingual responses in Japanese and English using Rasa conversational AI. CityTalk answers queries about hotels and restaurants in Tokyo Waterfront by searching knowledge graphs in Neo4j databases. To avoid unnatural switching between different voices, the same Japanese voice speaks both Japanese and English responses.  The video was shown at Japanese Society for Artificial Intelligence (JSAI 2023), Kumamoto, Japan, 2023.

Better responses with knowledge graph metadata (2022, 3 minutes)

A Furhat robot gives more intelligent responses using knowledge graph metadata. Asked about restaurants that serve European cuisine, CityTalk  uses subclass relationships in a taxonomy of cuisines and finds restaurants serving Italian cuisine. Asked about Asian cuisine, CityTalk finds restaurants serving Japanese, Chinese or Thai cuisines. The taxonomy was extracted from WikiData.  The video was shown at International Conference on Robot and Human Interactive Communication (RO-MAN 2022), Napoli, Italy, 2022.

Conversational AI and knowledge graphs (2022, 4 minutes)

The first demo of CityTalk on a Furhat social robot. Using Rasa conversational AI, the Virtual Furhat robot answers queries about restaurants and hotels in Tokyo Waterfront by accessing knowledge graphs in an external Neo4j graph database. The video was  shown at International Conference on Human-Robot Interaction (HRI 2022), Sapporo, Japan, 2022.

ERICA and WikiTalk (2019, 8 minutes)

A demo of ERICA and WikiTalk in the ERATO Ishiguro Symbiotic Human-Robot Interaction project at Kyoto University. In the video ERICA talks about ancient Greek, robots, artificial intelligence, intelligence in medieval philosophy, and androids. The demo was shown at International Joint Conference on Artificial Intelligence (IJCAI 2019), Macau, China, 2019.

CityTalk and Tokyo 2020 (2018, 10 minutes)

A robot gives information about hotels and restaurants in the Tokyo Waterfront area near the Tokyo 2020 Olympics. The robot tells about three hotels in different price ranges. When asked about restaurants, the robot switches domains and talks about several restaurants in different price ranges and serving different types of food. The video was shown at International Joint Conference on Artificial Intelligence (IJCAI 2018), Stockholm, Sweden, 2018.

English and Japanese WikiTalk (2015, 14 minutes)

A demo of WikiTalk's multilingual capabilities. The robot speaks English with an English speaker and Japanese with a Japanese speaker. It talks about Shakespeare and other topics using both English and Japanese Wikipedias, switching to Japanese when the Japanese speaker says Nihongo and back to English when the English speaker says English. The video was shown at SIGDIAL Conference on Discourse and Dialogue (SIGDIAL 2015), Prague, 2015.

Finnish WikiTalk (2015, 12 minutes)

A demo of Finnish WikiTalk in a domestic setting. The robot can speak English and Finnish, and talks in Finnish about a wide range of topics using Finnish Wikipedia. As well as making smooth topic shifts to related topics, the robot switches to a completely unrelated topic (a so-called awkward topic shift) when the user spells the first part of the name of the unrelated topic.

Furhat bilingual CityTalk first demo (2023, 4 minutes)

The first demo of bilingual Japanese/English CityTalk on Furhat robots. To avoid switching between two different voices, a Japanese voice speaks both Japanese and English responses. In this first demo, some mixed-language responses are produced. Digits are spoken as Japanese numbers when speaking English, and some English terms (such as price range) are used when speaking Japanese. This is fixed in a later demo. 

Furhat and WikiTalk first demo (2021, 13 minutes)

The first demo of WikiTalk with a Furhat robot. A Virtual Furhat talks for more than 10 minutes about social robots, Romeo and Juliet, Hamlet, Manchester United, Old Trafford, and artificial intelligence. The video was made for #FurhatChallenge 2021.

CityTalk Tokyo 2020 first demo (2018, 2 minutes)

The first demo of CityTalk with new hotel and restaurant databases for Tokyo 2020. The robot tells Dr Kristiina Jokinen of AI Research Center, AIST Tokyo Waterfront, about restaurants in the Tokyo Waterfront area, near the Tokyo 2020 Olympics and Paralympics.

CityTalk first demo (2018, 4 minutes)

The first demo of CityTalk. The robot tells about hotels and restaurants in Cambridge. When the person asks about hotels, the robot clarifies which part of town and what price range is suitable, then suggests a hotel using a local hotel database. When the person asks about a restaurant, the robot switches domains and asks which part of town and what kind of food they would like, then suggests two restaurants using a local restaurant database.

SamiTalk first demo (2016, 4 minutes) 

The world's first Sami-speaking robot talking about the situation of Sami languages in Finland. SamiTalk is a robot application that tells information from Sami Wikipedia in the North Sami language. This prototype was developed at University of Helsinki for the Academy of Finland DigiSami project. The video was shown at International Workshop on Spoken Dialogue Systems, Saariselkä, Finland, 2016. 

MoroTalk first demo (2015, 4 minutes)

MoroTalk is a news-reading robot application that tells local news from online news feeds. The application was developed at University of Helsinki for the future elder-care technology training unit of the SASKY Municipal Education and Training Consortium in Sastamala, Finland. The video shows the first demo of MoroTalk. The robot tells local news in Finnish about Tampere city in Finland using online news feeds of the local Aamulehti newspaper.

WikiTalk first demo (2012, 9 minutes)

The first ever demo of WikiTalk on a Nao robot at the International Summer Workshop on Multimodal Interfaces in Metz, France in 2012. Although the speech recognition on this older robot was sometimes difficult due to low confidence scores, the demo showed that WikiTalk could support successful information access dialogues with the talking robot. This version of WikiTalk was for English only.