# Digital Avatar for Cancer Treatment Silk Data took part in an ambitious and promising project regarding the usage of avatar technologies in healthcare. We applied our vast expertise in **[NLP-based development](http://silkdata.tech/natural-language-processing)** and just a couple of months later deployed a digital avatar able to guide through the latest data on cancer treatment. ## Challenge The customer was about to take part in American Society of Clinical Oncology 2025 annual meeting. The meeting regarded the current state of the cancer treatment. Their main request was to create a full-fledged digital avatar that would inform those interested in the cancer Treatment. The avatar was planned to attract the users with special conversation hooks (i.e. invitation phrases) and then answer their questions on the key aspects of cancer and its treatment. ## About the Project Our client is a respectful US-based cancer treatment clinic, a member of American Society of Clinical Oncology (ASCO) who has decades of experience in treating cancer. The research data and expertise info provided by the client became the basis for creating the logic and working principles of the future avatar. ###### Region USA ###### Time 2 month ###### Team 1 specialist ## Solution Work on the visual avatar was divided between Silk Data and our partner. They took responsibility for everything connected with ‘text-to-speech' and ‘speech-to-text' technologies along with making the avatar look the way it should. Our team, the contrary, worked on technical part, which included coding, testing and integration of the final product. In addition, the project required understanding of the key principles of the client company’s ideas and mechanisms they wanted to see in their digital assistant. ## The Project Specification Both our company and our partner were provided with a set of client instructions and initial wishes, which became the basis for further project specification. The specification contained the information about the main aspects of the future avatar’s working principles. ### Avatar profile The avatar or the **Clinical Trial Patient Ambassador**, named **Reggie**, will be an older adult male representing a Patient Ambassador group that offers input on clinical trial design. ### Assumptions The avatar will be speaking out into the conference space to ‘hook’ passersby into conversation. Once engaged (after a delegate starts speaking to Reggie), the avatar will **prompt the delegate to select specific cancer types** from the installed iPad. ### User interaction specifics The avatar will provide information and options for further conversation. Delegate questions and statements will be recorded and mapped to prescribed responses. In addition to these key phrases, Reggie should also be able to respond to random questions from delegates. However, **avatar will only discuss pre-identified topics** and will redirect delegates from non-approved themes. ### A discussion summary The results of the conversation between the conference delegate and the avatar will appear on a screen as it occurs, so the delegate can read the information if needed. **Note.** All the conversation lines provided by the avatar are pre-scripted. As most of the avatar answers require a high level of knowledge and expertise regarding both the healthcare industry and cancer treatment topics, the scripts were provided by the client company. In essence, the solution is an AI chatbot that was smoothly integrated into a hologram. ## Reggie’s conversation initiation There are also three key aspects regarding Reggie’s conversation initiation. ### Hooks The Avatar was programmed to use three conversation hooks to draw in delegates to converse with him, when not being actively engaged in the conversation. They are the following: The avatar should switch between these three hooks every 20 seconds. ### IPad prompts The installed iPad will prompt delegates to select one of the following categories of cancer trials: - Non-small cell lung cancer - Head and neck squamous cell cancer - Prostate cancer - Colorectal cancer - Bladder cancer - Solid tumors Each category has a few trial options, and Reggie will be able to talk about each of them. In addition, **if there are multiple trials in the category, the delegate will need to select which trial they want to know more** about on the iPad. If there’s only one trial in the category, **the avatar will immediately start talking about it** without any intermediate stage. ### Guard rails The client also wished to preserve the avatar from dealing with off-topic talks. For that purpose, a special line was created: **‘That’s an interesting topic but what I really want to discuss is how clinical trials are shaping the future of cancer treatment’** The avatar will provide this line to any question that doesn’t refer to the pre-scripted categories. Furthermore, if the user’s question refers to the cancer treatment topic, but goes beyond the dozens of pre-scripted questions and responses (like ‘Can you explain the science behind this?’ or ‘How do I connect with a trial site?’), the virtual avatar’s answer will be the following: **‘This question is better answered by a representative at the Medical Information Desk. I can tell you about \[list of remaining topics\]’** ## The Conversation Map ### The conversation body One of the main aspects of Reggie digital avatar’s work is the conversation map – a conversation system of topics, questions and responses set and fine-tuned within the avatar. At the specification stage only three notes were defined: #### Avoid repetition Reggie should know not to ask the same question about the same trial twice in one conversation. #### Pre-scripted responses The avatar should provide word-for-word pre-scripted responses to prevent misunderstanding or false information issues. #### Pronouncation support There are words anticipated to be difficult to pronounce, so they will have accompanied pronunciation audio. As we mentioned before, there are six main categories of cancer trials, and each category contains 1 to 5 trials to offer. Then, there are 8 key topics categories regarding each of the trials. These are: - Investigational drug. - Disease state. - Phase. - Trial design. - Recruitment locations and number of sites. - Inclusion criteria. - Exclusion criteria. - Endpoints. The avatar can percept up to a dozen questions (either direct or paraphrased) referring to each topic. However, only one answer is prepared for each topic in each trial. This was made **with intention to avoid possible ambiguity in responses.** ### The conversation trail There were also three main rules defined and implemented as the conversation trail functionality. - **Ending each answer with a trail.** Every avatar’s response regarding a particular trial would be ended with the sentence ‘Would you like to know anything else about \[trial name\]?’. - **Non-response reaction.** If the avatar gets no reaction to the first question another line is on: ‘I could tell you more about our other trials. We have trials for patients with \[lists disease states\]’. - **Reaction to non-specific questions.** If the user provides a positive answer but gives no specification, the avatar says ‘What would you like to hear about \[trial name\]? I can tell you about \[remaining topic 1, remaining topic 2, ... or remaining topic n\]’. ## Results and Conclusions At the end, the final product aligned with all the initial requirements. The avatar was successfully deployed, implemented and used during the American Society of Clinical Oncology 2025 meeting. Reggie provided consultation and assistance for hundreds of visitors for the week the conference lasted. ## Our Expert "This project clearly illustrates that advanced digital avatars are the reality that would gain more and more popularity. Now they answer your questions on the pre-scripted scenarios, and ten years later they’ll be following you everywhere. Sure, such a prediction is too optimistic, but the research made in one of our blogposts dedicated to AI avatars shows some prominent examples that already exist in retail, manufacturing and entertainment industries. We believe that further innovations are only a matter of time" Yuri Svirid, PhD. — CEO Silk Data ## Frequently Asked Questions ### What is a digital avatar? A digital avatar is a graphical representation of a real person or character built and deployed within a digital environment. Though the term can be referred even to simple 2D icons, most people refer to digital avatars as intricate, interactive and customizable 3D models used in movies, video games or special software. ### How to create a digital avatar? Creation of an advanced digital avatar is a rather complex process. It involves using specialized software for modeling, sculpting, rigging, motion and voice capture and much more. The most advanced avatars can also leverage machine learning technologies for self-learning and adaptation. All these abilities can turn a digital avatar into real digital AI assistants. ### What are digital avatars used for? Digital avatars are used across numerous applications. They represent users in video games, virtual reality meetings and social media platforms. In addition, both primitive and advanced digital avatars are crucial for the metaverse, virtual collaboration and digital storytelling. These days they also have growing use cases in AI and robotics as the visual interface for autonomous systems. ## Project Info ## Have a Look at Our Related Projects [Plagiarix](https://plagiarix.com) [Machine Learning](http://silkdata.tech/case-studies/machine-learning) [Plagiarix](https://plagiarix.com) [Machine Learning](http://silkdata.tech/case-studies/machine-learning) [ Plagiarism detection system is specifically designed for proctoring providers and users, LMS providers and integrators, as well as universities and educational organizations. With our advanced technology, you can process several hundreds of exam scripts in just a matter of hours. ](https://plagiarix.com) [AI-assisted Search](http://silkdata.tech/ai-assisted-search) [Machine Learning](http://silkdata.tech/case-studies/machine-learning) [AI-assisted Search](http://silkdata.tech/ai-assisted-search) [Machine Learning](http://silkdata.tech/case-studies/machine-learning) [ At our company, we are always on the lookout for the latest and greatest technology to help our clients achieve their goals. Our AI-assisted text search system takes text search to the next level. Unlike traditional text search systems that rely solely on keywords, our system takes into account the semantics, or meaning, of the text. This means that the search results are not only more accurate, but also more relevant to what the user is looking for. Our AI-assisted text search system uses advanced algorithms to analyze the text and identify the most relevant results, even if the keywords used in the search query are not an exact match. ](http://silkdata.tech/ai-assisted-search) [Contract Analysis](http://silkdata.tech/ai-contract-review) [Machine Learning](http://silkdata.tech/case-studies/machine-learning) [Contract Analysis](http://silkdata.tech/ai-contract-review) [Machine Learning](http://silkdata.tech/case-studies/machine-learning) [ The AI Contract Analysis web interface leverages the power of Artificial Intelligence and advanced algorithms to analyze contracts, agreements, and other legal documents. It can quickly identify key terms, clauses, and provisions, and provide a comprehensive summary of the document. This can help businesses save time and reduce the risk of errors or oversights that can lead to costly legal disputes. Our AI Contract Analysis web interface is user-friendly and easy to use, making it accessible to businesses of all sizes. ](http://silkdata.tech/ai-contract-review) [Text Summarization](http://silkdata.tech/text-summarizer) [Machine Learning](http://silkdata.tech/case-studies/machine-learning) [Text Summarization](http://silkdata.tech/text-summarizer) [Machine Learning](http://silkdata.tech/case-studies/machine-learning) [ Text Summarization is a powerful tool that can help companies extract valuable insights from large volumes of information. By using Artificial Intelligence and advanced algorithms, our tool can condense lengthy, complex texts into shorter, more concise summaries that are easier to understand and digest. The benefits of Text Summarization go beyond simply saving time and reducing the amount of information that needs to be reviewed. This can lead to more informed decision-making, better customer insights, and improved product development. ](http://silkdata.tech/text-summarizer) [Semantic Map](http://silkdata.tech/semantic-map) [Machine Learning](http://silkdata.tech/case-studies/machine-learning) [Semantic Map](http://silkdata.tech/semantic-map) [Machine Learning](http://silkdata.tech/case-studies/machine-learning) [ At Silk Data, we pride ourselves on using the latest technology to enhance the way we work with data. Our Semantic Map is a game-changer when it comes to making sense of large volumes of text data. With this technology, we are able to automatically identify the relationships between texts, map them to semantically related keywords, and then create a visual representation of the data. This means that you can quickly and easily understand the relationships between different pieces of information, and make informed decisions based on the insights you gain. ](http://silkdata.tech/semantic-map)