Imagine asking your virtual assistant, “Hey Google/Alexa, tell me the lyrics to ‘Beautiful People’ by Ed Sheeran.” Voice User Interface You could possibly receive the information you need within seconds. Cancer doctors and researchers face the challenge of exploring and interpreting cancer genomic data, which resembles a huge library with billions of pieces in different categories. What if you had an Alexa-like tool that could answer questions about the data within seconds?
Traditionally, researchers have used computer programming and interactive websites with point-and-click capabilities to analyze cancer genomic data. Researchers agree that these methods are not only time-consuming, but also often require advanced technical knowledge that not all clinicians and researchers possess. Scientists from Singapore and the United States have collaborated to develop a conversational virtual assistant to navigate the vast library of cancer genomes. They named this assistant Melvin. Their goal was to make relevant information quickly available to all users, regardless of technical expertise.
The scientists described Melvin as a software tool that allows users to interact with cancer genomic data through simple conversations with Amazon Alexa. It incorporates familiar Alexa features, such as the ability to understand and speak everyday English and the ability for researchers to initiate a conversation by saying the name “Alexa.” Additionally, the scientists incorporated a knowledge base containing genomic data for 33 types of cancer from a global cancer database. The Cancer Genome AtlasIt contains a variety of data, including gene expression data, mutations known to increase the risk of developing cancer, etc. It also incorporates secondary information from each database, such as the definition and location of human genes, protein information, and anti-cancer drug efficacy records, to help users effectively interpret the results.
The scientists collected nearly 24,000 pronunciation samples for cancer genes, cancer types, mutations, types of genomic data, and synonyms of all terms in these categories from nine cancer experts at the Cancer Science Institute of Singapore. These experts were from Singapore, Indonesia, Sri Lanka, the United States, and India, which was needed to increase the diversity of Melvin’s accents. The scientists said that due to the lengthy data collection time, the pronunciations did not cover all known cancer genes and traits.
The scientists explained that a voice user interface works well if it correctly hears and understands the user, including the context of the conversation. Because cancer terms differ from regular English vocabulary, the researchers trained Melvin to learn cancer vocabulary using a machine learning process that gives meaning to previously unknown words. Out-of-Vocabulary Mapper Service Design.
Additionally, the researchers developed a web portal where users can submit pronunciations of certain cancer features that Melvin may not initially recognize. This will allow Melvin to know what the user means when he hears those words. To address users’ potential security concerns about the recordings, the researchers noted that users can avoid data storage by deleting the recordings by following the instructions in their Amazon Alexa account. The researchers discussed opportunities to expand Melvin’s capabilities through crowdsourcing for pronunciation improvements. The researchers hope that these pronunciations will provide more data to match regional and national accents so that Melvin can understand and speak.
The scientists say Melvin will work with any device that supports Alexa and will be able to ” Gene Name” and “What percentage of lung cancer patients have a mutation in that gene?” Melvin reported that within seconds it processes these questions and returns responses in audio and visual form.
They also reported being able to ask follow-up questions based on previous conversations. They described the difficulty of getting valuable information from a single question and highlighted the value of Melvin’s ability to maintain context through incremental questioning. The scientists asserted that this design makes it easy for users to explore multiple relevant questions in a single conversation. They also demonstrated that Melvin performs advanced analytical tasks, such as comparing mutations of specific genes across different cancer types and analyzing how gene expression changes.
The scientists concluded that MELVIN can accelerate scientific discoveries in cancer research and help translate research results into solutions that clinicians can apply to patients. They acknowledged that while MELVIN’s framework is currently centered on cancer genes, it can be expanded to support more characteristics of cancer. The team plans to enhance MELVIN by adding more valuable datasets and features based on user feedback..
Post View: 203
Source: sciworthy.com
