REVELIO
AI Driven Voice Intelligence Enhancement Interface
SETTING THE SCENE
Have you ever been captivated by a CIA thriller on the big screen? Undoubtedly, you've witnessed those suspenseful moments when a CIA agent needed to discreetly place a bug within the antagonist’s hideout. Yet, have you wondered about the vast troves of audio data from those bugs, potentially spanning days, which offer these agents with vital insights into the antagonist's intentions and future actions? Enter the world of Voice Language Analysts, who diligently decipher and scrutinize such audio data on a daily basis, unearthing critical information essential for our nation's security.
About Revelio
Introducing Revelio, a user friendly interface designed to empower Voice Language Analysts gather robust and reliable intelligence. This interface aims to help users triage, visualize and analyze audio data that contains pivotal intelligence for the nation's security and well-being. By collaborating this interface with Artificial Intelligence, Revelio offers analysts the means to uncover the broader context and enhance their awareness of current events within their specialized domains.
Group Project in collaboration with Laboratory of Analytic Sciences.
Members - Diksha Bahirwani, Isha Parate, Kayla Rondinelli
PROJECT TYPE:
February 2023 to April 2023
DURATION:
Figma
Adobe After Effects
Procreate
Miro
iMovies
SOFTWARE USED:
While it was a group project, we discussed the project as a group but worked on our own solo concepts. The best features of each concepts was later combined into one final interface.
Questions for User Interviews and Note takingBuilding Persona and As-Is ScenarioWireframing and Sketching 3 Solo ConceptCreating Low Fidelity Interface of 1 Solo ConceptTurning Lo-Fi into High FidelityCombining best features of other group member concepts into one and solely sketching out final workflowWorking on Hi-Fi final interface screensWriting Script for AudioSolely creating Animations and Interactions for a 5 minute Scenario Video in Figma
MY ROLE:
PROJECT SPOTLIGHT
Featured in an article by the Laboratory of Analytic Sciences, highlighting their collaboration with NC State University. The article delved into my perspectives and experiences concerning this significant project.
RESEARCH QUESTION
How might the design of an interface use the affordances of Machine Learning to enable Voice Language Analysts to quickly produce robust and reliable intelligence that accurately conveys content, intent and context?
ASSUMPTIONS AND LIMITATIONS
Due to the nature of this project being secretive, no real names or pictures of people, countries or softwares have been used throughout the project. We were also assuming the existence of a country called “Kobia” who’s president was Richard Nixon. There is also a neighboring country called “Zoolandia”. Our persona, Sloane, is based out of the Kobian office and is not an expert with the language of Zoolandia. We also used Richard Nixon’s White House tapes as our audio dataset for this project assuming it was not in English language.
Since the interfaces and softwares used by Voice Language Analysts are confidential, we based some of our features on the answers provided during User Interviews.
PERSONA OVERVIEW
The role of a Voice Language Analyst was divided into three positions - Scanners, Translators and Quality Controllers(QCer’s). Scanners are responsible for listening to the audio files in their original language and understanding what cuts from the audio files are important while putting together the bigger picture of what threats the country could be facing. Translators are in charge of translating the important cuts highlighted by the scanners and create a report. These reports are passed through the QCer’s desk to make sure everything is correct before heading out to the concerned parties.
PERSONA
My group was responsible for creating an interface for the role of Scanners and understanding the affordances Artificial Intelligence could bring to such a system.
AS-IS SCENERIO
IDENTIFIED PAIN POINTS
Cognitive overload - switching between 19 different tabs
Struggles to prioritize cuts relevant to the storyline
Unable to efficiently overview audio cuts and access content
Needs more than 2 screen to see relevant information at the same time.
SCENARIO VIDEO
WIDER IMPLICATIONS
Individualized workspace supports and directs scanner workflow
AI anticipates and serves useful content to language analysts as they move through each phase of their workflow
Audio Visualization allow analysts to overview and leapfrog through essential information.
REFLECTIONS
How could the AI generated visualization system assist with bundling and sending multiple files?
What else might the AI decent in audio beyond emotions, non speech sounds and fidelity?
Process Work
CONCEPT 1 THEME: ML SCREEN LAYOUT
What pain points is this addressing?Solves issues of cognitive overload by only having important stuff on screenAbility to have a customizable monitor display/turn off these suggestionsAssists in jumping back into work after a break/new day because the stuff they see is reflective of the most important taskThe most urgent issues will be reflected in the visual hierarchy of the screen layout as the ML sees fit
How will this unfold over across the whole workflow?ML is making it easier for the scanner to remain focused and keyed into what they are doing so they can be more efficient throughout their whole workdayEye tracking helps users focus their visual attention on what they engage with where the ML could then make certain tabs or layouts better for the user based on thisSuggests what is most important based on the queue that the whole team hasHas the tabs open that are predicting what you would need nextIt could reflect what the other team members are doing -QC-er comes across a really important incoming news and Sloane's ML Screen Layout could pop up that tab etc.Organize the queue based on what other team members are viewing
CONCEPT 1 : BASIC SKETCH
CONCEPT 2 THEME: ML IDEA GENERATOR
What pain points is this addressing?This addresses not knowing what to look for/helps to jump into the day quicklyThis will help in instances where users don’t know much background on a topicA better version of a word cloud tool (this one is organized with regards to the importance/gender of the speaker- the more important their keywords will look)It could organize data by emotion and predicted reliability of infoIt could suggest queries (similar to a semantic search) that are similar based upon the queue, its knowledge of Kobian lingo & slangIt could provide complete opposite words to your key words to prompt the scanner to think about the situation in a new way/spark ideas and promote understanding (user can turn it off to reduce distractions)
How will this unfold over across the whole workflow?The ML keeps the scanner thinking about new and unexpected things throughout the dayKeeps their thinking more creative,Assists in getting the scanner to use the knowledge they have to make choices about where and what to look forSuggests unexpected or very opposite things to promote a different understanding of the unfolding situation
CONCEPT 2 : BASIC SKETCH
CONCEPT 3 THEME: ML SOURCE ORGANIZER
What pain points is this addressing?It would understand context and scan for which sources would be more helpfulLearn what sources the scanner does and does not trust/favorite news site or a helpful blog vs a videoReorganizes how users see social media and news feeds based on predictable reliabilityOrganizes and pushes for the type of media the scanner prefers receiving information best inDo not disturb mode (ML would filter the types of notifications that the scanner will receive based on sender/importance of message/relevance to task etc)
How will this unfold over across the whole workflow?News sources and other background info is presented in a way that the scanner cuts down on stuff she doesn’t want to see/will get distracted by/never trustsIf the databases and social media was filtered for Sloane at all times based on how she likes seeing info she is more likely to want to interact with those sourcesIt could present data with a cultural lens (this is the Kobian military's favorite newscaster etc)It could present info that other team members have flagged relevant to the specific queue topic being worked on
CONCEPT 3 : BASIC SKETCH
ML SCREEN LAYOUT CONCEPT WIREFRAME
ML SCREEN LAYOUT CONCEPT LO-FI
ML SCREEN LAYOUT CONCEPT HI-FI
FINAL CONCEPT WIREFRAMES
