Team Members

Charlotte Wright

Kalila Shapiro

Matthew Hoke

Ravi Jain

Elmer Rodriguez

Website: https://sites.google.com/u.rochester.edu/crowdspeak

CrowdSpeak

Crowdsourced†Feedback†on†Accuracy†of†ESL†Students’†Speeches

Problem†Statement

English is widely regarded as one of the more difficult languages to learn as a

second language, yet is also the most widely-known second language in the world.

Today, people who learn English have access to a variety of resources to have their

written language analyzed and corrected; however, little exists in regards to having

people’s speaking and pronunciation corrected. It is entirely possible for someone to

become “fluent” in regards to reading and writing English but still struggle to speak it.

Though there are many reasons why people have issues becoming proficient in

English speech, the main issue arises because of a lack of viable, on-demand tools to

practice and receive feedback in a singular context. It is difficult to find friends or

colleagues to consistently help you improve your speaking skills. Additionally, there are

not many cost-effective methods to solicit help and feedback outside of a class setting.

Anyone who is seeking to learn English speaking independently will run into these

issues whereas writing is something that is more formulaic and well-documented.

We seek to mitigate this problem through the creation of a crowdsourced-based

online auto-transcription tool. The primary goal of this mission is to provide individuals a

service for feedback that will generate a record of what they spoke and feedback

focused on tone, volume, cadence, etc. The final “smart” transcript from each video then

can processed in more depth using processes like Speakoo .

Goal∫†Make auto transcriptions algorithms better using crowdsourced user input.

1

Our product will produce a human sounding transcript with (time added, value, pause,

pace).

NeedFinding

Interviews