PhaseDuo

Bring your learning to PhaseDuo.

You're watching a neat video teaching you nuclear physics or complicated calculus, but the speaker goes on and on, using jargon that you don't understand or going too fast for you to keep up.

PhaseDuo is our solution to lengthy learning videos that leave you in the dust.

You give us a video from YouTube that you want to learn from, and we'll do the work. We'll take the video from speech to text, analyze the transcripts, and give you a summarized version next to the video on our site.

So how does PhaseDuo work?

We use IBM Watson to convert audio to text, then use natural language processing to analyze the text. From there, we find defined domain-specific names, create flash cards of related topics, and display a summary of the video. All of this info is placed near the video for easy viewing, so you can watch and learn all in one screen!

Natural language processing?

Yup! PhaseDuo use natural language processing algorithms to bring you the simplified information.

• TextRank - with Google's PageRank algorithm, we give weights to each word in a document corpus, allowing us to obtain a ranked list of important keywords on a corpus.

• NER - for an even fuzzier match, we use named-entity recognition tagging to obtain noun entities and get respective definitions for those noun entities (whether they be nouns or noun phrases).

• Dependency Parsing – this allows us to obtain a tree-like structure of word/phrase relations in a given corpus - verb-noun agreements, use of modifying verbs/adverbs, pronoun to noun relationships and etc.

Built With

• bluemix
• express.js
• flask
• ibm-watson
• node.js
• numpy
• pattern
• python
• scipy
• socket.io
• spacy
• wordnet