Academic Lecture Recording Segmentation

Authors:

• Chengyan Qi (chengyqi)
• Eddie Ye (edwardye)
• Ryan Cesiel (ryances)
• Yunke Cao (ykcao)

File Information

• /: contains all files for TextTiling, our different features, and the evaluation of these files

• The "Segmentation Models" section below describes running files within this folder.

• /data: this folder contains our annotated data. Below are links to the unnannotated data:

• Engineering Dynamics (Fall 2011)

• Introduction to Psychology (Fall 2011)

• Principles of Microeconomics (Fall 2011)

• /results: this folder contains data on testing different TextTiling parameters, testing the weighting scheme of our different features, output from automatic speech recognition systems, and cue word research.

• /web: this folder contains the information retrieval web app prototype of our system.

• The "Web Application Prototype" section below describes running this application.

Segmentation Models

We provide five different versions of TextTiling to run on different lectures, including:

• TextTiling with optimal parameters
• TextTiling with optimal parameters & Cue word Filter
• TextTiling with optimal parameters & Noun Phrase Feature
• TextTiling with optimal parameters & Verbs Feature
• TextTiling with optimal parameters & N-Gram Feature

Dependency Installation

First install dependencies using our requirements.txt file:

pip install -r requirements.txt

Running Different Models

If you only want to use TextTiling with optimal parameters, then just run

$ python lecture_segmentation.py  <filename>

For example,

$ python lecture_segmentation.py  data/Engineering_Dynamics/EngDyn_15

If you want to include extra features, the usage is

$ python lecture_segmentation.py  <filename> <Extra_feature>[Optional]

For example,

$ python lecture_segmentation.py  data/Engineering_Dynamics/EngDyn_15 Verb

If you want to set the weight of the new feature, the usage is

$ python lecture_segmentation.py  <filename> <Extra_feature>[Optional] <feature_weight>[Optional]

For example,

$ python lecture_segmentation.py  data/Engineering_Dynamics/EngDyn_15 Verb 0.2

Web Application Prototype

The web application is contained within web/ folder.

Installation

First install dependencies using our requirements.txt file:

pip install -r requirements.txt

Deployment

To run our website locally:

python run.py

This will start the website locally on port 5000. You can visit localhost:5000 and query the second lecture from the "Introduction to Psychology" course.

Acknowledgements

In addition to our "Related Works" section of our final report, our software relies heavily on the NLP open-source libraries:

• NLTK
• TextBlob