1/15/18
- Python skills development - we will be doing a lot with Python, so
complete the following:
- Create an account at
Computer Science Circles, complete lessons 0 thru 18, then show your
progress page to Dr. Way when done
- Supplement your learning at any time with exercises at
LearnPython.org
- Review Sentiment Analysis materials:
1/29/18
- Setup research & development environment
2/07/18
- Planning meeting (2:30 pm, CS dept lunch room)
- Caitlin Berner, Matt Mador in attendance. Matt Krause updated
via email.
- Discussed short-term plans
- Once initial Python tutorial is complete, students will each
develop a static sentiment analyzer
- After that, additional enhancements will be implemented,
selected from Project Ideas list
- Next meeting TBD. Possible date Thur. March 1, 2018 at 10:00 am
- Project Ideas
- Static sentiment analyzer - use a polarity dictionary that
contains positive and negative words, run some input text through
the program and calculate overall sentiment using the dictionary
- Twitter analyzer - use static sentiment analyzer on acquired
Twitter data
- Analyzer comparison - implement two or more simple sentiment
analyzers using approaches found in online searches, compare for
accuracy
- Retargetable analyzer - find a different polarity dictionary
that classifies sentiment in different ways (liberal vs.
conservative, pro vs. anti some position, degrees of happiness,
etc.) and report on its effectiveness
- Assisted improvable analyzer - use machine learning to build or
enhance a polarity dictionary, for example adding additional
polarity terms to the dictionary when a sentiment is “correct”
according to a human supervisor.
- Automatic improvable analyzer - figure out how to get the
improvable analyzer to learn automatically.
- Polarity dictionary builder - craft a way to harvest polarity
terms, perhaps using online searches, and determine if this approach
is feasible.
3/01/18
- Project meeting and demostration (10:30am, Dr. Way's office)
- Currently developed version using pre-coded, small lists of polarity
words (pos, neu, neg)
- Next steps discussed are:
- Find and use sentiment dictionaries to train the classifier (read
from text files, use lists of pos, neu and neg words)
- Find or create content to analyze, such as a text file of tweets or
movie reviewes (one on each line) or from an NLTK corpus
- Next meeting: tentatively 3/20/18 at 3:00 pm
3/12/18
- Check-in, report progress, help with any roadblocks
3/20/18
3/26/18
4/09/18
- Select final project (enhancements to earlier projects)
- Development
4/23/18
- Development
- Results write-up
- Evaluation
Final
- Submission of completed research work
|