Time and place TuTh 2:55pm-4:10pm, Phillips Hall 101
Instructor: Prof. Cristian Danescu-Niculescu-Mizil --- Office hours: Tu 4:30-5:30pm, you need to make an appointment
PhD TAs: Tom Davidson, Minsu Park
Graduate TAs: Annie Cheng, Ilan Filonenko
Undergrad TAs listed on Piaza
Office hours schedule listed on Piazza (Resources -> Staff)
Course homepage http://www.cs.cornell.edu/Courses/cs4300/2018sp/
Summary How to make sense of the vast amounts of information available online, and how to relate it and to the social context in which it appears? This course introduces basic tools for retrieving and analyzing unstructured textual infordia. Applications include information retrieval (with human feedback), sentiment analysis and social analysis of text. The coursework will include programming projects that play on the interaction between knowledge and social factors.
Programming proficiency: CS 2110 or equivalent and good Python skills.
Date | Lecture | Agenda | Assignments |
---|---|---|---|
Th, Jan 25, 2018 |
#1 |
Intro: Dimensions of Information Systems Conversational Behavior and Social information Related material:Linguistic Coordination Toolkit (Beta: feedback welcome) NPR Story: Before The Internet, Librarians Would 'Answer Everything' — And Still Do -- References:Cristian Danescu-Niculescu-Mizil, Lillian Lee, Bo Pang and Jon Kleinberg. Echoes of power: Language effects and power differences in social interaction. Cristian Danescu-Niculescu-Mizil, Michael Gamon and Susan Dumais. Mark my words! Linguistic style accommodation in social media. Proceedings of WWW, 2011. Kate G. Niederhoffer and James W. Pennebaker. Linguistic Style Matching in Social Interaction. Journal of Language and Social Psychology 2002 21: 337. |
Setup Quiz out (on CMS) Assignment 0 out (on CMS) |
Tu, Jan 30, 2018 |
#2 |
Text similarity measures: Minimum Edit Distance Edit Distance worksheet (includes sketch of the Wagner Fisher algorithm we used in class) Related materialReadings:J&M Chapters 3.11 |
Assignment 1 out (on CMS) |
Th, Feb 1, 2018 |
#3 |
Basic text processing concepts: Sentence Splitting, Word Tokenization, Types, Tokens Text similarity measures: Type Overlap, Jaccard similarity Classic (ad hoc) information retrieval systems Vector space model: binary representation In-class demo: Proto Information Retrieval System: IPython notebook and html Vector space model cheatsheet (useful to keep track of notation) Related material:Readings:J&M Chapters 3.8 and 23.1.1 |
|
Tu, Feb 6, 2018 |
#4 |
Vector Space Model: geometric intuition Cosine similarity Inverse document frequency (IDF) TF-IDF weighting In-class demo: (continued and updated) IPython notebook and html Readings:MRS Chapters 6.2, 6.3, 6.4.1 and 6.4.4 |
Assignment 2 out (on CMS) |
Th, Feb 8, 2018 |
#5 |
Efficient retrieval Inverted Index Posting merge algorithm Boolean search In-class demo: (continued and updated) IPython notebook and html Related Material:Numpy tutorial and linear algebra refresher (IPython notebook on Piazza) Readings:MRS Chapter 1 |
|
Tu, Feb 13, 2018 |
#6 |
Efficient cosine similarity scoring using the inverted index (algorithm) Fast cosine retrieval worksheet (includes sketch of the algorithm using the inverted index) Related Material:Inspiration for Assignment 3: QUOTUS project and interactive visualization Readings:MRS Chapter 6.3.3 |
Assignment 3 out (on CMS) |
Th, Feb 15, 2018 |
#7 |
Efficient cosine similarity scoring using the inverted index (implementation) In-class demo: (continued and updated) IPython notebook and html Before optimizing retrieval with inverted indexes (one query on a collection of 40,000 reality TV utterances): After optimizing retrieval with inverted indexes (one query on a collection of 40,000 reality TV utterances): |
|
Th, Feb 22, 2018 |
#8 |
Evaluation of ranked retrieval systems: Precision@k, Precision-recall curve, Mean Average Precision Thinking about evaluation metrics worksheet In-class demo: IPython notebook and html Readings:MSR Chapter 8 |
Assignment 4 out (on CMS) |
Tu, Feb 27, 2018 |
#9 |
Pooling, Annotation, K-statistic Relevance feedback, Rocchio's method for query rewriting Query update using relevance feedback worksheet (includes the Rocchio query update rule) Related material Readings:MSR Chapters 9 |
Assignment 5 out (on CMS) |
Th, Mar 1, 2018 |
#10 |
Geometric interpretation for query rewriting, Pseudo-relevance feedback Query expansion, Co-occurence matrix, Scikit Learn basics Readings:MSR Chapters 9 |
|
Tu, Mar 6, 2018 |
#11 |
Pointwise Mutual Information Project ideas brainstorming Readings:MSR Chapters 9 |
Project Milestone 0 out |
Th, Mar 8, 2018 |
#12 |
Wrapping up Ad-hoc IR, Midterm practice |
|
Tu, Mar 13, 2018 |
#13 |
MIDTERM - in class |
|
Th, Mar 15, 2018 |
#14 |
Lecture topics:Midterm discussion |
|
Tu, Mar 20, 2018 |
#15 |
Lecture topics:Text mining, Classifiers, Feature Represeantaion |
|
Th, Mar 22, 2018 |
#16 |
Lecture topics:Naive Bayes, Generative Models, Smoothing |
|
Tu, Mar 27, 2018 |
#17 |
Lecture topics:Practical unsupervised learning on textual data: Singular Value Decomposition (SVD) In-class demo: IPython notebooks Data exploration, SVD |
|
Th, Mar 28, 2018 |
#18 |
Lecture topics:Practical unsupervised learning on textual data: Latent semandic indexing and topic modeling In-class demo: IPython notebooks Kickstarter success prediction Related material:Indexing by latent semantic analysis. Deerwester, Dumais and Harshman 1990 |
|
Tu, Apr 10, 2018 |
#19 |
Lecture topics:Applications of SVD: Question typologies Lexicons and off the shelf NLP tools (listed on Piazza) Related material: |
|
Th, Apr 12, 2018 |
#20 |
Lecture topics:Opinions and Trust: Link Analysis, Hubs and Authorities, Spectral Analysis Related material:NetworkX python package for link analysis Reading: |
|
Tu, Apr 17, 2018 |
#21 |
Project Prototype Madness |
|
Th, Apr 19, 2018 |
#22 |
Project Prototype Madness |
|
Tu, Apr 24, 2018 |
#23 |
Opinions and Trust: Sentiment Analysis, Lexicon Expansion, Pivot features Related material:In-class demo: Building sentiment lexicon with supervision notebook and html In-class demo: Building sentiment lexicon without supervision notebook and html Thumbs up? Sentiment Classification using Machine Learning Techniques |
|
Th, Apr 26, 2018 |
#24 |
Opinions and Trust: Using social information for sentiment analysis, Helpfulness, Deception Related Material:
|
|
Tu, May 1, 2018 |
#25 |
Project presentations |
|
Th, May 3, 2018 |
#26 |
Project presentations |