Advanced Natural Language Processing
Course facts
- Payment: Free, no certificate
- Credits: Non Credit
- Learning style: Self-paced, start anytime
- Resources: Lecture Notes; Problem Sets
- Before you start: MIT courses 6.034 and 6.046J, or permission of instructor
MIT OpenCourseWare details
- As taught in: Fall 2005
- Department: Electrical Engineering and Computer Science
| Course Details | ||||||
|---|---|---|---|---|---|---|
|
||||||
| Outcomes | ||||||
|
||||||
| Price & Terms | ||||||
|
||||||
| Technical | ||||||
|
||||||
| Specification | ||||||
|
||||||
Graduate NLP with statistical methods: parsing, the EM algorithm, tagging, coreference and machine translation, with six data-driven homeworks.
This graduate introduction studies human language computationally, emphasizing machine learning and corpus-based methods. It covers parsing, smoothed language models, the EM algorithm, log-linear models and tagging, then lexical similarity, discourse and coreference. Applications include statistical machine translation, dialogue and summarization. Six homeworks supply data for implementing the algorithms, such as a Viterbi tagger.
What You Will Achieve
- Build smoothed language models from counts
- Apply the EM algorithm in NLP
- Implement a tagger using the Viterbi algorithm
- Explain statistical machine translation and dialogue systems
Program Overview
Language Models and Parsing
The first lectures establish statistical methods.
- Parsing and syntax
- Smoothed estimation, and language modeling
- The EM algorithm
Tagging and Similarity
The next lectures extend machine learning.
- Lexical similarity
- Log-linear models
- Tagging and history-based models
- Grammar induction
Discourse and Applications
The later lectures turn to larger systems.
- Text segmentation
- Local coherence and coreference
- Machine translation
- Dialogue processing
Course Completion
Job Outlook
These foundations support research in natural language processing and work in search, speech and language technology.
Good for: Graduate students who have completed MIT’s 6.034 and 6.046J or similar and want the statistical side of NLP.
Keep in mind: Its textbooks date from 1999 and 2000, so treat it as background to current practice. BoosTexter must run on UNIX, and no solutions are listed.
Bottom line: A record of statistical NLP, with data sets that make the homeworks usable.
Frequently Asked Questions
Which textbooks are suggested?
Speech and Language Processing by Jurafsky and Martin, and Foundations of Statistical Natural Language Processing by Manning and Schutze.
What do the homework files include?
Counts for language modeling, part-of-speech corpora, development data for testing tag trigrams and Viterbi, coreference samples, and German-English corpora for translation probabilities.
Does it include a guest lecture?
Yes. Stephanie Seneff gives a guest lecture following the two lectures on dialogue processing, near the end of the term.
User Reviews
Write your review
You must be logged in to post a review.
Reviews can only be posted by registered learners. Log in or create a free account, then you will return here to post.
There are no reviews yet.