Lecture 2__Questions Search and Supporting Passages.pdf

(1475 KB) Pobierz
LECTURE 2: QUESTIONS
+ HOMEWORK
April 2014
Wlodek Zadrozny
•
Class process questions
•
Watson and NLP-related questions
•
General technology questions
•
Personal questions
Class related questions
(1) From me to you:
“did you do your homework?”
(2) From you to me: longer list
Practicum – elements of the system
•
Information retrieval: Indri, Lucene/Solr, Google Search
API, new: Bing API
•
Machine learning: Weka
•
Question answering: replicating some of the Watson
capabilities
•
Question analysis – regular expressions
•
Candidate generation through search -- above
•
Scoring – your creativity is the limit
•
Learning how to score – Weka
•
Data preparation
•
Understanding of the theoretical material.
•
To check your understanding of data prep and counting: go through
NLTK Book exercises in Ch.1-3.
•
Make sure you understand the Bayes theorem
•
Watch the lectures (mine or e.g. on Coursera)
•
Familiarity with Eclipse and Java
•
Access to a computer with 4G RAM, and 100G space on
a drive.
Homework for March and April
(1) Watch Lecture 1 from 2013, and optionally Lecture 1 from 2012.
•
Make sure you understand Bayes theorem
•
Search (on-line) for new applications of NLP, describe the architecture of
one of them (input, output, and what you think might be going on)
•
Review the architecture of the IBM Watson system
(2) Go through Google Search API, create a
custom search engine
and
run a 100 queries.
(3) Do the same as in (2) for BingAPI [instructions are provided below]
(4) Play with search and evidence retrieval

Plik z chomika:

Inne pliki z tego folderu:

Inne foldery tego chomika:

Zgłoś jeśli naruszono regulamin