Term Projects:
Guidelines:
1. Projects can be done in groups of
atmost two students.
2. The purpose of the project is to do research work on recent topics
in AI.
Research publications are highly encouraged.
3. Groups should present their project proposals before the midsem. If
you need to discuss with me for discussing your ideas, either mail me a
draft of your idea or talk to me after the class.
4. Groups should email the group composition project title and an
abstract to TA by September 15th.
5. It is desirable that the projects have a web page.
6. A report on the project should be submitted before the endsem.
Guidelines for preparing the report will be provided.
7. A final presentation on the project will be held before the endsem.
8. Projects will be evaluated based on, quality of work, novelty,
presentation and report.
9. Plagiarism in any form will be strictly penalised.
List of Candidate
Projects:
(Students may also propose projects not mentioned in this list.)
Emotion Recognition from Text
and/or speech
Modeling Human Visual Attention
in a Scene
Knowledge Management in
Education/Health/Law
Named Entity Recognition in Text
Spock Challenge
Focussed Crawling
Drug
Design
Title: Heuristic
Search Algorithms for In-Silico Drug Design
Description:
Drug molecules are proteins which act by binding with some other
proteins causing the disease. Binding (or docking) is determined by the
3-dimensional structure of the protein. Drug design in computers (in
silico) involve finding out structures which can bind with a disease
causing molecule. This can be posed as a search problem.
The problems bears similarity with robot path planning.
Goal of the project:
1. Formulate drug design as a search/CSP problem.
2. Design good heuristic functions.
3. Design efficient search algorithm.
4. Run the search algorithm to design drugs which can bind to some
benchmark molecule.
References:
Search techniques for
rational drug design, Finn, Kavraki, Latombe, Motwani, and
Venkarasubramaniam.
RAPID:
Randomized Pharmacophore Identification for Drug Design, Kavraki,
Latombe et al
QA
Title: Question
Answering in Restricted Domains
Description:
Question answering was one of the earliest AI tasks considered.
Although, no general purpose QA system exist till date. QA systems for
restricted domains like law, medical system achieved considerable
success. Components of a QA system includes, representing and building
a knowledge base, logical inferencing and natural language interface.
Goal of the project:
1. To build a knowledge base for a restricted domain, (say, academic
rules at IIT Kharagpur, Cricket World Cup, railway travel etc) in a
suitable framework (e.g., XML). The knowedge may be extracted from web
pages containing institute academic rules/cricket.org.
2. Build an inference engine.
3. Build an user interface.
4. Integrate the systems to build a Web-based Virtual Faculty
Advisor/Cricket expert etc
References:
ACL 2004
Workshop on Question Answering in Restricted Domain
Text
REtrieval Conference (TREC) QA Data
Is Question Answering
an Acquired Skill?
GGP
Title: General
Game Playing
Description:
General game players are computer systems able to
accept formal
descriptions of arbitrary games and able to play those games
effectively without human intervention. General game playing systems
are characterized by their use of general cognitive
information-processing technologies (such as knowledge representation,
reasoning, learning, and rational behavior). Unlike specialized game
playing systems (such as Deep Blue), they do not rely on algorithms
designed in advance for specific games.
Goal of the Project:
1. To particpate in the AAAI 2005 GGP Competition
Reference:
http://www.aaai.org/Conferences/National/2005/games05.html
TA
Title: Automated
Trading Agent
Description:
Building an agent for automated trading in a stock market is a
challenging AI task. It is an optimization problem in a multiagent
dynamic environment. Currently popular approaches involve using game
theory and mechanism design techniques.
Goal of the Project:
1. To design an agent to operate in the Penn-Lehmann ATA
environment.
References:
1. http://www.cis.upenn.edu/~mkearns/projects/plat.html
2. The Penn-Lehman Automated Trading
Project. Michael Kearns and Luis Ortiz. IEEE Intelligent
Systems, 2003. [PDF]
Mozilla Autocomplete
Title: Machine
Learning for Mozilla Autocomplete
Description:
Most browsers support autocomplete features. Mozilla has been fostering
an open source project on using machine learning algorithms for
enhanced autocompletion. A beta version is available with the latest
Firefox release. Mozilla also plans to use machine learning in other
aspects of the browser.
Goal of the Project:
1. To build a Mozilla plugin which uses machine learning for
autocomplete.
Reference:
http://www.mozilla.org/projects/ml/autocomplete/
KDD Cup
Title: KDD
Cup Query Categorization Task
Description:
The knowledge discovery cup is a competition held every year on
data mining problems. This years task involves categorizing search
engine queries. The task is to categorize
800,000 queries into 67 predefined categories. The
meaning and intention of search queries is subjective. A search query
"Saturn"
might mean Saturn car to some people and Saturn the planet to others.
We will
use multiple human editors to classify a subset of queries selected
from the
total set given to you. The collection of human editors is assumed to
have the
most complete knowledge about internet as compared with any individual
end
user.
A portion of the editor labeled queries is given to you
(CategorizedQuerySample.txt in the zip file for downloading) and the
rest will
be held back for evaluation.
You will not know which queries will be used
for evaluation and are asked to categorize all queries given.
Goal of the Project:
1. To build a query categorizer.
Reference:
http://kdd05.lac.uic.edu/kddcup.html
Computer
Vision
Title: Computer
Vision: Where am I?
Description:
Contestants are given a collection of color images taken by a
calibrated digital camera. The photographs have been taken at
various locations and often share overlapping fields of view or certain
objects in common. The GPS locations for a subset of the images
are provided. The goal of the contest is to guess, as accurately
as possible, the GPS locations of the un-labeled images.
Contestants are free to use whatever combination of programs they
wish, including existing imaging libraries. The compiled
executable must read a descriptor file that contains a list of images
and associated GPS locations (for a subset of the images), as well as
open and process the JPEG images listed in this file. Its output
must be a similar descriptor file, with the missing GPS locations
filled in, which is then sent to the evaluation system for scoring.
The system may also be used for locating players/balls in sports
events.
Goal of the Project:
1. To design and implement an algorithm for the above task.
Reference:
http://research.microsoft.com/iccv2005/Contest/
ADG
Title: Automatic
Deduction in Geometry
Description:
Automated theorem proving has been widely studied in AI literature. It
is interesting to study automated reasoning/deduction systems for
proving geometry theorems expressed in diagramatic form.
Goal of the Project:
1. Develop a suitable representation of (plane) geometry theorems.
2. Develop a theorem proving system (Prolog or other inference/planning
engines may be used)
References:
http://www.risc.uni-linz.ac.at/about/conferences/adg2002/
Preference
Elictitation
Title: Preference
Elicitation System
Description:
Preference elicitation systems are a part of decision support systems
which collect information about user's preferences and guides the user
towards a desired goal. Examples include a internet shopping assistant
(a primitive version of which is provided by Amazon).
Goal of the Project:
Use AI techniques to develop a preference elictitation system for
a book shop.
References:
Survey of Preference
Elicitation Methods
A POMDP Formulation of Preference
Elicitation Problems
Cognition, AI and Creativity, AI in Music, AI in Game Playing