Teadmiste otsing, formaliseerimine ja hoidmine

Allikas: Lambda
Mine navigeerimisribale Mine otsikasti
teadmised
teadmised

Code: ITV0060
Link: http://www.lambda.ee/index.php/Teadmiste_otsing,_formaliseerimine_ja_hoidmine or http://www.lambda.ee/index/itv0060
Lecturer: Tanel Tammet
Contact: tanel.tammet@ttu.ee, 6203457, TTÜ AK223
Archives of previous years: 2012, 2011, 2010, 2009, 2008, 2007, 2006, [http:/cs.ttu.ee/kursused/wav4130/ older].


Time, place, result

Semester: spring
Grading: exam
Points: 3.5

Lectures: every Monday 16:00-17.30, room IT-140.
Practical work: Thursdays on odd weeks 16:00-17:30 room IT-213D, IT-213H.
Practical work will give 40% and exam 60% of points underlying the final grade.


Practical work

There are three labs. All labs have to be presented to the prof and all students present at labwork.

First lab 2013

The main goal is to select a domain and a task for the second lab. What to search, from where, etc. You should experiment with scraping, search and apis to find out what you can really get, in practice. The ideas, experiments and the final plan should be presented. Might be useful to have a look at:


Deadline: 21. February.

Second lab 2013

The main goal is to develop an app which scrapes/searches potentially uncertain data from the web and creates a representation of that data in a machine-processable form (json, csv, xml etc). You can scrape html pages, google, bing, wolfram alpha results and/or use any apis like twitter, google places, google plus etc. It is best to have a combination of these.

Deadline: end of March.

Third lab 2013

The main goal of the third lab is to develop a ruleset for generating new interesting facts from the data obtained from the second lab. The ruleset has to be encoded in a format suitable for some first-order based logical prover and then used with the prover to actually derive new facts.

Deadline: end of the spring semester.


Old versions: 2012

The old versions are here for inspiration and initial ideas.

Old version, not for 2013: First lab

The main goal of the first lab is to web-scrape data from the public sources using search engines and simple text analysis/statistics.

Details in Teadmised first lab 2012. The lab should be completed by March, later presentation will cut the points to half.

Old version, not for 2013: Second lab

The main goal of the second lab is to augment data obtained from the first lab by writing and using probabilistic rules with a rule engine.

Details about second lab will be available on beginning of April. Previous year lab can be seen here: Teadmised second lab 2012. You may also have a look at the 2011 version of the lab Teadmised second lab 2011.

Old version, not for 2013: Optional third lab

The optional third lab can be done in parallel with the second. It will be a relatively complex lab: successful passing will give additional 20 points, thus the weight of the exam is diminished from 60% to 40% of the overall score.

What to do: use Attempto to create a system such that you can write rules and/or facts and/or queries for the second lab in restricted english. These restricted english sentences have to be automatically converted to the form usable by Otter. It is a good idea to avoid probabilities in these rules, since it would be very hard to encode probabilities using attempto. See: http://attempto.ifi.uzh.ch/site/ , http://attempto.ifi.uzh.ch/ape/

Lectures

1. Declarative and procedural representations

Lecture materials:

2. Databases: meaning and representation of facts

Lecture materials:


3. NoSQL databases


4. HTML annotations. RDFa, reading and using RDF data from web pages. Starting RDFs

Lecture materials:

Understand main parts:

RDF, RDFS, OWL and conversion to 1st order logic

Lecture material:

Understand RDF and RDFS:

Understand basics of owl:


RDF relation to logic

Notes from lecture:

Understand a bit of:

Semantics of ordinary relational base


Rules in logic: apps in planning, robotics, ...

Lecture material:

See also:

First order logic in query engines and planning


Data extraction

Web mining by Nell:

Restricted english jms

Uncertain knowledge logic

Indexes

Traditional database indexes incl B+ tree:

Bitmap indexes:

Hash indexes:

Fulltext indexes:

Term indexes: Understand basics of discrimination tree trie-type term indexes and path indexes from

Course structure

We start with representing and converting facts, continue to rules, first monotonic, then probabilistic and nonmonotonic, then look into search and indexing of complex information, and finally, some learning.

Intro and paradigms of knowledge representation

Lecture materials:

For KR, declarative and procedural rep see:

Web mining by Nell:

Databases: meaning and representation of facts

Lecture material:

See also about databases and logic:

See also about RDF and related:


Rules in logic: apps in planning, robotics, ...

Lecture material:

Topics:

  • propositional, first order, higher order
  • semantics/models
  • proofs and deriving new information
  • automated reasoning

See also:

Various rule/logic languages

Lecture material:

Several languages:

  • RDFS
  • OWL
  • KIF
  • CL
  • ontologies
  • wordnet
  • cyc
  • restricted english systems

See also:

Scraping semistructured data

Time, context, metainformation

Probabilistic and nonmonotonic reasoning

See also:

Personalized search and recommendations

Fast search of facts and rules using indexing

Learning

See also:


Pärit leheküljelt "http://www.lambda.ee/w/index.php?title=Teadmiste_otsing,_formaliseerimine_ja_hoidmine&oldid=10047"