Genetic Learning for Information Retrieval

1
Genetic Learning forInformation Retrieval

• Andrew Trotman
• Computer Science
• 365 24 60 / 40 13,140

2
Genetic Learning

• The Core Algorithm
• Crossover, Mutation, Reproduction
• Fitness proportionate selection
• Genetic Algorithms
• Chromosome is an array
• Genetic Programming
• Chromosome isan abstract syntax tree

A B C D E F X 1 2 3 4 5 6
3
Information Retrieval (Text)

• Online Systems
• Dialog, LexisNexis, etc.
• Web Systems
• Alta Vista, Excite, Google, etc.
• Scientific Literature Systems
• CiteSeer, PubMed, BioMedNet, etc.
• Question
• How should scientific literature be ranked?
• Less time searching / More time researching
• Higher exposure for good work

4
How Google Works

• PageRank
• Document ranking from PageRank
• A documents PageRank is some factor (d) of the
  rank of incoming citations
• A documents influence is some factor of its rank
  and its outgoing citations
• Characteristics of Scientific Literature
• Citations unidirectional (backwards in time)
• 12 month publication cycle
• Scientific citation cliques

5
How IR works

• Indexing
• Build the dictionary
• Construct the Postings (ltd,fgt pairs)
• Searching
• Look up terms in dictionary
• Boolean resolution
• Rank on density (probability, vector space, etc.)
• Performance
• Recall and precision

6
Structured-IR

• Sci-Lit documents have structure
• Title, abstract, conclusions, etc.
• ltd,fgt becomes ltd,p,fgt

7
Using Structure in Ranking

• Documents have structure
• Title, Abstract, Conclusions, etc.
• Weight each structure on importance
• Title higher than Abstract higher than
• How to choose the weights
• Specified in the query (XIRQL)
• Query feedback
• Learn with a Genetic Algorithm
• Adapt ranking model to use structure
• Each tree node is a locus
• Weights are genes

8
Experiment

• 50 training queries
• 50 evaluation queries
• 25 generations
• Probabilistic IR
• Vector Space IR

Results

• PROBABILISTIC IR
• 75.5 queries improved
• 6.7 increase in MAP (8.8 max)
• VECTOR SPACE IR
• 61 queries improved
• 4.7 increase in MAP (5.4 max)

9
Ranking Algorithms

• Multitude exist
• Probability, vector space, Boolean
• Several published nomenclatures
• Over 100,000 published algorithms
• Purpose
• Put relevant documents first
• Sorting
• Performance measures with precision
• Sources
• Some guy thought it up

10
Experiment

• 50 training queries
• 50 evaluation queries
• 31 runs
• Weekend time limit
• Compare to Probabilistic

Results

• 67 queries improved
• 15 increase in MAP

11
Function Comparison
Vector Space
Probability
Learned
wdqStÎq(((((((((U / sqrt(sqrt(nt))) / (mq /
sqrt((((Lq / (sqrt(sqrt(Ld)) / sqrt((U / nc))))
min(mq, N)) / sqrt(((((((Tmax / sqrt(U)) /
sqrt((((log2(sqrt(nt)) / sqrt(nt)) / sqrt(Umax))
/ (M / nc)))) / sqrt((U / nc))) - uq) / mq) /
sqrt(nt))))))) / sqrt((log(Tmax) / nc))) /
sqrt(nt)) / sqrt(nt)) / sqrt((Lq /
sqrt(((sqrt((sqrt(sqrt(Ld)) / sqrt((min(mq,
sqrt((((log(Tmax) / nc) / sqrt(Umax)) / (mq /
sqrt(((N min((sqrt(nc) / sqrt(U)), Ld)) /
sqrt(N))))))) / sqrt(Ld))))) / sqrt((Tmax / nc)))
/ sqrt(nt)))))) / sqrt((min(mq, N) / nc))) /
sqrt((log(Tmax) / nc))) / sqrt(nt))
12
Conclusions

• Using document structure improved ranking
• Structure weights can be learned with a GA
• GP can be used to learn ranking functions
• Speculation
• Combining GA and GP to learn a structure ranking
  algorithm will better GA and GP alone

13
Questions?
14
Random NumbersAre your results an artifact of
your random number generator?