BeeSpace Informatics Research ChengXiang (“Cheng”) Zhai Department of Computer Science Institute for Genomic Biology Statistics Graduate School of Library & Information Science University of Illinois at Urbana-Champaign BeeSpace Workshop, May 22, 2009
Overview of BeeSpace Technology Users … Task Support Gene Summarizer Function Annotator Space Navigation Space/Region Manager, Navigation Support Search Engine Text Miner Relational Database Words/Phrases Entities Content Analysis Natural Language Understanding Meta Data Literature Text
Part 1: Content Analysis
Natural Language Understanding …We have cloned and sequenced a cDNA encoding Apis mellifera ultraspiracle (AMUSP) and examined its responses to … NP VP Gene Gene
Sample Technique 1: Automatic Gene Recognition Syntactic clues: Capitalization (especially acronyms) Numbers (gene families) Punctuation: -, /, :, etc. Contextual clues: Local: surrounding words such as “gene”, “encoding”, “regulation”, “expressed”, etc. Global: same noun phrase occurs several times in the same article
Maximum Entropy Model for Gene Tagging Given an observation (a token or a noun phrase), together with its context, denoted as x Predict y {gene, non-gene} Maximum entropy model: P(y|x) = K exp(ifi(x, y)) Typical f: y = gene & candidate phrase starts with a capital letter y = gene & candidate phrase contains digits Estimate i with training data
Domain overfitting problem When a learning based gene tagger is applied to a domain different from the training domain(s), the performance tends to decrease significantly. The same problem occurs in other types of text, e.g., named entities in news articles. Training domain Test domain F1 mouse 0.541 fly 0.281 Reuters 0.908 WSJ 0.643
Observation I Overemphasis on domain-specific features in the trained model wingless daughterless eyeless apexless … fly “suffix –less” weighted high in the model trained from fly data
Observation II Generalizable features: generalize well in all domains …decapentaplegic and wingless are expressed in analogous patterns in each primordium of… (fly) …that CD38 is expressed by both neurons and glial cells…that PABPC5 is expressed in fetal brain and in a range of adult tissues. (mouse)
Observation II Generalizable features: generalize well in all domains …decapentaplegic and wingless are expressed in analogous patterns in each primordium of… (fly) …that CD38 is expressed by both neurons and glial cells…that PABPC5 is expressed in fetal brain and in a range of adult tissues. (mouse) “wi+2 = expressed” is generalizable
Generalizability-based feature ranking training data fly mouse D3 … Dm 1 2 3 4 5 6 7 8 … -less expressed 1 2 3 4 5 6 7 8 … expressed -less 1 2 3 4 5 6 7 8 … expressed -less … 1 2 3 4 5 6 7 8 … expressed -less … expressed -less … 0.125 0.167
Adapting Biological Named Entity Recognizer test data T1 Tm training data … learning entity recognizer d = λ0d0 + (1 – λ0) (λ1d1 + … + λmdm) d features λ0, λ1, … , λm testing O1 Om … individual domain feature ranking domain-specific features feature re-ranking O’ generalizable features feature selection for D1 feature selection for D0 top d0 features for D0 top d1 features for D1 feature selection for Dm top dm features for Dm …
Effectiveness of Domain Adaptation Exp Method Precision Recall F1 F+M→Y Baseline 0.557 0.466 0.508 Domain 0.575 0.516 0.544 % Imprv. +3.2% +10.7% +7.1% F+Y→M 0.571 0.335 0.422 0.582 0.381 0.461 +1.9% +13.7% +9.2% M+Y→F 0.583 0.097 0.166 0.591 0.139 0.225 +1.4% +43.3% +35.5% Text data from BioCreAtIvE (Medline) 3 organisms (Fly, Mouse, Yeast)
Gene Recognition in V3 A variation of the basic maximum entropy Classes: {Begin, Inside, Outside} Features: syntactical features, POS tags, class labels of previous two tokens Post-processing to exploit global features Leverage existing toolkit: BMR
Part 2: Navigation Support
Space-Region Navigation … Topic Regions Intersection, Union,… My Regions/Topics Bird Singing EXTRACT Fly Rover EXTRACT Bee Forager MAP MAP … Bee Bird Fly My Spaces SWITCHING Intersection, Union,… Behavior Literature Spaces
MAP: Topic/RegionSpace MAP: Use the topic/region description as a query to search a given space Retrieval algorithm: Query word distribution: p(w|Q) Document word distribution: p(w|D) Score a document based on similarity of Q and D Leverage existing retrieval toolkits: Lemur/Indri
EXTRACT: Space Topic/Region Assume k topics, each being represented by a word distribution Use a k-component mixture model to fit the documents in a given space (EM algorithm) The estimated k component word distributions are taken as k topic regions Likelihood: Maximum likelihood estimator: Bayesian estimator:
A Sample Topic & Corresponding Space Word Distribution (language model) labels Meaningful labels actin filaments flight muscle flight muscles filaments 0.0410238 muscle 0.0327107 actin 0.0287701 z 0.0221623 filament 0.0169888 myosin 0.0153909 thick 0.00968766 thin 0.00926895 sections 0.00924286 er 0.00890264 band 0.00802833 muscles 0.00789018 antibodies 0.00736094 myofibrils 0.00688588 flight 0.00670859 images 0.00649626 Example documents actin filaments in honeybee-flight muscle move collectively arrangement of filaments and cross-links in the bee flight muscle z disk by image analysis of oblique sections identification of a connecting filament protein in insect fibrillar flight muscle the invertebrate myosin filament subfilament arrangement of the solid filaments of insect flight muscles structure of thick filaments from insect flight muscle
Incorporating Topic Priors Either topic extraction or clustering: User exploration: usually has preference. E.g., want one topic/cluster is about foraging behavior Use prior to guild topic extraction Prior as a simple language model E.g. forage 0.2; foraging 0.3; food 0.05; etc.
Incorporating a Topic Prior Original EM: EM with Prior:
Incorporating Topic Priors: Sample Topic 1 age 0.0672687 division 0.0551497 labor 0.052136 colony 0.038305 foraging 0.0357817 foragers 0.0236658 workers 0.0191248 task 0.0190672 behavioral 0.0189017 behavior 0.0168805 older 0.0143466 tasks 0.013823 old 0.011839 individual 0.0114329 ages 0.0102134 young 0.00985875 genotypic 0.00963096 social 0.00883439 Prior: labor 0.2 division 0.2
Incorporating Topic Priors: Sample Topic 2 behavioral 0.110674 age 0.0789419 maturation 0.057956 task 0.0318285 division 0.0312101 labor 0.0293371 workers 0.0222682 colony 0.0199028 social 0.0188699 behavior 0.0171008 performance 0.0117176 foragers 0.0110682 genotypic 0.0106029 differences 0.0103761 polyethism 0.00904816 older 0.00808171 plasticity 0.00804363 changes 0.00794045 Prior: behavioral 0.2 maturation 0.2
Exploit Prior for Concept Switching foraging 0.142473 foragers 0.0582921 forage 0.0557498 food 0.0393453 nectar 0.03217 colony 0.019416 source 0.0153349 hive 0.0151726 dance 0.013336 forager 0.0127668 information 0.0117961 feeder 0.010944 rate 0.0104752 recruitment 0.00870751 individual 0.0086414 reward 0.00810706 flower 0.00800705 dancing 0.00794827 behavior 0.00789228 foraging 0.290076 nectar 0.114508 food 0.106655 forage 0.0734919 colony 0.0660329 pollen 0.0427706 flower 0.0400582 sucrose 0.0334728 source 0.0319787 behavior 0.0283774 individual 0.028029 rate 0.0242806 recruitment 0.0200597 time 0.0197362 reward 0.0196271 task 0.0182461 sitter 0.00604067 rover 0.00582791 rovers 0.00306051
Part 3: Task Support
Gene Summarization Task: Automatically generate a text summary for a given gene Challenge: Need to summarize different aspects of a gene Standard summarization methods would generate an unstructured summary Solution: A new method for generating semi-structured summaries
An Ideal Gene Summary http://flybase.bio.indiana.edu/.bin/fbidq.html?FBgn0000017 GP EL SI GI MP WFPI
Semi-structured Text Summarization
Summary example (Abl)
A General Entity Summarizer Task: Given any entity and k aspects to summarize, generate a semi-structured summary Assumption: Training sentences available for each aspect Method: Train a recognizer for each aspect Given an entity, retrieve sentences relevant to the entity Classify each sentence into one of the k aspects Choose the best sentences in each category
Summary All the methods we developed are General Scalable The problems are hard, but good progress has been made in all the directions The V3 system has only incorporated the basic research results More advanced technologies are available for immediate implementation Better tokenization for retrieval Domain adaptation techniques Automatic topic labeling General entity summarizer More research to be done in Entity & relation extraction Graph mining/question answering Domain adaptation Active learning
Looking Ahead: X-Space… Users … Task Support Gene Summarizer Function Annotator Space Navigation Space/Region Manager, Navigation Support Search Engine Text Miner Relational Database Words/Phrases Entities Content Analysis Natural Language Understanding Meta Data Literature Text
Thank You! Questions?