LDA on social bookmarking systems

1. LDA on Social Bookmarking Systems: an experiment on CiteUlike Introduction to Natural Language Processing, CS2731 Professor Rebecca Hwa University of Pittsburgh Denis Parra-Santander December 16th 2009 1

2. Outline 2 Outline Topic ModelingJoke (check mood of people…) LDAIntroduction Sorry I’m nervous… Motivation Smart Statement… Definitions Monte Carlo: a great place pass your vacations DIRICHLET: [diʀiˈkleː] Uuuh… Uuuh… Experiments Results END Evaluation method

3. Topic modeling: Evolution  LSA [Deerwester et al. 90]: find “latent” structure or “concepts” in a text corpus.: ◦ Compare texts using a vector-based representation that is learned from a corpus. Relies on SVD (for dimensionality reduction)  PLSA [Hoffman 99] extends LSA by adding the idea of mixture decomposition derived from a latent class model.  LDA [Blei et al. 2004] : extends PLSA by using a generative model, in particular, by adding a Dirichlet prior. 3

4. Document 22 LDA : Generative Model* (I/II) 4 Words: Information About catalog pricing changes 2009 welcome looking hands-on science ideas try kitchen • LDA assumes that each word in the document was generated by a distribution of topics over words. Topic 15: science experiment learning ideas practice information Topic 9: catalog shopping buy internet checkout cart • Paired with an inference mechanism (Gibbs sampling), learns per- document distribution over topics, per-topic distributions over words … *Original slide by Daniel Ramage, Stanford University

5. LDA I/II : Graphical Model 5  Graphical model representations  Compact notation: Cat w1 w2 w3 w4 wn … *Original slide by Roger Levy, UCSD Cat w1 n “generate a word from Cat n times” a “plate”

6. LDA II/II : Graphical Model 6 Nd D zi wi θ (d) φ (j) α β θ (d) ∼ Dirichlet(α) zi ∼ Discrete(θ (d) )φ(j) ∼ Dirichlet(β) wi ∼ Discrete(φ(zi) ) T distribution over topics for each document topic assignment for each word distribution over words for each topic word generated from assigned topic Dirichlet priors *Original slide by Roger Levy, UCSD

7. Learning the parameters 7  Maximum likelihood estimation (EM) ◦ e.g. Hofmann (1999)  Deterministic approximate algorithms ◦ variational EM; Blei, Ng & Jordan (2001; 2003) ◦ expectation propagation; Minka & Lafferty (2002)  Markov chain Monte Carlo ◦ full Gibbs sampler; Pritchard et al. (2000) ◦ collapsed Gibbs sampler; Griffiths & Steyvers (2004) *Original slide by Roger Levy, UCSD

8. My Experiments  IdentifyTopics in a collection of documents from a social bookmarking system (citeULike) [Ramage et al. 2008]  Objective: Clusterise documents by LDA  QUESTION: If the documents have, in addition to title and text, USERTAGS… how can they help/influence/improve topic identification/clustering? 8

9. Tools available  Many implementations of LDA based on Gibbs sampling:  LingPipe (Java)  Mallet (Java)  STMT (Scala) – I chose this one 9

10. The Dataset  Initially ◦ Corpus: ~45k documents, ◦ Definition of 99 topics (queries) ◦ Gold-std : Identification document-topic by expert feedback, defining a ground-truth  But then, then gold-standard and RAM… ◦ Not all documents were relevant ◦ Unable to train model with 45k, 20k and10k  And then, the tags: not all the documents in gold-standard had associated tags (#>2) ◦ Finally:Training with 1.1k documents ◦ Experiments on 212 documents 10

11. Evaluation: Pair-wise precision / recall 11*Original slide by Daniel Ramage, Stanford University

12. … results … 12

13. Perplexity 13 # of topics 38 52 99 Content Tags 1860.7642 1880.7974 1270.8032 Title + text 2526.7589 2447.5477 2755.1329 Using Stanford Topic Modeling Toolbox (STMT) Training with ~1.1k documents, 80% training, 20% to calculate pp.

14. F1 ( & precision/recall)  F-1, in parenthesis precision and recall 14 # of topics 38 52 99 Tags 0.139 (0.118/0.167) 0.168 (0.187/0.152) 0.215 (0.267/0.18) Title + text 0.1252 (0.122/0.128) 0.157 (0.151/0.163) 0.156 (0.198/0.129)

15. Conclusions  Results are not the same than “motivational” paper, though are consistent with their conclusions (dataset is very domain-specific)  Pending: combining tags and documents, in particular MM-LDA  Importance to NLP: extensions of the model have been used to: ◦ learn syntactic and semantic factors that guide word choice ◦ Identify authorship ◦ Many others () 15

16. … and to finish … Thanks! And… 16

17. “Invent new worlds and watch your word; The adjective, when it doesn’t give life, kills…” Ars Poetica Vicente Huidobro “Inventa nuevos mundos y cuida tu palabra; El adjetivo, cuando no da vida, mata…” 17

18. References Heinrich, G. (2008). Parameter estimation for text analysis,.Technical report, University of Leipzig. Ramage, D., P. Heymann, C. D. Manning, and H. G. Molina (2009). Clustering the tagged web. In WSDM '09: Proceedings of the Second ACM International Conference onWeb Search and Data Mining, NewYork, NY, USA, pp. 54-63. ACM. Steyvers, M. and T. Griffiths (2007). Probabilistic Topic Models. Lawrence Erlbaum Associates. 18

19. Backup Slides 19

20. LSA: 3 claims (2 match with LDA)  Semantic Information can be derived from a word-document co-ocurrence matrix  Dimensionality reduction is an essential part of this derivation  Words and documents can be represented as points in an Euclidean Space => different than LDA: semantic properties of words and docs are expressed in terms of probabilistic topics 20

21. 21 Parameter estimation and Gibbs Sampling (3 Slides)

22. Inverting the generative model  Maximum likelihood estimation (EM) ◦ e.g. Hofmann (1999)  Deterministic approximate algorithms ◦ variational EM; Blei, Ng & Jordan (2001; 2003) ◦ expectation propagation; Minka & Lafferty (2002)  Markov chain Monte Carlo ◦ full Gibbs sampler; Pritchard et al. (2000) ◦ collapsed Gibbs sampler; Griffiths & Steyvers (2004)

23. The collapsed Gibbs sampler  Using conjugacy of Dirichlet and multinomial distributions, integrate out continuous parameters  Defines a distribution on discrete ensembles z ΦΦΦ= ∫ ∆ dpPP T W )(),|()|( zwzw ΘΘΘ= ∫ ∆ dpPP D T )()|()( zz ∑ = z zzw zzw wz )()|( )()|( )|( PP PP P ∏ ∑ ∏ = +Γ Γ Γ +Γ = T j w j w W w j w n Wn 1 )( )( )( )( )( )( β β β β ∏ ∑ ∏ = +Γ Γ Γ +Γ = D d j d j T j d j n Tn 1 )( )( )( )( )( )( α α α α

24. The collapsed Gibbs sampler  Sample each zi conditioned on z-i  This is nicer than your average Gibbs sampler: ◦ memory: counts can be cached in two sparse matrices ◦ optimization: no special functions, simple arithmetic ◦ the distributions on Φ and Θ are analytic given z and w, and can later be found for each sample α α β β Tn n Wn n zP i i i i i d d j z z w ii + + + + ∝ •• − )( )( )( )( ),|( zw

25. Gibbs Sampling from PTM paper 25

26. 26 Extensions and Applications

27. Nu U zi wi θ(u) φ (j) α β θ(u)|su=0 ∼ Delta(θ(u-1)) θ(u)|su=1 ∼ Dirichlet(α) zi ∼ Discrete(θ (u) ) φ(j) ∼ Dirichlet(β) wi ∼ Discrete(φ(zi) ) T Extension: a model for meetings su θ(u-1) … (Purver, Kording, Griffiths, & Tenenbaum, 2006)

28. Sample of ICSI meeting corpus (25 meetings)  no it's o_k.  it's it'll work.  well i can do that.  but then i have to end the presentation in the middle so i can go back to open up javabayes.  o_k fine.  here let's see if i can.  alright.  very nice.  is that better.  yeah.  o_k.  uh i'll also get rid of this click to add notes.  o_k. perfect  NEW TOPIC (not supplied to algorithm)  so then the features we decided or we decided we were talked about.  right.  uh the the prosody the discourse verb choice.  you know we had a list of things like to go and to visit and what not.  the landmark-iness of uh.  i knew you'd like that.

29. Topic segmentation applied to meetings Inferred Segmentation Inferred Topics

30. Comparison with human judgments Topics recovered are much more coherent than those found using random segmentation, no segmentation, or an HMM

31. Learning the number of topics  Can use standard Bayes factor methods to evaluate models of different dimensionality ◦ e.g. importance sampling via MCMC  Alternative: nonparametric Bayes ◦ fixed number of topics per document, unbounded number of topics per corpus (Blei, Griffiths, Jordan, & Tenenbaum, 2004) ◦ unbounded number of topics for both (the hierarchical Dirichlet process) (Teh, Jordan, Beal, & Blei, 2004)

32. The Author-Topic model (Rosen-Zvi, Griffiths,Smyth, & Steyvers, 2004) Nd D zi wi θ (a) φ (j) α β θ (a) ∼ Dirichlet(α) zi ∼ Discrete(θ (xi) ) φ(j) ∼ Dirichlet(β) wi ∼ Discrete(φ(zi) ) T xi A xi ∼ Uniform(A(d) ) each author has a distribution over topics the author of each word is chosen uniformly at random

33. Four example topics from NIPS WORD PROB. WORD PROB. WORD PROB. WORD PROB. LIKELIHOOD 0.0539 RECOGNITION 0.0400 REINFORCEMENT 0.0411 KERNEL 0.0683 MIXTURE 0.0509 CHARACTER 0.0336 POLICY 0.0371 SUPPORT 0.0377 EM 0.0470 CHARACTERS 0.0250 ACTION 0.0332 VECTOR 0.0257 DENSITY 0.0398 TANGENT 0.0241 OPTIMAL 0.0208 KERNELS 0.0217 GAUSSIAN 0.0349 HANDWRITTEN 0.0169 ACTIONS 0.0208 SET 0.0205 ESTIMATION 0.0314 DIGITS 0.0159 FUNCTION 0.0178 SVM 0.0204 LOG 0.0263 IMAGE 0.0157 REWARD 0.0165 SPACE 0.0188 MAXIMUM 0.0254 DISTANCE 0.0153 SUTTON 0.0164 MACHINES 0.0168 PARAMETERS 0.0209 DIGIT 0.0149 AGENT 0.0136 REGRESSION 0.0155 ESTIMATE 0.0204 HAND 0.0126 DECISION 0.0118 MARGIN 0.0151 AUTHOR PROB. AUTHOR PROB. AUTHOR PROB. AUTHOR PROB. Tresp_V 0.0333 Simard_P 0.0694 Singh_S 0.1412 Smola_A 0.1033 Singer_Y 0.0281 Martin_G 0.0394 Barto_A 0.0471 Scholkopf_B 0.0730 Jebara_T 0.0207 LeCun_Y 0.0359 Sutton_R 0.0430 Burges_C 0.0489 Ghahramani_Z 0.0196 Denker_J 0.0278 Dayan_P 0.0324 Vapnik_V 0.0431 Ueda_N 0.0170 Henderson_D 0.0256 Parr_R 0.0314 Chapelle_O 0.0210 Jordan_M 0.0150 Revow_M 0.0229 Dietterich_T 0.0231 Cristianini_N 0.0185 Roweis_S 0.0123 Platt_J 0.0226 Tsitsiklis_J 0.0194 Ratsch_G 0.0172 Schuster_M 0.0104 Keeler_J 0.0192 Randlov_J 0.0167 Laskov_P 0.0169 Xu_L 0.0098 Rashid_M 0.0182 Bradtke_S 0.0161 Tipping_M 0.0153 Saul_L 0.0094 Sackinger_E 0.0132 Schwartz_A 0.0142 Sollich_P 0.0141 TOPIC 19 TOPIC 24 TOPIC 29 TOPIC 87

34. Who wrote what? A method1 is described which like the kernel1 trick1 in support1 vector1 machines1 SVMs1 lets us generalize distance1 based2 algorithms to operate in feature1 spaces usually nonlinearly related to the input1 spaceThis is done by identifying a class of kernels1 which can be represented as norm1 based2 distances1 in Hilbert spaces It turns1 out that common kernel1 algorithms such as SVMs1 and kernel1 PCA1 are actually really distance1 based2 algorithms and can be run2 with that class of kernels1 too As well as providing1 a useful new insight1 into how these algorithms work the present2 work can form the basis1 for conceiving new algorithms This paper presents2 a comprehensive approach for model2 based2 diagnosis2 which includes proposals for characterizing and computing2 preferred2 diagnoses2 assuming that the system2 description2 is augmented with a system2 structure2 a directed2 graph2 explicating the interconnections between system2 components2 Specifically we first introduce the notion of a consequence2 which is a syntactically2 unconstrained propositional2 sentence2 that characterizes all consistency2 based2 diagnoses2 and show2 that standard2 characterizations of diagnoses2 such as minimal conflicts1 correspond to syntactic2 variations1 on a consequence2 Second we propose a new syntactic2 variation on the consequence2 known as negation2 normal form NNF and discuss its merits compared to standard variationsThird we introduce a basic algorithm2 for computing consequences in NNF given a structured system2 description We show that if the system2 structure2 does not contain cycles2 then there is always a linear size2 consequence2 in NNF which can be computed in linear time2 For arbitrary1 system2 structures2 we show a precise connection between the complexity2 of computing2 consequences and the topology of the underlying system2 structure2 Finally we present2 an algorithm2 that enumerates2 the preferred2 diagnoses2 characterized by a consequence2 The algorithm2 is shown1 to take linear time2 in the size2 of the consequence2 if the preference criterion1 satisfies some general conditions Written by (1) Scholkopf_B Written by (2) Darwiche_A

35. Analysis of PNAS abstracts  Test topic models with a real database of scientific papers from PNAS  All 28,154 abstracts from 1991-2001  All words occurring in at least five abstracts, not on “stop” list (20,551)  Total of 3,026,970 tokens in corpus (Griffiths & Steyvers, 2004)

36. FORCE SURFACE MOLECULES SOLUTION SURFACES MICROSCOPY WATER FORCES PARTICLES STRENGTH POLYMER IONIC ATOMIC AQUEOUS MOLECULAR PROPERTIES LIQUID SOLUTIONS BEADS MECHANICAL HIV VIRUS INFECTED IMMUNODEFICIENCY CD4 INFECTION HUMAN VIRAL TAT GP120 REPLICATION TYPE ENVELOPE AIDS REV BLOOD CCR5 INDIVIDUALS ENV PERIPHERAL MUSCLE CARDIAC HEART SKELETAL MYOCYTES VENTRICULAR MUSCLES SMOOTH HYPERTROPHY DYSTROPHIN HEARTS CONTRACTION FIBERS FUNCTION TISSUE RAT MYOCARDIAL ISOLATED MYOD FAILURE STRUCTURE ANGSTROM CRYSTAL RESIDUES STRUCTURES STRUCTURAL RESOLUTION HELIX THREE HELICES DETERMINED RAY CONFORMATION HELICAL HYDROPHOBIC SIDE DIMENSIONAL INTERACTIONS MOLECULE SURFACE NEURONS BRAIN CORTEX CORTICAL OLFACTORY NUCLEUS NEURONAL LAYER RAT NUCLEI CEREBELLUM CEREBELLAR LATERAL CEREBRAL LAYERS GRANULE LABELED HIPPOCAMPUS AREAS THALAMIC A selection of topics TUMOR CANCER TUMORS HUMAN CELLS BREAST MELANOMA GROWTH CARCINOMA PROSTATE NORMAL CELL METASTATIC MALIGNANT LUNG CANCERS MICE NUDE PRIMARY OVARIAN

37. Cold topics Hot topics 2 SPECIES GLOBAL CLIMATE CO2 WATER ENVIRONMENTAL YEARS MARINE CARBON DIVERSITY OCEAN EXTINCTION TERRESTRIAL COMMUNITY ABUNDANCE 134 MICE DEFICIENT NORMAL GENE NULL MOUSE TYPE HOMOZYGOUS ROLE KNOCKOUT DEVELOPMENT GENERATED LACKING ANIMALS REDUCED 179 APOPTOSIS DEATH CELL INDUCED BCL CELLS APOPTOTIC CASPASE FAS SURVIVAL PROGRAMMED MEDIATED INDUCTION CERAMIDE EXPRESSION 37 CDNA AMINO SEQUENCE ACID PROTEIN ISOLATED ENCODING CLONED ACIDS IDENTITY CLONE EXPRESSED ENCODES RAT HOMOLOGY 289 KDA PROTEIN PURIFIED MOLECULAR MASS CHROMATOGRAPHY POLYPEPTIDE GEL SDS BAND APPARENT LABELED IDENTIFIED FRACTION DETECTED 75 ANTIBODY ANTIBODIES MONOCLONAL ANTIGEN IGG MAB SPECIFIC EPITOPE HUMAN MABS RECOGNIZED SERA EPITOPES DIRECTED NEUTRALIZING

38. 38 The effect of Alpha and beta as hyperparameters

39. Effects of hyperparameters  α and β control the relative sparsity of Φ and Θ ◦ smaller α, fewer topics per document ◦ smaller β, fewer words per topic  Good assignments z compromise in sparsity logΓ(x) x ∏ ∑ ∏ = +Γ Γ Γ +Γ = T j w j w W w j w n Wn P 1 )( )( )( )( )( )( )|( β β β β zw ∏ ∑ ∏ = +Γ Γ Γ +Γ = D d j d j T j d j n Tn P 1 )( )( )( )( )( )( )( α α α α z

40. Varying α decreasing α increases sparsity

41. Varying β decreasing β increases sparsity ?

42. Multi-Multinomial LDA (MM-LDA) 42

43. Ramage 2009 results 43