Learning Conference Reviewer Assignments Adith Swaminathan Guide : Prof. Soumen Chakrabarti Department of Computer Science and Engineering, Indian Institute of Technology, Bombay

Future Work (from BTP1) Given WWW2010’s assignments, learn Affinity_Param, Topic_Param and Irritation Citations as edge features Load-Constrained Partial Assignments Better estimation of Assignment Quality

Background Conference Reviewer-Paper Assignment as a Many-Many- matching [1] Minimum Cost Network Flow (MCF)

Conference Reviewer Assignment Set of Reviewers, R, max #papers = L_i Set of Papers, P, min #reviews = K Assumption : Only require #reviews, not quality Suppose we have cost function A_ij(y) for

ILP -> Assumption -> MCF

Two problems Integer Linear Programs are NP-Hard! – Relax? – More assumptions? How to determine A_ij? – M * N ~ – Multimodal clues

ILP -> Assumption -> MCF Enforce structure on A_ij – Better model multimodality – Fewer parameters to fix “Learn” A_ij using Structured Learning Techniques A_ij = w T Φ(R_i, P_j, y_ij)

Ramifications of Structured Costs Costs decompose over pairs – Decomposable Preference Auction – Polynomial Algorithms for DPAs [2] Restricted notion of optimality – Per-reviewer/Per-paper constraint could be combinatorial – Stability?

ILP -> Assumption -> MCF

Minimum Cost Network Flow Directed graph G=(V,E), capacities u(E)>= 0, costs c(E) Nodes have numbers b(V) : Sum(b(V)) = 0 Task : Find a function f: E->R + which satisfies the b-flow at minimum cost Successive Shortest Path Algorithm

Node features and Edge features Reviewer ProfileTopics Paper ContentsTopics Affinity Bid Topic Overlap Cites

The Loss Function L_ij = w_1 * exp(-Affinity_ij) + w_2 * [[1 – Topic_Overlap_ij]] + w_3 * Bid_Cost Bid_Cost = Potential(R_i, P_j, y_ij) Irritation (I) and Disappointment (D) needs to be set

Assignment Quality Measures Number of Bids Violated? – Not a reliable measure. +ve Bids Violated –ve Bids Violated Assignments satisfying Topic Match Confidence?

Confidence == Quality? Very sparse – Fewer than 5% observed – Extrapolated Confidence? Reliable – Bids as a precursor of Confidence [3] – Confidence-Augmented Loss?

Learning w’s Transductive Ordinal Regression – Assume : Assignments are independent (Naïve) – Heuristic : Augment observed dataset – Extrapolate observed Confidence [4] – Learn w over extrapolated dataset Support Vector Machine for Structured Outputs – Cast as soft-margin SVM formulation [5] – Upper-bound objective with a convex fn (Optimality?) – Minimize, using Cutting Plane (Approximate)

Transductive Ordinal Regression [6]

SVM Struct. [7] Loss Augmented Inference ~ Most Violated Constraint Loss is decomposable -> Modified MCF

PARA : Paper Assignment to Reviewers Apparatus

Results

Bimodal Behaviour Reviewer either gets few or L_i papers Load Penalties [8] Introduce more parameters Infer using modified MCF Learning parameters? Load Rebalancing Tradeoff between MCF optimum and old assignment

Penalise Reviewer Loads

Load Constrained Assignments

Avenues for Future Work Document Modelling for Affinity Scores Objective Assignment Evaluation Transitive Citation Scores Load Penalty Parameter Estimation

References 1. The Conference Paper Assignment Problem, J. Goldsmith, R.H. Sloan, MultiAgent Systems: Algorithmic, Game-Theoretic, and Logical Foundations, Y. Shoham, K. Leyton- Brown, Automating the Assignment of Submitted Manuscripts to Reviewers, S.T. Dumais, J. Nielson, Semisupervised Regression with cotraining algorithms, Z. Zhou, M. Li, 2007

References – contd. 5. Learning structured prediction models : A Large Margin Approach, B. Taskar, et al, Ologit : Ordinal Logistic Regression for Zelig, G. King, et al, SVM Learning for Interdependant and Structured Output Spaces, I. Tsochantaridis, et al, Word Alignment via Quadratic Assignment, S. Lacoste-Julien, et al, 2006