Search

We review the task of Sentence Pair Scoring, popular in the literature in various forms - viewed as Answer Sentence Selection, Semantic Text Scoring, Next Utterance Ranking, Recognizing Textual Entailment, Paraphrasing or e.g. a component of Memory Networks. We argue that all such tasks are similar from the model perspective and propose new baselines by comparing the performance of common IR metrics and popular convolutional, recurrent and attention-based neural models across many Sentence Pair Scoring tasks and datasets. We discuss the problem of evaluating randomized models, propose a statistically grounded methodology, and attempt to improve comparisons by releasing new datasets that are much harder than some of the currently used well explored benchmarks. We introduce a unified open source software framework with easily pluggable models and tasks, which enables us to experiment with multi-task reusability of trained sentence model. We set a new state-of-art in performance on the Ubuntu Dialogue dataset., Comment: submitted as paper to CoNLL 2016

We study the computational complexity of graph planarization via edge contraction. The problem CONTRACT asks whether there exists a set $S$ of at most $k$ edges that when contracted produces a planar graph. We work with a more general problem called $P$-RESTRICTEDCONTRACT in which $S$, in addition, is required to satisfy a fixed MSOL formula $P(S,G)$. We give an FPT algorithm in time $O(n^2 f(k))$ which solves $P$-RESTRICTEDCONTRACT, where $P(S,G)$ is (i) inclusion-closed and (ii) inert contraction-closed (where inert edges are the edges non-incident to any inclusion minimal solution $S$). As a specific example, we can solve the $\ell$-subgraph contractibility problem in which the edges of a set $S$ are required to form disjoint connected subgraphs of size at most $\ell$. This problem can be solved in time $O(n^2 f'(k,\ell))$ using the general algorithm. We also show that for $\ell \ge 2$ the problem is NP-complete.