Learning to Rank Answer Candidates for Automatic Resolution of Crossword Puzzles

Abstract

In this paper, we study the impact of relational and syntactic representations for an interesting and challenging task: the automatic resolution of crossword puzzles. Automatic solvers are typically based on two answer retrieval modules: (i) a web search engine, e.g., Google, Bing, etc. and (ii) a database (DB) system for accessing previously resolved crossword puzzles. We show that learning to rank models based on relational syntactic structures defined between the clues and the answer can improve both modules above. In particular, our approach accesses the DB using a search engine and reranks its output by modeling paraphrasing. This improves on the MRR of previous system up to 53% in ranking answer candidates and greatly impacts on the resolution accuracy of crossword puzzles up to 15%.

LiMoSINe Pipeline: Multilingual UIMA-based NLP Platform

“…Since the beginning of the LiMoSINe project, the platform has been used for providing robust preprocessing for a variety of high-level tasks. Thus, we have recently shown how structural representations, extracted with our pipeline, improve multilingual opinion mining on YouTube (Severyn et al, 2015) or crossword puzzle resolution (Barlacchi et al, 2014).…”

Solving and Generating Chinese Character Riddles

“…And the rule-based module and the dictionary module are mentioned in his work. The tree kernel is used to rerank the candidates proposed by Barlacchi et al (2014) on Chinese language cultures, such as the couplet generation and the poem generation. A statistical machine translation (SMT) framework is proposed to generate Chinese couplets and classic Chinese poetry (He et al, 2012;Zhou et al, 2009;Jiang and Zhou, 2008).…”

“…And the rule-based module and the dictionary module are mentioned in his work. The tree kernel is used to rerank the candidates proposed by Barlacchi et al (2014) for automatic resolution of crossword puzzles.…”

Distributional Neural Networks for Automatic Resolution of Crossword Puzzles

“…Then, both query and candidates are represented by shallow syntactic structures (generated by running a set of NLP parsers) and traditional similarity features which are fed to a kernelbased reranker. Hereafter, we give a brief description of our models for clue reranking whereas the reader can refer to our previous work (Barlacchi et al, 2014a;Barlacchi et al, 2014b) for more specific details.…”

“…In (Barlacchi et al, 2014a), we proposed the BM25 retrieval model to generate clue lists, which were further refined by applying our reranking models. The latter promote the most similar, which are probably associated with the same answer of the query clue, to the top.…”