Scaling a Natural Language Generation System

Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics, pages 1148–1157, Berlin, Germany, August 7-12, 2016. c⃝2016 Association for Computational Linguistics Scaling a Natural Language Generation System Jonathan Pfeil Department of EECS Case Western Reserve University Cleveland, OH, USA jonathan.pfeil@case.edu Soumya Ray Department of EECS Case Western Reserve University Cleveland, OH, USA sray@case.edu Abstract A key goal in natural language genera- tion (NLG) is to enable fast generation even with large vocabularies, grammars and worlds. In this work, we build upon a recently proposed NLG system, Sentence Tree Realization with UCT (STRUCT). We describe four enhancements to this system: (i) pruning the grammar based on the world and the communicative goal, (ii) intelligently caching and pruning the com- binatorial space of semantic bindings, (iii) reusing the lookahead search tree at differ- ent search depths, and (iv) learning and us- ing a search control heuristic. We evaluate the resulting system on three datasets of increasing size and complexity, the largest of which has a vocabulary of about 10K words, a grammar of about 32K lexical- ized trees and a world with about 11K enti- ties and 23K relations between them. Our results show that the system has a median generation time of 8.5s and finds the best sentence on average within 25s. These re- sults are based on a sequential, interpreted implementation and are significantly bet- ter than the state of the art for planning- based NLG systems. 1 Introduction and Related Work We consider the restricted natural language gen- eration (NLG) problem (Reiter and Dale, 1997): given a grammar, lexicon, world and a commu- nicative goal, output a valid sentence that satis- fies this goal. Though restricted, this problem is still challenging when the NLG system has to deal with the large probabilistic grammars of natural language, large knowledge bases representing re- alistic worlds with many entities and relations be- tween them, and complex communicative goals. Prior work has approach NLG from two di- rections. One strategy is over-generation and ranking, in which an intermediate structure gen- erates many candidate sentences which are then ranked according to how well they match the goal. This includes systems built on chart parsers (Shieber, 1988; Kay, 1996; White and Baldridge, 2003), systems that use forest architectures such as HALogen/Nitrogen, (Langkilde-Geary, 2002), systems that use tree conditional random fields (Lu et al., 2009), and newer systems that use recur- rent neural networks (Wen et al., 2015b; Wen et al., 2015a). Another strategy formalizes NLG as a goal-directed planning problem to be solved using an automated planner. This plan is then semanti- cally enriched, followed by surface realization to turn it into natural language. This is often viewed as a pipeline generation process (Reiter and Dale, 1997). An alternative to pipeline generation is inte- grated generation, in which the sentence plan- ning and surface realization tasks happen simul- taneously (Reiter and Dale, 1997). CRISP (Koller and Stone, 2007) and PCRISP (Bauer and Koller, 2010) are two such systems. These generators en- code …

Content truncated to 500 words