BRAINSUP: Brainstorming Support for Creative Sentence Generation

Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics, pages 1446–1455, Sofia, Bulgaria, August 4-9 2013. c⃝2013 Association for Computational Linguistics BRAINSUP: Brainstorming Support for Creative Sentence Generation G¨ozde ¨Ozbal FBK-irst Trento, Italy gozbalde@gmail.com Daniele Pighin Google Inc. Z¨urich, Switzerland daniele.pighin@gmail.com Carlo Strapparava FBK-irst Trento, Italy strappa@fbk.eu Abstract We present BRAINSUP, an extensible framework for the generation of creative sentences in which users are able to force several words to appear in the sen- tences and to control the generation pro- cess across several semantic dimensions, namely emotions, colors, domain related- ness and phonetic properties. We evalu- ate its performance on a creative sentence generation task, showing its capability of generating well-formed, catchy and effec- tive sentences that have all the good qual- ities of slogans produced by human copy- writers. 1 Introduction A variety of real-world scenarios involve talented and knowledgable people in a time-consuming process to write creative, original sentences gen- erated according to well-defined requisites. For instance, to advertise a new product it could be desirable to have its name appearing in a punchy sentence together with some keywords relevant for marketing, e.g. “fresh”, or “thirst” for the adver- tisement of a drink. Besides, it could be interesting to characterize the sentence with respect to a spe- cific color, like “blue” to convey the idea of fresh- ness, or to a color more related to the brand of the company, e.g. “red” for a new Ferrari. Moreover, making the slogan evoke “joy” or “satisfaction” could make the advertisement even more catchy for customers. On the other hand, there are many examples of provocative slogans in which copy- writers try to impress their readers by suscitating strong negative feelings, as in the case of anti- smoke campaigns (e.g., “there are cooler ways to die than smoking” or “cancer cures smoking”), or the famous beer motto “Guinness is not good for you”. As another scenario, creative sentence gen- eration is also a useful teaching device. For ex- ample, the keyword or linkword method used for second language learning links the translation of a foreign (target) word to one or more keywords in the native language which are phonologically or lexically similar to the target word (Sagarra and Alba, 2006). To illustrate, for teaching the Ital- ian word “tenda”, which means “curtain” in En- glish, the learners are asked to imagine “rubbing a tender part of their leg with a curtain”. These words should co-occur in the same sentence, but constructing such sentences by hand can be a dif- ficult and very time-consuming process. ¨Ozbal and Strapparava (2011), who attempted to auto- mate the process, conclude that the inability to re- trieve from the web a good sentence for all cases is a major bottleneck. Although state of the art computational mod- els of creativity often produce remarkable results, e.g., Manurung et al. (2008), Greene et al. (2010), Guerini et al. (2011), Colton et al. (2012) just to name a few, to our best knowledge there is …

Content truncated to 500 words