Charmanteau: Character Embedding Models For Portmanteau Creation

Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, pages 2917–2922 Copenhagen, Denmark, September 7–11, 2017. c⃝2017 Association for Computational Linguistics CharManteau: Character Embedding Models For Portmanteau Creation Varun Gangal ∗, Harsh Jhamtani ∗, Graham Neubig, Eduard Hovy, Eric Nyberg Language Technologies Institute, Carnegie Mellon University {vgangal,jharsh,gneubig,hovy,ehn}@cs.cmu.edu Abstract Portmanteaus are a word formation phe- nomenon where two words are com- bined to form a new word. We pro- pose character-level neural sequence-to- sequence (S2S) methods for the task of portmanteau generation that are end-to- end-trainable, language independent, and do not explicitly use additional phonetic information. We propose a noisy-channel- style model, which allows for the incorpo- ration of unsupervised word lists, improv- ing performance over a standard source- to-target model. This model is made pos- sible by an exhaustive candidate genera- tion strategy specifically enabled by the features of the portmanteau task. Ex- periments find our approach superior to a state-of-the-art FST-based baseline with respect to ground truth accuracy and hu- man evaluation. 1 Introduction Portmanteaus (or lexical blends Algeo (1977)) are novel words formed from parts of multiple root words in order to refer to a new concept which can’t otherwise be expressed concisely. Portman- teaus have become frequent in modern-day social media, news reports and advertising, one popu- lar example being Brexit (Britain + Exit). Petri (2012). These are found not only in English but many other languages such as Bahasa Indone- sia Dardjowidjojo (1979), Modern Hebrew Bat- El (1996); Berman (1989) and Spanish Pi˜neros (2004). Their short length makes them ideal for headlines and brandnames (Gabler, 2015). Unlike better-defined morphological phenomenon such as inflection and derivation, portmanteau generation ∗* denotes equal contribution Figure 1: A sketch of our BACKWARD, noisy- channel model. The attentional S2S model with bidirectional encoder gives P(x|y) and next- character model gives P(y), where y (spime) is the portmanteau and x = concat(x(1), “;”, x(2)) are the concatenated root words (space and time). is difficult to capture using a set of rules. For instance, Shaw et al. (2014) state that the com- position of the portmanteau from its root words depends on several factors, two important ones being maintaining prosody and retaining charac- ter segments from the root words, especially the head. An existing work by Deri and Knight (2015) aims to solve the problem of predicting portman- teau using a multi-tape FST model, which is data- driven, unlike prior approaches. Their methods rely on a grapheme to phoneme converter, which takes into account the phonetic features of the lan- guage, but may not be available or accurate for non-dictionary words, or low resource languages. Prior works, such as Faruqui et al. (2016), have demonstrated the efficacy of neural approaches for morphological tasks such as inflection. We hy- pothesize that such neural methods can (1) pro- vide a simpler and more integrated end-to-end framework than multiple FSTs used in the previ- ous work, and (2) automatically capture features such as phonetic similarity through the use of char- acter embeddings, removing …

Content truncated to 500 words