Coming to Terms: Automatic Formation of Neologisms in Hebrew

Findings of the Association for Computational Linguistics: EMNLP 2020, pages 4918–4929 November 16 - 20, 2020. c⃝2020 Association for Computational Linguistics 4918 Coming to Terms: Automatic Formation of Neologisms in Hebrew Moran Mizrahi ∗, Stav Yardeni Seelig ∗, Dafna Shahaf The Hebrew University of Jerusalem {moranmiz,stav.yardeni,dshahaf}@cs.huji.ac.il Abstract Spoken languages are ever-changing, with new words entering them all the time. However, coming up with new words (neologisms) to- day relies exclusively on human creativity. In this paper we propose a system to automati- cally suggest neologisms. We focus on the He- brew language as a test case due to the unusual regularity of its noun formation. User studies comparing our algorithm to experts and non- experts demonstrate that our algorithm is capa- ble of generating high-quality outputs, as well as enhance human creativity. More broadly, we seek to inspire more computational work around the topic of linguistic creativity, which we believe offers numerous unexplored oppor- tunities. 1 Introduction Human languages are always changing, evolving, and adapting to the needs of their speakers. New words regularly enter our vocabulary, while others disappear. For example, the word “selfie” (self- portrait digital photo, typically taken with a smart- phone) has recently become part of everyday En- glish, even spawning variations such as helfie (a selfie of one’s hair), welfie (a selfie taken during a workout), and drelfie (a selfie taken while being drunk) (Christiansen and Chater, 2016). Newly coined words or expressions are termed neologisms. There are many neologism formation mechanisms; common ones include loanwords bor- rowed from another language (kindergarten), mor- phological derivation (socialize, simplify), com- pounding (football, breakwater), blending (spoon + fork = spork), and acronyms (laser). Importantly, the coining of novel words relies on human creativity, with the new terms often con- veying a lot of information in an inventive way. In ∗ Both authors contributed equally to this paper. this work, we set out to explore the possibility of automating some of this inherently-human, cre- ative linguistic process. In other words, we ask whether computers can generate high-quality, novel words on their own, or alternatively help inspire people to find better words. We focus on automatic generation of neologisms in the Hebrew language. Hebrew has several prop- erties which make it particularly interesting for our goal: first, modern Hebrew was revived after a long period of time (Rabin, 1963; Fellman, 1973), which is unique. There are no other cases of a natural lan- guage without any native speakers subsequently acquiring millions of native speakers. For this rea- son, foreign words are very common in Hebrew, and many terms need to be coined. Another reason for focusing on Hebrew is its unusual regularity of noun formation. While port- manteaus (word blends), word combinations and other formation mechanisms do exist in Hebrew, most words are created by combination of root and pattern. To the best of our knowledge, this method of word generation was not explored before in a computational context. Our contributions are: • We propose a novel task, automating the …

Content truncated to 500 words