Interactive Word Completion for Morphologically Complex Languages

Proceedings of the 28th International Conference on Computational Linguistics, pages 4600–4611 Barcelona, Spain (Online), December 8-13, 2020 4600 Interactive Word Completion for Morphologically Complex Languages William Lane and Steven Bird Northern Institute Charles Darwin University Abstract Text input technologies for low-resource languages support literacy, content authoring, and lan- guage learning. However, tasks such as word completion pose a challenge for morphologically complex languages thanks to the combinatorial explosion of possible words. We have devel- oped a method for morphologically-aware text input in Kunwinjku, a polysynthetic language of northern Australia. We modify an existing finite state recognizer to map input morph prefixes to morph completions, respecting the morphosyntax and morphophonology of the language. We demonstrate the portability of the method by applying it to Turkish. We show that the space of proximal morph completions is many orders of magnitude smaller than the space of full word completions for Kunwinjku. We provide a visualization of the morph completion space to enable the text completion parameters to be fine-tuned. Finally, we report on a web services deployment, along with a web interface which helps users enter morphologically complex words and which retrieves corresponding entries from the lexicon. 1 Introduction Indigenous communities in Australia’s far north are engaged in a variety of efforts to maintain their lan- guages. In the Kunwinjku-speaking communities of West Arnhem, the challenges are to maintain orality and develop literacy. Technological solutions can support community efforts to create language resources and make them accessible. These solutions include dictionaries, digital archives, and collaborative ap- plications for documentation and language learning (Hunt et al., 2019; Bow et al., 2014; Fillmore et al., 2019). All of these technologies require text input, posing a challenge for communities whose languages are primarily oral, and where writing—especially writing that conforms to an orthography standard—is not an established practice. In the context of morphologically complex languages, the term “intelligent dictionary” has been ap- plied to systems which morphologically analyze the query in the course of retrieving relevant lexical entries (Johnson et al., 2013; Arppe et al., 2016). This is an important feature for speakers of polysyn- thetic languages like Kunwinjku, which have multiple morph slots with sometimes large or unbounded sets of fillers. However, effective use of an intelligent dictionary still requires the ability to enter the fully-inflected surface form. For speakers and learners of morphologically complex languages, produc- ing well-formed queries may pose a significant challenge. Littell et al. make a similar observation in their work on dictionary interfaces for North American languages, where few speakers are “both fluent in the language and trained in a systematic orthography” (Littell et al., 2017). While they address the problem of retrieving dictionary entries given a noisy query, we go further, guiding users as they type by suggesting completions up to the next morph boundary. The typical definition of “autocomplete” implies word-level completion (Cai et al., 2016; Dunlop and Crossan, 2000), but this is impractical for polysynthetic languages where single words are capable of expressing the meaning of an …

Content truncated to 500 words