A Pointer Network Architecture for Joint Morphological Segmentation and Tagging

Findings of the Association for Computational Linguistics: EMNLP 2020, pages 4368–4378 November 16 - 20, 2020. c⃝2020 Association for Computational Linguistics 4368 A Pointer Network Architecture for Joint Morphological Segmentation and Tagging Amit Seker Open University, Ra’anana, Israel aseker00@gmail.com Reut Tsarfaty Bar Ilan University, Israel reut.tsarfaty@biu.ac.il Abstract Morphologically Rich Languages (MRLs) such as Arabic, Hebrew and Turkish often require Morphological Disambiguation (MD), i.e., the prediction of the correct morphologi- cal decomposition of tokens into morphemes, early in the pipeline. Neural MD may be ad- dressed as a simple pipeline, where segmen- tation is followed by sequence tagging, or as an end-to-end model, predicting morphemes from raw tokens. Both approaches are sub- optimal; the former is heavily prone to er- ror propagation, and the latter does not enjoy explicit access to the basic processing units called morphemes. This paper offers an MD ar- chitecture that combines the symbolic knowl- edge of morphemes with the learning capacity of neural end-to-end modeling. We propose a new, general and easy-to-implement Pointer Network model where the input is a morpho- logical lattice and the output is a sequence of indices pointing at a single disambiguated path of morphemes. We demonstrate the ef- ficacy of the model on segmentation and tag- ging, for Hebrew and Turkish texts, based on their respective Universal Dependencies (UD) treebanks. Our experiments show that with complete lattices, our model outperforms all shared-task results on segmenting and tagging these languages. On the SPMRL treebank, our model outperforms all previously reported re- sults for Hebrew MD in realistic scenarios. 1 Introduction In Morphologically Rich Languages (MRLs) (Tsar- faty et al., 2010), raw tokens are morphologically ambiguous, complex, and consist of sub-token units referred to as morphemes.1 Morphological Disambiguation (MD) is the task of decompos- ing the tokens into their constituent morphemes, 1In Universal Dependencies terms, these are called syntac- tic words, to be distinguished from raw input tokens. to be used as the basic processing units for NLP tasks down the pipeline (Mueller et al., 2013; More and Tsarfaty, 2016). As opposed to the commonly known scenario of morphological tagging (Bohnet et al., 2013), where every input token is assigned a single morphological signature (containing its lemma, part-of-speech tag, and morphological fea- tures such as gender, number, person, tense, etc.), in the MD scenario internally-complex input to- kens may consist of multiple distinct units, each of which gets assigned its own morphological signa- ture. Pre-neural statistical approaches for MD (Bar- haim et al., 2008; Adler and Elhadad, 2006a; Lee et al., 2011; Habash et al., 2013) typically used weighted finite-state machines to unravel the pos- sible morphological decompositions, and classic machine learning models to select the most likely decomposition. Current neural models, however, take radically different paths. One neural approach to MD employs pipeline, where a predicted segmentation of words into mor- phemes is passed on to sequence labeling com- ponent that performs tagging of each segment in context. This segmentation-first scenario employs sequence tagging to assign a single morphological tag to each …

Content truncated to 500 words