Neural Arabic Text Diacritization: State-of-the-Art Results and a Novel Approach for Arabic NLP Downstream Tasks

In this work, we present several deep learning models for the automatic diacritization of Arabic text. Our models are built using two main approaches, viz. Feed-Forward Neural Network (FFNN) and Recurrent Neural Network (RNN), with several enhancements such as 100-hot encoding, embeddings, Conditional Random Field (CRF), and Block-Normalized Gradient (BNG). The models are tested on the only freely available benchmark dataset and the results show that our models are either better or on par with other models even those requiring human-crafted language-dependent post-processing steps, unlike ours. Moreover, we show how diacritics in Arabic can be used to enhance the models of downstream NLP tasks such as Machine Translation (MT) and Sentiment Analysis (SA) by proposing novel Translation over Diacritization (ToD) and Sentiment over Diacritization (SoD) approaches.

A Context-Enhanced De-identification System

Many modern entity recognition systems, including the current state-of-the-art de-identification systems, are based on bidirectional long short-term memory (biLSTM) units augmented by a conditional random field (CRF) sequence optimizer. These systems process the input sentence by sentence. This approach prevents the systems from capturing dependencies over sentence boundaries and makes accurate sentence boundary detection a prerequisite. Since sentence boundary detection can be problematic especially in clinical reports, where dependencies and co-references across sentence boundaries are abundant, these systems have clear limitations. In this study, we built a new system on the framework of one of the current state-of-the-art de-identification systems, NeuroNER, to overcome these limitations. This new system incorporates context embeddings through forward and backward n -grams without using sentence boundaries. Our context-enhanced de-identification (CEDI) system captures dependencies over sentence boundaries and bypasses the sentence boundary detection problem altogether. We enhanced this system with deep affix features and an attention mechanism to capture the pertinent parts of the input. The CEDI system outperforms NeuroNER on the 2006 i2b2 de-identification challenge dataset, the 2014 i2b2 shared task de-identification dataset, and the 2016 CEGS N-GRID de-identification dataset ( p < 0.01 ). All datasets comprise narrative clinical reports in English but contain different note types varying from discharge summaries to psychiatric notes. Enhancing CEDI with deep affix features and the attention mechanism further increased performance.

A Modified Markov Based Maximum-entropy Model for POS Tagging of Odia Text

POS (Parts of Speech) tagging, a vital step in diverse Natural Language Processing (NLP) tasks has not drawn much attention in case of Odia a computationally under-developed language. The proposed hybrid method suggests a robust POS tagger for Odia. Observing the rich morphology of the language and unavailability of sufficient annotated text corpus a combination of machine learning and linguistic rules is adopted in the building of the tagger. The tagger is trained on tagged text corpus from the domain of tourism and is capable of obtaining a perceptible improvement in the result. Also an appreciable performance is observed for news articles texts of varied domains. The performance of proposed algorithm experimenting on Odia language shows its manifestation in dominating over existing methods like rule based, hidden Markov model (HMM), maximum entropy (ME) and conditional random field (CRF).

Deep Learning Approach for the Morphological Synthesis in Malayalam and Tamil at the Character Level

Morphological synthesis is one of the main components of Machine Translation (MT) frameworks, especially when any one or both of the source and target languages are morphologically rich. Morphological synthesis is the process of combining two words or two morphemes according to the Sandhi rules of the morphologically rich language. Malayalam and Tamil are two languages in India which are morphologically abundant as well as agglutinative. Morphological synthesis of a word in these two languages is challenging basically because of the following reasons: (1) Abundance in morphology; (2) Complex Sandhi rules; (3) The possibilty in Malayalam to form words by combining words that belong to different syntactic categories (for example, noun and verb); and (4) The construction of a sentence by combining multiple words. We formulated the task of the morphological generation of nouns and verbs of Malayalam and Tamil as a character-to-character sequence tagging problem. In this article, we used deep learning architectures like Recurrent Neural Network (RNN) , Long Short-Term Memory Networks (LSTM) , Gated Recurrent Unit (GRU) , and their stacked and bidirectional versions for the implementation of morphological synthesis at the character level. In addition to that, we investigated the performance of the combination of the aforementioned deep learning architectures and the Conditional Random Field (CRF) in the morphological synthesis of nouns and verbs in Malayalam and Tamil. We observed that the addition of CRF to the Bidirectional LSTM/GRU architecture achieved more than 99% accuracy in the morphological synthesis of Malayalam and Tamil nouns and verbs.

Towards Tokenization and Part-of-Speech Tagging for Khmer: Data and Discussion

As a highly analytic language, Khmer has considerable ambiguities in tokenization and part-of-speech (POS) tagging processing. This topic is investigated in this study. Specifically, a 20,000-sentence Khmer corpus with manual tokenization and POS-tagging annotation is released after a series of work over the last 4 years. This is the largest morphologically annotated Khmer dataset as of 2020, when this article was prepared. Based on the annotated data, experiments were conducted to establish a comprehensive benchmark on the automatic processing of tokenization and POS-tagging for Khmer. Specifically, a support vector machine, a conditional random field (CRF) , a long short-term memory (LSTM) -based recurrent neural network, and an integrated LSTM-CRF model have been investigated and discussed. As a primary conclusion, processing at morpheme-level is satisfactory for the provided data. However, it is intrinsically difficult to identify further grammatical constituents of compounds or phrases because of the complex analytic features of the language. Syntactic annotation and automatic parsing for Khmer will be scheduled in the near future.