Search

A number of natural language processing tools for Urdu language processing have been developed in the past few years for word segmentation, part of speech tagging, chunking, named entity recognition and parsing. Corpora, especially treebanks, are essential data resources for language processing. This work presents the development and evaluation of an Urdu treebank, the CLE-UTB and a statistical parser. The treebank has been annotated with phrase structure annotation. Part of speech tagging has been performed semi-automatically by using an existing tagger and incorrect tags were corrected manually by annotators. The syntactic annotation has been performed in the Penn Treebank style to mark phrases. The annotation scheme also adds functional labels for grammatical roles. Currently, the treebank contains 7854 annotated sentences and 148,575 tokens. Completeness and correctness of the syntactic labels have been checked automatically after manual annotation. To ensure the annotation consistency of the resource, a grammar-based evaluation and an automatic consistency checking tool have been used to detect linguistically implausible constituents. The inter-annotator agreement is greater than 90%. We have developed a bidirectional long-short term memory (BiLSTM) based parser and a POS tagger which have been trained on the final version of the treebank. We have improved our results by training the word embeddings on a large Urdu text corpus. Our parser produced an f-score of 88.1% and the POS tagger performed with an accuracy of 96.3%.

This article presents an analysis of experiments with statistical and neural parsing techniques for Urdu, a widely spoken South Asian language. We demonstrate state of the art constituency parsing results for an Urdu treebank. Urdu is a morphologically rich and is characterized by free word order. Language representation (e.g. input type, lemmatization, word clusters), part of speech tag set, phrase labels and the size of a training corpus are crucial for parsing such languages. In this article, probabilistic context-free grammars, data-oriented parsing, and recursive neural network based models have been experimented with several linguistic features which show improvements in the parsing results. Features include syntactic sub-categorization of POS tags, empirically learned horizontal and vertical markovizations and lexical head words. These features enable dependency information for case markers and add phrasal and lexical context to the parse trees. The data-oriented parsing and recursive neural network model give an f-score of 87.1 by considering gold POS tags in the test set, on textual input, they show a performance with f-scores of 83.4 and 84.2, respectively. To overcome the issue of data sparsity due to the morphological richness, lemmatization and unsupervised word clustering have been performed. A treebank should cover most probable word orders of the language so that models can learn various orders accurately. To analyze the order coverage of the treebank and learning capability of different parsers, a test set has been prepared conditioning different word orders. This test set is evaluated with the best performing parsing models and with gold POS tags, f-scores are above 90 and on textual input, the average f-score is 87.6.

In the original publication of the article the column headers of the Tables 17 and 18 were incorrectly published. The corrected version of Tables 17 and 18 are provided with this Correction.