Neural models command state-of-the-art performance across NLP tasks, including ones involving “reasoning”. Models claiming to reason about the evidence presented to them should attend to the correct parts of the input while avoiding spurious patterns therein, be self-consistent in their predictions across inputs, and be immune to biases derived from their pre-training in a nuanced, context- sensitive fashion. Do the prevalent *BERT- family of models do so? In this paper, we study this question using the problem of reasoning on tabular data. Tabular inputs are especially well-suited for the study—they admit systematic probes targeting the properties listed above. Our experiments demonstrate that a RoBERTa-based model, representative of the current state-of-the-art, fails at reasoning on the following counts: it (a) ignores relevant parts of the evidence, (b) is over- sensitive to annotation artifacts, and (c) relies on the knowledge encoded in the pre-trained language model rather than the evidence presented in its tabular inputs. Finally, through inoculation experiments, we show that fine- tuning the model on perturbed data does not help it overcome the above challenges.

Code-switching is a phenomenon of mixing grammatical structures of two or more languages under varied social constraints. The code-switching data differ so radically from the benchmark corpora used in NLP community that the application of standard technologies to these data degrades their performance sharply. Unlike standard corpora, these data often need to go through additional processes such as language identification, normalization and/or back-transliteration for their efficient processing. In this paper, we investigate these indispensable processes and other problems associated with syntactic parsing of code-switching data and propose methods to mitigate their effects. In particular, we study dependency parsing of code-switching data of Hindi and English multilingual speakers from Twitter. We present a treebank of Hindi-English code-switching tweets under Universal Dependencies scheme and propose a neural stacking model for parsing that efficiently leverages the part-of-speech tag and syntactic tree annotations in the code-switching treebank and the preexisting Hindi and English treebanks. We also present normalization and back-transliteration models with a decoding process tailored for code-switching data. Results show that our neural stacking parser is 1.5% LAS points better than the augmented parsing model and 3.8% LAS points better than the one which uses first-best normalization and/or back-transliteration.

This paper describes our system (SLT-Interactions) for the CoNLL 2018 shared task: Multilingual Parsing from Raw Text to Universal Dependencies. Our system performs three main tasks: word segmentation (only for few treebanks), POS tagging and parsing. While segmentation is learned separately, we use neural stacking for joint learning of POS tagging and parsing tasks. For all the tasks, we employ simple neural network architectures that rely on long short-term memory (LSTM) networks for learning task-dependent features. At the basis of our parser, we use an arc-standard algorithm with Swap action for general non-projective parsing. Additionally, we use neural stacking as a knowledge transfer mechanism for cross-domain parsing of low resource domains. Our system shows substantial gains against the UDPipe baseline, with an average improvement of 4.18% in LAS across all languages. Overall, we are placed at the 12th position on the official test sets.

In this paper, we propose efficient and less resource-intensive strategies for parsing of code-mixed data. These strategies are not constrained by in-domain annotations, rather they leverage pre-existing monolingual annotated resources for training. We show that these methods can produce significantly better results as compared to an informed baseline. Due to lack of an evaluation set for code-mixed structures, we also present a data set of 450 Hindi and English code-mixed tweets of Hindi multilingual speakers for evaluation.

We investigate the problem of parsing conversational data of morphologically-rich languages such as Hindi where argument scrambling occurs frequently. We evaluate a state-of-the-art non-linear transition-based parsing system on a new dataset containing 506 dependency trees for sentences from Bollywood (Hindi) movie scripts and Twitter posts of Hindi monolingual speakers. We show that a dependency parser trained on a newswire treebank is strongly biased towards the canonical structures and degrades when applied to conversational data. Inspired by Transformational Generative Grammar (Chomsky, 1965), we mitigate the sampling bias by generating all theoretically possible alternative word orders of a clause from the existing (kernel) structures in the treebank. Training our parser on canonical and transformed structures improves performance on conversational data by around 9% LAS over the baseline newswire parser.

This paper describes our efforts for the development of a Proposition Bank for Urdu, an Indo-Aryan language. Our primary goal is the labeling of syntactic nodes in the existing Urdu dependency Treebank with specific argument labels. In essence, it involves annotation of predicate argument structures of both simple and complex predicates in the Treebank corpus. We describe the overall process of building the PropBank of Urdu. We discuss various statistics pertaining to the Urdu PropBank and the issues which the annotators encountered while developing the PropBank. We also discuss how these challenges were addressed to successfully expand the PropBank corpus. While reporting the Inter-annotator agreement between the two annotators, we show that the annotators share similar understanding of the annotation guidelines and of the linguistic phenomena present in the language. The present size of this Propbank is around 180,000 tokens which is double-propbanked by the two annotators for simple predicates. Another 100,000 tokens have been annotated for complex predicates of Urdu.

In Computational Linguistics, Hindi and Urdu are not viewed as a monolithic entity and have received separate attention with respect to their text processing. From part-of-speech tagging to machine translation, models are separately trained for both Hindi and Urdu despite the fact that they represent the same language. The reasons mainly are their divergent literary vocabularies and separate orthographies, and probably also their political status and the social perception that they are two separate languages. In this article, we propose a simple but efficient approach to bridge the lexical and orthographic differences between Hindi and Urdu texts. With respect to text processing, addressing the differences between the Hindi and Urdu texts would be beneficial in the following ways: (a) instead of training separate models, their individual resources can be augmented to train single, unified models for better generalization, and (b) their individual text processing applications can be used interchangeably under varied resource conditions. To remove the script barrier, we learn accurate statistical transliteration models which use sentence-level decoding to resolve word ambiguity. Similarly, we learn cross-register word embeddings from the harmonized Hindi and Urdu corpora to nullify their lexical divergences. As a proof of the concept, we evaluate our approach on the Hindi and Urdu dependency parsing under two scenarios: (a) resource sharing, and (b) resource augmentation. We demonstrate that a neural network-based dependency parser trained on augmented, harmonized Hindi and Urdu resources performs significantly better than the parsing models trained separately on the individual resources. We also show that we can achieve near state-of-the-art results when the parsers are used interchangeably.

Kashmiri is a resource poor language with very less computational and language resources available for its text processing. As the main contribution of this paper, we present an initial version of the Kashmiri Dependency Treebank. The treebank consists of 1,000 sentences (17,462 tokens), annotated with part-of-speech (POS), chunk and dependency information. The treebank has been manually annotated using the Paninian Computational Grammar (PCG) formalism (Begum et al., 2008; Bharati et al., 2009). This version of Kashmiri treebank is an extension of its earlier verion of 500 sentences (Bhat, 2012), a pilot experiment aimed at defining the annotation guidelines on a small subset of Kashmiri corpora. In this paper, we have refined the guidelines with some significant changes and have carried out inter-annotator agreement studies to ascertain its quality. We also present a dependency parsing pipeline, consisting of a tokenizer, a stemmer, a POS tagger, a chunker and an inter-chunk dependency parser. It, therefore, constitutes the first freely available, open source dependency parser of Kashmiri, setting the initial baseline for Kashmiri dependency parsing.