In the last few years, a large number of automatic evaluation metrics have been proposed for evaluating Natural Language Generation (NLG) systems. The rapid development and adoption of such automatic evaluation metrics in a relatively short time has created the need for a survey of these metrics. In this survey, we (i) highlight the challenges in automatically evaluating NLG systems, (ii) propose a coherent taxonomy for organising existing evaluation metrics, (iii) briefly describe different existing metrics, and finally (iv) discuss studies criticising the use of automatic evaluation metrics. We then conclude the article highlighting promising future directions of research.

There is an increasing focus on model-based dialog evaluation metrics such as ADEM, RUBER, and the more recent BERT-based metrics. These models aim to assign a high score to all relevant responses and a low score to all irrelevant responses. Ideally, such models should be trained using multiple relevant and irrelevant responses for any given context. However, no such data is publicly available, and hence existing models are usually trained using a single relevant response and multiple randomly selected responses from other contexts (random negatives). To allow for better training and robust evaluation of model-based metrics, we introduce the DailyDialog++ dataset, consisting of (i) five relevant responses for each context and (ii) five adversarially crafted irrelevant responses for each context. Using this dataset, we first show that even in the presence of multiple correct references, n-gram based metrics and embedding based metrics do not perform well at separating relevant responses from even random negatives. While model-based metrics perform better than n-gram and embedding based metrics on random negatives, their performance drops substantially when evaluated on adversarial examples. To check if large scale pretraining could help, we propose a new BERT-based evaluation metric called DEB, which is pretrained on 727M Reddit conversations and then finetuned on our dataset. DEB significantly outperforms existing models, showing better correlation with human judgments and better performance on random negatives (88.27% accuracy). However, its performance again drops substantially when evaluated on adversarial responses, thereby highlighting that even large-scale pretrained evaluation models are not robust to the adversarial examples in our dataset. The dataset 1 and code 2 are publicly available.

The task of Reading Comprehension with Multiple Choice Questions, requires a human (or machine) to read a given {passage, question} pair and select one of the n given options. The current state of the art model for this task first computes a questionaware representation for the passage and then selects the option which has the maximum similarity with this representation. However, when humans perform this task they do not just focus on option selection but use a combination of elimination and selection. Specifically, a human would first try to eliminate the most irrelevant option and then read the passage again in the light of this new information (and perhaps ignore portions corresponding to the eliminated option). This process could be repeated multiple times till the reader is finally ready to select the correct option. We propose ElimiNet, a neural network-based model which tries to mimic this process. Specifically, it has gates which decide whether an option can be eliminated given the {passage, question} pair and if so it tries to make the passage representation orthogonal to this eliminated option (akin to ignoring portions of the passage corresponding to the eliminated option). The model makes multiple rounds of partial elimination to refine the passage representation and finally uses a selection module to pick the best option. We evaluate our model on the recently released large scale RACE dataset and show that it outperforms the current state of the art model on 7 out of the 13 question types in this dataset. Further, we show that taking an ensemble of our elimination-selection based method with a selection based method gives us an improvement of 3.1% over the best-reported performance on this dataset.