An official website of the United States government
Discourse parsing has proven to be useful for a number of NLP tasks that require complex reasoning. However, over a decade since the advent of the Penn Discourse Treebank, predicting implicit discourse relations in text remains challenging. There are several possible reasons for this, and we hypothesize that models should be exposed to more context as it plays an important role in accurate human annotation; meanwhile adding uncertainty measures can improve model accuracy and calibration. To thoroughly investigate this phenomenon, we perform a series of experiments to determine 1) the effects of context on human judgments, and 2) the effect of quantifying uncertainty with annotator confidence ratings on model accuracy and calibration (which we measure using the Brier score (Brier et al, 1950)). We find that including annotator accuracy and confidence improves model accuracy, and incorporating confidence in the model’s temperature function can lead to models with significantly better-calibrated confidence measures. We also find some insightful qualitative results regarding human and model behavior on these datasets.
more »
« less
Uncertainty over Uncertainty: Investigating the Assumptions, Annotations, and Text Measurements of Economic Policy Uncertainty
Methods and applications are inextricably linked in science, and in particular in the domain of text-as-data. In this paper, we examine one such text-as-data application, an established economic index that measures economic policy uncertainty from keyword occurrences in news. This index, which is shown to correlate with firm investment, employment, and excess market returns, has had substantive impact in both the private sector and academia. Yet, as we revisit and extend the original authors’ annotations and text measurements we find interesting text-as-data methodological research questions: (1) Are annotator disagreements a reflection of ambiguity in language? (2) Do alternative text measurements correlate with one another and with measures of external predictive validity? We find for this application (1) some annotator disagreements of economic policy uncertainty can be attributed to ambiguity in language, and (2) switching measurements from keyword-matching to supervised machine learning classifiers results in low correlation, a concerning implication for the validity of the index.
more »
« less
- Award ID(s):
- 1845576
- PAR ID:
- 10281911
- Date Published:
- Journal Name:
- Proceedings of the Fourth Workshop on Natural Language Processing and Computational Social Science
- Page Range / eLocation ID:
- 116-131
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
When groups of people are tasked with making a judgment, the issue of uncertainty often arises. Existing methods to reduce uncertainty typically focus on iteratively improving specificity in the overall task instruction. However, uncertainty can arise from multiple sources, such as ambiguity of the item being judged due to limited context, or disagreements among the participants due to different perspectives and an under-specified task. A one-size-fits-all intervention may be ineffective if it is not targeted to the right source of uncertainty. In this paper we introduce a new workflow, Judgment Sieve, to reduce uncertainty in tasks involving group judgment in a targeted manner. By utilizing measurements that separate different sources of uncertainty during an initial round of judgment elicitation, we can then select a targeted intervention adding context or deliberation to most effectively reduce uncertainty on each item being judged. We test our approach on two tasks: rating word pair similarity and toxicity of online comments, showing that targeted interventions reduced uncertainty for the most uncertain cases. In the top 10% of cases, we saw an ambiguity reduction of 21.4% and 25.7%, and a disagreement reduction of 22.2% and 11.2% for the two tasks respectively. We also found through a simulation that our targeted approach reduced the average uncertainty scores for both sources of uncertainty as opposed to uniform approaches where reductions in average uncertainty from one source came with an increase for the other.more » « less
-
Abstract We investigate how disagreement in natural language inference (NLI) annotation arises. We developed a taxonomy of disagreement sources with 10 categories spanning 3 high- level classes. We found that some disagreements are due to uncertainty in the sentence meaning, others to annotator biases and task artifacts, leading to different interpretations of the label distribution. We explore two modeling approaches for detecting items with potential disagreement: a 4-way classification with a “Complicated” label in addition to the three standard NLI labels, and a multilabel classification approach. We found that the multilabel classification is more expressive and gives better recall of the possible interpretations in the data.more » « less
-
As text generated by large language models proliferates, it becomes vital to understand how humans engage with such text, and whether or not they are able to detect when the text they are reading did not originate with a human writer. Prior work on human detection of generated text focuses on the case where an entire passage is either human-written or machine-generated. In this paper, we study a more realistic setting where text begins as human-written and transitions to being generated by state-of-the-art neural language models. We show that, while annotators often struggle at this task, there is substantial variance in annotator skill and that given proper incentives, annotators can improve at this task over time. Furthermore, we conduct a detailed comparison study and analyze how a variety of variables (model size, decoding strategy, fine-tuning, prompt genre, etc.) affect human detection performance. Finally, we collect error annotations from our participants and use them to show that certain textual genres influence models to make different types of errors and that certain sentence-level features correlate highly with annotator selection. We release the RoFT dataset: a collection of over 21,000 human annotations paired with error classifications to encourage future work in human detection and evaluation of generated text.more » « less
-
As text generated by large language models proliferates, it becomes vital to understand how humans engage with such text, and whether or not they are able to detect when the text they are reading did not originate with a human writer. Prior work on human detection of generated text focuses on the case where an entire passage is either human-written or machine-generated. In this paper, we study a more realistic setting where text begins as human-written and transitions to being generated by state-of-the-art neural language models. We show that, while annotators often struggle at this task, there is substantial variance in annotator skill and that given proper incentives, annotators can improve at this task over time. Furthermore, we conduct a detailed comparison study and analyze how a variety of variables (model size, decoding strategy, fine-tuning, prompt genre, etc.) affect human detection performance. Finally, we collect error annotations from our participants and use them to show that certain textual genres influence models to make different types of errors and that certain sentence-level features correlate highly with annotator selection. We release the RoFT dataset: a collection of over 21,000 human annotations paired with error classifications to encourage future work in human detection and evaluation of generated text.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/http://dx.doi.org/10.18653/v1/2020.nlpcss-1.13
