User Response Driven Content Understanding with Causal Inference

Tan, Fei; Wei, Zhi; Pani, Abhishek; Yan, Zhen

doi:10.1109/icdm.2019.00168

Cited by 2 publications

(4 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this work, our focus is understanding the effect of specific characteristics of text on the outcome of interest. Previous work in this area has studied various text characteristics and outcomes, such as effect of words on sentiment classification (Paul, 2017), effect of presence of theorems on the acceptance rate of papers and the effect of gender on the popularity of social media posts (Veitch et al, 2020), and the effect of specific content features on the user response (Tan et al, 2019;. These work focus on identifying the effect of textual features on the outcome.…”

Section: Related Workmentioning

confidence: 99%

“…π(x i ) = p(t i = 1|x i ). We employ multi-layer neural networks to approximate propensity scores (Tan et al, 2019). The propensity scoring model is trained using the assigned treatment t i corresponding to the observed covariates x i with cross entropy loss.…”

Section: Causal Features Identificationmentioning

confidence: 99%

“…This provides an estimate of which text features have the most impact on the outcome (target) metric. 2 The ATE of continuous treatment features can be estimated in a similar fashion, assuming a normal treatment distribution (Tan et al, 2019).…”

Section: Causal Features Identificationmentioning

confidence: 99%

“…These engagement metrics (hereafter, metrics) are proxy for target goals. Based on historical content, textual content characteristics that successfully achieve the desired metrics can be assessed (Tan et al, 2019;. Guiding text generation models by these signals is important for meeting the required goals.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

CaM-Gen: Causally Aware Metric-Guided Text Generation

Goyal¹,

Paneri²,

Agarwal³

et al. 2022

Findings of the Association for Computational Linguistics: ACL 2022

View full text Add to dashboard Cite

Content is created for a well-defined purpose, often described by a metric or signal represented in the form of structured information. The relationship between the goal (metrics) of target content and the content itself is nontrivial. While large-scale language models show promising text generation capabilities, guiding the generated text with external metrics is challenging. These metrics and content tend to have inherent relationships and not all of them may be of consequence. We introduce CaM-Gen: Causally aware Generative Networks guided by user-defined target metrics incorporating the causal relationships between the metric and content features. We leverage causal inference techniques to identify causally significant aspects of a text that lead to the target metric and then explicitly guide generative models towards these by a feedback mechanism. We propose this mechanism for variational autoencoder and Transformer-based generative models. The proposed models beat baselines in terms of the target metric control while maintaining fluency and language quality of the generated text. To the best of our knowledge, this is one of the early attempts at controlled generation incorporating a metric guide using causal inference.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Causal Features Identificationmentioning

confidence: 99%

Section: Causal Features Identificationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations