Global autoregressive models for data-efficient sequence learning

@inproceedings{parshakova-etal-2019-global,
    title = "Global Autoregressive Models for Data-Efficient Sequence Learning",
    author = "Parshakova, Tetiana  and
      Andreoli, Jean-Marc  and
      Dymetman, Marc",
    booktitle = "Proceedings of the 23rd Conference on Computational Natural Language Learning (CoNLL)",
    month = nov,
    year = "2019",
    address = "Hong Kong, China",
    publisher = "Association for Computational Linguistics",
    url = "https://www.aclweb.org/anthology/K19-1084",
    doi = "10.18653/v1/K19-1084",
    pages = "900--909",
    abstract = "Standard autoregressive seq2seq models are easily trained by max-likelihood, but tend to show poor results under small-data conditions. We introduce a class of seq2seq models, GAMs (Global Autoregressive Models), which combine an autoregressive component with a log-linear component, allowing the use of global {\textbackslash}textit{a priori} features to compensate for lack of data. We train these models in two steps. In the first step, we obtain an {\textbackslash}emph{unnormalized} GAM that maximizes the likelihood of the data, but is improper for fast inference or evaluation. In the second step, we use this GAM to train (by distillation) a second autoregressive model that approximates the {\textbackslash}emph{normalized} distribution associated with the GAM, and can be used for fast inference and evaluation. Our experiments focus on language modelling under synthetic conditions and show a strong perplexity reduction of using the second autoregressive model over the standard one.",
}