Dialogue

A general multimodal attention unit for any number of modalities. Graphical models inspire it, i.e., it infers several attention beliefs via aggregated interaction messages.

ERNIE-GEN is a multi-flow sequence to sequence pre-training and fine-tuning framework which bridges the discrepancy between training and inference with an infilling generation mechanism and a noise-aware generation method. To make generation closer to human writing patterns, this framework introduces a span-by-span generation flow that trains the model to predict semantically-complete spans consecutively rather than predicting word by word. Unlike existing pre-training methods, ERNIE-GEN incorporates multi-granularity target sampling to construct pre-training data, which enhances the correlation between encoder and decoder.

Factorized Dense Synthesized Attention is a synthesized attention mechanism, similar to dense synthesized attention, but we factorize the outputs to reduce parameters and prevent overfitting. It was proposed as part of the Synthesizer architecture. The factorized variant of the dense synthesizer can be expressed as follows: $$A, B = F\_{A}\left(X\_{i}\right), F\_{B}\left(X\_{i}\right)$$ where $F\_{A}\left(.\right)$ projects input $X\_{i}$ into $a$ dimensions, $F\_B\left(.\right)$ projects $X\_{i}$ to $b$ dimensions, and $a \text{ x } b = l$. The output of the factorized module is now written as: $$ Y = \text{Softmax}\left(C\right)G\left(X\right) $$ where $C = H\_{A}\left(A\right) * H\_{B}\left(B\right)$, where $H\_{A}$, $H\_{B}$ are tiling functions and $C \in \mathbb{R}^{l \text{ x } l}$. The tiling function simply duplicates the vector $k$ times, i.e., $\mathbb{R}^{l} \rightarrow \mathbb{R}^{lk}$. In this case, $H\_{A}\left(\right)$ is a projection of $\mathbb{R}^{a} \rightarrow \mathbb{R}^{ab}$ and $H\_{B}\left(\right)$ is a projection of $\mathbb{R}^{b} \rightarrow \mathbb{R}^{ba}$. To avoid having similar values within the same block, we compose the outputs of $H\_{A}$ and $H\_{B}$.

Factorized Random Synthesized Attention, introduced with the Synthesizer architecture, is similar to factorized dense synthesized attention but for random synthesizers. Letting $R$ being a randomly initialized matrix, we factorize $R$ into low rank matrices $R\_{1}, R\_{2} \in \mathbb{R}^{l\text{ x}k}$ in the attention function: $$ Y = \text{Softmax}\left(R\_{1}R\_{2}^{T}\right)G\left(X\right) . $$ Here $G\left(.\right)$ is a parameterized function that is equivalent to $V$ in Scaled Dot-Product Attention. For each head, the factorization reduces the parameter costs from $l^{2}$ to $2\left(lk\right)$ where $k << l$ and hence helps prevent overfitting. In practice, we use a small value of $k = 8$. The basic idea of a Random Synthesizer is to not rely on pairwise token interactions or any information from individual token but rather to learn a task-specific alignment that works well globally across many samples.

Fast Schema Guided Tracker, or FastSGT, is a fast and robust BERT-based model for state tracking in goal-oriented dialogue systems. The model employs carry-over mechanisms for transferring the values between slots, enabling switching between services and accepting the values offered by the system during dialogue. It also uses multi-head attention projections in some of the decoders to have a better modelling of the encoder outputs. The model architecture is illustrated in the Figure. It consists of four main modules: 1-Utterance Encoder, 2-Schema Encoder, 3-State Decoder, and 4-State Tracker. The first three modules constitute the NLU component and are based on neural networks, whereas the state tracker is a rule-based module. BERT was used for both encoders in the model. The Utterance Encoder is a BERT model which encodes the user and system utterances at each turn. The Schema Encoder is also a BERT model which encodes the schema descriptions of intents, slots, and values into schema embeddings. These schema embeddings help the decoders to transfer or share knowledge between different services by having some language understanding of each slot, intent, or value. The schema and utterance embeddings are passed to the State Decoder - a multi-task module. This module consists of five sub-modules producing the information necessary to track the state of the dialogue. Finally, the State Tracker module takes the previous state along with the current outputs of the State Decoder and predicts the current state of the dialogue by aggregating and summarizing the information across turns.

PanGu-$α$ is an autoregressive language model (ALM) with up to 200 billion parameters pretrained on a large corpus of text, mostly in Chinese language. The architecture of PanGu-$α$ is based on Transformer, which has been extensively used as the backbone of a variety of pretrained language models such as BERT and GPT. Different from them, there's an additional query layer developed on top of Transformer layers which aims to explicitly induce the expected output.

TransferQA is a transferable generative QA model, built upon T5 that combines extractive QA and multi-choice QA via a text-to-text transformer framework, and tracks both categorical slots and non-categorical slots in DST. In addition, it introduces two effective ways to construct unanswerable questions, namely, negative question sampling and context truncation, which enable the model to handle “none” value slots in the zero-shot DST setting.

The Triplet Entropy Loss (TEL) training method aims to leverage both the strengths of Cross Entropy Loss (CEL) and Triplet loss during the training process, assuming that it would lead to better generalization. The TEL method though does not contain a pre-training step, but trains simultaneously with both CEL and Triplet losses.