With the power of large pretrained language models, various research works have integrated knowledge into dialogue systems. The traditional techniques treat knowledge as part of the input sequence for the dialogue system, prepending a set of knowledge statements in front of dialogue history.However, such a mechanism forces knowledge sets to be concatenated in an ordered manner, making models implicitly pay imbalanced attention to the sets during training.In this paper, we first investigate how the order of the knowledge set can influence autoregressive dialogue systems’ responses. We conduct experiments on two commonly used dialogue datasets with two types of transformer-based models and find that models view the input knowledge unequally. To this end, we propose a simple and novel technique to alleviate the order effect by modifying the position embeddings of knowledge input in these models. With the proposed position embedding method, the experimental results show that each knowledge statement is uniformly considered to generate responses.

Conversational assistants are ubiquitous among the general population, however, these systems have not had an impact on people with disabilities, or speech and language disorders, for whom basic day-to-day communication and social interaction is a huge struggle. Language model technology can play a huge role in empowering these users and help them interact with others with less effort via interaction support. To enable this population, we build a system that can represent them in a social conversation and generate responses that can be controlled by the users using cues/keywords. We build models that can speed up this communication by suggesting relevant cues in the dialog response context. We also introduce a keyword-loss to lexically constrain the model response output. We present automatic and human evaluation of our cue/keyword predictor and the controllable dialog system to show that our models perform significantly better than models without control. Our evaluation and user study shows that keyword-control on end-to-end response generation models is powerful and can enable and empower users with degenerative disorders to carry out their day-to-day communication.

Intelligent conversational assistants have become an integral part of our lives for performing simple tasks. However, such agents, for example, Google bots, Alexa and others are yet to have any social impact on minority population, for example, for people with neurological disorders and people with speech, language and social communication disorders, sometimes with locked-in states where speaking or typing is a challenge. Language model technologies can be very powerful tools in enabling these users to carry out daily communication and social interactions. In this work, we present a system that users with varied levels of disabilties can use to interact with the world, supported by eye-tracking, mouse controls and an intelligent agent Cue-bot, that can represent the user in a conversation. The agent provides relevant controllable ‘cues’ to generate desirable responses quickly for an ongoing dialog context. In the context of usage of such systems for people with degenerative disorders, we present automatic and human evaluation of our cue/keyword predictor and the controllable dialog system and show that our models perform significantly better than models without control and can also reduce user effort (fewer keystrokes) and speed up communication (typing time) significantly.

Incremental meeting temporal summarization, summarizing relevant information of partial multi-party meeting dialogue, is emerging as the next challenge in summarization research. Here we examine the extent to which human abstractive summaries of the preceding increments (context) can be combined with extractive meeting dialogue to generate abstractive summaries. We find that previous context improves ROUGE scores. Our findings further suggest that contexts begin to outweigh the dialogue. Using keyphrase extraction and semantic role labeling (SRL), we find that SRL captures relevant information without overwhelming the the model architecture. By compressing the previous contexts by ~70%, we achieve better ROUGE scores over our baseline models. Collectively, these results suggest that context matters, as does the way in which context is presented to the model.

Our senses individually work in a coordinated fashion to express our emotional intentions. In this work, we experiment with modeling modality-specific sensory signals to attend to our latent multimodal emotional intentions and vice versa expressed via low-rank multimodal fusion and multimodal transformers. The low-rank factorization of multimodal fusion amongst the modalities helps represent approximate multiplicative latent signal interactions. Motivated by the work of (CITATION) and (CITATION), we present our transformer-based cross-fusion architecture without any over-parameterization of the model. The low-rank fusion helps represent the latent signal interactions while the modality-specific attention helps focus on relevant parts of the signal. We present two methods for the Multimodal Sentiment and Emotion Recognition results on CMU-MOSEI, CMU-MOSI, and IEMOCAP datasets and show that our models have lesser parameters, train faster and perform comparably to many larger fusion-based architectures.

Building multimodal dialogue understanding capabilities situated in the in-cabin context is crucial to enhance passenger comfort in autonomous vehicle (AV) interaction systems. To this end, understanding passenger intents from spoken interactions and vehicle vision systems is an important building block for developing contextual and visually grounded conversational agents for AV. Towards this goal, we explore AMIE (Automated-vehicle Multimodal In-cabin Experience), the in-cabin agent responsible for handling multimodal passenger-vehicle interactions. In this work, we discuss the benefits of multimodal understanding of in-cabin utterances by incorporating verbal/language input together with the non-verbal/acoustic and visual input from inside and outside the vehicle. Our experimental results outperformed text-only baselines as we achieved improved performances for intent detection with multimodal approach.

Understanding Affect from video segments has brought researchers from the language, audio and video domains together. Most of the current multimodal research in this area deals with various techniques to fuse the modalities, and mostly treat the segments of a video independently. Motivated by the work of (Zadeh et al., 2017) and (Poria et al., 2017), we present our architecture, Relational Tensor Network, where we use the inter-modal interactions within a segment (intra-segment) and also consider the sequence of segments in a video to model the inter-segment inter-modal interactions. We also generate rich representations of text and audio modalities by leveraging richer audio and linguistic context alongwith fusing fine-grained knowledge based polarity scores from text. We present the results of our model on CMU-MOSEI dataset and show that our model outperforms many baselines and state of the art methods for sentiment classification and emotion recognition.