Machine translation is used in e-commerce to translate second-language queries into the primary language of the store, to be matched by the search system against the product catalog.However, many queries contain spelling mistakes.We first present an analysis of the spelling-robustness of a population of MT systems, quantifying how spelling variations affect MT output, the list of returned products, and ultimately user behavior.We then present two sets of practical experiments illustrating how spelling-robustness may be specifically improved.For MT, reducing the number of BPE operations significantly improves spellingrobustness in six language pairs.In end-toend e-commerce, the inclusion of a dedicated spelling correction model, and the augmentation of that model's training data with languagerelevant phenomena, each improve robustness and consistency of search results.

This paper proposes a neural network-based user simulator that can provide a multimodal interactive environment for training Reinforcement Learning (RL) agents in collaborative tasks involving multiple modes of communication. The simulator is trained on the existing ELDERLY-AT-HOME corpus and accommodates multiple modalities such as language, pointing gestures, and haptic-ostensive actions. The paper also presents a novel multimodal data augmentation approach, which addresses the challenge of using a limited dataset due to the expensive and time-consuming nature of collecting human demonstrations. Overall, the study highlights the potential for using RL and multimodal user simulators in developing and improving domestic assistive robots.

Named Entity Recognition (NER) is a vital task in various NLP applications. However, in many real-world scenarios (e.g., voice-enabled assistants) new named entities are frequently introduced, entailing re-training NER models to support these new entities. Re-annotating the original training data for the new entities could be costly or even impossible when storage limitations or security concerns restrict access to that data, and annotating a new dataset for all of the entities becomes impractical and error-prone as the number of entities increases. To tackle this problem, we introduce a novel Continual Learning approach for NER, which requires new training material to be annotated only for the new entities. To preserve the existing knowledge previously learned by the model, we exploit the Knowledge Distillation (KD) framework, where the existing NER model acts as the teacher for a new NER model (i.e., the student), which learns the new entity by using the new training material and retains knowledge of old entities by imitating the teacher's outputs on this new training set. Our experiments show that this approach allows the student model to ``progressively'' learn to identify new entities without forgetting the previously learned ones. We also present a comparison with multiple strong baselines to demonstrate that our approach is superior for continually updating an NER model.

In a collaborative task and the interaction that accompanies it, the participants often take on distinct roles, and dynamically switch the roles as the task requires. A domestic assistive robot thus needs to have similar capabilities. Using our previously proposed Multimodal Interaction Manager (MIM) framework, this paper investigates how role switching for a robot can be implemented. It identifies a set of primitive subtasks that encode common interaction patterns observed in our data corpus and that can be used to easily construct complex task models. It also describes an implementation on the NAO robot that, together with our original work, demonstrates that the robot can take on different roles. We provide a detailed analysis of the performance of the system and discuss the challenges that arise when switching roles in human-robot interactions.

With rapid advances in social robotics, humanoids and autonomy, robot assistants appear to be within reach. However, robots are still unable to effectively interact with humans in activities of daily living. One of the challenges is in the frequent use of multiple communication modalities when humans engage in collaborative activities. In this paper, we propose a Multimodal Interaction Manager, a framework for an assistive robot that maintains an active multimodal interaction with a human partner while performing physical collaborative tasks. The heart of our framework is a Hierarchical Bipartite Action-Transition Network (HBATN), which allows the robot to infer the state of the task and the dialogue given spoken utterances and observed pointing gestures from a human partner, and to plan its next actions. Finally, we implemented this framework on a robot to provide preliminary evidence that the robot can successfully participate in a task-oriented multimodal interaction.

Software projects produce large quantities of data such as feature requests, requirements, design artifacts, source code, tests, safety cases, release plans, and bug reports. If leveraged effectively, this data can be used to provide project intelligence that supports diverse software engineering activities such as release planning, impact analysis, and software analytics. However, project stakeholders often lack skills to formulate complex queries needed to retrieve, manipulate, and display the data in meaningful ways. To address these challenges we introduce TiQi, a natural language interface, which allows users to express software-related queries verbally or written in natural language. TiQi is a web-based tool. It visualizes available project data as a prompt to the user, accepts Natural Language (NL) queries, transforms those queries into SQL, and then executes the queries against a centralized or distributed database. Raw data is stored either directly in the database or retrieved dynamically at runtime from case tools and repositories such as Github and Jira. The transformed query is visualized back to the user as SQL and augmented UML, and raw data results are returned. Our tool demo can be found on YouTube at the following link:http://tinyurl.com/TIQIDemo.

Software Traceability is a critical element in all safety critical software systems. Trace links are created across diverse artifacts such as requirements, design, code, test cases, and hazards - either manually or with the help of supporting tools. The links are then used to support a range of software engineering activities including impact analysis, compliance verification, and safety inspections. For traceability to effectively support these activities it is important for the meaning and rationale of each link to be clearly communicated. It is often insuficient to know that one artifact satisfies, realizes, or complies to another. Instead, it is important to know why and how it does so. Terms and phrases used to describe artifacts are connected through composition, synonymic, and generalization relationships which often can only be interpreted by domain experts. In this RE:Next! paper we propose a novel approach for utilizing domain-specific knowledge bases to generate trace link rationales. We illustrate our approach with examples of automatically generated rationales taken from the domain of Communication and Control of a Transportation system, and from a Medical Infusion pump domain.

State-of-the-art software trace retrieval techniques are unable to perform the complex reasoning that a human analyst follows in order to create accurate trace links between artifacts such as regulatory codes and requirements. As a result, current algorithms often generate imprecise links. To address this problem, we present the Domain-Contextualized Intelligent Traceability Solution (DoCIT), designed to mimic some of the higher level reasoning that a human trace analyst performs. We focus our efforts on the complex domain of communication and control in a transportation system. DoCIT includes rules for extracting ``action units'' from software artifacts, a domain-specific knowledge base for relating semantically similar concepts across action units, and a set of link-creation heuristics which utilize the action units to establish meaningful trace links across pairs of artifacts. Our approach significantly improves the quality of the generated trace links. We illustrate and evaluate DoCIT with examples and experiments from the control and communication sector of a transportation domain.