Search

Fairness-related assumptions about what constitute appropriate NLG system behaviors range from invariance, where systems are expected to behave identically for social groups, to adaptation, where behaviors should instead vary across them. To illuminate tensions around invariance and adaptation, we conduct five case studies, in which we perturb different types of identity-related language features (names, roles, locations, dialect, and style) in NLG system inputs. Through these cases studies, we examine people's expectations of system behaviors, and surface potential caveats of these contrasting yet commonly held assumptions. We find that motivations for adaptation include social norms, cultural differences, feature-specific information, and accommodation; in contrast, motivations for invariance include perspectives that favor prescriptivism, view adaptation as unnecessary or too difficult for NLG systems to do appropriately, and are wary of false assumptions. Our findings highlight open challenges around what constitute "fair" or "good" NLG system behaviors., Comment: 36 pages, 24 figures, NAACL 2024

The recent increase in the volume of online meetings necessitates automated tools for managing and organizing the material, especially when an attendee has missed the discussion and needs assistance in quickly exploring it. In this work, we propose a novel end-to-end framework for generating interactive questionnaires for preference-based meeting exploration. As a result, users are supplied with a list of suggested questions reflecting their preferences. Since the task is new, we introduce an automatic evaluation strategy. Namely, it measures how much the generated questions via questionnaire are answerable to ensure factual correctness and covers the source meeting for the depth of possible exploration.

Pre-trained language models (LMs) have been shown to memorize a substantial amount of knowledge from the pre-training corpora; however, they are still limited in recalling factually correct knowledge given a certain context. Hence, they tend to suffer from counterfactual or hallucinatory generation when used in knowledge-intensive natural language generation (NLG) tasks. Recent remedies to this problem focus on modifying either the pre-training or task fine-tuning objectives to incorporate knowledge, which normally require additional costly training or architecture modification of LMs for practical applications. We present Knowledge Infused Decoding (KID) -- a novel decoding algorithm for generative LMs, which dynamically infuses external knowledge into each step of the LM decoding. Specifically, we maintain a local knowledge memory based on the current context, interacting with a dynamically created external knowledge trie, and continuously update the local memory as a knowledge-aware constraint to guide decoding via reinforcement learning. On six diverse knowledge-intensive NLG tasks, task-agnostic LMs (e.g., GPT-2 and BART) armed with KID outperform many task-optimized state-of-the-art models, and show particularly strong performance in few-shot scenarios over seven related knowledge-infusion techniques. Human evaluation confirms KID's ability to generate more relevant and factual language for the input context when compared with multiple baselines. Finally, KID also alleviates exposure bias and provides stable generation quality when generating longer sequences. Code for KID is available at https://github.com/microsoft/KID., Comment: In ICLR 2022

Recent work has focused on compressing pre-trained language models (PLMs) like BERT where the major focus has been to improve the in-distribution performance for downstream tasks. However, very few of these studies have analyzed the impact of compression on the generalizability and robustness of compressed models for out-of-distribution (OOD) data. Towards this end, we study two popular model compression techniques including knowledge distillation and pruning and show that the compressed models are significantly less robust than their PLM counterparts on OOD test sets although they obtain similar performance on in-distribution development sets for a task. Further analysis indicates that the compressed models overfit on the shortcut samples and generalize poorly on the hard ones. We further leverage this observation to develop a regularization strategy for robust model compression based on sample uncertainty. Experimental results on several natural language understanding tasks demonstrate that our bias mitigation framework improves the OOD generalization of the compressed models, while not sacrificing the in-distribution task performance., Comment: Accepted by EACL 2023

Global models are trained to be as generalizable as possible, with user invariance considered desirable since the models are shared across multitudes of users. As such, these models are often unable to produce personalized responses for individual users, based on their data. Contrary to widely-used personalization techniques based on few-shot learning, we propose UserIdentifier, a novel scheme for training a single shared model for all users. Our approach produces personalized responses by adding fixed, non-trainable user identifiers to the input data. We empirically demonstrate that this proposed method outperforms the prefix-tuning based state-of-the-art approach by up to 13%, on a suite of sentiment analysis datasets. We also show that, unlike prior work, this method needs neither any additional model parameters nor any extra rounds of few-shot fine-tuning.

We study the problem of multilingual automated reply suggestions (RS) model serving many languages simultaneously. Multilingual models are often challenged by model capacity and severe data distribution skew across languages. While prior works largely focus on monolingual models, we propose Conditional Generative Matching models (CGM), optimized within a Variational Autoencoder framework to address challenges arising from multi-lingual RS. CGM does so with expressive message conditional priors, mixture densities to enhance multi-lingual data representation, latent alignment for language discrimination, and effective variational optimization techniques for training multi-lingual RS. The enhancements result in performance that exceed competitive baselines in relevance (ROUGE score) by more than 10\% on average, and 16\% for low resource languages. CGM also shows remarkable improvements in diversity (80\%) illustrating its expressiveness in representation of multi-lingual data.

Reply suggestion models help users process emails and chats faster. Previous work only studies English reply suggestion. Instead, we present MRS, a multilingual reply suggestion dataset with ten languages. MRS can be used to compare two families of models: 1) retrieval models that select the reply from a fixed set and 2) generation models that produce the reply from scratch. Therefore, MRS complements existing cross-lingual generalization benchmarks that focus on classification and sequence labeling tasks. We build a generation model and a retrieval model as baselines for MRS. The two models have different strengths in the monolingual setting, and they require different strategies to generalize across languages. MRS is publicly available at https://github.com/zhangmozhi/mrs., Comment: ACL 2021

We consider the problem of scaling automated suggested replies for Outlook email system to multiple languages. Faced with increased compute requirements and low resources for language expansion, we build a single universal model for improving the quality and reducing run-time costs of our production system. However, restricted data movement across regional centers prevents joint training across languages. To this end, we propose a multi-task continual learning framework, with auxiliary tasks and language adapters to learn universal language representation across regions. The experimental results show positive cross-lingual transfer across languages while reducing catastrophic forgetting across regions. Our online results on real user traffic show significant gains in CTR and characters saved, as well as 65% training cost reduction compared with per-language models. As a consequence, we have scaled the feature in multiple languages including low-resource markets.

The combination of multilingual pre-trained representations and cross-lingual transfer learning is one of the most effective methods for building functional NLP systems for low-resource languages. However, for extremely low-resource languages without large-scale monolingual corpora for pre-training or sufficient annotated data for fine-tuning, transfer learning remains an under-studied and challenging task. Moreover, recent work shows that multilingual representations are surprisingly disjoint across languages, bringing additional challenges for transfer onto extremely low-resource languages. In this paper, we propose MetaXL, a meta-learning based framework that learns to transform representations judiciously from auxiliary languages to a target one and brings their representation spaces closer for effective transfer. Extensive experiments on real-world low-resource languages - without access to large-scale monolingual corpora or large amounts of labeled data - for tasks like cross-lingual sentiment analysis and named entity recognition show the effectiveness of our approach. Code for MetaXL is publicly available at github.com/microsoft/MetaXL., Comment: 2021 Annual Conference of the North American Chapter of the Association for Computational Linguistics (NAACL 2021)

Email remains one of the most frequently used means of online communication. People spend a significant amount of time every day on emails to exchange information, manage tasks and schedule events. Previous work has studied different ways for improving email productivity by prioritizing emails, suggesting automatic replies or identifying intents to recommend appropriate actions. The problem has been mostly posed as a supervised learning problem where models of different complexities were proposed to classify an email message into a predefined taxonomy of intents or classes. The need for labeled data has always been one of the largest bottlenecks in training supervised models. This is especially the case for many real-world tasks, such as email intent classification, where large scale annotated examples are either hard to acquire or unavailable due to privacy or data access constraints. Email users often take actions in response to intents expressed in an email (e.g., setting up a meeting in response to an email with a scheduling request). Such actions can be inferred from user interaction logs. In this paper, we propose to leverage user actions as a source of weak supervision, in addition to a limited set of annotated examples, to detect intents in emails. We develop an end-to-end robust deep neural network model for email intent identification that leverages both clean annotated data and noisy weak supervision along with a self-paced learning mechanism. Extensive experiments on three different intent detection tasks show that our approach can effectively leverage the weakly supervised data to improve intent detection in emails., Comment: 10 pages, 3 figures

We study the basic operation of set union in the global model of differential privacy. In this problem, we are given a universe $U$ of items, possibly of infinite size, and a database $D$ of users. Each user $i$ contributes a subset $W_i \subseteq U$ of items. We want an ($\epsilon$,$\delta$)-differentially private algorithm which outputs a subset $S \subset \cup_i W_i$ such that the size of $S$ is as large as possible. The problem arises in countless real world applications; it is particularly ubiquitous in natural language processing (NLP) applications as vocabulary extraction. For example, discovering words, sentences, $n$-grams etc., from private text data belonging to users is an instance of the set union problem. Known algorithms for this problem proceed by collecting a subset of items from each user, taking the union of such subsets, and disclosing the items whose noisy counts fall above a certain threshold. Crucially, in the above process, the contribution of each individual user is always independent of the items held by other users, resulting in a wasteful aggregation process, where some item counts happen to be way above the threshold. We deviate from the above paradigm by allowing users to contribute their items in a $\textit{dependent fashion}$, guided by a $\textit{policy}$. In this new setting ensuring privacy is significantly delicate. We prove that any policy which has certain $\textit{contractive}$ properties would result in a differentially private algorithm. We design two new algorithms, one using Laplace noise and other Gaussian noise, as specific instances of policies satisfying the contractive properties. Our experiments show that the new algorithms significantly outperform previously known mechanisms for the problem., Comment: 23 pages, 7 figures

Email is an integral part of people's work and life, enabling them to perform activities such as communicating, searching, managing tasks and storing information. Modern email clients take a step forward and help improve users' productivity by automatically creating reminders, tasks or responses. The act of reading is arguably the only activity that is in common in most -- if not all -- of the interactions that users have with their emails. In this paper, we characterize how users read their enterprise emails, and reveal the various contextual factors that impact reading time. Our approach starts with a reading time analysis based on the reading events from a major email platform, followed by a user study to provide explanations for some discoveries. We identify multiple temporal and user contextual factors that are correlated with reading time. For instance, email reading time is correlated with user devices: on desktop reading time increases through the morning and peaks at noon but on mobile it increases through the evening till midnight. The reading time is also negatively correlated with the screen size. We have established the connection between user status and reading time: users spend more time reading emails when they have fewer meetings and busy hours during the day. In addition, we find that users also reread emails across devices. Among the cross-device reading events, 76% of reread emails are first visited on mobile and then on desktop. Overall, our study is the first to characterize enterprise email reading time on a very large scale. The findings provide insights to develop better metrics and user models for understanding and improving email interactions.

We consider the problem of diversifying automated reply suggestions for a commercial instant-messaging (IM) system (Skype). Our conversation model is a standard matching based information retrieval architecture, which consists of two parallel encoders to project messages and replies into a common feature representation. During inference, we select replies from a fixed response set using nearest neighbors in the feature space. To diversify responses, we formulate the model as a generative latent variable model with Conditional Variational Auto-Encoder (M-CVAE). We propose a constrained-sampling approach to make the variational inference in M-CVAE efficient for our production system. In offline experiments, M-CVAE consistently increased diversity by ~30-40% without significant impact on relevance. This translated to a 5% gain in click-rate in our online production system.

Email triage involves going through unhandled emails and deciding what to do with them. This familiar process can become increasingly challenging as the number of unhandled email grows. During a triage session, users commonly defer handling emails that they cannot immediately deal with to later. These deferred emails, are often related to tasks that are postponed until the user has more time or the right information to deal with them. In this paper, through qualitative interviews and a large-scale log analysis, we study when and what enterprise email users tend to defer. We found that users are more likely to defer emails when handling them involves replying, reading carefully, or clicking on links and attachments. We also learned that the decision to defer emails depends on many factors such as user's workload and the importance of the sender. Our qualitative results suggested that deferring is very common, and our quantitative log analysis confirms that 12% of triage sessions and 16% of daily active users had at least one deferred email on weekdays. We also discuss several deferral strategies such as marking emails as unread and flagging that are reported by our interviewees, and illustrate how such patterns can be also observed in user logs. Inspired by the characteristics of deferred emails and contextual factors involved in deciding if an email should be deferred, we train a classifier for predicting whether a recently triaged email is actually deferred. Our experimental results suggests that deferral can be classified with modest effectiveness. Overall, our work provides novel insights about how users handle their emails and how deferral can be modeled.

Federated search (federated information retrieval or distributed information retrieval) is a technique for searching multiple text collections simultaneously. Queries are submitted to a subset of collections that are most likely to return relevant answers. The results returned by selected collections are integrated and merged into a single list. Federated search is preferred over centralized search alternatives in many environments. For example, commercial search engines such as Google cannot easily index uncrawlable hidden web collections while federated search systems can search the contents of hidden web collections without crawling. In enterprise environments, where each organization maintains an independent search engine, federated search techniques can provide parallel search over multiple collections. There are three major challenges in federated search. For each query, a subset of collections that are most likely to return relevant documents are selected. This creates the collection selection problem. To be able to select suitable collections, federated search systems need to acquire some knowledge about the contents of each collection, creating the collection representation problem. The results returned from the selected collections are merged before the final presentation to the user. This final step is the result merging problem. The goal of this work is to provide a comprehensive summary of the previous research on the federated search challenges described above.

Federated search (federated information retrieval or distributed information retrieval) is a technique for searching multiple text collections simultaneously. Queries are submitted to a subset of collections that are most likely to return relevant answers. The results returned by selected collections are integrated and merged into a single list. Federated search is preferred over centralized search alternatives in many environments. For example, commercial search engines such as Google cannot easily index uncrawlable hidden web collections while federated search systems can search the contents of hidden web collections without crawling. In enterprise environments, where each organization maintains an independent search engine, federated search techniques can provide parallel search over multiple collections. There are three major challenges in federated search. For each query, a subset of collections that are most likely to return relevant documents are selected. This creates the collection selection problem. To be able to select suitable collections, federated search systems need to acquire some knowledge about the contents of each collection, creating the collection representation problem. The results returned from the selected collections are merged before the final presentation to the user. This final step is the result merging problem. The goal of this work is to provide a comprehensive summary of the previous research on the federated search challenges described above.

In order to bridge the “Semantic gap”, a number of relevance feedback (RF) mechanisms have been applied to content-based image retrieval (CBIR). However current RF techniques in most existing CBIR systems still lack satisfactory user interaction although some work has been done to improve the interaction as well as the search accuracy. In this paper, we propose a four-factor user interaction model and investigate its effects on CBIR by an empirical evaluation. Whilst the model was developed for our research purposes, we believe the model could be adapted to any content-based search system.

In order to bridge the “Semantic gap”, a number of relevance feedback (RF) mechanisms have been applied to content-based image retrieval (CBIR). However current RF techniques in most existing CBIR systems still lack satisfactory user interaction although some work has been done to improve the interaction as well as the search accuracy. In this paper, we propose a four-factor user interaction model and investigate its effects on CBIR by an empirical evaluation. Whilst the model was developed for our research purposes, we believe the model could be adapted to any content-based search system.

In order to bridge the “Semantic gap”, a number of relevance feedback (RF) mechanisms have been applied to content-based image retrieval (CBIR). However current RF techniques in most existing CBIR systems still lack satisfactory user interaction although some work has been done to improve the interaction as well as the search accuracy. In this paper, we propose a four-factor user interaction model and investigate its effects on CBIR by an empirical evaluation. Whilst the model was developed for our research purposes, we believe the model could be adapted to any content-based search system.

In order to bridge the “Semantic gap”, a number of relevance feedback (RF) mechanisms have been applied to content-based image retrieval (CBIR). However current RF techniques in most existing CBIR systems still lack satisfactory user interaction although some work has been done to improve the interaction as well as the search accuracy. In this paper, we propose a four-factor user interaction model and investigate its effects on CBIR by an empirical evaluation. Whilst the model was developed for our research purposes, we believe the model could be adapted to any content-based search system.

Modern techniques for distributed information retrieval use a set of documents sampled from each server, but these samples have been underutilised in server selection. We describe a new server selection algorithm, SUSHI, which unlike earlier algorithms can make full use of the text of each sampled document and which does not need training data. SUSHI can directly optimise for many common cases, including high precision retrieval, and by including a simple stopping condition can do so while reducing network traffic. Our experiments compare SUSHI with alternatives and show it achieves the same effectiveness as the best current methods while being substantially more efficient, selecting as few as 20% as many servers.

In order to bridge the “Semantic gap”, a number of relevance feedback (RF) mechanisms have been applied to content-based image retrieval (CBIR). However current RF techniques in most existing CBIR systems still lack satisfactory user interaction although some work has been done to improve the interaction as well as the search accuracy. In this paper, we propose a four-factor user interaction model and investigate its effects on CBIR by an empirical evaluation. Whilst the model was developed for our research purposes, we believe the model could be adapted to any content-based search system.

In order to bridge the “Semantic gap”, a number of relevance feedback (RF) mechanisms have been applied to content-based image retrieval (CBIR). However current RF techniques in most existing CBIR systems still lack satisfactory user interaction although some work has been done to improve the interaction as well as the search accuracy. In this paper, we propose a four-factor user interaction model and investigate its effects on CBIR by an empirical evaluation. Whilst the model was developed for our research purposes, we believe the model could be adapted to any content-based search system.