arXiv subjects

Arash Einolghozati

Publications and source records attributed to Arash Einolghozati.

At least 19 recordsLinked to original sources

arXiv2024

CoDi: Conversational Distillation for Grounded Question Answering

Distilling conversational skills into Small Language Models (SLMs) with approximately 1 billion parameters presents significant challenges. Firstly, SLMs have limited capacity in their model parameters to learn extensive knowledge compared to larger models. Secondly, high-quality conversational datasets are often scarce, small, and domain-specific. Addressing these challenges, we introduce a novel data distillation framework named CoDi (short for Conversational Distillation, pronounced "Cody"), allowing us to synthesize large-scale, assistant-style datasets in a steerable and diverse manner. Specifically, while our framework is task agnostic at its core, we explore and evaluate the potential of CoDi on the task of conversational grounded reasoning for question answering. This is a typical on-device scenario for specialist SLMs, allowing for open-domain model responses, without requiring the model to "memorize" world knowledge in its limited weights. Our evaluations show that SLMs trained with CoDi-synthesized data achieve performance comparable to models trained on human-annotated data in standard metrics. Additionally, when using our framework to generate larger datasets from web data, our models surpass larger, instruction-tuned models in zero-shot conversational grounded reasoning tasks.

cs.CL↗

arXiv2024

Small But Funny: A Feedback-Driven Approach to Humor Distillation

The emergence of Large Language Models (LLMs) has brought to light promising language generation capabilities, particularly in performing tasks like complex reasoning and creative writing. Consequently, distillation through imitation of teacher responses has emerged as a popular technique to transfer knowledge from LLMs to more accessible, Small Language Models (SLMs). While this works well for simpler tasks, there is a substantial performance gap on tasks requiring intricate language comprehension and creativity, such as humor generation. We hypothesize that this gap may stem from the fact that creative tasks might be hard to learn by imitation alone and explore whether an approach, involving supplementary guidance from the teacher, could yield higher performance. To address this, we study the effect of assigning a dual role to the LLM - as a "teacher" generating data, as well as a "critic" evaluating the student's performance. Our experiments on humor generation reveal that the incorporation of feedback significantly narrows the performance gap between SLMs and their larger counterparts compared to merely relying on imitation. As a result, our research highlights the potential of using feedback as an additional dimension to data when transferring complex language abilities via distillation.

cs.CL↗

arXiv2024

Improving Faithfulness of Abstractive Summarization by Controlling Confounding Effect of Irrelevant Sentences

Lack of factual correctness is an issue that still plagues state-of-the-art summarization systems despite their impressive progress on generating seemingly fluent summaries. In this paper, we show that factual inconsistency can be caused by irrelevant parts of the input text, which act as confounders. To that end, we leverage information-theoretic measures of causal effects to quantify the amount of confounding and precisely quantify how they affect the summarization performance. Based on insights derived from our theoretical results, we design a simple multi-task model to control such confounding by leveraging human-annotated relevant sentences when available. Crucially, we give a principled characterization of data distributions where such confounding can be large thereby necessitating the use of human annotated relevant sentences to generate factual summaries. Our approach improves faithfulness scores by 20\% over strong baselines on AnswerSumm \citep{fabbri2021answersumm}, a conversation summarization dataset where lack of faithfulness is a significant issue due to the subjective nature of the task. Our best method achieves the highest faithfulness score while also achieving state-of-the-art results on standard metrics like ROUGE and METEOR. We corroborate these improvements through human evaluation.

cs.CL↗

arXiv2023

A Study on the Efficiency and Generalization of Light Hybrid Retrievers

Hybrid retrievers can take advantage of both sparse and dense retrievers. Previous hybrid retrievers leverage indexing-heavy dense retrievers. In this work, we study "Is it possible to reduce the indexing memory of hybrid retrievers without sacrificing performance"? Driven by this question, we leverage an indexing-efficient dense retriever (i.e. DrBoost) and introduce a LITE retriever that further reduces the memory of DrBoost. LITE is jointly trained on contrastive learning and knowledge distillation from DrBoost. Then, we integrate BM25, a sparse retriever, with either LITE or DrBoost to form light hybrid retrievers. Our Hybrid-LITE retriever saves 13X memory while maintaining 98.0% performance of the hybrid retriever of BM25 and DPR. In addition, we study the generalization capacity of our light hybrid retrievers on out-of-domain dataset and a set of adversarial attacks datasets. Experiments showcase that light hybrid retrievers achieve better generalization performance than individual sparse and dense retrievers. Nevertheless, our analysis shows that there is a large room to improve the robustness of retrievers, suggesting a new research direction.

cs.IR↗

arXiv2021

EASE: Extractive-Abstractive Summarization with Explanations

Current abstractive summarization systems outperform their extractive counterparts, but their widespread adoption is inhibited by the inherent lack of interpretability. To achieve the best of both worlds, we propose EASE, an extractive-abstractive framework for evidence-based text generation and apply it to document summarization. We present an explainable summarization system based on the Information Bottleneck principle that is jointly trained for extraction and abstraction in an end-to-end fashion. Inspired by previous research that humans use a two-stage framework to summarize long documents (Jing and McKeown, 2000), our framework first extracts a pre-defined amount of evidence spans as explanations and then generates a summary using only the evidence. Using automatic and human evaluations, we show that explanations from our framework are more relevant than simple baselines, without substantially sacrificing the quality of the generated summary.

cs.CL↗

arXiv2021

El Volumen Louder Por Favor: Code-switching in Task-oriented Semantic Parsing

Being able to parse code-switched (CS) utterances, such as Spanish+English or Hindi+English, is essential to democratize task-oriented semantic parsing systems for certain locales. In this work, we focus on Spanglish (Spanish+English) and release a dataset, CSTOP, containing 5800 CS utterances alongside their semantic parses. We examine the CS generalizability of various Cross-lingual (XL) models and exhibit the advantage of pre-trained XL language models when data for only one language is present. As such, we focus on improving the pre-trained models for the case when only English corpus alongside either zero or a few CS training instances are available. We propose two data augmentation methods for the zero-shot and the few-shot settings: fine-tune using translate-and-align and augment using a generation model followed by match-and-filter. Combining the few-shot setting with the above improvements decreases the initial 30-point accuracy gap between the zero-shot and the full-data settings by two thirds.

cs.CL↗

arXiv2020

Sound Natural: Content Rephrasing in Dialog Systems

We introduce a new task of rephrasing for a more natural virtual assistant. Currently, virtual assistants work in the paradigm of intent slot tagging and the slot values are directly passed as-is to the execution engine. However, this setup fails in some scenarios such as messaging when the query given by the user needs to be changed before repeating it or sending it to another user. For example, for queries like 'ask my wife if she can pick up the kids' or 'remind me to take my pills', we need to rephrase the content to 'can you pick up the kids' and 'take your pills' In this paper, we study the problem of rephrasing with messaging as a use case and release a dataset of 3000 pairs of original query and rephrased query. We show that BART, a pre-trained transformers-based masked language model with auto-regressive decoding, is a strong baseline for the task, and show improvements by adding a copy-pointer and copy loss to it. We analyze different tradeoffs of BART-based and LSTM-based seq2seq models, and propose a distilled LSTM-based seq2seq as the best practical model.

cs.CL↗

arXiv2019

Likelihood Ratios and Generative Classifiers for Unsupervised Out-of-Domain Detection In Task Oriented Dialog

The task of identifying out-of-domain (OOD) input examples directly at test-time has seen renewed interest recently due to increased real world deployment of models. In this work, we focus on OOD detection for natural language sentence inputs to task-based dialog systems. Our findings are three-fold: First, we curate and release ROSTD (Real Out-of-Domain Sentences From Task-oriented Dialog) - a dataset of 4K OOD examples for the publicly available dataset from (Schuster et al. 2019). In contrast to existing settings which synthesize OOD examples by holding out a subset of classes, our examples were authored by annotators with apriori instructions to be out-of-domain with respect to the sentences in an existing dataset. Second, we explore likelihood ratio based approaches as an alternative to currently prevalent paradigms. Specifically, we reformulate and apply these approaches to natural language inputs. We find that they match or outperform the latter on all datasets, with larger improvements on non-artificial OOD benchmarks such as our dataset. Our ablations validate that specifically using likelihood ratios rather than plain likelihood is necessary to discriminate well between OOD and in-domain data. Third, we propose learning a generative classifier and computing a marginal likelihood (ratio) for OOD detection. This allows us to use a principled likelihood while at the same time exploiting training-time labels. We find that this approach outperforms both simple likelihood (ratio) based and other prior approaches. We are hitherto the first to investigate the use of generative classifiers for OOD detection at test-time.

cs.CL↗

arXiv2019

Improving Robustness of Task Oriented Dialog Systems

Task oriented language understanding in dialog systems is often modeled using intents (task of a query) and slots (parameters for that task). Intent detection and slot tagging are, in turn, modeled using sentence classification and word tagging techniques respectively. Similar to adversarial attack problems with computer vision models discussed in existing literature, these intent-slot tagging models are often over-sensitive to small variations in input -- predicting different and often incorrect labels when small changes are made to a query, thus reducing their accuracy and reliability. However, evaluating a model's robustness to these changes is harder for language since words are discrete and an automated change (e.g. adding `noise') to a query sometimes changes the meaning and thus labels of a query. In this paper, we first describe how to create an adversarial test set to measure the robustness of these models. Furthermore, we introduce and adapt adversarial training methods as well as data augmentation using back-translation to mitigate these issues. Our experiments show that both techniques improve the robustness of the system substantially and can be combined to yield the best results.

cs.CL↗

arXiv2019

Improving Semantic Parsing for Task Oriented Dialog

Semantic parsing using hierarchical representations has recently been proposed for task oriented dialog with promising results [Gupta et al 2018]. In this paper, we present three different improvements to the model: contextualized embeddings, ensembling, and pairwise re-ranking based on a language model. We taxonomize the errors possible for the hierarchical representation, such as wrong top intent, missing spans or split spans, and show that the three approaches correct different kinds of errors. The best model combines the three techniques and gives 6.4% better exact match accuracy than the state-of-the-art, with an error reduction of 33%, resulting in a new state-of-the-art result on the Task Oriented Parsing (TOP) dataset.

cs.CL↗

arXiv2018

CERES: Distantly Supervised Relation Extraction from the Semi-Structured Web

The web contains countless semi-structured websites, which can be a rich source of information for populating knowledge bases. Existing methods for extracting relations from the DOM trees of semi-structured webpages can achieve high precision and recall only when manual annotations for each website are available. Although there have been efforts to learn extractors from automatically-generated labels, these methods are not sufficiently robust to succeed in settings with complex schemas and information-rich websites. In this paper we present a new method for automatic extraction from semi-structured websites based on distant supervision. We automatically generate training labels by aligning an existing knowledge base with a web page and leveraging the unique structural characteristics of semi-structured websites. We then train a classifier based on the potentially noisy and incomplete labels to predict new relation instances. Our method can compete with annotation-based techniques in the literature in terms of extraction quality. A large-scale experiment on over 400,000 pages from dozens of multi-lingual long-tail websites harvested 1.25 million facts at a precision of 90%.

cs.AI↗

arXiv2017

soc2seq: Social Embedding meets Conversation Model

While liking or upvoting a post on a mobile app is easy to do, replying with a written note is much more difficult, due to both the cognitive load of coming up with a meaningful response as well as the mechanics of entering the text. Here we present a novel textual reply generation model that goes beyond the current auto-reply and predictive text entry models by taking into account the content preferences of the user, the idiosyncrasies of their conversational style, and even the structure of their social graph. Specifically, we have developed two types of models for personalized user interactions: a content-based conversation model, which makes use of location together with user information, and a social-graph-based conversation model, which combines content-based conversation models with social graphs.

cs.SI↗

arXiv2014

Relaying in Diffusion-Based Molecular Communication

Molecular communication between biological entities is a new paradigm in communications. Recently, we studied molecular communication between two nodes formed from synthetic bacteria. Due to high randomness in behavior of bacteria, we used a population of them in each node. The reliability of such communication systems depends on both the maximum concentration of molecules that a transmitter node is able to produce at the receiver node as well as the number of bacteria in each nodes. This maximum concentration of molecules falls with distance which makes the communication to the far nodes nearly impossible. In order to alleviate this problem, in this paper, we propose to use a molecular relaying node. The relay node can resend the message either by the different or the same type of molecules as the original signal from the transmitter. We study two scenarios of relaying. In the first scenario, the relay node simply senses the received concentration and forwards it to the receiver. We show that this sense and forward scenario, depending on the type of molecules used for relaying, results in either increasing the range of concentration of molecules at the receiver or increasing the effective number of bacteria in the receiver node. For both cases of sense and forward relaying, we obtain the resulting improvement in channel capacity. We conclude that multi-type molecular relaying outperforms the single-type relaying. In the second scenario, we study the decode and forward relaying for the M-ary signaling scheme. We show that this relaying strategy increases the reliability of M-ary communication significantly.

cs.ET↗

arXiv2014

Design and Analysis of Wireless Communication Systems Using Diffusion-Based Molecular Communication Among Bacteria

The design of biologically-inspired wireless communication systems using bacteria as the basic element of the system is initially motivated by a phenomenon called \emph{Quorum Sensing}. Due to high randomness in the individual behavior of a bacterium, reliable communication between two bacteria is almost impossible. Therefore, we have recently proposed that a population of bacteria in a cluster is considered as a bio node in the network capable of molecular transmission and reception. This proposition enables us to form a reliable bio node out of many unreliable bacteria. In this paper, we study the communication between two nodes in such a network where information is encoded in the concentration of molecules by the transmitter. The molecules produced by the bacteria in the transmitter node propagate through the diffusion channel. Then, the concentration of molecules is sensed by the bacteria population in the receiver node which would decode the information and output light or fluorescent as a result. The uncertainty in the communication is caused by all three components of communication, i.e., transmission, propagation and reception. We study the theoretical limits of the information transfer rate in the presence of such uncertainties. Finally, we consider M-ary signaling schemes and study their achievable rates and corresponding error probabilities.

cs.IT↗

arXiv2012

BPRS: Belief Propagation Based Iterative Recommender System

In this paper we introduce the first application of the Belief Propagation (BP) algorithm in the design of recommender systems. We formulate the recommendation problem as an inference problem and aim to compute the marginal probability distributions of the variables which represent the ratings to be predicted. However, computing these marginal probability functions is computationally prohibitive for large-scale systems. Therefore, we utilize the BP algorithm to efficiently compute these functions. Recommendations for each active user are then iteratively computed by probabilistic message passing. As opposed to the previous recommender algorithms, BPRS does not require solving the recommendation problem for all the users if it wishes to update the recommendations for only a single active. Further, BPRS computes the recommendations for each user with linear complexity and without requiring a training period. Via computer simulations (using the 100K MovieLens dataset), we verify that BPRS iteratively reduces the error in the predicted ratings of the users until it converges. Finally, we confirm that BPRS is comparable to the state of art methods such as Correlation-based neighborhood model (CorNgbr) and Singular Value Decomposition (SVD) in terms of rating and precision accuracy. Therefore, we believe that the BP-based recommendation algorithm is a new promising approach which offers a significant advantage on scalability while providing competitive accuracy for the recommender systems.

cs.LG↗

arXiv2012

Molecular Communication Between Two Populations of Bacteria

Molecular communication is an expanding body of research. Recent advances in biology have encouraged using genetically engineered bacteria as the main component in the molecular communication. This has stimulated a new line of research that attempts to study molecular communication among bacteria from an information-theoretic point of view. Due to high randomness in the individual behavior of the bacterium, reliable communication between two bacteria is almost impossible. Therefore, we recently proposed that a population of bacteria in a cluster is considered as a node capable of molecular transmission and reception. This proposition enables us to form a reliable node out of many unreliable bacteria. The bacteria inside a node sense the environment and respond accordingly. In this paper, we study the communication between two nodes, one acting as the transmitter and the other as the receiver. We consider the case in which the information is encoded in the concentration of molecules by the transmitter. The molecules produced by the bacteria in the transmitter node propagate in the environment via the diffusion process. Then, their concentration sensed by the bacteria in the receiver node would decode the information. The randomness in the communication is caused by both the error in the molecular production at the transmitter and the reception of molecules at the receiver. We study the theoretical limits of the information transfer rate in such a setup versus the number of bacteria per node. Finally, we consider M-ary modulation schemes and study the achievable rates and their error probabilities.

cs.IT↗

arXiv2012

Data Gathering in Networks of Bacteria Colonies: Collective Sensing and Relaying Using Molecular Communication

The prospect of new biological and industrial applications that require communication in micro-scale, encourages research on the design of bio-compatible communication networks using networking primitives already available in nature. One of the most promising candidates for constructing such networks is to adapt and engineer specific types of bacteria that are capable of sensing, actuation, and above all, communication with each other. In this paper, we describe a new architecture for networks of bacteria to form a data collecting network, as in traditional sensor networks. The key to this architecture is the fact that the node in the network itself is a bacterial colony; as an individual bacterium (biological agent) is a tiny unreliable element with limited capabilities. We describe such a network under two different scenarios. We study the data gathering (sensing and multihop communication) scenario as in sensor networks followed by the consensus problem in a multi-node network. We will explain as to how the bacteria in the colony collectively orchestrate their actions as a node to perform sensing and relaying tasks that would not be possible (at least reliably) by an individual bacterium. Each single bacterium in the colony forms a belief by sensing external parameter (e.g., a molecular signal from another node) from the medium and shares its belief with other bacteria in the colony. Then, after some interactions, all the bacteria in the colony form a common belief and act as a single node. We will model the reception process of each individual bacteria and will study its impact on the overall functionality of a node. We will present results on the reliability of the multihop communication for data gathering scenario as well as the speed of convergence in the consensus scenario.

cs.IT↗

arXiv2012

Collective Sensing-Capacity of Bacteria Populations

The design of biological networks using bacteria as the basic elements of the network is initially motivated by a phenomenon called quorum sensing. Through quorum sensing, each bacterium performs sensing the medium and communicating it to others via molecular communication. As a result, bacteria can orchestrate and act collectively and perform tasks impossible otherwise. In this paper, we consider a population of bacteria as a single node in a network. In our version of biological communication networks, such a node would communicate with one another via molecular signals. As a first step toward such networks, this paper focuses on the study of the transfer of information to the population (i.e., the node) by stimulating it with a concentration of special type of a molecules signal. These molecules trigger a chain of processes inside each bacteria that results in a final output in the form of light or fluorescence. Each stage in the process adds noise to the signal carried to the next stage. Our objective is to measure (compute) the maximum amount of information that we can transfer to the node. This can be viewed as the collective sensing capacity of the node. The molecular concentration, which carries the information, is the input to the node, which should be estimated by observing the produced light as the output of the node (i.e., the entire population of bacteria forming the node). We focus on the noise caused by the random process of trapping molecules at the receptors as well as the variation of outputs of different bacteria in the node. The capacity variation with the number of bacteria in the node and the number of receptors per bacteria is obtained. Finally, we investigated the collective sensing capability of the node when a specific form of molecular signaling concentration is used.

cs.IT↗