Chandra Khatri

Vivek Dahiya, Sunny Nehra, Vipul Dholariya et al.

We evaluate whether frontier LLMs are ready for cybersecurity through a dual-mode benchmark: white-box function-level vulnerability detection (VulnLLM-R, across C/Java/Python) and black-box web application security testing (five production-style applications with 118 ground-truth vulnerabilities across 20+ CWE families, which we will open-source). We test six frontier models (GPT-5.4, Codex~5.3, Claude Opus~4.6, Sonnet~4.6, Gemini~3.1~Pro and Gemini~3~Flash) and two domain-specialized models across four testing paradigms. Our findings are sobering: (1)~every frontier model produces 10-50% false positive rates in white-box detection, systematically over-predicting vulnerabilities; (2)~in black-box testing, frontier models achieve only 4-8% ground-truth coverage, improving to just 10-19% even with external security tools (Playwright MCP, Burp Suite MCP); (3)~structured penetration-testing methodology encoded in domain-specialized agents raises per-family detection above 50%, demonstrating that methodology, not scale, is the primary lever; and (4)~a domain-specialized defense model achieves the highest precision (0.904) and lowest false positive rate (9.7%) among all models, on a single GPU. We identify the absence of structured security testing traces end-to-end request/response sequences, failure-heavy data, and multi-step attack chains as the fundamental training data bottleneck, and propose self-play security testing as a data generation strategy. Our results make the case for vertical foundation models purpose-built for cybersecurity.

Rahul Kumar, Shubham Kakde, Divyansh Rajput et al.

We present a novel approach to data preparation for developing multilingual Indic large language model. Our meticulous data acquisition spans open-source and proprietary sources, including Common Crawl, Indic books, news articles, and Wikipedia, ensuring a diverse and rich linguistic representation. For each Indic language, we design a custom preprocessing pipeline to effectively eliminate redundant and low-quality text content. Additionally, we perform deduplication on Common Crawl data to address the redundancy present in 70% of the crawled web pages. This study focuses on developing high-quality data, optimizing tokenization for our multilingual dataset for Indic large language models with 3B and 7B parameters, engineered for superior performance in Indic languages. We introduce a novel multilingual tokenizer training strategy, demonstrating our custom-trained Indic tokenizer outperforms the state-of-the-art OpenAI Tiktoken tokenizer, achieving a superior token-to-word ratio for Indic languages.

Suranjan Goswami, Abhinav Ravi, Raja Kolla et al.

Despite significant advances in document understanding, determining the correct orientation of scanned or photographed documents remains a critical pre-processing step in the real world settings. Accurate rotation correction is essential for enhancing the performance of downstream tasks such as Optical Character Recognition (OCR) where misalignment commonly arises due to user errors, particularly incorrect base orientations of the camera during capture. In this study, we first introduce OCR-Rotation-Bench (ORB), a new benchmark for evaluating OCR robustness to image rotations, comprising (i) ORB-En, built from rotation-transformed structured and free-form English OCR datasets, and (ii) ORB-Indic, a novel multilingual set spanning 11 Indic mid to low-resource languages. We also present a fast, robust and lightweight rotation classification pipeline built on the vision encoder of Phi-3.5-Vision model with dynamic image cropping, fine-tuned specifically for 4-class rotation task in a standalone fashion. Our method achieves near-perfect 96% and 92% accuracy on identifying the rotations respectively on both the datasets. Beyond classification, we demonstrate the critical role of our module in boosting OCR performance: closed-source (up to 14%) and open-weights models (up to 4x) in the simulated real-world setting.

Neel Prabhanjan Rachamalla, Aravind Konakalla, Gautam Rajeev et al.

The effectiveness of Large Language Models (LLMs) depends heavily on the availability of high-quality post-training data, particularly instruction-tuning and preference-based examples. Existing open-source datasets, however, often lack multilingual coverage, cultural grounding, and suffer from task diversity gaps that are especially pronounced for Indian languages. We introduce a human-in-the-loop pipeline that combines translations with synthetic expansion to produce reliable and diverse Indic post-training data. Using this pipeline, we curate two datasets: Pragyaan-IT (22.5K) and Pragyaan-Align (100K) across 10 Indian languages covering 13 broad and 56 sub-categories, leveraging 57 diverse datasets. Our dataset protocol incorporates several often-overlooked dimensions and emphasize task diversity, multi-turn dialogue, instruction fidelity, safety alignment, and preservation of cultural nuance, providing a foundation for more inclusive and effective multilingual LLMs.

Guduru Manoj, Neel Prabhanjan Rachamalla, Ashish Kulkarni et al.

In the context of pretraining of Large Language Models (LLMs), synthetic data has emerged as an alternative for generating high-quality pretraining data at scale. This is particularly beneficial in low-resource language settings where the benefits of recent LLMs have been unevenly distributed across languages. In this work, we present a systematic study on the generation and evaluation of synthetic multilingual pretraining data for Indic languages, where we construct a large-scale synthetic dataset BhashaKritika, comprising 540B tokens using 5 different techniques for 10 languages. We explore the impact of grounding generation in documents, personas, and topics. We analyze how language choice, both in the prompt instructions and document grounding, affects data quality, and we compare translations of English content with native generation in Indic languages. To support scalable and language-sensitive evaluation, we introduce a modular quality evaluation pipeline that integrates script and language detection, metadata consistency checks, n-gram repetition analysis, and perplexity-based filtering using KenLM models. Our framework enables robust quality control across diverse scripts and linguistic contexts. Empirical results through model runs reveal key trade-offs in generation strategies and highlight best practices for constructing effective multilingual corpora.

Shaharukh Khan, Ali Faraz, Abhinav Ravi et al.

Multimodal research has predominantly focused on single-image reasoning, with limited exploration of multi-image scenarios. Recent models have sought to enhance multi-image understanding through large-scale pretraining on interleaved image-text datasets. However, most Vision-Language Models (VLMs) are trained primarily on English datasets, leading to inadequate representation of Indian languages. To address this gap, we introduce the Chitrakshara dataset series, covering 11 Indian languages sourced from Common Crawl. It comprises (1) Chitrakshara-IL, a large-scale interleaved pretraining dataset with 193M images, 30B text tokens, and 50M multilingual documents, and (2) Chitrakshara-Cap, which includes 44M image-text pairs with 733M tokens. This paper details the data collection pipeline, including curation, filtering, and processing methodologies. Additionally, we present a comprehensive quality and diversity analysis to assess the dataset's representativeness across Indic languages and its potential for developing more culturally inclusive VLMs.

Ali Faraz, Akash, Shaharukh Khan et al.

Vision-language models (VLMs) have demonstrated impressive generalization across multimodal tasks, yet most evaluation benchmarks remain Western-centric, leaving open questions about their performance in culturally diverse and multilingual settings. To address this gap, we introduce IndicVisionBench, the first large-scale benchmark centered on the Indian subcontinent. Covering English and 10 Indian languages, our benchmark spans 3 multimodal tasks, including Optical Character Recognition (OCR), Multimodal Machine Translation (MMT), and Visual Question Answering (VQA), covering 6 kinds of question types. Our final benchmark consists of a total of ~5K images and 37K+ QA pairs across 13 culturally grounded topics. In addition, we release a paired parallel corpus of annotations across 10 Indic languages, creating a unique resource for analyzing cultural and linguistic biases in VLMs. We evaluate a broad spectrum of 8 models, from proprietary closed-source systems to open-weights medium and large-scale models. Our experiments reveal substantial performance gaps, underscoring the limitations of current VLMs in culturally diverse contexts. By centering cultural diversity and multilinguality, IndicVisionBench establishes a reproducible evaluation framework that paves the way for more inclusive multimodal research.

Souvik Rana, Arul Menezes, Ashish Kulkarni et al.

Tokenizers play a crucial role in determining the performance, training efficiency, and the inference cost of Large Language Models (LLMs). Designing effective tokenizers for multilingual LLMs is particularly challenging due to diverse scripts and rich morphological variation. While subword methods such as Byte Pair Encoding (BPE) are widely adopted, their effectiveness in multilingual settings remains underexplored. We present IndicSuperTokenizer, a tokenizer for Indic multilingual LLMs, that combines both subword and multi-word tokenization, along with language-specific pre-tokenization, leading to more linguistically aligned tokens and achieving a new state-of-the-art in fertility score. Evaluated across English, 22 Indian languages and code data, our tokenizer improves the average fertility score by 39.5% over LLaMA4 and by 18% over Sutra (the current best). This translates to 44% improvement in inference throughput over LLaMA4 while maintaining comparable performance on English and Indic benchmarks. We also present detailed ablations across tokenizer training data size, vocabulary size, merging techniques, and pre-tokenization strategies, demonstrating the robustness of our design choices.

Shaharukh Khan, Ayush Tarun, Abhinav Ravi et al.

Recent multimodal foundation models are primarily trained on English or high resource European language data, which hinders their applicability to other medium and low-resource languages. To address this limitation, we introduce Chitrarth (Chitra: Image; Artha: Meaning), an inclusive Vision-Language Model (VLM), specifically targeting the rich linguistic diversity and visual reasoning across 10 prominent Indian languages. Our model effectively integrates a state-of-the-art (SOTA) multilingual Large Language Model (LLM) with a vision module, primarily trained on multilingual image-text data. Furthermore, we also introduce BharatBench, a comprehensive framework for evaluating VLMs across various Indian languages, ultimately contributing to more diverse and effective AI systems. Our model achieves SOTA results for benchmarks across low resource languages while retaining its efficiency in English. Through our research, we aim to set new benchmarks in multilingual-multimodal capabilities, offering substantial improvements over existing models and establishing a foundation to facilitate future advancements in this arena.

Shaharukh Khan, Ayush Tarun, Ali Faraz et al.

In this work, we provide the system description of our submission as part of the English to Lowres Multimodal Translation Task at the Workshop on Asian Translation (WAT2024). We introduce Chitranuvad, a multimodal model that effectively integrates Multilingual LLM and a vision module for Multimodal Translation. Our method uses a ViT image encoder to extract visual representations as visual token embeddings which are projected to the LLM space by an adapter layer and generates translation in an autoregressive fashion. We participated in all the three tracks (Image Captioning, Text only and Multimodal translation tasks) for Indic languages (ie. English translation to Hindi, Bengali and Malyalam) and achieved SOTA results for Hindi in all of them on the Challenge set while remaining competitive for the other languages in the shared task.

Aditya Kallappa, Palash Kamble, Abhinav Ravi et al.

India is a diverse society with unique challenges in developing AI systems, including linguistic diversity, oral traditions, data accessibility, and scalability. Existing foundation models are primarily trained on English, limiting their effectiveness for India's population. Indic languages comprise only 1 percent of Common Crawl corpora despite India representing 18 percent of the global population, leading to linguistic biases. Thousands of regional languages, dialects, and code mixing create additional representation challenges due to sparse training data. We introduce Krutrim LLM, a 2 trillion token multilingual model designed for India's linguistic landscape. It incorporates the largest known Indic dataset, mitigating data scarcity and ensuring balanced performance across dialects. Krutrim outperforms or matches state-of-the-art models on Indic benchmarks while maintaining competitive English performance. Despite being significantly smaller in training flops, Krutrim LLM matches or exceeds models like LLAMA-2 on 10 out of 16 tasks, with an average score of 0.57 versus 0.55. This evidences Krutrim's flexible multilingual fluency across diverse linguistic contexts. Krutrim is integrated with real-time search to improve factual accuracy in conversational AI applications. This enhances accessibility for over 1 billion users worldwide. Through intentional design choices addressing data imbalances, Krutrim LLM signifies meaningful progress in building ethical, globally representative AI models.

Dhruv Jain, Harshit Shukla, Gautam Rajeev et al.

Large-scale Speech Language Models (SpeechLMs) have enabled voice assistants capable of understanding natural spoken queries and performing complex tasks. However, existing speech benchmarks primarily focus on isolated capabilities such as transcription, or question-answering, and do not systematically evaluate agentic scenarios encompassing multilingual and cultural understanding, as well as adversarial robustness. To address this, we introduce VoiceAgentBench, a comprehensive benchmark designed to evaluate SpeechLMs in realistic spoken agentic settings. It comprises over 5,500 synthetic spoken queries, including dialogues grounded in Indian context, covering single-tool invocations, multi-tool workflows, multi-turn interactions, and safety evaluations. The benchmark supports English, Hindi, and 5 other Indian languages, reflecting real-world linguistic and cultural diversity. We simulate speaker variability using a novel sampling algorithm that selects audios for TTS voice conversion based on its speaker embeddings, maximizing acoustic and speaker diversity. Our evaluation measures tool selection accuracy, structural consistency, and the correctness of tool invocations, including adversarial robustness. Our experiments reveal significant gaps in contextual tool orchestration tasks, Indic generalization, and adversarial robustness, exposing critical limitations of current SpeechLMs.

Mehrnoosh Mirtaheri, Nikhil Varghese, Chandra Khatri et al.

Task-oriented dialogue systems rely on predefined conversation schemes (dialogue flows) often represented as directed acyclic graphs. These flows can be manually designed or automatically generated from previously recorded conversations. Due to variations in domain expertise or reliance on different sets of prior conversations, these dialogue flows can manifest in significantly different graph structures. Despite their importance, there is no standard method for evaluating the quality of dialogue flows. We introduce FuDGE (Fuzzy Dialogue-Graph Edit Distance), a novel metric that evaluates dialogue flows by assessing their structural complexity and representational coverage of the conversation data. FuDGE measures how well individual conversations align with a flow and, consequently, how well a set of conversations is represented by the flow overall. Through extensive experiments on manually configured flows and flows generated by automated techniques, we demonstrate the effectiveness of FuDGE and its evaluation framework. By standardizing and optimizing dialogue flows, FuDGE enables conversational designers and automated techniques to achieve higher levels of efficiency and automation.

Pranav Guruprasad, Negar Mokhberian, Nikhil Varghese et al.

Intent discovery is crucial for both building new conversational agents and improving existing ones. While several approaches have been proposed for intent discovery, most rely on clustering to group similar utterances together. Traditional evaluation of these utterance clusters requires intent labels for each utterance, limiting scalability. Although some clustering quality metrics exist that do not require labeled data, they focus solely on cluster geometry while ignoring the linguistic nuances present in conversational transcripts. In this paper, we introduce Keyword-based Utterance Level Clustering Quality (KULCQ), an unsupervised metric that leverages keyword analysis to evaluate clustering quality. We demonstrate KULCQ's effectiveness by comparing it with existing unsupervised clustering metrics and validate its performance through comprehensive ablation studies. Our results show that KULCQ better captures semantic relationships in conversational data while maintaining consistency with geometric clustering principles.

Amol Kelkar, Rohan Relan, Vaishali Bhardwaj et al.

To access data stored in relational databases, users need to understand the database schema and write a query using a query language such as SQL. To simplify this task, text-to-SQL models attempt to translate a user's natural language question to corresponding SQL query. Recently, several generative text-to-SQL models have been developed. We propose a novel discriminative re-ranker to improve the performance of generative text-to-SQL models by extracting the best SQL query from the beam output predicted by the text-to-SQL generator, resulting in improved performance in the cases where the best query was in the candidate list, but not at the top of the list. We build the re-ranker as a schema agnostic BERT fine-tuned classifier. We analyze relative strengths of the text-to-SQL and re-ranker models across different query hardness levels, and suggest how to combine the two models for optimal performance. We demonstrate the effectiveness of the re-ranker by applying it to two state-of-the-art text-to-SQL models, and achieve top 4 score on the Spider leaderboard at the time of writing this article.

Yue Weng, Sai Sumanth Miryala, Chandra Khatri et al.

The quality of automatic speech recognition (ASR) is critical to Dialogue Systems as ASR errors propagate to and directly impact downstream tasks such as language understanding (LU). In this paper, we propose multi-task neural approaches to perform contextual language correction on ASR outputs jointly with LU to improve the performance of both tasks simultaneously. To measure the effectiveness of this approach we used a public benchmark, the 2nd Dialogue State Tracking (DSTC2) corpus. As a baseline approach, we trained task-specific Statistical Language Models (SLM) and fine-tuned state-of-the-art Generalized Pre-training (GPT) Language Model to re-rank the n-best ASR hypotheses, followed by a model to identify the dialog act and slots. i) We further trained ranker models using GPT and Hierarchical CNN-RNN models with discriminatory losses to detect the best output given n-best hypotheses. We extended these ranker models to first select the best ASR output and then identify the dialogue act and slots in an end to end fashion. ii) We also proposed a novel joint ASR error correction and LU model, a word confusion pointer network (WCN-Ptr) with multi-head self-attention on top, which consumes the word confusions populated from the n-best. We show that the error rates of off the shelf ASR and following LU systems can be reduced significantly by 14% relative with joint models trained using small amounts of in-domain data.

Andrea Madotto, Mahdi Namazifar, Joost Huizinga et al.

This work presents an exploration and imitation-learning-based agent capable of state-of-the-art performance in playing text-based computer games. Text-based computer games describe their world to the player through natural language and expect the player to interact with the game using text. These games are of interest as they can be seen as a testbed for language understanding, problem-solving, and language generation by artificial agents. Moreover, they provide a learning environment in which these skills can be acquired through interactions with an environment rather than using fixed corpora. One aspect that makes these games particularly challenging for learning agents is the combinatorially large action space. Existing methods for solving text-based games are limited to games that are either very simple or have an action space restricted to a predetermined set of admissible actions. In this work, we propose to use the exploration approach of Go-Explore for solving text-based games. More specifically, in an initial exploration phase, we first extract trajectories with high rewards, after which we train a policy to solve the game by imitating these trajectories. Our experiments show that this approach outperforms existing solutions in solving text-based games, and it is more sample efficient in terms of the number of interactions with the environment. Moreover, we show that the learned policy can generalize better than existing solutions to unseen games without using any restriction on the action space.

Alexandros Papangelis, Mahdi Namazifar, Chandra Khatri et al.

As the field of Spoken Dialogue Systems and Conversational AI grows, so does the need for tools and environments that abstract away implementation details in order to expedite the development process, lower the barrier of entry to the field, and offer a common test-bed for new ideas. In this paper, we present Plato, a flexible Conversational AI platform written in Python that supports any kind of conversational agent architecture, from standard architectures to architectures with jointly-trained components, single- or multi-party interactions, and offline or online training of any conversational agent component. Plato has been designed to be easy to understand and debug and is agnostic to the underlying learning frameworks that train each component.

Sanghyun Yi, Rahul Goel, Chandra Khatri et al.

Encoder-decoder based neural architectures serve as the basis of state-of-the-art approaches in end-to-end open domain dialog systems. Since most of such systems are trained with a maximum likelihood~(MLE) objective they suffer from issues such as lack of generalizability and the generic response problem, i.e., a system response that can be an answer to a large number of user utterances, e.g., "Maybe, I don't know." Having explicit feedback on the relevance and interestingness of a system response at each turn can be a useful signal for mitigating such issues and improving system quality by selecting responses from different approaches. Towards this goal, we present a system that evaluates chatbot responses at each dialog turn for coherence and engagement. Our system provides explicit turn-level dialog quality feedback, which we show to be highly correlated with human evaluation. To show that incorporating this feedback in the neural response generation models improves dialog quality, we present two different and complementary mechanisms to incorporate explicit feedback into a neural response generation model: reranking and direct modification of the loss function during training. Our studies show that a response generation model that incorporates these combined feedback mechanisms produce more engaging and coherent responses in an open-domain spoken dialog setting, significantly improving the response quality using both automatic and human evaluation.

Alessandra Cervone, Chandra Khatri, Rahul Goel et al.

Current approaches to Natural Language Generation (NLG) for dialog mainly focus on domain-specific, task-oriented applications (e.g. restaurant booking) using limited ontologies (up to 20 slot types), usually without considering the previous conversation context. Furthermore, these approaches require large amounts of data for each domain, and do not benefit from examples that may be available for other domains. This work explores the feasibility of applying statistical NLG to scenarios requiring larger ontologies, such as multi-domain dialog applications or open-domain question answering (QA) based on knowledge graphs. We model NLG through an Encoder-Decoder framework using a large dataset of interactions between real-world users and a conversational agent for open-domain QA. First, we investigate the impact of increasing the number of slot types on the generation quality and experiment with different partitions of the QA data with progressively larger ontologies (up to 369 slot types). Second, we perform multi-task learning experiments between open-domain QA and task-oriented dialog, and benchmark our model on a popular NLG dataset. Moreover, we experiment with using the conversational context as an additional input to improve response generation quality. Our experiments show the feasibility of learning statistical NLG models for open-domain QA with larger ontologies.

Chandra Khatri, Behnam Hedayatnia, Anu Venkatesh et al.

Building open domain conversational systems that allow users to have engaging conversations on topics of their choice is a challenging task. Alexa Prize was launched in 2016 to tackle the problem of achieving natural, sustained, coherent and engaging open-domain dialogs. In the second iteration of the competition in 2018, university teams advanced the state of the art by using context in dialog models, leveraging knowledge graphs for language understanding, handling complex utterances, building statistical and hierarchical dialog managers, and leveraging model-driven signals from user responses. The 2018 competition also included the provision of a suite of tools and models to the competitors including the CoBot (conversational bot) toolkit, topic and dialog act detection models, conversation evaluators, and a sensitive content detection model so that the competing teams could focus on building knowledge-rich, coherent and engaging multi-turn dialog systems. This paper outlines the advances developed by the university teams as well as the Alexa Prize team to achieve the common goal of advancing the science of Conversational AI. We address several key open-ended problems such as conversational speech recognition, open domain natural language understanding, commonsense reasoning, statistical dialog management, and dialog evaluation. These collaborative efforts have driven improved experiences by Alexa users to an average rating of 3.61, the median duration of 2 mins 18 seconds, and average turns to 14.6, increases of 14%, 92%, 54% respectively since the launch of the 2018 competition. For conversational speech recognition, we have improved our relative Word Error Rate by 55% and our relative Entity Error Rate by 34% since the launch of the Alexa Prize. Socialbots improved in quality significantly more rapidly in 2018, in part due to the release of the CoBot toolkit.

Chandra Khatri, Behnam Hedayatnia, Rahul Goel et al.

As open-ended human-chatbot interaction becomes commonplace, sensitive content detection gains importance. In this work, we propose a two stage semi-supervised approach to bootstrap large-scale data for automatic sensitive language detection from publicly available web resources. We explore various data selection methods including 1) using a blacklist to rank online discussion forums by the level of their sensitiveness followed by randomly sampling utterances and 2) training a weakly supervised model in conjunction with the blacklist for scoring sentences from online discussion forums to curate a dataset. Our data collection strategy is flexible and allows the models to detect implicit sensitive content for which manual annotations may be difficult. We train models using publicly available annotated datasets as well as using the proposed large-scale semi-supervised datasets. We evaluate the performance of all the models on Twitter and Toxic Wikipedia comments testsets as well as on a manually annotated spoken language dataset collected during a large scale chatbot competition. Results show that a model trained on this collected data outperforms the baseline models by a large margin on both in-domain and out-of-domain testsets, achieving an F1 score of 95.5% on an out-of-domain testset compared to a score of 75% for models trained on public datasets. We also showcase that large scale two stage semi-supervision generalizes well across multiple classes of sensitivities such as hate speech, racism, sexual and pornographic content, etc. without even providing explicit labels for these classes, leading to an average recall of 95.5% versus the models trained using annotated public datasets which achieve an average recall of 73.2% across seven sensitive classes on out-of-domain testsets.

Chandra Khatri, Rahul Goel, Behnam Hedayatnia et al.

Accurate prediction of conversation topics can be a valuable signal for creating coherent and engaging dialog systems. In this work, we focus on context-aware topic classification methods for identifying topics in free-form human-chatbot dialogs. We extend previous work on neural topic classification and unsupervised topic keyword detection by incorporating conversational context and dialog act features. On annotated data, we show that incorporating context and dialog acts leads to relative gains in topic classification accuracy by 35% and on unsupervised keyword detection recall by 11% for conversational interactions where topics frequently span multiple utterances. We show that topical metrics such as topical depth is highly correlated with dialog evaluation metrics such as coherence and engagement implying that conversational topic models can predict user satisfaction. Our work for detecting conversation topics and keywords can be used to guide chatbots towards coherent dialog.

Chandra Khatri, Gyanit Singh, Nish Parikh

Sequence to sequence (Seq2Seq) learning has recently been used for abstractive and extractive summarization. In current study, Seq2Seq models have been used for eBay product description summarization. We propose a novel Document-Context based Seq2Seq models using RNNs for abstractive and extractive summarizations. Intuitively, this is similar to humans reading the title, abstract or any other contextual information before reading the document. This gives humans a high-level idea of what the document is about. We use this idea and propose that Seq2Seq models should be started with contextual information at the first time-step of the input to obtain better summaries. In this manner, the output summaries are more document centric, than being generic, overcoming one of the major hurdles of using generative models. We generate document-context from user-behavior and seller provided information. We train and evaluate our models on human-extracted-golden-summaries. The document-contextual Seq2Seq models outperform standard Seq2Seq models. Moreover, generating human extracted summaries is prohibitively expensive to scale, we therefore propose a semi-supervised technique for extracting approximate summaries and using it for training Seq2Seq models at scale. Semi-supervised models are evaluated against human extracted summaries and are found to be of similar efficacy. We provide side by side comparison for abstractive and extractive summarizers (contextual and non-contextual) on same evaluation dataset. Overall, we provide methodologies to use and evaluate the proposed techniques for large document summarization. Furthermore, we found these techniques to be highly effective, which is not the case with existing techniques.

Anirudh Raju, Behnam Hedayatnia, Linda Liu et al.

Statistical language models (LM) play a key role in Automatic Speech Recognition (ASR) systems used by conversational agents. These ASR systems should provide a high accuracy under a variety of speaking styles, domains, vocabulary and argots. In this paper, we present a DNN-based method to adapt the LM to each user-agent interaction based on generalized contextual information, by predicting an optimal, context-dependent set of LM interpolation weights. We show that this framework for contextual adaptation provides accuracy improvements under different possible mixture LM partitions that are relevant for both (1) Goal-oriented conversational agents where it's natural to partition the data by the requested application and for (2) Non-goal oriented conversational agents where the data can be partitioned using topic labels that come from predictions of a topic classifier. We obtain a relative WER improvement of 3% with a 1-pass decoding strategy and 6% in a 2-pass decoding framework, over an unadapted model. We also show up to a 15% relative improvement in recognizing named entities which is of significant value for conversational ASR systems.

Chandra Khatri

In current study, a mechanism to extract traffic related information such as congestion and incidents from textual data from the internet is proposed. The current source of data is Twitter. As the data being considered is extremely large in size automated models are developed to stream, download, and mine the data in real-time. Furthermore, if any tweet has traffic related information then the models should be able to infer and extract this data. Currently, the data is collected only for United States and a total of 120,000 geo-tagged traffic related tweets are extracted, while six million geo-tagged non-traffic related tweets are retrieved and classification models are trained. Furthermore, this data is used for various kinds of spatial and temporal analysis. A mechanism to calculate level of traffic congestion, safety, and traffic perception for cities in U.S. is proposed. Traffic congestion and safety rankings for the various urban areas are obtained and then they are statistically validated with existing widely adopted rankings. Traffic perception depicts the attitude and perception of people towards the traffic. It is also seen that traffic related data when visualized spatially and temporally provides the same pattern as the actual traffic flows for various urban areas. When visualized at the city level, it is clearly visible that the flow of tweets is similar to flow of vehicles and that the traffic related tweets are representative of traffic within the cities. With all the findings in current study, it is shown that significant amount of traffic related information can be extracted from Twitter and other sources on internet. Furthermore, Twitter and these data sources are freely available and are not bound by spatial and temporal limitations. That is, wherever there is a user there is a potential for data.

Anu Venkatesh, Chandra Khatri, Ashwin Ram et al.

Conversational agents are exploding in popularity. However, much work remains in the area of non goal-oriented conversations, despite significant growth in research interest over recent years. To advance the state of the art in conversational AI, Amazon launched the Alexa Prize, a 2.5-million dollar university competition where sixteen selected university teams built conversational agents to deliver the best social conversational experience. Alexa Prize provided the academic community with the unique opportunity to perform research with a live system used by millions of users. The subjectivity associated with evaluating conversations is key element underlying the challenge of building non-goal oriented dialogue systems. In this paper, we propose a comprehensive evaluation strategy with multiple metrics designed to reduce subjectivity by selecting metrics which correlate well with human judgement. The proposed metrics provide granular analysis of the conversational agents, which is not captured in human ratings. We show that these metrics can be used as a reasonable proxy for human judgment. We provide a mechanism to unify the metrics for selecting the top performing agents, which has also been applied throughout the Alexa Prize competition. To our knowledge, to date it is the largest setting for evaluating agents with millions of conversations and hundreds of thousands of ratings from users. We believe that this work is a step towards an automatic evaluation process for conversational AIs.

Fenfei Guo, Angeliki Metallinou, Chandra Khatri et al.

Dialog evaluation is a challenging problem, especially for non task-oriented dialogs where conversational success is not well-defined. We propose to evaluate dialog quality using topic-based metrics that describe the ability of a conversational bot to sustain coherent and engaging conversations on a topic, and the diversity of topics that a bot can handle. To detect conversation topics per utterance, we adopt Deep Average Networks (DAN) and train a topic classifier on a variety of question and query data categorized into multiple topics. We propose a novel extension to DAN by adding a topic-word attention table that allows the system to jointly capture topic keywords in an utterance and perform topic classification. We compare our proposed topic based metrics with the ratings provided by users and show that our metrics both correlate with and complement human judgment. Our analysis is performed on tens of thousands of real human-bot dialogs from the Alexa Prize competition and highlights user expectations for conversational bots.

Ashwin Ram, Rohit Prasad, Chandra Khatri et al.

Conversational agents are exploding in popularity. However, much work remains in the area of social conversation as well as free-form conversation over a broad range of domains and topics. To advance the state of the art in conversational AI, Amazon launched the Alexa Prize, a 2.5-million-dollar university competition where sixteen selected university teams were challenged to build conversational agents, known as socialbots, to converse coherently and engagingly with humans on popular topics such as Sports, Politics, Entertainment, Fashion and Technology for 20 minutes. The Alexa Prize offers the academic community a unique opportunity to perform research with a live system used by millions of users. The competition provided university teams with real user conversational data at scale, along with the user-provided ratings and feedback augmented with annotations by the Alexa team. This enabled teams to effectively iterate and make improvements throughout the competition while being evaluated in real-time through live user interactions. To build their socialbots, university teams combined state-of-the-art techniques with novel strategies in the areas of Natural Language Understanding, Context Modeling, Dialog Management, Response Generation, and Knowledge Acquisition. To support the efforts of participating teams, the Alexa Prize team made significant scientific and engineering investments to build and improve Conversational Speech Recognition, Topic Tracking, Dialog Evaluation, Voice User Experience, and tools for traffic management and scalability. This paper outlines the advances created by the university teams as well as the Alexa Prize team to achieve the common goal of solving the problem of Conversational AI.