Opens in a new tab<\/span><\/p>\n Breakthroughs in Artificial Intelligence (AI) have typically shown that AI systems are good at solving specific tasks that have a well-defined goal, such as Speech Recognition, Image Captioning, Games like Poker\/Go\/Jeopardy, among others. However, as AI systems become ubiquitous, it is not enough for them to solve specific tasks; rather they will have to continuously interact with human-users as well as other AI systems in a rapidly evolving environment. These systems will have to continuously review and evolve their interaction strategies during the ongoing interaction. Goals may not be defined in advance, and might evolve dynamically. The systems have to ensure that apart from solving the primary task, the user receives a pleasant and professional experience that is\u202f”socio-culturally appropriate”. In other words, we are quickly moving towards a world where AI systems have to go far beyond functional intelligence\u202f \u2013 they have to be socio-culturally adept and behaviourally intelligent. We refer to this phenomenon as \u201cArtificial Social Intelligence<\/strong>\u201d and the consequent systems as Socially Intelligent Agents<\/em>.<\/p>\n As one can imagine, ASI is an extremely multi-disciplinary endeavor, where one needs inputs not only from AI and Machine-Learning researchers, but also linguists, social scientists, HCI, design and vision researchers.<\/p>\n The above figure shows a hypothetical interaction between a chatbot \u201cbotty<\/em>\u201d and a young boy \u201cchhota bheem<\/em>\u201d in Hinglish to demonstrate the importance of ASI. While botty<\/em> is able to interpret sentences and generate responses perfectly, it misses the fact that \u201ca princess hat\u201d is not a culturally appropriate birthday gift for chhota bheem\u2019<\/em>s mother. Thus, recommender systems (if you imagine botty had a gift recommender system embedded within it) need to take into account larger as well as user-specific socio-cultural contexts into account while making recommendations. Further, one might observe that botty has used \u201cuske\u201d (non-honorific pronoun) for Chhota Bheem\u2019<\/em>s mother and \u201cunko\u201d (the honorific pronoun) for referring to his younger sister, though the conversation etiquettes and pragmatics of Hindi demands the pronouns to be used the other way round.<\/p>\n ASI is an emergent field. While there has been research on some specific aspects of ASI, the parts are yet to come together and coalesce into a field or an interactive AI agent. We believe there are four fundamental sets of problems within the broad scope of ASI, which though can be dealt with independently, at the end should feed into each other:<\/p>\n Proposals spanning any of the above sub-areas of ASI are welcome.<\/p>\n References<\/strong><\/p>\n [1] Mark my words! Linguistic style accommodation in social media. (opens in new tab)<\/span><\/a><\/p>\n [2] Understanding Language Preference for Expression of Opinion and Sentiment: What do Hindi-English Speakers do on Twitter? (opens in new tab)<\/span><\/a><\/p>\n [3] Managing human-robot engagement with forecasts and \u2026 um \u2026 Hesitations. (opens in new tab)<\/span><\/a><\/p>\n [4] Fairness, Accountability, and Transparency in Machine Learning Workshop Series (opens in new tab)<\/span><\/a><\/p>\n [5] https:\/\/www.technologyreview.com\/s\/602692\/chatbots-with-social-skills-will-convince-you-to-buy-something\/?set=602726 (opens in new tab)<\/span><\/a><\/p>\n [6] Strategy and policy learning for non-task oriented conversational bots (opens in new tab)<\/span><\/a><\/p>\n [7] On data driven parametric backchannel synthesis for expressing attentiveness in conversational agents (opens in new tab)<\/span><\/a><\/p>\n [8] Deciphering the Silent Participant: On the Use of Audio-Visual Cues for the Classification of Listener Categories in Group Discussions (opens in new tab)<\/span><\/a><\/p>\n [9] Conversational involvement and synchronous nonverbal behaviour (opens in new tab)<\/span><\/a><\/p>\n [10] Towards the automatic detection of involvement in conversation (opens in new tab)<\/span><\/a><\/p>\n [11] Modeling ethnicity in\/with technology: Using virtual agents to understand sociolinguistic variation (opens in new tab)<\/span><\/a><\/p>\n [12] Negotiated Collusion: Modeling Social Language and its Relationship Effects in Intelligent Agents (opens in new tab)<\/span><\/a><\/p>\n [13] ‘How about this weather?’ Social Dialog with Embodied Conversational Agents (opens in new tab)<\/span><\/a>Opens in a new tab<\/span><\/p>\n The MSR Summer Workshop on Artificial Social Intelligence (ASI) will be an intense project-based research endeavor to enable intelligent systems to be more socially and culturally aware. Proposals are invited from faculty members in Indian universities and Indian start-ups in topics including but not limited to:<\/p>\n Proposals could suggest building a Socially Intelligent Agent or generic platforms for architecting such agents; we also invite proposals that seek to conduct large-scale socio-cultural studies using computational methods that will guide architecting ASI systems. Projects done in the summer workshop should lead to either a working system, dataset or an in-depth, large-scale study leading to publishable work.<\/p>\n All faculty and students selected for the workshop are expected not to pursue parallel work during the workshop, since this is intended to be an intense 4 week effort leading to publishable work and significant progress in the field.<\/p>\n The submitted proposals along with the nominations for PhD and\/or Postdocs will be selected through a thorough evaluation and revision process by the Program Committee. In parallel, the undergrad and masters students for the workshop will be selected by the Organizing Committee through a different process.<\/p>\n The proposals will be evaluated by the members of the program committee on the basis of following criteria: innovativeness<\/em>, generality,<\/em> usability<\/em> (of the proposed system\/study), feasibility<\/em> (of completing the project in 4 weeks), and preciseness of the goals<\/em> and success metrics<\/em>.<\/p>\n After initial round of evaluation, proposals will be shortlisted and the proposers will be invited to Microsoft Research India to give a presentation. Post this 4 to 6 projects will be selected for the summer workshop. Proposals will be refined and may be merged with other proposals before the final selection.<\/p>\n Proposals are invited from faculty in Indian institutions in the following format:<\/p>\n Proposals should be no longer than 2000 words (including diagrams, tables etc.), and can be sent in doc or pdf format.<\/p>\n Along with the proposal, please also send<\/p>\n The completed proposals should be emailed to\u00a0msriasi2017@microsoft.com, cc-ing Monojit Choudhury at monojitc@microsoft.com no later than 11:59 pm IST 28th<\/sup> February 2017.<\/p>\n Please note that we cannot guarantee the selection of the students nominated in the proposal even if the proposal is selected, because we are seeking UG\/masters student nominations through other means as well.<\/p>\n Deadline for submitting proposals: Feb 28, 2017<\/strong> Faculty\/Researchers: <\/strong>Faculty members from universities in India are invited to submit proposals. They will be asked to present the proposal in person at Bangalore, if shortlisted. Projects can also have ONE collaborator from industry\/other academic institutions from India or abroad. Selected faculty should be willing to work with other faculty\/researchers if their proposals get merged. Faculty members can nominate ONE Post Docs\/PhD student and up to TWO UG\/master students in their proposal. Faculty members should also be willing to give lectures and tutorials in their areas of expertise during the workshop.<\/p>\n The Principal proposer of each selected project will receive an amount of Rs. 1.5 Lakhs as remuneration for the research efforts in the project. This amount is also expected to cover their travel to\/from Bangalore, meals outside of those provided during the workshop and other incidental expenses. Faculty will have to make their own travel arrangements. Microsoft Research will provide accommodation with breakfast, and meals during working hours.<\/p>\n Important<\/strong>: The above remuneration will only be paid to the principal faculty proposer. Collaborators, if any, will receive accommodation and meals.<\/p>\n Post Doc\/PhD students: <\/strong>Post Doc and PhD students CANNOT apply directly. They need to be nominated by the faculty member writing the proposal. Each proposal can have ONE Post Doc or PhD student nomination, from which the program committee will make the final selection.<\/p>\n Postdoc\/PhD students will be provided with accommodation with breakfast, and meals during working hours. For outstation students, we will provide a travel allowance that should cover a large part of the estimated airfare (if booked well in advance), meals outside working hours and any other incidental expenses. They do not need to submit any bills or tickets to get the travel allowance. It will be a flat amount based on the city where their college is located. If they spend less than the allowance, they do not need to return the funds to us and if they spend more, we hope they can put in money from their own sources to cover the balance.<\/p>\n Masters\/Undergraduate students: <\/strong>Masters and Undergraduate students will be selected by the program committee and assigned to projects based on their interests and skills. There are two routes through which these students can apply:<\/p>\n UG\/Masters students will be provided with accommodation (on twin sharing basis) with breakfast, and meals during working hours. For outstation students, we will provide a travel allowance that should cover a large part of the estimated airfare (if booked well in advance), meals outside working hours and any other incidental expenses. They do not need to submit any bills or tickets to get the travel allowance. It will be a flat amount based on the city where their college is located. If they spend less than the allowance, they do not need to return the funds to us and if they spend more, we hope they can put in money from their own sources to cover the balance.<\/p>\n Garage style workstations will be provided to each team in Microsoft Research India premises. Team members are expected to work on their own laptops. However, we will provide Internet connection as well as Azure accounts to each team for running large scale experiments. Access to certain MS proprietary datasets and development environment could also be provided based on need and availability.<\/p>\n Each project will be assigned a research collaborator as well as an engineer from MSR\/Microsoft, who will actively involve in the research as well as the development aspects of the project.<\/p>\n All projects will be presented and there will be a competition at the end of the workshop to choose the winning project. The winning project will be awarded seed funding of INR 700,000\/- from Microsoft Research India to continue the work.<\/p>\n Based on their performance in the workshop, the UG\/PG students participating in the workshop will be considered for internship at MSR, if they wish so.<\/p>\n For any queries on the proposal \/ eligibility, etc. please write to : msriasi2017@microsoft.com, cc-ing Monojit Choudhury at monojitc@microsoft.comOpens in a new tab<\/span><\/p>\n For any queries on the proposal \/ eligibility, etc. please write to : msriasi2017@microsoft.com, cc-ing Monojit Choudhury at monojitc@microsoft.comOpens in a new tab<\/span><\/p>\n \u00a01. Accent Adaptation in ASR Systems. \u00a0<\/strong><\/p>\n Proposer –\u00a0 Prof. Preethi Jyothi, IIT Bombay<\/strong><\/p>\n Abstract:<\/strong> Voice-driven automated agents such as personal assistants\u00a0<\/span>are becoming increasingly popular. However, in a multi-lingual and\u00a0<\/span>multi-cultural country like India, deploying such agents to engage\u00a0<\/span>with large sections of the population is highly challenging. A major\u00a0<\/span>hindrance in this regard is the difficulty the agents would face in\u00a0<\/span>understanding varying speech accents of the users. Even when the\u00a0<\/span>language of interaction with the underlying automatic speech\u00a0<\/span>recognition (ASR) system is restricted to a lingua franca (such as\u00a0<\/span>English), the accent of the speaker can vary dramatically based on\u00a0<\/span>their cultural and linguistic background, posing a fundamental\u00a0<\/span>challenge for ASR systems. Tackling this challenge will be a necessary\u00a0<\/span>first step towards building socially accepted and commercially successful agents in the Indian context. <\/span><\/p>\n The main focus of this project will be to take this first step, by\u00a0<\/span>improving state-of-the-art performance of ASR systems on accented\u00a0<\/span>speech – specifically, speech with Indian accents. We shall develop\u00a0<\/span> deep neural network based acoustic models that will be trained using\u00a0<\/span>not only accented speech data but also speech in the native languages\u00a0<\/span>associated with the accent. We shall also develop a tool that will be\u00a0<\/span>trained to identify various Indian accents automatically. Finally, we\u00a0<\/span>shall investigate how accented-speech-ASR can be effectively\u00a0<\/span>incorporated into intelligent agents to help them act in socio-culturally appropriate ways.<\/span><\/p>\n 2. \u00a0HollyChat: Domain Specific Conversation Systems. <\/strong><\/p>\n Proposer –\u00a0Prof. Mitesh Khapra, IIT Madras<\/strong><\/p>\n Abstract: <\/strong>Most of the AI systems today are driven by three key components (i) data (ii) common sense knowledge and (iii) powerful learning algorithms which can harness this data and knowledge to learn task specific meaningful patterns. Recently there has been a lot of interest in domain-specific dialog systems with several downstream use cases such as shopping assistants, customer support, tour guides, etc. Most of the existing dialog systems are partly in line with the trend mentioned above – in the sense that they are data driven and use powerful algorithms (deep recurrent neural networks and their variants). However, we are nowhere close to building deployable domain-specific conversation systems. One of the primary reasons for this shortfall is that such agents do not exploit any common sense or real-world knowledge, and thereby are unable to maintain a richer context of the conversation. We propose to focus on domain specific conversation systems which use domain specific knowledge graphs as external memory. The idea is to use a domain-specific knowledge graph to discover the latent intent of the user and keep the conversation coherent with this intent. For example, the knowledge graph driven intention network could map the user’s utterance textit{“I really liked the action sequences in Inception (movie)”} to all tuples containing the entity textit{“Inception”} and keep the conversation restricted to concepts linked to this entity. An appropriate response in this case could be textit{“Yes, indeed, movies directed by Christopher Nolan are known for their action”} which contains entities and predicates linked to textit{“Inception”}. This would help in the task of dialog planning and also address the problem of keeping track of large contexts (which would be required for longer conversations containing many turns). In this case, the model could learn to abstract out the context in terms of entities and predicates in the knowledge base and just track these and their immediate neighborhood in the knowledge graph.\u00a0<\/span><\/p>\n 3. Detection of Aggressive Behavior on Social Media. <\/strong><\/p>\n Proposer: Prof. Ritesh Kumar, Ambedkar University<\/strong><\/p>\n Abstract: <\/strong>As the interaction over the web has increased, incidents of aggression and related events like trolling, cyberbullying, flaming, hate speech, etc. too have increased manifold across the globe. While most of these behaviour like bullying or hate speech have predated the Internet, the reach and extent of the Internet has given these an unprecedented power and influence to affect the lives of billions of people. It has been observed that the incidents of aggression and unratified verbal behaviour has not remained just a minor nuisance but has acquired the form of a major criminal activity that affects a large number of people. These have not only created mental and psychological agony to the users of the web but has in fact forced people to deactivate their accounts and in rare instances also commit suicides. So it is of utmost significance and importance that some preventive measures be taken to safeguard the interests of the people using the web as well as of the web itself such that it remains a viable medium of communication and connection, in general.<\/p>\n The aim of the project is to develop a prototype that could automatically tell ratified (both aggressive as well as non-aggressive) linguistic behaviour from unratified (aggressive) ones (recognised by varied names like flaming, aggression, trolling, hate speech, cyberbullying, etc.) on the online forums (especially social media and news\/opinion websites\/blogs). I propose to develop the system using supervised text classification methods combined with sequence models that would be trained using a dataset annotated with labels pertaining to aggression in Hindi and Hindi-English code-mixed data collected from different kinds of Facebook Pages including those of news\/media organisations, support\/help groups, celebrity pages and other similar pages as well as from certain focussed topics\/themes on Twitter.<\/p>\n 4.\u00a0Utilising Social Media for Disaster Relief: Linguistic Analysis of Resource Requests and Offers on Twitter. <\/strong><\/p>\n Proposer: Prof. Saptarshi Ghosh, IIT Kharagpur<\/strong><\/p>\n Abstract: <\/strong>Effective coordination of post-disaster relief operations depends critically on the availability of reliable situational information, as well as on citizen participation in the operations. The advent of online social media (e.g., Twitter, Facebook) and the widespread availability of mobile Internet today enable regular citizens to contribute to the relief operations, even if they are themselves stuck in the disaster effected zones. The aim of this project is to develop mechanisms for utilizing online social media for helping post-disaster relief operations. Specifically, our goal is to develop tools that help coordinate resource requests and resource offerings, to ensure optimal resource utilization during the disaster. Opens in a new tab<\/span><\/p>\n![\"\"](\"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-content\/uploads\/2017\/01\/botty-1-300x188.png\")
<\/p>\n\n
Update:\u00a0 We have shortlisted the final proposals along with the proposers (faculty) and the students for the Summer workshop. We will not accept any more proposals\/student nominations. <\/b><\/h2>\n
Call for Proposals<\/h2>\n
\n
Selection Process<\/h2>\n
![\"MSR](\"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-content\/uploads\/2017\/01\/selection_process.png\")
<\/p>\nSubmission Guidelines<\/h2>\n
\n
\n
\n
Important dates<\/h2>\n
\nNotification of shortlisted proposals: Mar 20, 2017<\/strong>
\nPresentation at MSRI: 1st week of April 2017<\/strong>
\nNotification of selection: April 30, 2017<\/strong>
\nSummer School: June 5-30, 2017<\/strong> (tentative)<\/p>\nWho can participate and How?<\/h2>\n
\n
Computational Infrastructure<\/h2>\n
Best Project Award<\/h2>\n
Contact<\/h2>\n
Program Committee<\/h2>\n
\n
Organizing committee<\/h2>\n
\n
Contact<\/h2>\n
\nTo this end, we first propose to analyze the linguistic characteristics of resource requests and resource offerings posted on Twitter during various disaster scenarios. This analysis is likely to yield insights into how different people phrase requests and offers for resources, in various languages. Next, we plan to utilize the insights obtained from the linguistic analysis, to build systems that will help coordinate the resource requests and offerings. Specifically, we envision building an automated bot that responds appropriately to resource requests and offers, and then matches corresponding requests and offers.<\/span><\/p>\n\n\n
\n Date<\/strong><\/td>\n 09:30 \u2013 11:00<\/strong>\u00a0<\/strong><\/td>\n 11:30 \u2013 13:00<\/strong><\/td>\n 14:30 \u2013 17:00<\/strong><\/td>\n<\/tr>\n \n 5-Jun<\/strong><\/td>\n \u00a0Welcome & Introductions<\/td>\n ASI: Monojit Choudhury<\/td>\n Tutorial: Azure & Azure ML: Gopal Srinivasa<\/td>\n<\/tr>\n \n 6-Jun<\/strong><\/td>\n Dialogue Systems: Alan Black<\/td>\n Chat Bots: Alan Black<\/td>\n Tutorial \u2013 Building Chatbots: Mitesh Khapra<\/td>\n<\/tr>\n \n 7-Jun<\/strong><\/td>\n NLP & Social Media: Monojit Choudhury<\/td>\n <\/td>\n <\/td>\n<\/tr>\n \n 8-Jun<\/strong><\/td>\n User Studies: Indrani Medhi<\/td>\n Data Ethics & Privacy: Kalika Bali<\/td>\n <\/td>\n<\/tr>\n \n 9-Jun<\/strong><\/td>\n Speech recognition: Preethi Jyothi & Sriram Ganapathy<\/td>\n <\/td>\n Tutorial \u2013 Kaldi: Preethi Jyothi & Sriram Ganapathy<\/td>\n<\/tr>\n \n \u00a0<\/strong><\/td>\n <\/td>\n <\/td>\n <\/td>\n<\/tr>\n \n 12-Jun<\/strong><\/td>\n <\/td>\n <\/td>\n Computational Sociolinguistics: A. Seza Do\u011fru\u00f6z<\/td>\n<\/tr>\n \n 13-Jun<\/strong><\/td>\n Multiligualism: Kalika Bali<\/td>\n <\/td>\n <\/td>\n<\/tr>\n \n 14-Jun<\/strong><\/td>\n Computational Pragmatics: Ritesh Kumar & Monojit Choudhury<\/td>\n <\/td>\n <\/td>\n<\/tr>\n \n 15-Jun<\/strong><\/td>\n Text Analytics: Thamar Solorio<\/td>\n <\/td>\n <\/td>\n<\/tr>\n \n 16-Jun<\/strong><\/td>\n Social Media Analytics: Saptarshi Ghosh<\/td>\n <\/td>\n <\/td>\n<\/tr>\n \n \u00a0<\/strong><\/td>\n <\/td>\n <\/td>\n <\/td>\n<\/tr>\n \n 19-Jun<\/strong><\/td>\n <\/td>\n Status Update<\/td>\n <\/td>\n<\/tr>\n \n 20-Jun<\/strong><\/td>\n Error Analysis: Sunayana Sitaram<\/td>\n <\/td>\n <\/td>\n<\/tr>\n \n 21-Jun<\/strong><\/td>\n ML for ASI – 1: Prateek Jain<\/td>\n <\/td>\n <\/td>\n<\/tr>\n \n 22-Jun<\/strong><\/td>\n ML for ASI – 2: Praneeth Netrapalli<\/td>\n <\/td>\n <\/td>\n<\/tr>\n \n