| { |
| "paper_id": "2022", |
| "header": { |
| "generated_with": "S2ORC 1.0.0", |
| "date_generated": "2023-01-19T03:45:53.994210Z" |
| }, |
| "title": "A Minimal Computational Improviser Based on Oral Thought", |
| "authors": [ |
| { |
| "first": "Nick", |
| "middle": [], |
| "last": "Montfort", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "MIT Trope Tank", |
| "location": { |
| "addrLine": "77 Mass Ave", |
| "postCode": "14E-316, 02139", |
| "settlement": "Cambridge", |
| "region": "MA", |
| "country": "USA" |
| } |
| }, |
| "email": "nickm@nickm.com" |
| }, |
| { |
| "first": "Sebastian", |
| "middle": [ |
| "Bartlett" |
| ], |
| "last": "Fernandez", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "MIT Trope Tank", |
| "location": { |
| "addrLine": "77 Mass Ave", |
| "postCode": "14E-316, 02139", |
| "settlement": "Cambridge", |
| "region": "MA", |
| "country": "USA" |
| } |
| }, |
| "email": "" |
| } |
| ], |
| "year": "", |
| "venue": null, |
| "identifiers": {}, |
| "abstract": "We describe our system for playing a minimal improvisational game in a group. In Chain Reaction, players collectively build a chain of word pairs or solid compounds. The game emphasizes memory and rapid improvisation, while absurdity and humor increases during play. Our approach is unique in that we have grounded our work in the principles of oral culture according to Walter Ong, an early scholar of orature. We show how a simple computer model can be designed to embody many aspects of oral poetics, suggesting design directions for other work in oral improvisation and poetics. The opportunities for our system's further development include creating culturally specific automated players; situating play in different temporal, physical, and social contexts; and constructing a more elaborate improvisor.", |
| "pdf_parse": { |
| "paper_id": "2022", |
| "_pdf_hash": "", |
| "abstract": [ |
| { |
| "text": "We describe our system for playing a minimal improvisational game in a group. In Chain Reaction, players collectively build a chain of word pairs or solid compounds. The game emphasizes memory and rapid improvisation, while absurdity and humor increases during play. Our approach is unique in that we have grounded our work in the principles of oral culture according to Walter Ong, an early scholar of orature. We show how a simple computer model can be designed to embody many aspects of oral poetics, suggesting design directions for other work in oral improvisation and poetics. The opportunities for our system's further development include creating culturally specific automated players; situating play in different temporal, physical, and social contexts; and constructing a more elaborate improvisor.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Abstract", |
| "sec_num": null |
| } |
| ], |
| "body_text": [ |
| { |
| "text": "We developed a prototype computer system to play the game Chain Reaction. This is both a memory game and an oral improvisational game, and may be the simplest such game. The game is best introduced with an example. Consider four players sitting in a circle, uttering the following:", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "\u2022 Player 1, beginning the game: Post office. can't think of anything! \u2022 Player 2: Shelf life! \u2022 Player 1: Too late! Let's start again. Each player's task is to continue the chain by reciting it quickly, without hesitation, and to immediately add a word that will create a coherent pair with the word before it. \"Coherent\" simply means that the pair must refer to a meaningful single item or concept. In written language we could call this a collocation or a bigram, but a pair could also form a solid compound such as \"chairman\" or \"manchild.\" There is no requirement that the added word make any sense when joined together with any earlier words, only that the last word and the added word together constitute \"a thing.\" As seen in the case of band/banned, words are considered to be oral units and it is fine to continue the chain while treating the previous word as a homophone. Typically, there is a prohibition on re-using words within the same game. Although this example doesn't show it, it's fine to use verb phrases (e.g., \"slide off,\" \"run up\"), which also constitute \"a thing.\" Many of these are easy to continue (e.g., \"off brand,\" \"up hill\").", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "What exactly is \"a thing\"? Like obscenity, we know it when we see it, or in this case, when we hear it. \"A thing\" is determined by consensus. It is whatever the group accepts as a suitable two-word phrase or compound word. In the example given, Player 2 might have made a first move that involved uttering \"post office plant\" (an office plant being similar to a house plant, but for the office). This would have been a less obvious phrase, yet probably acceptable. Even so, the continuations \"office salad\" and \"office sky\" would probably not have been accepted, even if the player could have spun a story about how there are such things (a salad stored in the office kitchen's fridge and consumed at one's desk or the view of the sky from one's office, for instance). If a player has to stop to explain the phrase, the game's continuity is broken, so such phrases are, at least, not good moves.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "New players usually find it surprisingly easy to recite long chains. There are reasons for this related to oral practices. A fundamental reason involves shared cultural and linguistic expectations. Each word pair in the chain, after all, was almost instantly thought up by a player.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "Oral Improvisation?", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Why Build Computational Models of", |
| "sec_num": "2" |
| }, |
| { |
| "text": "There has been intriguing work done to implement interactive improvisational games for language learning (Morgado da Costa and Sio, 2020). One game, Forced Links, is even related to chain reaction, in that it involves the formation of word chains, although not a word at a time. Our work differs from this project because it does not have any goal extrinsic to the game itself, and because it is strongly based on oral practices and principles. Oral cultures and thought encompass a broad span of human history prior to the development of writing and the establishment of cultures grounded in literacy. Aspects of oral thought persist today, as do new oral practices that are now situated in a culture suffused with manuscript, print, and electronic practices. A prominent and innovative one, for instance, is freestyle rap.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Why Build Computational Models of", |
| "sec_num": "2" |
| }, |
| { |
| "text": "There are many approaches to understanding complex phenomena such as orality. We chose the epistemological approach of developing a computational model. This allows us to explore the nature and consequence of orality in three distinct ways. First, in the process of building the system we can discover constituent elements of oral thought in accordance with the computational-imperative principle: \"any model of human intelligence should introduce only computational capabilities that enable observed behaviors without enabling unobserved behaviors.\" (Winston and Holmes, 2018) . Second, we can examine the system's functions and connections as a map towards understanding the functions and structures of orality in action. Last, by engaging with the built system, we can explore surprises and unexpected (yet sometimes positive) behavior, which can be used as further input to generate new questions to explore.", |
| "cite_spans": [ |
| { |
| "start": 551, |
| "end": 577, |
| "text": "(Winston and Holmes, 2018)", |
| "ref_id": "BIBREF16" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Why Build Computational Models of", |
| "sec_num": "2" |
| }, |
| { |
| "text": "There have been remarkable systems for oral improvisation and poetics, including a physical robot which engages in an agonistic, improvised singing practice, bertsolaritza, that is traditional in the Basque Country. (Astigarraga et al., 2013) This \"Bertsobot\" does a complex sort of improvisation and is a multimodal system with an elaborate architecture, as opposed to the system we describe. The interface is speech-based. However the underlying generation of verses, based on a vector-space model of sentences, is not based on oral principles. The way language is characterized by researchers (as \"sentences\" rather than \"utterances\") indicates a literate-culture orientation in the system's design and development. Bertsobot's sentence selection is also based on a corpus that consists of some preexisting, transcribed bertsos, but mostly of text from a newspaper, Berria -a literate, written source. More recently, a physical robot that can engage in freestyle rap battles, Shimon, has been developed (Savery et al., 2020) . Again, this is an elaborate system that multimodally performs a complex type of improvisation. Shimon was trained on a database of rap lyrics (and, in a different condition, metal music lyrics), but these lyrics were composed rather than improvised. Researchers studied flow and drew on how-to books about rapping, but it not clear that the project was informed by the ethnographic literature on freestyle or a significant neuroscience study (Liu et al., 2012) .", |
| "cite_spans": [ |
| { |
| "start": 216, |
| "end": 242, |
| "text": "(Astigarraga et al., 2013)", |
| "ref_id": "BIBREF1" |
| }, |
| { |
| "start": 1006, |
| "end": 1027, |
| "text": "(Savery et al., 2020)", |
| "ref_id": "BIBREF14" |
| }, |
| { |
| "start": 1472, |
| "end": 1490, |
| "text": "(Liu et al., 2012)", |
| "ref_id": "BIBREF6" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Why Build Computational Models of", |
| "sec_num": "2" |
| }, |
| { |
| "text": "Both Bertsobot and Shimon were engineered to perform optimally and to be evaluated positively by people. They were developed with clear awareness of bertsolaritza and freestyle rap. Still, the language generation in these systems did not seem completely anchored to any explicit and general theory of oral poetics. While our project may seem trivial by comparison to these very elaborate ones, it does have such a basis. And by focusing on a minimal improvisational situation, and paring down our system to its essence, we hope to show what oral poetic design principles apply most widely.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Why Build Computational Models of", |
| "sec_num": "2" |
| }, |
| { |
| "text": "More than 110 years ago the technique of word association was developed to allow patients to surface unconscious ideas. (Jung, 1910) The experimenter instructs the subject: \"Answer as quickly as possible the first word that occurs to your mind.\" Then, a series of words are uttered and the response to each is recorded. In word association, it is not nec-essary to make a Chain-Reaction-style completion, a pair or \"thing.\" The reply can be an antonym, a synonym, a hypernym, a hyponym, a meronym, or anything else. The relationship of this technique to Chain Reaction is that players in the game must also utter a continuation as quickly as possible, often the first thing that comes to mind.", |
| "cite_spans": [ |
| { |
| "start": 120, |
| "end": 132, |
| "text": "(Jung, 1910)", |
| "ref_id": "BIBREF5" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "History of the Game", |
| "sec_num": "3.1" |
| }, |
| { |
| "text": "The concept of word association is culturally understood as having incongruous and humorous potential and has been employed in comedy, for instance in the in the 1975 Saturday Night Live sketch \"Word Association\" with Chevy Chase and Richard Pryor (Saucier et al., 2016) and in the 1989/1990 Monty Python's Flying Circus sketch \"Word Association Football\" (Aarons, 2012) . These are scripted performances, not improvisational verbal games, but help to illustrate that there can be humor in instantaneous responses to a word.", |
| "cite_spans": [ |
| { |
| "start": 248, |
| "end": 270, |
| "text": "(Saucier et al., 2016)", |
| "ref_id": "BIBREF13" |
| }, |
| { |
| "start": 356, |
| "end": 370, |
| "text": "(Aarons, 2012)", |
| "ref_id": "BIBREF0" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "History of the Game", |
| "sec_num": "3.1" |
| }, |
| { |
| "text": "In January 1980 a game show devised by Bob Stewart premiered on US television. The show, hosted by Bill Cullen, was called Chain Reaction and required that players guess a chain of words connecting a given first word with a given last word. In early rounds, the chain was eight words long. (Nedeff, 2013) The first run of the show was shortlived, but it has been revived several times.", |
| "cite_spans": [ |
| { |
| "start": 290, |
| "end": 304, |
| "text": "(Nedeff, 2013)", |
| "ref_id": "BIBREF10" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "History of the Game", |
| "sec_num": "3.1" |
| }, |
| { |
| "text": "We do not find the first mention of the verbal game Chain Reaction, with rules similar to those we provided in the introduction, until the early 21st Century. It is described in a book of games to be played in the car by children. (Gladstone, 2004) While Gladstone (or whoever devised this game) likely had the game show in mind, contestants on TV had to find predetermined intermediate words in a chain of fixed length. The verbal game involves the same sort of chaining, but allows the continuations to be determined by players. In the verbal game, the chain is not of fixed length, but can grow indefinitely.", |
| "cite_spans": [ |
| { |
| "start": 231, |
| "end": 248, |
| "text": "(Gladstone, 2004)", |
| "ref_id": "BIBREF3" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "History of the Game", |
| "sec_num": "3.1" |
| }, |
| { |
| "text": "There are other types of verbal chain games, for instance, some that involve connecting one word to another that are used as improv theater warm-ups. And there is a children's verbal game that involves the memory component, I'm Going to a Picnic and I'm Bringing..., where players each add an item of food and have to remember the entire list. But Gladstone's is the only precise description of Chain Reaction, under any name, that we have found.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "History of the Game", |
| "sec_num": "3.1" |
| }, |
| { |
| "text": "We have participated in many Chain Reaction games in different contexts, offering us a diversity of experience, although certainly not the distance for impartial observation. Fundamentally, our play of chain reaction serves as a \"reality check\" to demonstrate that the game as explained in this recent book is indeed playable and (for us and many participants) can be fun. Our informal but frequent experience of play offers us some insights as observers of others and allows us to reflect on our own cognition during play, as we discuss later.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "History of the Game", |
| "sec_num": "3.1" |
| }, |
| { |
| "text": "At first glance Chain Reaction seems like a competitive game. It is possible for a player to lose, as Player 4 did in the example game transcript. A player can lose by forgetting part of the chain or by failing to come up with a new word to extend the chain. In this case, we could say that everyone else in the circle gets a point for that game. If a group were playing like this, the group would want to ensure that a continuation of the chain actually existed, so the player that was unable to come up with a word might have the right to challenge the previous player (e.g., \"... plastic spatula ... uh ... come on, spatula what? There's no way to continue that!\"). The game as presented by Gladstone allows for a player to win by connecting the end of the chain to the first word uttered by Player 1. For instance, if playing with this rule, Player 2 might have said: \"post office lamp\" and Player 3 might have triumphantly declared \"post office lamp post -I win!\" Perhaps some number of points greater than one could be given to the player who achieves the rare loop of this sort.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Seemingly Competitive, but Cooperative", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "While we can discuss fine details of how this game can be scored and thus won or lost, the truth is, in social circumstances observed by authorparticipants, the game is not really played competitively. It is a cooperative game such as footbag (the trademarked name of which is Hackeysack). In this game, a circle of players kick a small bag filled with sand into the air with their feet, trying to keep it from hitting the ground. In this social game, a good footbag player is one who keeps the bag in the air and makes it possible for others to also play, extending the fun of the game. The same is true in Chain Reaction.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Seemingly Competitive, but Cooperative", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "Much of the fun in Chain Reaction arises from the relationship between individuals, the cultural commonality in the group, and the cultural differences between players. Different players have some cognitive differences. They draw from varying life experiences and find different ways to continue chains.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Context, Individuality, Cultures", |
| "sec_num": "3.3" |
| }, |
| { |
| "text": "Players can come from different cultures, which might influence the way they follow up particular words: American and Canadian players might choose \"station eleven\" (the name of an American TV series based on a Canadian novel) while British players might be more inclined to say \"station stop\" (a chiefly British way to indicate a train station). Yet players do share a language, as well as the commonality of broadly defined human experience.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Context, Individuality, Cultures", |
| "sec_num": "3.3" |
| }, |
| { |
| "text": "Gameplay is influenced by the embodied nature of the game. This brings the world of social interactions, context, and the physical environment of play into the dynamics of the game, whether consciously or unconsciously. A game may vary greatly, for instance, depending on whether it is played in a classroom, during dinner, or in a park.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Context, Individuality, Cultures", |
| "sec_num": "3.3" |
| }, |
| { |
| "text": "It would be simple enough to download a large textual corpus, create a list of word pairs from this data, then sample uniformly to create an automated player for Chain Reaction. The player so constructed could be given a higher or lower temperature parameter so that it utters more common or more unusual words to continue the chain. This player, however, would lack the nuance humans will have when playing, and would have no basis in the principles of orature or cognitive operational constraints.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The Obvious Approach Misses the Point", |
| "sec_num": "4" |
| }, |
| { |
| "text": "Human speakers have not perfectly memorized libraries of digital and digitalized text, and even if they had, speech and writing represent different registers. A corpus will not represent how the a language is spoken unless it consists entirely of accurate transcripts, given the many differences between speech and writing noted and theorized by Ong, Halliday (1989) and others (e.g. Lukin et al., 2011) . These have often been automatically distinguished using computational linguistics systems (e.g. Murata and Isahara, 2002; Ortmann and Dipper, 2019) .", |
| "cite_spans": [ |
| { |
| "start": 351, |
| "end": 366, |
| "text": "Halliday (1989)", |
| "ref_id": "BIBREF4" |
| }, |
| { |
| "start": 384, |
| "end": 403, |
| "text": "Lukin et al., 2011)", |
| "ref_id": "BIBREF7" |
| }, |
| { |
| "start": 502, |
| "end": 527, |
| "text": "Murata and Isahara, 2002;", |
| "ref_id": "BIBREF9" |
| }, |
| { |
| "start": 528, |
| "end": 553, |
| "text": "Ortmann and Dipper, 2019)", |
| "ref_id": "BIBREF12" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The Obvious Approach Misses the Point", |
| "sec_num": "4" |
| }, |
| { |
| "text": "Additionally, humans do not uniformly sample from a fixed database of possible word pairs or compound words in their minds. Chain continuations are uttered based on common speech, idioms and expressions, individual backgrounds and history, cultural backgrounds, current events, the physical setting of the game, previous conversations players have had with each other, and so on. In addition to factors that are not conscious, explicit decisionmaking plays a role. Players may sometimes choose uncommon continuations as \"curve-balls\" to make the game more interesting. And finally, the store of pairs within the mind may be large, but certainly is constrained.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The Obvious Approach Misses the Point", |
| "sec_num": "4" |
| }, |
| { |
| "text": "Chain Reaction Time (CRT) is a simple prototype system, developed in Python, for playing Chain Reaction. In developing CRT we were not interested in providing a surrogate social environment and having a single human player use the computer as a partner in this game. The point of developing CRT, and any future automated players of games like Chain Reaction, is to be able to add these computer players to groups of several humans and make the dynamics of a multi-player game even more unusual and fun.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "CRT, a Bot to Play Chain Reaction with Humans", |
| "sec_num": "5" |
| }, |
| { |
| "text": "Readying CRT for play begins by formulating one or more lists of valid word pairs that can be used by computer players in-game. This is achieved via a learning mechanism that gleans word pairs from arbitrary textual corpora. As a data-transforming mechanism, the learning module has no innate preference toward textual or oral modes of thought. Although oral culture is primarily maintained in our system as discussed in the analysis of Ong's characteristics of orality, we have chosen to also embody this aspect by only using a casual spokenword corpus, the Santa Barbara Corpus of Spoken American English (Du Bois et al., 2000) . At the cost of omitting new and emerging word pairs, we use this method to ensure that pairs have a real-life oral basis and reflect the orally-based cognition used in casual daily conversations. CRT has a model of pairs that allows for the system to treat words and symbols as oral tokens rather than textual ones, although of course, like any computer system, it is based on inscribed, recorded that relates more to literature culture. The existing interface can also, unfortunately, highlight the lexical, written aspects of the system.", |
| "cite_spans": [ |
| { |
| "start": 611, |
| "end": 629, |
| "text": "Bois et al., 2000)", |
| "ref_id": "BIBREF2" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "System Architecture", |
| "sec_num": "5.1" |
| }, |
| { |
| "text": "Nevertheless, the pairs model not only functionally achieves orally-based responses, but acts as a core component of the oral nature of the broader system. To achieve this, the pairs model first transforms the ending word of the chain into a list of all its possible phonetic representations given by the CMU Pronouncing Dictionary (Weide, 1998). Pairs then uses a dictionary of homophones to find all lexemes of all pronunciations of the word in question. Lastly, a continuation is chosen from the list of all possible continuations found for all lexemes. This results in an oral system that is functionally ambivalent to the rigid textual representation of a word and can naturally switch between word meanings, lexemes, and homophones.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "System Architecture", |
| "sec_num": "5.1" |
| }, |
| { |
| "text": "Traditional human-computer I/O follows almost exclusively our textual and literate traditions. (CRT was developed before the smart speakers became as pervasive as they are today, but even these systems only do a little to diminish the influence of literate culture in HCI overall.) One fundamental characteristic of orality is the ephemerality of the word and its inescapable attachment to the present moment. A spoken word can be heard for only the moment it sounds in the air. Textual systems embody the polar opposite of this characteristic: written or printed text (on paper or on a computer terminal) endures and can be read again.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Player Interface", |
| "sec_num": "5.2" |
| }, |
| { |
| "text": "Although we have not released CRT yet, we wanted to build it on accessible free/libre/open source technologies, and sought to implement a minimum viable system. These inclinations led us, in our early work, to use the textual display of words on the screen and a keyboard interface for input. Although not ideal, we specialized this interface to present some of the most important features of oral exchange.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Player Interface", |
| "sec_num": "5.2" |
| }, |
| { |
| "text": "The Python curses library was used to achieve the key oral element of ephemerality as well as present-moment-attachment of the word, allowing us to emulate those aspects of speech. This interface is currently named OralTyping. Although text is still used, the timing and visual presentation of the text aims to recapture key elements of orality lost in standard textual displays. For interface systems without an auditory component, OralTyping performs well in its intended purpose of representing the ephemerality of oral poetic production in text. Of course, a two-way verbal interface would make for a better play experience, and we do plan to implement one.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Player Interface", |
| "sec_num": "5.2" |
| }, |
| { |
| "text": "As a step toward this, we developed a second interface using the pyttsx3 text to speech module. This allows CRT to speak the chain aloud when it is the system's turn to play. Without having conducted any formal evaluation, it is clear that this interface improves play in addition to being more faithful to the underlying oral principles of Chain Reaction. It is important to note that this system currently only provides speech output; human players must still inform computerized players of the latest chain additions before the computer can consider and output a continuation to the chain. In future versions of the system, we plan to use modern machine learning models for both speech synthesis (such as Tacotron 2), as well as speech recognition (such as wav2vec) in order to remove current text input limitations.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Player Interface", |
| "sec_num": "5.2" |
| }, |
| { |
| "text": "In Orality and Literacy, Walter Ong outlines nine unique characteristics that distinguish primary oral thought processes from textually based cognition (Ong, 2002) . While Ong is certainly not the last word on orality -he is, rather, the first major scholar and theorist of it -we believe his principles remain critical to understanding oral improvisation and oral culture. Our selection of Chain Reaction and development of CRT to play it embodies these characteristics as follows.", |
| "cite_spans": [ |
| { |
| "start": 152, |
| "end": 163, |
| "text": "(Ong, 2002)", |
| "ref_id": "BIBREF11" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Principles of Oral Poetics as They Apply to Chain Reaction and CRT", |
| "sec_num": "6" |
| }, |
| { |
| "text": "Oral works tend to use linear, additive grammatical structures, whereas literate texts employ more complex structures to give contextual information that would otherwise be found in real-world oration. The Chain Reaction game is effectively the most direct functional implementation of additive structure, since its core mechanic is the retention of active chain followed by the addition of one word forming a new ending pair. CRT does not modify this mechanic, and thus exhibits this oral characteristic.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Additive rather than subordinate", |
| "sec_num": "6.1" |
| }, |
| { |
| "text": "In orature, descriptive archetype pairs occur often as small-scale formulas. As Ong states, \"The elements of orally based thought and expression tend to be not so much simple integers as clusters of integers, such as parallel terms or phrases or clauses, antithetical terms or phrases or clauses, epithets.\" Given that Chain Reaction is entirely founded on culturally memorable consecutive word pairs (\"things\"), the game prioritizes these pairs as the underlying root structure of thinking-in-play. Although these pairs have greater constraints than called for by the general aggregative trait of orality, their existence as the cognitive root of Chain Reaction (and CRT by extension) indicate a strong oral foundation.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Aggregative rather than analytic", |
| "sec_num": "6.2" |
| }, |
| { |
| "text": "Oral works repeat many story elements previously stated including events, characters, and associated descriptions. The requirement that players orally repeat the entire chain is a direct example of this redundancy. Indeed, many of the underlying motives for this redundancy in oral cultures can be experienced firsthand in Chain Reaction gameplay.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Redundant or \"copious\"", |
| "sec_num": "6.3" |
| }, |
| { |
| "text": "In particular, Ong states, \"Not everyone in a large audience understands every word a speaker utters, if only because of acoustical problems. It is advantageous for the speaker to say the same thing, or equivalently the same thing, two or three times.\"", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Redundant or \"copious\"", |
| "sec_num": "6.3" |
| }, |
| { |
| "text": "A primary objective of oral works is to preserve the knowledge gained by the culture, given that in primary oral cultures the spoken word is the sole method of record-keeping. In Chain Reaction gameplay, the creation of truly novel word pairs is effectively prohibited, since one criterion for a pair to be valid is that it must be recognized as preexisting. This ensures that although chains created may be new, their constituent elements are conservative by definition and rule. From a different perspective, given that words in the chain need not have any logical relation other than being consecutive word pairs, this can result in chains that grow more bizarre and more amusing with each expansion, which can be seen as a source of novelty. Even so, this novelty is grounded in a locally conservative requirement.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Conservative or traditionalist", |
| "sec_num": "6.4" |
| }, |
| { |
| "text": "As Ong states, \"In the absence of elaborate analytic categories that depend on writing to structure knowledge at a distance from lived experience, oral cultures must conceptualize and verbalize all their knowledge with more or less close reference to the human lifeworld, assimilating the alien, objective world to the more immediate, familiar interaction of human beings.\" Pairs that come to mind in Chain Reaction are likely to stem from everday human experience. While there is no design element of CRT that works toward this principle, the use of an orally-based corpus is meant to help address it.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Close to the human lifeworld", |
| "sec_num": "6.5" |
| }, |
| { |
| "text": "Continuations are likely to draw on the immediate context of the player, with more abstract elements being less likely. Players' physical location, the time of day, the weather, the social setting, dynamics between players, and other factors will lead gameplay to be simultaneously grounded in the active, present lives of the players (6.5) and situationally based by context (6.6). We have observed players looking at and even pointing to others in the circle as they go around recalling words in the chain, and players have told us that they sometimes try to remember who said each word to aid their recollection. Although not presently implemented, we recognize this element of orality and intend to eventually model some aspects of it in CRT.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Situational rather than abstract", |
| "sec_num": "6.6" |
| }, |
| { |
| "text": "A distinctive signature of oral works, Ong states, is the embedding of the work and spoken word in an agonistic context of struggle. In Chain Reaction, the act of playing can be seen as a struggle between players to properly recite the growing chain while adding on to it. As discussed earlier, the strong formulation of agonistic struggle is tempered by the essential nature of this game as cooperative. Thus, there is an additional dynamic at play: An implicit goal of continuing the game as long as possible, rather than purposefully trying to gain an upper hand on other players. So while this game has an element of being agonistically toned (reflected in CRT as well), it is in the service of cooperation and group enjoyment.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Agonistically toned", |
| "sec_num": "6.7" |
| }, |
| { |
| "text": "This characteristic ties in strongly with (6.5) Close to the human lifeworld, in that it indicates an innate connection between the work and the speaker or speakers. In the case of Chain Reaction, the participatory nature of the game itself, along with its human context, encourages empathetic response through the creation of additional pairs.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Empathetic and participatory rather than objectively distanced", |
| "sec_num": "6.8" |
| }, |
| { |
| "text": "Ong characterizes the homeostatic nature of oral cultures by stating that they \"live very much in a present which keeps itself in equilibrium or homeostasis by sloughing off memories which no longer have present relevance.\" The nature of orality precludes the use of dictionaries or other written references to learn about the past: All there is is the present. Chain Reaction is not played with reference books, of course. And our pairs knowledge bases derive not from such texts but from modern transcripts of oral conversations. It would work against this principle to use textual data derived from historical written books or dictionaries, which could result in anachronistic responses from computer players as well as an unrealistic store of information that would not accurately model the thought processes of human players in Chain Reaction.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Homeostatic", |
| "sec_num": "6.9" |
| }, |
| { |
| "text": "An example transcript of a play session with two human players and CRT is given in Appendix A.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Reflections on Play with CRT", |
| "sec_num": "7" |
| }, |
| { |
| "text": "We have not undertaken any formal evaluation of CRT, but can share some reflections based on our informal participation in and observation of the game when CRT participates. The CRT system does work and can participate in play with human players. Including one or more computerized players in a group of human players can inflect gameplay in an interesting way, as the computer player is able to participate, but in a way that is sometimes noticeably nonhuman. Its pace of recitation is unusually regular. It is capable of forming good continuations, but currently, not ones that are influenced by pairs earlier in the chain or by the human context. Generally, the current system's ability to form continuations is not currently near a human level, so games often end with the automated players unable to find continuations that human players have in mind. The limited but noticeable success of CRT proves that even a simple system can embody almost all major the aspects of oral poetics as theorized by Ong. It highlights the aspects that are most challenging to model, including those that rely on physical, temporal, and social contexts of play. We hope that this design directions for other work in oral improvisation and poetics, specifically, pointing out (1) what aspects of orature are easy to model and should be implemented in related systems, and (2) helping to show what aspects are more challenging and should be a focus for future research.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Reflections on Play with CRT", |
| "sec_num": "7" |
| }, |
| { |
| "text": "We believe that learning from additional corpora and forming a larger pairs knowledge base would be the simplest way to address current limitations. A significant concern is how to integrate corpora of written texts in a way that is suitable for this orallygrounded project. We of course do not have access to the smart speaker recordings of human conversation collected by Amazon and Google -sparing us any ethical dilemmas regarding the use of this data. We may be able to use available written corpora that are contemporary and vernacular, and address the written origins of this textual data by filtering it so that offhand and more or less improvisational writings, rather than ones that involved the consultation of sources and revision, are emphasized.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Improving Chain Continuation", |
| "sec_num": "7.1" |
| }, |
| { |
| "text": "One important limitation of the current learning mechanism is that of detecting valid pairs written as solid compounds, e.g., \"signpost\" and \"newspaper.\" The learning submodule does not detect this. This highlights an underlying structural difference in how words are perceived orally versus textually. \"Written words are residue. Oral tradition has no such residue or deposit.\" (Ong, 2002) Ong argues that in authentic oral cultures, \"signpost\" and \"sign post\" are delivered in the same way; even attempting to formulate a distinction between them is impossible within an oral framework. This is not an impairment of orality; this equality of word representation reveals some of the ontological commitments textual cultures have made in order to enable literate representation of the spoken word. From an a priori perspective, it is not evident which unitary ideas should be written separately and which should be written as compound words. Further evidence of this can be seen in the common shifting of these boundaries, in such cases as \"to morrow,\" \"after noon,\" \"mail box,\" and \"class room.\" This is one example of an element of orality directly encountered in the development of CRT, manifested as a concrete problem that must be solved. A partial solution to this problem may be to include a list of valid compounds (based on an auto-generated list, but edited) in the pairs model.", |
| "cite_spans": [ |
| { |
| "start": 379, |
| "end": 390, |
| "text": "(Ong, 2002)", |
| "ref_id": "BIBREF11" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Solid Compounds", |
| "sec_num": "7.2" |
| }, |
| { |
| "text": "Computerized players currently do not have a model of fallible memory; that is, an automated player can only fail by being unable to find a con-tinuation. We plan to implement a temporal model whereby computerized players may forget parts of the chain as a function of increasing time as well as the difficulty or atypicality of pairs (judged by relative frequency of a given pair across learning data). This would model the human tendency to forget, and thus provide a more accurate experience that can be tweaked via parameters.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Making Memory Worse", |
| "sec_num": "7.3" |
| }, |
| { |
| "text": "Having the system forget a chain is not a high priority for us, even though memory is an important aspect of Chain Reaction. Generally, it is not fun for the game to end, and it may leave players disappointed if an automated player were to stumble and end the game. An exception to this may be when there is a very tenuous continuation at some point in the chain, or when the game is progressing poorly because many players have made obscure continuations. In this case, players can be relieved and appreciate the ability to start over, and to form a better chain. Still, it is not essential that an automated player be the one to end such a game.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Making Memory Worse", |
| "sec_num": "7.3" |
| }, |
| { |
| "text": "CRT could be developed to explain the reasoning behind a particular continuation after the end of a game, as a human would reasonably be expected to do if asked. Without developing general AI, it would be straightforward to implement a mechanism by which CRT could justify its chain continuations as influenced by categories such as \"food,\" \"furniture,\" and \"expression.\" When questioned about a chosen continuation, a computer player could respond adequately by stating the association it was reminded of and the word that caused this association to activate. Work on this aspect of the system might help us develop better methods of continuing the chain.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Explaining Continuations", |
| "sec_num": "7.4" |
| }, |
| { |
| "text": "Perhaps most interesting to us would be elaborating CRT to have custom pairs databases, with custom weights, to model of cultural and individual differences. In alignment with our earlier comments, the computational model of a Canadian or American player might give more weight to \"station eleven\" while a British player model gave more to \"station stop.\" Such cultural differences would be the first step of in modeling different sorts of players. Beyond this, we might be able to model how certain individuals watched a good deal of TV and others none at all, which would affect the weighting of two-word or compound TV show titles. With these sorts of nuances, computer players could introduce additional interesting twists into games of Chain Reaction.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Embodying Cultural and Individual Differences", |
| "sec_num": "7.5" |
| }, |
| { |
| "text": "We have in mind ways to relax the constraint on what a pair can be and have played sample games with different rules. Even with a very minimal framework for oral improvisation and poetics, we are likely to determine a different game that allows for more types of creativity, and enhance our system to play this game. From there, we could elaborate the system to utter rhymed couplets in response to previous utterances. This would allow it to participate in a minimal rap cypher. By building up from a seemingly trivial system and maintaining a connection to principles of orature throughout the process, we would be taking a very different approach from that of complex multimodal systems that devise lyrics as part of their improvisations.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Allowing More Elaborate Improvisation", |
| "sec_num": "7.6" |
| }, |
| { |
| "text": "Complex systems can be impressive, but they can also face difficulties. If they succeed, it is hard to know the basis for their success: Which of the many components of the system were crucial to the positive improvisational performance? If they fail, it is, similarly, hard to know why. By taking a bottom-up, step-by-step approach, we hope to be able to answer questions that more elaborate bots, although impressive and admirable in many ways, have not been able to address.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Allowing More Elaborate Improvisation", |
| "sec_num": "7.6" |
| } |
| ], |
| "back_matter": [ |
| { |
| "text": "The work described was done in Fall 2019 at the MIT Trope Tank in conversation with Angela Chang and Judy Heflin, both of whom contributed important insights into Ong's thinking and its connection to Chain Reaction. We had the opportunity to play this improvisational game at MIT and also with international groups of poets, rappers, computational writers, and other friends and family.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Acknowledgements", |
| "sec_num": null |
| }, |
| { |
| "text": "The following is a transcript of play undertaken by two human players and CRT on May 10, 2022.\u2022 ", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "A Chain Reaction Played with CRT", |
| "sec_num": null |
| } |
| ], |
| "bib_entries": { |
| "BIBREF0": { |
| "ref_id": "b0", |
| "title": "Jokes and the linguistic mind. Routledge", |
| "authors": [ |
| { |
| "first": "Debra", |
| "middle": [], |
| "last": "Aarons", |
| "suffix": "" |
| } |
| ], |
| "year": 2012, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Debra Aarons. 2012. Jokes and the linguistic mind. Routledge, New York.", |
| "links": null |
| }, |
| "BIBREF1": { |
| "ref_id": "b1", |
| "title": "Bertsobot: the first minstrel robot", |
| "authors": [ |
| { |
| "first": "Aitzol", |
| "middle": [], |
| "last": "Astigarraga", |
| "suffix": "" |
| }, |
| { |
| "first": "Manex", |
| "middle": [], |
| "last": "Agirrezabal", |
| "suffix": "" |
| }, |
| { |
| "first": "Elena", |
| "middle": [], |
| "last": "Lazkano", |
| "suffix": "" |
| }, |
| { |
| "first": "Ekaitz", |
| "middle": [], |
| "last": "Jauregi", |
| "suffix": "" |
| }, |
| { |
| "first": "Basilio", |
| "middle": [], |
| "last": "Sierra", |
| "suffix": "" |
| } |
| ], |
| "year": 2013, |
| "venue": "2013 6th International Conference on Human System Interactions (HSI)", |
| "volume": "", |
| "issue": "", |
| "pages": "129--136", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Aitzol Astigarraga, Manex Agirrezabal, Elena Lazkano, Ekaitz Jauregi, and Basilio Sierra. 2013. Bertsobot: the first minstrel robot. In 2013 6th International Conference on Human System Interactions (HSI), pages 129-136. IEEE.", |
| "links": null |
| }, |
| "BIBREF2": { |
| "ref_id": "b2", |
| "title": "Santa barbara corpus of spoken american english. CD-ROM. Philadelphia: Linguistic Data Consortium", |
| "authors": [ |
| { |
| "first": "John W Du", |
| "middle": [], |
| "last": "Bois", |
| "suffix": "" |
| }, |
| { |
| "first": "Wallace", |
| "middle": [ |
| "L" |
| ], |
| "last": "Chafe", |
| "suffix": "" |
| }, |
| { |
| "first": "Charles", |
| "middle": [], |
| "last": "Meyer", |
| "suffix": "" |
| }, |
| { |
| "first": "A", |
| "middle": [], |
| "last": "Sandra", |
| "suffix": "" |
| }, |
| { |
| "first": "Nii", |
| "middle": [], |
| "last": "Thompson", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Martey", |
| "suffix": "" |
| } |
| ], |
| "year": 2000, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "John W Du Bois, Wallace L Chafe, Charles Meyer, Sandra A Thompson, and Nii Martey. 2000. Santa barbara corpus of spoken american english. CD- ROM. Philadelphia: Linguistic Data Consortium.", |
| "links": null |
| }, |
| "BIBREF3": { |
| "ref_id": "b3", |
| "title": "Gladstone's Games to Go: Verbal Volleys, Coin Contests, Dot Duels, and Other Games for Boredom-Free Days", |
| "authors": [ |
| { |
| "first": "Jim", |
| "middle": [], |
| "last": "Gladstone", |
| "suffix": "" |
| } |
| ], |
| "year": 2004, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Jim Gladstone. 2004. Gladstone's Games to Go: Verbal Volleys, Coin Contests, Dot Duels, and Other Games for Boredom-Free Days. Quirk Books, Philadelphia.", |
| "links": null |
| }, |
| "BIBREF4": { |
| "ref_id": "b4", |
| "title": "Language, context, and text: Aspects of language in a social-semiotic perspective", |
| "authors": [ |
| { |
| "first": "Michael Alexander Kirkwood", |
| "middle": [], |
| "last": "Halliday", |
| "suffix": "" |
| }, |
| { |
| "first": "Ruqaiya", |
| "middle": [], |
| "last": "Hasan", |
| "suffix": "" |
| } |
| ], |
| "year": 1989, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Michael Alexander Kirkwood Halliday, Ruqaiya Hasan, et al. 1989. Language, context, and text: Aspects of language in a social-semiotic perspective.", |
| "links": null |
| }, |
| "BIBREF5": { |
| "ref_id": "b5", |
| "title": "The association method. The American journal of psychology", |
| "authors": [ |
| { |
| "first": "G", |
| "middle": [], |
| "last": "Carl", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Jung", |
| "suffix": "" |
| } |
| ], |
| "year": 1910, |
| "venue": "", |
| "volume": "21", |
| "issue": "", |
| "pages": "219--269", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Carl G Jung. 1910. The association method. The Amer- ican journal of psychology, 21(2):219-269.", |
| "links": null |
| }, |
| "BIBREF6": { |
| "ref_id": "b6", |
| "title": "Neural correlates of lyrical improvisation: an fmri study of freestyle rap", |
| "authors": [ |
| { |
| "first": "Siyuan", |
| "middle": [], |
| "last": "Liu", |
| "suffix": "" |
| }, |
| { |
| "first": "Ming", |
| "middle": [], |
| "last": "Ho", |
| "suffix": "" |
| }, |
| { |
| "first": "Yisheng", |
| "middle": [], |
| "last": "Chow", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Xu", |
| "suffix": "" |
| }, |
| { |
| "first": "Katherine", |
| "middle": [ |
| "E" |
| ], |
| "last": "Michael G Erkkinen", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Swett", |
| "suffix": "" |
| }, |
| { |
| "first": "W", |
| "middle": [], |
| "last": "Michael", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Eagle", |
| "suffix": "" |
| }, |
| { |
| "first": "A", |
| "middle": [], |
| "last": "Daniel", |
| "suffix": "" |
| }, |
| { |
| "first": "Allen R", |
| "middle": [], |
| "last": "Rizik-Baer", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Braun", |
| "suffix": "" |
| } |
| ], |
| "year": 2012, |
| "venue": "Scientific reports", |
| "volume": "2", |
| "issue": "1", |
| "pages": "1--8", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Siyuan Liu, Ho Ming Chow, Yisheng Xu, Michael G Erkkinen, Katherine E Swett, Michael W Eagle, Daniel A Rizik-Baer, and Allen R Braun. 2012. Neu- ral correlates of lyrical improvisation: an fmri study of freestyle rap. Scientific reports, 2(1):1-8.", |
| "links": null |
| }, |
| "BIBREF7": { |
| "ref_id": "b7", |
| "title": "Halliday's model of register revisited and explored", |
| "authors": [ |
| { |
| "first": "Annabelle", |
| "middle": [], |
| "last": "Lukin", |
| "suffix": "" |
| }, |
| { |
| "first": "Alison", |
| "middle": [ |
| "R" |
| ], |
| "last": "Moore", |
| "suffix": "" |
| }, |
| { |
| "first": "Maria", |
| "middle": [], |
| "last": "Herke", |
| "suffix": "" |
| }, |
| { |
| "first": "Rebekah", |
| "middle": [], |
| "last": "Wegener", |
| "suffix": "" |
| }, |
| { |
| "first": "Canzhong", |
| "middle": [], |
| "last": "Wu", |
| "suffix": "" |
| } |
| ], |
| "year": 2011, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Annabelle Lukin, Alison R Moore, Maria Herke, Re- bekah Wegener, and Canzhong Wu. 2011. Halliday's model of register revisited and explored.", |
| "links": null |
| }, |
| "BIBREF8": { |
| "ref_id": "b8", |
| "title": "CALLIG: Computer assisted language learning using improvisation games", |
| "authors": [ |
| { |
| "first": "Lu\u00eds", |
| "middle": [], |
| "last": "Morgado Da Costa", |
| "suffix": "" |
| }, |
| { |
| "first": "Joanna Ut-Seong", |
| "middle": [], |
| "last": "Sio", |
| "suffix": "" |
| } |
| ], |
| "year": 2020, |
| "venue": "Workshop on Games and Natural Language Processing", |
| "volume": "", |
| "issue": "", |
| "pages": "49--58", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Lu\u00eds Morgado da Costa and Joanna Ut-Seong Sio. 2020. CALLIG: Computer assisted language learning using improvisation games. In Workshop on Games and Natural Language Processing, pages 49-58, Mar- seille, France. European Language Resources Asso- ciation.", |
| "links": null |
| }, |
| "BIBREF9": { |
| "ref_id": "b9", |
| "title": "Automatic extraction of differences between spoken and written languages, and automatic translation from the written to the spoken language", |
| "authors": [ |
| { |
| "first": "Masaki", |
| "middle": [], |
| "last": "Murata", |
| "suffix": "" |
| }, |
| { |
| "first": "Hitoshi", |
| "middle": [], |
| "last": "Isahara", |
| "suffix": "" |
| } |
| ], |
| "year": 2002, |
| "venue": "Proceedings of the Third International Conference on Language Resources and Evaluation (LREC'02)", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Masaki Murata and Hitoshi Isahara. 2002. Automatic extraction of differences between spoken and writ- ten languages, and automatic translation from the written to the spoken language. In Proceedings of the Third International Conference on Language Re- sources and Evaluation (LREC'02), Las Palmas, Ca- nary Islands -Spain. European Language Resources Association (ELRA).", |
| "links": null |
| }, |
| "BIBREF10": { |
| "ref_id": "b10", |
| "title": "Quizmaster: The Life and Times and Fun and Games of Bill Cullen. BearManor Media", |
| "authors": [ |
| { |
| "first": "Adam", |
| "middle": [], |
| "last": "Nedeff", |
| "suffix": "" |
| } |
| ], |
| "year": 2013, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Adam Nedeff. 2013. Quizmaster: The Life and Times and Fun and Games of Bill Cullen. BearManor Me- dia, Albany, GA.", |
| "links": null |
| }, |
| "BIBREF11": { |
| "ref_id": "b11", |
| "title": "Orality and literacy. Routledge", |
| "authors": [ |
| { |
| "first": "J", |
| "middle": [], |
| "last": "Walter", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Ong", |
| "suffix": "" |
| } |
| ], |
| "year": 2002, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Walter J Ong. 2002. Orality and literacy. Routledge, New York.", |
| "links": null |
| }, |
| "BIBREF12": { |
| "ref_id": "b12", |
| "title": "Variation between different discourse types: Literate vs. oral", |
| "authors": [ |
| { |
| "first": "Katrin", |
| "middle": [], |
| "last": "Ortmann", |
| "suffix": "" |
| }, |
| { |
| "first": "Stefanie", |
| "middle": [], |
| "last": "Dipper", |
| "suffix": "" |
| } |
| ], |
| "year": 2019, |
| "venue": "Proceedings of the Sixth Workshop on NLP for Similar Languages, Varieties and Dialects", |
| "volume": "", |
| "issue": "", |
| "pages": "64--79", |
| "other_ids": { |
| "DOI": [ |
| "10.18653/v1/W19-1407" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "Katrin Ortmann and Stefanie Dipper. 2019. Variation between different discourse types: Literate vs. oral. In Proceedings of the Sixth Workshop on NLP for Sim- ilar Languages, Varieties and Dialects, pages 64-79, Ann Arbor, Michigan. Association for Computational Linguistics.", |
| "links": null |
| }, |
| "BIBREF13": { |
| "ref_id": "b13", |
| "title": "The bad, the good, the misunderstood: The social effects of racial humor", |
| "authors": [ |
| { |
| "first": "A", |
| "middle": [], |
| "last": "Donald", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Saucier", |
| "suffix": "" |
| }, |
| { |
| "first": "J", |
| "middle": [], |
| "last": "Conor", |
| "suffix": "" |
| }, |
| { |
| "first": "Megan", |
| "middle": [ |
| "L" |
| ], |
| "last": "O'dea", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Strain", |
| "suffix": "" |
| } |
| ], |
| "year": 2016, |
| "venue": "Translational Issues in Psychological Science", |
| "volume": "2", |
| "issue": "1", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Donald A Saucier, Conor J O'Dea, and Megan L Strain. 2016. The bad, the good, the misunderstood: The social effects of racial humor. Translational Issues in Psychological Science, 2(1):75.", |
| "links": null |
| }, |
| "BIBREF14": { |
| "ref_id": "b14", |
| "title": "Shimon the rapper: A real-time system for human-robot interactive rap battles", |
| "authors": [ |
| { |
| "first": "Richard", |
| "middle": [], |
| "last": "Savery", |
| "suffix": "" |
| }, |
| { |
| "first": "Lisa", |
| "middle": [], |
| "last": "Zahray", |
| "suffix": "" |
| }, |
| { |
| "first": "Gil", |
| "middle": [], |
| "last": "Weinberg", |
| "suffix": "" |
| } |
| ], |
| "year": 2020, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": { |
| "arXiv": [ |
| "arXiv:2009.09234" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "Richard Savery, Lisa Zahray, and Gil Weinberg. 2020. Shimon the rapper: A real-time system for human-robot interactive rap battles. arXiv preprint arXiv:2009.09234.", |
| "links": null |
| }, |
| "BIBREF15": { |
| "ref_id": "b15", |
| "title": "The cmu pronouncing dictionary", |
| "authors": [ |
| { |
| "first": "", |
| "middle": [], |
| "last": "Robert L Weide", |
| "suffix": "" |
| } |
| ], |
| "year": 1998, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Robert L Weide. 1998. The cmu pronouncing dictionary. URL: http://www. speech. cs. cmu. edu/cgibin/cmudict.", |
| "links": null |
| }, |
| "BIBREF16": { |
| "ref_id": "b16", |
| "title": "The genesis enterprise: Taking artificial intelligence to another level via a computational account of human story understanding", |
| "authors": [ |
| { |
| "first": "Patrick", |
| "middle": [], |
| "last": "Henry Winston", |
| "suffix": "" |
| }, |
| { |
| "first": "Dylan", |
| "middle": [], |
| "last": "Holmes", |
| "suffix": "" |
| } |
| ], |
| "year": 2018, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Patrick Henry Winston and Dylan Holmes. 2018. The genesis enterprise: Taking artificial intelligence to another level via a computational account of human story understanding. Technical report, CSAIL, MIT.", |
| "links": null |
| } |
| }, |
| "ref_entries": { |
| "FIGREF0": { |
| "num": null, |
| "text": "Post office chair. \u2022 Player 3: Post office chair man [chairman]. \u2022 Player 4: Post office chair man child [manchild]. \u2022 Player 1: Post office chair man child labor. \u2022 Player 2: Post office chair man child labor law. \u2022 Player 3: Post office chair man child labor law school. \u2022 Player 4: Post office chair man child labor law school boy [schoolboy]. \u2022 Player 1: Post office chair man child labor law school boy band. \u2022 Player 2: Post office chair man child labor law school boy band [\"banned\"] book. \u2022 Player 3: Post office chair man child labor law school boy banned book shelf [bookshelf]. \u2022 Player 4: Post office chair man child labor law school boy banned book shelf ... uh, I", |
| "type_str": "figure", |
| "uris": null |
| } |
| } |
| } |
| } |
