acl acl2010 acl2010-224 knowledge-graph by maker-knowledge-mining
Source: pdf
Author: Tina Kluwer ; Peter Adolphs ; Feiyu Xu ; Hans Uszkoreit ; Xiwen Cheng
Abstract: This paper describes the KomParse system, a natural-language dialog system in the three-dimensional virtual world Twinity. In order to fulfill the various communication demands between nonplayer characters (NPCs) and users in such an online virtual world, the system realizes a flexible and hybrid approach combining knowledge-intensive domainspecific question answering, task-specific and domain-specific dialog with robust chatbot-like chitchat.
Reference: text
sentIndex sentText sentNum sentScore
1 de Abstract This paper describes the KomParse system, a natural-language dialog system in the three-dimensional virtual world Twinity. [sent-5, score-0.626]
2 1 Introduction In recent years multi-user online games in virtual worlds such as Second Life or World of Warcraft have attracted large user communities. [sent-7, score-0.405]
3 Such virtual online game worlds provide new social and economic platforms for people to work and interact in. [sent-8, score-0.453]
4 Furthermore, virtual worlds open new perspectives for research in the social, behavioral, and economic sciences, as well as in human-centered computer science (Bainbridge, 2007). [sent-9, score-0.318]
5 Depending on the game type, non-player characters (NPCs) are often essential for supporting the game plot, for making the artificial world more vivid and ultimately for making it more immersive. [sent-10, score-0.249]
6 In addition, NPCs are useful to populate new worlds by carrying out jobs the user-led characters come in touch with. [sent-11, score-0.117]
7 This shortcoming creates a strong need for progress in areas such as AI and NLP, especially their planning and dialog systems. [sent-13, score-0.357]
8 The KomParse system, described in this paper, provides NPCs for a virtual online world named Twinity, a product of the Berlin startup company Metaversum1 . [sent-14, score-0.296]
9 The KomParse NPCs offer various services through conversation with game users using question-answering and dialog functionality. [sent-15, score-0.666]
10 The utilization of Semantic Web technology with RDF-encoded generic and domain-specific ontologies furthermore enables semantic search and inference. [sent-16, score-0.077]
11 This paper is organized as follows: Section 2 presents the NPC modelling and explains the application scenarios. [sent-17, score-0.044]
12 Section 3 details the knowledge representation and semantic inference in our system. [sent-18, score-0.104]
13 Section 4 explains the system architecture and its key components. [sent-19, score-0.047]
14 2 Application Scenario and NPC Modelling The online game Twinity extends the Second Life idea by mirroring an urban part of the real world. [sent-22, score-0.101]
15 At the time of this writing, the simulated section of reality already contains 3D models of the cities of Berlin, Singapore and London and it keeps grow- ing. [sent-23, score-0.026]
16 Users can log into the virtual world, where they can meet other users and communicate with them using the integrated chat function or talk to each other via Voice-over-IP. [sent-24, score-0.31]
17 They can style their virtual appearance, can rent or buy their own flats and decorate them as to their preferences and tastes. [sent-25, score-0.237]
18 Out of many types of NPCs useful for this application such as pedestrians, city guides and personnel in stores, restaurants and bars, we start with two specific characters: a female furniture sales agent and a male bartender. [sent-26, score-0.636]
19 The furniture seller is designed for helping users furnish their virtual apartments. [sent-27, score-0.559]
20 Users can buy pieces of furniture and room decoration from the NPC by describing their demands and wishes in a text chat. [sent-28, score-0.389]
21 In the second scenario, the bartender sells virtual drinks. [sent-34, score-0.372]
22 He can talk about cocktails with users, but moreover, he can also entertain his guests by providing trivia-type information about popular celebrities and various relations among them. [sent-35, score-0.186]
23 We chose these two characters not only because of their value for the Twinity application but also for our research goals. [sent-36, score-0.044]
24 First of all, the furniture sales agent is controlled by a complex task model including ontology-driven and data-driven components to guide the conversation. [sent-38, score-0.636]
25 This agent also possesses a much more fine-grained action model, which allows several different actions to cover the potential conversation situations for the selling task. [sent-39, score-0.372]
26 The bartender agent on the other hand is designed not to fulfill one strict task because his clients do not follow a specific goal except ordering drinks. [sent-40, score-0.408]
27 Our bartender has the role of a conversation companion and is able to entertain clients with his broad knowledge. [sent-41, score-0.405]
28 Thus, he is allowed to access to several knowledge bases and is able to handle questions (and later conversations) about a much larger domain called the “gossip domain” which enables conversation about pop stars, movie actors and other celebrities as well as the relations between these people. [sent-42, score-0.324]
29 In order to achieve a high robustness, we integrate a chatbot into the bar- tender agent to catch chitchat utterances we cannot handle. [sent-43, score-0.223]
30 Figure 2: Our bartender NPC in his bar in Twinity 3 Knowledge Representation and Semantic Inference Semantic Web technology is employed for modelling the knowledge of the NPCs. [sent-44, score-0.24]
31 An RDF statement is a binary relation instance between two individuals, that is a triple of a predicate and two arguments, called the subject and the object, and written as subj pred obj (e. [sent-46, score-0.045]
32 All objects and properties the NPC can talk about are modelled in this way. [sent-49, score-0.128]
33 Therefore the knowledge base has to reflect the physical properties of the virtual objects in Twinity as faithfully as possible. [sent-50, score-0.362]
34 For instance, specific pieces of furniture are described by their main color, material or style, whereas cocktails are characterized by their ingredients, color, consistence and taste. [sent-51, score-0.349]
35 Furthermore, references to the 3D models of the objects are stored in order to create, find and remove such objects in the virtual world. [sent-52, score-0.356]
36 The concepts and individuals of the particular domain are structured and organized in domainspecific ontologies. [sent-53, score-0.104]
37 These ontologies are modelled in the Web Ontology Language (OWL). [sent-54, score-0.072]
38 Our ontologies are defined by the freely available ontology editor Prot e´g e´ 4. [sent-56, score-0.147]
39 The advantage ofusing an ontology for structuring the domain knowledge is 2http : //protege . [sent-58, score-0.161]
40 edu/, 27 Oct 2009 as accessed 37 TSweirnvietyrTCwTliCwineTlCi nwientl i ynitnetynit yKoSmerPvaer seCoCnovCnAeovgrnAe svgnaAre st grnaioest n aioat inloanlal Figure 3: Overall System Architecture – Server/Client Architecture for NPC Control the modular non-redundant encoding. [sent-60, score-0.027]
41 This knowledge base con- • • tains 95,258 triples, 123 furniture classes, 20 color classes, 243 color instances and various classes defining styles and similar concepts. [sent-63, score-0.506]
42 A cocktail ontology, defining 13 cocktail cAlas csoecsk wtaiitlh ingredients aenfidn tnagste 1s3 3in c 21,880 triples. [sent-64, score-0.206]
43 A biographical ontology, the “gossip ontology”, defining biographical “agnods careerspecific concepts for people. [sent-65, score-0.152]
44 This ontology is accompanied by a huge database of celebrities, which has been automatically acquired from the Web and covers nearly 600,000 persons and relations between these people like family relationships, marriages and professional relations. [sent-66, score-0.1]
45 , 2010) The furniture ontology is the only knowledge base for the furniture sales agent, whereas the bartender NPC has access to both the cocktail as well as the gossip knowledge base. [sent-68, score-1.263]
46 We use SwiftOwlim3 for storing and querying the data. [sent-69, score-0.05]
47 SwiftOwlim is a “triple store”, a kind of database which is specifically built for storing and querying RDF data. [sent-70, score-0.05]
48 It provides a forwardchaining inference engine which evaluates the domain definitions when loading the knowledge repository, and makes implicit knowledge explicit by asserting triples that must also hold true according to the ontology. [sent-71, score-0.155]
49 Once the reasoner is finished, the triple store can be queried directly using the RDF query language SPARQL. [sent-72, score-0.108]
50 com/ owl im/ 4 Overall System Architecture Figure 3 shows the overall system architecture. [sent-75, score-0.052]
51 Twinity is a server/client application, in which the server hosts the virtual world and coordinates the user interactions. [sent-76, score-0.541]
52 In order to use Twinity, users have to download the Twinity client. [sent-77, score-0.067]
53 The client allows the user to control the physical representation of the user’s character in the virtual world, also called the “avatar”. [sent-78, score-0.476]
54 Thus the client is responsible for displaying the graphics, calculating the effects of physical interactions, handling the user’s input and synchronizing the 3D data and user actions with the Twinity server. [sent-79, score-0.25]
55 Each NPC comprises two major parts: whereas its avatar is the physical appearance of the NPC in the virtual world, the “conversational agent” provides the actual control logic which controls the avatar autonomously. [sent-80, score-0.485]
56 It is in particular able to hold a conversation with Twinity users in that it reacts to a user’s presence, interprets user’s utterances in dialog context and generates adequate responses. [sent-81, score-0.598]
57 The KomParse server is a multi-client, multithreaded server written in Java that hosts the conversational agents for the NPCs (section 5). [sent-82, score-0.397]
58 We utilize the Python interface provided by the Twinity client to call our own plugin which opens a bidirectional socket connection to the KomParse server. [sent-84, score-0.128]
59 The plugin is started together with the Twinity client and serves as a mediator between the Twinity server and the KomParse server from then on (fig. [sent-85, score-0.42]
60 It sends all in-game events relevant to our system to the server and translates the commands sent by the server into Twinity-specific actions. [sent-87, score-0.292]
61 The integration architecture allows us to be maximally independent of the specific game platform. [sent-88, score-0.121]
62 Thus the KomParse system 38 can be naturally extended to other platforms since only the avatar interfaces have to be adapted. [sent-90, score-0.162]
63 The integration architecture also has the advantage that the necessary services can be easily distributed in a networked multi-platform environment. [sent-91, score-0.07]
64 The Twinity clients require a Microsoft Windows machine with a 3D graphics card supporting DirectX 9. [sent-92, score-0.094]
65 The triple store is run as a separate server process and is accessed by an XML-RPC interface. [sent-95, score-0.247]
66 2 GB RAM are required for loading our current knowledge base. [sent-97, score-0.07]
67 5 Conversational Agent: KomParse Dialog System Figure 4: Dialog System: Conversational Agent The KomParse dialog system, the main functionality of the conversational agent, consists of the following three major components: input analyzer, dialog manager and output generator (fig. [sent-98, score-0.934]
68 The input analyzer is responsible for the lin- guistic analysis of the user’s textual input including preprocessing such as string cleaning, part-ofspeech tagging, named entity recognition, parsing and semantic interpretation. [sent-100, score-0.093]
69 It yields a semantic representation which is the input for the dialog manager. [sent-101, score-0.425]
70 The dialog manager takes the result of the input analyzer and delivers an interpretation based on the dialog context and the available knowledge. [sent-102, score-0.878]
71 It also controls the task conversation chain and handles user requests. [sent-103, score-0.238]
72 The dialog manager determines the next system action based on the interpreted parameters. [sent-104, score-0.487]
73 The output generator realizes the action defined by the dialog manager with its multimodal generation competence. [sent-105, score-0.552]
74 As mentioned above, our dialog system has to deal with two different scenarios. [sent-107, score-0.357]
75 While the focal point of the bartender agent lies in the question answering functionality, the furniture sales agent is driven by a complex dialog task model based on a dialog graph. [sent-108, score-1.664]
76 Thus, the bartender agent relies mainly on question answering technology, in that it needs to understand questions and extract the right answer from our knowledge bases, whereas the sales agent has to accommodate various dialog situations with respect to a sales scenario. [sent-109, score-1.265]
77 It therefore has to understand the dialog acts intended by the user and trigger the corresponding reactions, such as presenting an object, memorizing user preferences, negotiating further sales goals, etc. [sent-110, score-0.745]
78 The task model for sales conversations is inspired by a corpus resulting from the annotation of a Wizard-of-Oz experiment in the furniture sales agent scenario carried out by our project partner at ZAS (Bertomeu and Benz, 2009). [sent-111, score-0.921]
79 In these experiments, 18 users spent one hour each on furnishing a virtual living room in a Twinity apartment by talking to a human wizard controlling the virtual sales agent. [sent-112, score-0.8]
80 The final corpus consists of 18 dialogs containing 3,171 turns with 4,3 13 utterances and 23,015 alpha-numerical strings (words). [sent-113, score-0.029]
81 Table 1: Example Conversation from the Wizardof-Oz Experiment The flow of the task-based conversation is controlled by a data-driven finite-state model, which is the backbone of the dialog manager. [sent-118, score-0.525]
82 During a sales conversation, objects and features of objects mentioned by the NPC and the user are extracted from the knowledge bases and added into the underspecified graph nodes and egdes at runtime. [sent-119, score-0.553]
83 This strategy keeps the finite-state graph as small as possible. [sent-120, score-0.026]
84 Discussed objects and their features are stored in a frame-based sub-component named ”form”. [sent-121, score-0.072]
85 The form contains entries which correspond to ontological concepts in the furni39 ture ontology. [sent-122, score-0.029]
86 This frame-based approach increases the flexibility of the dialog manager (McTear, 2002) and is particularly useful for a task-driven dialog system. [sent-124, score-0.808]
87 As long as the negotiated object is not yet fully specified, the form represents the underspecified object description according to the ontology concept. [sent-125, score-0.177]
88 Every time the user states a new preference or request, the form is enriched with additional features until the set of objects is small enough to be presented to the user for final selection. [sent-126, score-0.258]
89 Thus the actual flow of dialog according to the task model does not have to be expressed by the graph but can be derived on demand from the knowledge and handled by the form which in turn activates the appropriate dialog subgraphs. [sent-127, score-0.773]
90 This combination of graph-based dialog models and form-based task modelling effectively accounts for the interaction of sequential dialog strategies and the non-sequential nature of complex dialog goals. [sent-128, score-1.115]
91 Given a resolved semantic representation, the dialog manager triggers either a semantic search in the knowledge bases to deliver factual answers as needed in a gossip conversation or a further dialog response for example providing choices for the user in a sales domain. [sent-129, score-1.515]
92 In case that the semantic representation can neither be resolved in the task domain nor in the gossip domain, it gets passed to the embedded A. [sent-131, score-0.215]
93 chatbot that uses its own understanding and generation components (Wallace and Bush, 2001). [sent-136, score-0.067]
94 1 Semantic Representation The input understanding of the system is implemented as one single understanding pipeline. [sent-138, score-0.054]
95 The understanding pipeline delivers a semantic representation which is the basis for the decision of the dialog manager which action to perform next. [sent-139, score-0.613]
96 The representation’s structure is inspired by our knowledge representation design described in section 3 as well as by predicate logic. [sent-141, score-0.074]
97 The core of the representation is a predicate-argument structure limited to two arguments including message type and the whole syntactic information found by the analysis pipeline. [sent-142, score-0.038]
wordName wordTfidf (topN-words)
[('twinity', 0.389), ('dialog', 0.357), ('npc', 0.297), ('furniture', 0.28), ('komparse', 0.252), ('npcs', 0.229), ('virtual', 0.212), ('sales', 0.202), ('bartender', 0.16), ('agent', 0.154), ('server', 0.146), ('conversation', 0.145), ('avatar', 0.103), ('ontology', 0.1), ('gossip', 0.1), ('manager', 0.094), ('user', 0.093), ('client', 0.091), ('game', 0.074), ('worlds', 0.073), ('conversational', 0.072), ('objects', 0.072), ('rdf', 0.069), ('celebrities', 0.069), ('cocktail', 0.069), ('users', 0.067), ('color', 0.065), ('clients', 0.06), ('world', 0.057), ('gb', 0.055), ('berlin', 0.053), ('owl', 0.052), ('bases', 0.049), ('architecture', 0.047), ('ontologies', 0.047), ('adolphs', 0.046), ('apartment', 0.046), ('biographical', 0.046), ('cocktails', 0.046), ('triple', 0.045), ('characters', 0.044), ('modelling', 0.044), ('physical', 0.042), ('realizes', 0.04), ('entertain', 0.04), ('chatbot', 0.04), ('analyzer', 0.039), ('representation', 0.038), ('ingredients', 0.037), ('plugin', 0.037), ('selling', 0.037), ('knowledge', 0.036), ('action', 0.036), ('loading', 0.034), ('fulfill', 0.034), ('graphics', 0.034), ('reasoner', 0.034), ('platforms', 0.034), ('dfki', 0.034), ('partner', 0.034), ('economic', 0.033), ('hosts', 0.033), ('room', 0.032), ('talk', 0.031), ('delivers', 0.031), ('defining', 0.031), ('semantic', 0.03), ('life', 0.03), ('concepts', 0.029), ('utterances', 0.029), ('store', 0.029), ('ram', 0.029), ('demands', 0.029), ('styles', 0.029), ('underspecified', 0.029), ('functionality', 0.029), ('talking', 0.029), ('online', 0.027), ('understanding', 0.027), ('accessed', 0.027), ('querying', 0.026), ('keeps', 0.026), ('appearance', 0.025), ('modelled', 0.025), ('conversations', 0.025), ('domain', 0.025), ('buy', 0.025), ('individuals', 0.025), ('generator', 0.025), ('domainspecific', 0.025), ('interfaces', 0.025), ('object', 0.024), ('scenario', 0.024), ('storing', 0.024), ('responsible', 0.024), ('triples', 0.024), ('services', 0.023), ('flow', 0.023), ('pieces', 0.023), ('resolved', 0.022)]
simIndex simValue paperId paperTitle
same-paper 1 1.0000001 224 acl-2010-Talking NPCs in a Virtual Game World
Author: Tina Kluwer ; Peter Adolphs ; Feiyu Xu ; Hans Uszkoreit ; Xiwen Cheng
Abstract: This paper describes the KomParse system, a natural-language dialog system in the three-dimensional virtual world Twinity. In order to fulfill the various communication demands between nonplayer characters (NPCs) and users in such an online virtual world, the system realizes a flexible and hybrid approach combining knowledge-intensive domainspecific question answering, task-specific and domain-specific dialog with robust chatbot-like chitchat.
2 0.12272903 82 acl-2010-Demonstration of a Prototype for a Conversational Companion for Reminiscing about Images
Author: Yorick Wilks ; Roberta Catizone ; Alexiei Dingli ; Weiwei Cheng
Abstract: This paper describes an initial prototype demonstrator of a Companion, designed as a platform for novel approaches to the following: 1) The use of Information Extraction (IE) techniques to extract the content of incoming dialogue utterances after an Automatic Speech Recognition (ASR) phase, 2) The conversion of the input to Resource Descriptor Format (RDF) to allow the generation of new facts from existing ones, under the control of a Dialogue Manger (DM), that also has access to stored knowledge and to open knowledge accessed in real time from the web, all in RDF form, 3) A DM implemented as a stack and network virtual machine that models mixed initiative in dialogue control, and 4) A tuned dialogue act detector based on corpus evidence. The prototype platform was evaluated, and we describe this briefly; it is also designed to support more extensive forms of emotion detection carried by both speech and lexical content, as well as extended forms of machine learning.
3 0.089262113 167 acl-2010-Learning to Adapt to Unknown Users: Referring Expression Generation in Spoken Dialogue Systems
Author: Srinivasan Janarthanam ; Oliver Lemon
Abstract: We present a data-driven approach to learn user-adaptive referring expression generation (REG) policies for spoken dialogue systems. Referring expressions can be difficult to understand in technical domains where users may not know the technical ‘jargon’ names of the domain entities. In such cases, dialogue systems must be able to model the user’s (lexical) domain knowledge and use appropriate referring expressions. We present a reinforcement learning (RL) framework in which the sys- tem learns REG policies which can adapt to unknown users online. Furthermore, unlike supervised learning methods which require a large corpus of expert adaptive behaviour to train on, we show that effective adaptive policies can be learned from a small dialogue corpus of non-adaptive human-machine interaction, by using a RL framework and a statistical user simulation. We show that in comparison to adaptive hand-coded baseline policies, the learned policy performs significantly better, with an 18.6% average increase in adaptation accuracy. The best learned policy also takes less dialogue time (average 1.07 min less) than the best hand-coded policy. This is because the learned policies can adapt online to changing evidence about the user’s domain expertise.
4 0.077987559 199 acl-2010-Preferences versus Adaptation during Referring Expression Generation
Author: Martijn Goudbeek ; Emiel Krahmer
Abstract: Current Referring Expression Generation algorithms rely on domain dependent preferences for both content selection and linguistic realization. We present two experiments showing that human speakers may opt for dispreferred properties and dispreferred modifier orderings when these were salient in a preceding interaction (without speakers being consciously aware of this). We discuss the impact of these findings for current generation algorithms.
5 0.071524329 112 acl-2010-Extracting Social Networks from Literary Fiction
Author: David Elson ; Nicholas Dames ; Kathleen McKeown
Abstract: We present a method for extracting social networks from literature, namely, nineteenth-century British novels and serials. We derive the networks from dialogue interactions, and thus our method depends on the ability to determine when two characters are in conversation. Our approach involves character name chunking, quoted speech attribution and conversation detection given the set of quotes. We extract features from the social networks and examine their correlation with one another, as well as with metadata such as the novel’s setting. Our results provide evidence that the majority of novels in this time period do not fit two characterizations provided by literacy scholars. Instead, our results suggest an alternative explanation for differences in social networks.
6 0.068440065 64 acl-2010-Complexity Assumptions in Ontology Verbalisation
7 0.06734661 187 acl-2010-Optimising Information Presentation for Spoken Dialogue Systems
8 0.062363084 173 acl-2010-Modeling Norms of Turn-Taking in Multi-Party Conversation
9 0.058571011 178 acl-2010-Non-Cooperation in Dialogue
10 0.057009228 142 acl-2010-Importance-Driven Turn-Bidding for Spoken Dialogue Systems
11 0.050975155 129 acl-2010-Growing Related Words from Seed via User Behaviors: A Re-Ranking Based Approach
12 0.04602227 215 acl-2010-Speech-Driven Access to the Deep Web on Mobile Devices
13 0.043381121 35 acl-2010-Automated Planning for Situated Natural Language Generation
14 0.04309731 113 acl-2010-Extraction and Approximation of Numerical Attributes from the Web
15 0.042948816 128 acl-2010-Grammar Prototyping and Testing with the LinGO Grammar Matrix Customization System
16 0.042938046 168 acl-2010-Learning to Follow Navigational Directions
17 0.041168869 239 acl-2010-Towards Relational POMDPs for Adaptive Dialogue Management
18 0.039754514 47 acl-2010-Beetle II: A System for Tutoring and Computational Linguistics Experimentation
19 0.038638599 218 acl-2010-Structural Semantic Relatedness: A Knowledge-Based Method to Named Entity Disambiguation
20 0.036999717 202 acl-2010-Reading between the Lines: Learning to Map High-Level Instructions to Commands
topicId topicWeight
[(0, -0.098), (1, 0.074), (2, -0.024), (3, -0.079), (4, 0.002), (5, -0.098), (6, -0.075), (7, 0.042), (8, -0.038), (9, -0.018), (10, 0.018), (11, -0.028), (12, -0.023), (13, -0.01), (14, 0.034), (15, 0.006), (16, 0.037), (17, 0.072), (18, -0.019), (19, 0.037), (20, 0.002), (21, -0.092), (22, 0.016), (23, -0.031), (24, 0.009), (25, 0.046), (26, -0.03), (27, -0.029), (28, -0.028), (29, -0.03), (30, -0.021), (31, 0.003), (32, 0.021), (33, 0.083), (34, -0.073), (35, -0.052), (36, -0.07), (37, -0.004), (38, 0.024), (39, -0.0), (40, -0.107), (41, 0.073), (42, 0.055), (43, 0.071), (44, 0.019), (45, -0.023), (46, -0.097), (47, 0.007), (48, 0.012), (49, 0.201)]
simIndex simValue paperId paperTitle
same-paper 1 0.92938554 224 acl-2010-Talking NPCs in a Virtual Game World
Author: Tina Kluwer ; Peter Adolphs ; Feiyu Xu ; Hans Uszkoreit ; Xiwen Cheng
Abstract: This paper describes the KomParse system, a natural-language dialog system in the three-dimensional virtual world Twinity. In order to fulfill the various communication demands between nonplayer characters (NPCs) and users in such an online virtual world, the system realizes a flexible and hybrid approach combining knowledge-intensive domainspecific question answering, task-specific and domain-specific dialog with robust chatbot-like chitchat.
2 0.6480242 82 acl-2010-Demonstration of a Prototype for a Conversational Companion for Reminiscing about Images
Author: Yorick Wilks ; Roberta Catizone ; Alexiei Dingli ; Weiwei Cheng
Abstract: This paper describes an initial prototype demonstrator of a Companion, designed as a platform for novel approaches to the following: 1) The use of Information Extraction (IE) techniques to extract the content of incoming dialogue utterances after an Automatic Speech Recognition (ASR) phase, 2) The conversion of the input to Resource Descriptor Format (RDF) to allow the generation of new facts from existing ones, under the control of a Dialogue Manger (DM), that also has access to stored knowledge and to open knowledge accessed in real time from the web, all in RDF form, 3) A DM implemented as a stack and network virtual machine that models mixed initiative in dialogue control, and 4) A tuned dialogue act detector based on corpus evidence. The prototype platform was evaluated, and we describe this briefly; it is also designed to support more extensive forms of emotion detection carried by both speech and lexical content, as well as extended forms of machine learning.
3 0.58397967 64 acl-2010-Complexity Assumptions in Ontology Verbalisation
Author: Richard Power
Abstract: We describe the strategy currently pursued for verbalising OWL ontologies by sentences in Controlled Natural Language (i.e., combining generic rules for realising logical patterns with ontology-specific lexicons for realising atomic terms for individuals, classes, and properties) and argue that its success depends on assumptions about the complexity of terms and axioms in the ontology. We then show, through analysis of a corpus of ontologies, that although these assumptions could in principle be violated, they are overwhelmingly respected in practice by ontology developers.
4 0.51227885 142 acl-2010-Importance-Driven Turn-Bidding for Spoken Dialogue Systems
Author: Ethan Selfridge ; Peter Heeman
Abstract: Current turn-taking approaches for spoken dialogue systems rely on the speaker releasing the turn before the other can take it. This reliance results in restricted interactions that can lead to inefficient dialogues. In this paper we present a model we refer to as Importance-Driven Turn-Bidding that treats turn-taking as a negotiative process. Each conversant bids for the turn based on the importance of the intended utterance, and Reinforcement Learning is used to indirectly learn this parameter. We find that Importance-Driven Turn-Bidding performs better than two current turntaking approaches in an artificial collaborative slot-filling domain. The negotiative nature of this model creates efficient dia- logues, and supports the improvement of mixed-initiative interaction.
5 0.49157763 167 acl-2010-Learning to Adapt to Unknown Users: Referring Expression Generation in Spoken Dialogue Systems
Author: Srinivasan Janarthanam ; Oliver Lemon
Abstract: We present a data-driven approach to learn user-adaptive referring expression generation (REG) policies for spoken dialogue systems. Referring expressions can be difficult to understand in technical domains where users may not know the technical ‘jargon’ names of the domain entities. In such cases, dialogue systems must be able to model the user’s (lexical) domain knowledge and use appropriate referring expressions. We present a reinforcement learning (RL) framework in which the sys- tem learns REG policies which can adapt to unknown users online. Furthermore, unlike supervised learning methods which require a large corpus of expert adaptive behaviour to train on, we show that effective adaptive policies can be learned from a small dialogue corpus of non-adaptive human-machine interaction, by using a RL framework and a statistical user simulation. We show that in comparison to adaptive hand-coded baseline policies, the learned policy performs significantly better, with an 18.6% average increase in adaptation accuracy. The best learned policy also takes less dialogue time (average 1.07 min less) than the best hand-coded policy. This is because the learned policies can adapt online to changing evidence about the user’s domain expertise.
6 0.47384235 187 acl-2010-Optimising Information Presentation for Spoken Dialogue Systems
7 0.46871281 178 acl-2010-Non-Cooperation in Dialogue
8 0.44986039 259 acl-2010-WebLicht: Web-Based LRT Services for German
9 0.44331819 204 acl-2010-Recommendation in Internet Forums and Blogs
10 0.43649119 138 acl-2010-Hunting for the Black Swan: Risk Mining from Text
11 0.41423029 113 acl-2010-Extraction and Approximation of Numerical Attributes from the Web
12 0.41368961 239 acl-2010-Towards Relational POMDPs for Adaptive Dialogue Management
13 0.40918213 112 acl-2010-Extracting Social Networks from Literary Fiction
14 0.38487664 173 acl-2010-Modeling Norms of Turn-Taking in Multi-Party Conversation
15 0.38240963 176 acl-2010-Mood Patterns and Affective Lexicon Access in Weblogs
16 0.38105297 7 acl-2010-A Generalized-Zero-Preserving Method for Compact Encoding of Concept Lattices
17 0.36275306 254 acl-2010-Using Speech to Reply to SMS Messages While Driving: An In-Car Simulator User Study
18 0.35929224 235 acl-2010-Tools for Multilingual Grammar-Based Translation on the Web
19 0.34831083 199 acl-2010-Preferences versus Adaptation during Referring Expression Generation
20 0.34208545 128 acl-2010-Grammar Prototyping and Testing with the LinGO Grammar Matrix Customization System
topicId topicWeight
[(1, 0.013), (14, 0.013), (25, 0.564), (42, 0.03), (44, 0.012), (59, 0.044), (73, 0.032), (78, 0.028), (83, 0.06), (84, 0.022), (98, 0.073)]
simIndex simValue paperId paperTitle
same-paper 1 0.95010954 224 acl-2010-Talking NPCs in a Virtual Game World
Author: Tina Kluwer ; Peter Adolphs ; Feiyu Xu ; Hans Uszkoreit ; Xiwen Cheng
Abstract: This paper describes the KomParse system, a natural-language dialog system in the three-dimensional virtual world Twinity. In order to fulfill the various communication demands between nonplayer characters (NPCs) and users in such an online virtual world, the system realizes a flexible and hybrid approach combining knowledge-intensive domainspecific question answering, task-specific and domain-specific dialog with robust chatbot-like chitchat.
2 0.92656797 203 acl-2010-Rebanking CCGbank for Improved NP Interpretation
Author: Matthew Honnibal ; James R. Curran ; Johan Bos
Abstract: Once released, treebanks tend to remain unchanged despite any shortcomings in their depth of linguistic analysis or coverage of specific phenomena. Instead, separate resources are created to address such problems. In this paper we show how to improve the quality of a treebank, by integrating resources and implementing improved analyses for specific constructions. We demonstrate this rebanking process by creating an updated version of CCGbank that includes the predicate-argument structure of both verbs and nouns, baseNP brackets, verb-particle constructions, and restrictive and non-restrictive nominal modifiers; and evaluate the impact of these changes on a statistical parser.
3 0.91306454 28 acl-2010-An Entity-Level Approach to Information Extraction
Author: Aria Haghighi ; Dan Klein
Abstract: We present a generative model of template-filling in which coreference resolution and role assignment are jointly determined. Underlying template roles first generate abstract entities, which in turn generate concrete textual mentions. On the standard corporate acquisitions dataset, joint resolution in our entity-level model reduces error over a mention-level discriminative approach by up to 20%.
4 0.82944655 69 acl-2010-Constituency to Dependency Translation with Forests
Author: Haitao Mi ; Qun Liu
Abstract: Tree-to-string systems (and their forestbased extensions) have gained steady popularity thanks to their simplicity and efficiency, but there is a major limitation: they are unable to guarantee the grammaticality of the output, which is explicitly modeled in string-to-tree systems via targetside syntax. We thus propose to combine the advantages of both, and present a novel constituency-to-dependency translation model, which uses constituency forests on the source side to direct the translation, and dependency trees on the target side (as a language model) to ensure grammaticality. Medium-scale experiments show an absolute and statistically significant improvement of +0.7 BLEU points over a state-of-the-art forest-based tree-to-string system even with fewer rules. This is also the first time that a treeto-tree model can surpass tree-to-string counterparts.
5 0.7626524 237 acl-2010-Topic Models for Word Sense Disambiguation and Token-Based Idiom Detection
Author: Linlin Li ; Benjamin Roth ; Caroline Sporleder
Abstract: This paper presents a probabilistic model for sense disambiguation which chooses the best sense based on the conditional probability of sense paraphrases given a context. We use a topic model to decompose this conditional probability into two conditional probabilities with latent variables. We propose three different instantiations of the model for solving sense disambiguation problems with different degrees of resource availability. The proposed models are tested on three different tasks: coarse-grained word sense disambiguation, fine-grained word sense disambiguation, and detection of literal vs. nonliteral usages of potentially idiomatic expressions. In all three cases, we outper- form state-of-the-art systems either quantitatively or statistically significantly.
6 0.75838739 89 acl-2010-Distributional Similarity vs. PU Learning for Entity Set Expansion
7 0.75289381 23 acl-2010-Accurate Context-Free Parsing with Combinatory Categorial Grammar
8 0.56263393 71 acl-2010-Convolution Kernel over Packed Parse Forest
9 0.56163496 53 acl-2010-Blocked Inference in Bayesian Tree Substitution Grammars
10 0.55938607 75 acl-2010-Correcting Errors in a Treebank Based on Synchronous Tree Substitution Grammar
11 0.55678928 128 acl-2010-Grammar Prototyping and Testing with the LinGO Grammar Matrix Customization System
12 0.54844224 169 acl-2010-Learning to Translate with Source and Target Syntax
13 0.54506004 257 acl-2010-WSD as a Distributed Constraint Optimization Problem
14 0.54065555 118 acl-2010-Fine-Grained Tree-to-String Translation Rule Extraction
15 0.53212535 211 acl-2010-Simple, Accurate Parsing with an All-Fragments Grammar
16 0.52185494 46 acl-2010-Bayesian Synchronous Tree-Substitution Grammar Induction and Its Application to Sentence Compression
17 0.51776993 261 acl-2010-Wikipedia as Sense Inventory to Improve Diversity in Web Search Results
19 0.51349413 121 acl-2010-Generating Entailment Rules from FrameNet
20 0.50451231 82 acl-2010-Demonstration of a Prototype for a Conversational Companion for Reminiscing about Images