acl acl2013 acl2013-266 knowledge-graph by maker-knowledge-mining

266 acl-2013-PAL: A Chatterbot System for Answering Domain-specific Questions


Source: pdf

Author: Yuanchao Liu ; Ming Liu ; Xiaolong Wang ; Limin Wang ; Jingjing Li

Abstract: In this paper, we propose PAL, a prototype chatterbot for answering non-obstructive psychological domain-specific questions. This system focuses on providing primary suggestions or helping people relieve pressure by extracting knowledge from online forums, based on which the chatterbot system is constructed. The strategies used by PAL, including semantic-extension-based question matching, solution management with personal information consideration, and XML-based knowledge pattern construction, are described and discussed. We also conduct a primary test for the feasibility of our system.

Reference: text


Summary: the most important sentenses genereted by tfidf model

sentIndex sentText sentNum sentScore

1 School of public health, Harbin Medical University, Harbin, China , , { lyc ml iu wangxl , j j l} @ insun . [sent-2, score-0.062]

2 hit Abstract In this paper, we propose PAL, a prototype chatterbot for answering non-obstructive psychological domain-specific questions. [sent-3, score-0.462]

3 This system focuses on providing primary suggestions or helping people relieve pressure by extracting knowledge from online forums, based on which the chatterbot system is constructed. [sent-4, score-0.659]

4 The strategies used by PAL, including semantic-extension-based question matching, solution management with personal information consideration, and XML-based knowledge pattern construction, are described and discussed. [sent-5, score-0.28]

5 We also conduct a primary test for the feasibility of our system. [sent-6, score-0.044]

6 1 Introduction A wide variety of chatterbots and question-and-answer (Q&A;) systems have been proposed over the past decades, each with strengths that make them appropriate for particular applications. [sent-7, score-0.178]

7 With numerous advances in information construction, people increasingly aim to communicate with computers using natural language. [sent-8, score-0.064]

8 For example, chatterbots in some e-commerce Web sites can interact with customers and provide help similar to a real-life secretary (DeeAnna Merz Nagel, 2011; Yvette Colón, 2011). [sent-9, score-0.27]

9 In this paper, we propose PAL (Psychologist of Artificial Language), a chatterbot system for answering non-obstructive psychological questions. [sent-10, score-0.506]

10 Non-obstructive questions refer to problems on family, human relationships, marriage, life pressure, learning, work and so on. [sent-11, score-0.157]

11 In these cases, we expect the chatterbot to play an active role by providing tutoring, solution, support, advice, or even sympathy depending on the help needed by its users. [sent-12, score-0.32]

12 In the following sections, we will briefly discuss related work and then introduce our system and its main features. [sent-17, score-0.044]

13 2 Related Work A number of research work on chatterbots (Rafael E. [sent-18, score-0.178]

14 Several studies on the application of natural language processing technologies for non-obstructive psychological Q&A; systems have also been published (Hai-hu Shi, 2005). [sent-21, score-0.089]

15 Several online psychology counselling Web sites with service provided by human experts have also been established recently (DeeAnna Merz Nagel, 2011; Yvette Colón, 2011). [sent-22, score-0.226]

16 For these Web sites, when the visitors ask similar questions, the expert may provide the same or very similar answers repeatedly. [sent-23, score-0.178]

17 Based on this observation and consideration, we collected a large number of counselling Q&A; pairs to extract common knowledge for the construction of a chatterbot system. [sent-24, score-0.447]

18 Advances in automatic language analysis and processing are used as the bases for the emergence of a complex, task-oriented chatterbot system. [sent-25, score-0.32]

19 The basic control logic strategy is shown in Figure 3. [sent-29, score-0.156]

20 Basic Control Logic of PAL As shown in Figure 3, the initial state is set to welcome mode, and the system can select a sentence from the “sign on” list, which will then provide a response. [sent-31, score-0.107]

21 When users enter a question, the system will conduct the necessary analysis. [sent-32, score-0.128]

22 The system’s knowledge base is indexed by Clucene1 beforehand. [sent-33, score-0.183]

23 Thus, the knowledge index will be used to search the matched records quickly. [sent-34, score-0.182]

24 If the system can find the matched patterns directly and the answer is suitable for the current user, one answer will be randomly selected to generate the response. [sent-35, score-0.359]

25 Historical information and personal information will be analysed when necessary. [sent-36, score-0.113]

26 We mainly adopted the method of ELIZA2, which is an open-source program, to consider the historical information. [sent-37, score-0.086]

27 A “not found” response list is also set to deal with situations when no suitable answers can be identified. [sent-38, score-0.174]

28 Both system utterance and user input will be pushed into the stack as historical information. [sent-39, score-0.238]

29 Given that user questions are at times very simple, the combination with historical input may also be required to determine its meaning. [sent-40, score-0.351]

30 This step can also avoid the duplication of utterances. [sent-41, score-0.031]

31 5 Knowledge Construction Question Matching Method and We design P-XML to store the knowledge base for PAL, as shown in Figure 4. [sent-42, score-0.146]

32 The knowledge base for PAL is mainly derived from the Q&A; pairs in the BAIDU ZHIDAO community3. [sent-43, score-0.146]

33 com 69 An effective method of capturing the user’s meaning accurately is to create an extension for questions in the knowledge base. [sent-50, score-0.268]

34 In this paper, the extension is primarily a synonym expansion of the keywords of questions, with CILIN (Wanxiang Che, 2010) as extension knowledge source. [sent-51, score-0.166]

35 The questions are indexed by Clucene to improve the retrieval efficiency of the search for a matched entry in the knowledge base. [sent-52, score-0.337]

36 During the knowledge base searching step, both the index of the original form and the extension form of the problem are used to find the most possible matched record for the user’s question, as shown in algorithm 1. [sent-53, score-0.379]

37 Algorithm 1is used to examine the similarity between user input and the record returned by Clucene, including traditional and extension similarities. [sent-54, score-0.324]

38 6 Response Management Method One question usually has many corresponding answers in the knowledge base, and these answers differ in explanation quality. [sent-55, score-0.394]

39 Thus, the basic strategy employed by solution management is to select a reliable answer from the matched r ec ord as response, as shown in algorithm 2. [sent-56, score-0.244]

40 Based on these rules, if one answer contains personal information, it will be selected as the candidate answer only when the personal information is consistent with that of the current user. [sent-59, score-0.454]

41 Very concise answers that do not contain personal information can generally be selected as a candidate answer. [sent-60, score-0.248]

42 7 Experiments For the current implementation of PAL, the size of the knowledge base is approximately 1. [sent-61, score-0.146]

43 com, which is one of the largest Chinese online communities. [sent-65, score-0.059]

44 The criterion for choosing these six categories is also because they are the main topics in BAIDU communities about psychological problems. [sent-66, score-0.089]

45 Some information on the knowledge base is given in Table 1, in which “Percent of questions matched” denotes the number of similar questions found when 100 open questions are input (we suppose that if the similarity threshold is bigger than 0. [sent-67, score-0.89]

46 5, then a similar question will be deemed as “hit” in the knowledge base). [sent-68, score-0.124]

47 1, we examine the feasibility of using downloaded dialogue collection for constructing the knowledge base. [sent-70, score-0.21]

48 of unique Percent of questions Size(MB) length Terms in ques. [sent-79, score-0.157]

49 1 System Performance Evaluation Additional questions and their corresponding answers beyond the knowledge base are also used as a test set to evaluate system performance. [sent-96, score-0.482]

50 Concretely, suppose question Q has |A| answers in the test set. [sent-97, score-0.242]

51 Suppose the system output is O, we examine if one best answer exists among |A| answers that are very similar to O (the similarity is greater than threshold T3). [sent-99, score-0.438]

52 If yes, we then assume that one suitable answer has been found. [sent-100, score-0.114]

53 In this way, 70 precision can be calculated as the number of questions that have very similar answers in the system divided by the number of all input questions. [sent-101, score-0.375]

54 The horizon axis denotes the similarity threshold (T1 for sim1 and T2 for sim2) between a user’s input and the questions in the knowledge base. [sent-103, score-0.496]

55 Sim1 is the original similarity, whereas sim2 is the semantic extension similarity. [sent-104, score-0.055]

56 The similarity threshold T3 denotes the similarity between the answer in the test set and system output O. [sent-108, score-0.353]

57 Basically, when T3 is bigger, the system’s performance tends to decrease because a high T3 value denotes a strict evaluation standard. [sent-117, score-0.041]

58 When only index is used and both sim1 and sim2 are below the corresponding threshold T1 or T2, the system can still return record set RS2, but the returned answer may be inconsistent with user’s question. [sent-119, score-0.349]

59 Precision of PAL with different similarity thresholds T3 (The X axis denotes different thresholds for sim1 (T1) and sim2 (T2). [sent-134, score-0.241]

60 The Y axis stands for the precision value of different T1, T2 and T3 values for different areas). [sent-140, score-0.061]

61 2.. Some Interaction Examples Table 2 provides an interaction example of PAL in English. [sent-142, score-0.048]

62 As users input their questions, the system can output the corresponding responses. [sent-143, score-0.117]

63 Basically, when users describe their troubles, PAL can provide some explanation or even try to console the user. [sent-145, score-0.034]

64 Compared to other chatterbots like Eliza, PAL is domain-specific and may provide help to some extent. [sent-146, score-0.178]

65 Thus, the user could input the same question many times, with the system providing suggestions from different angles. [sent-148, score-0.252]

66 PAL can also output voice by using Xiao-kun and Microsoft Anna as the text-to-speech (TTS) engine. [sent-149, score-0.041]

67 The speed, volume and echo properties of the voice can be adjusted. [sent-150, score-0.041]

68 Dialog for Figure 1 (in English) PAL> Hello, you are welcome to communicate with me! [sent-151, score-0.127]

69 You can input from the left side and press enter to get answer! [sent-152, score-0.089]

70 I> How do I deal with the children who like to complain very frequently? [sent-153, score-0.068]

71 Don't’ let him experience the benefits of complaining very frequently. [sent-155, score-0.031]

72 I> Why do parents give children so much pressure PAL> Children have pressure indeed; but parents also have a lot of pressure. [sent-156, score-0.411]

73 I> Why am Iunable to control my emotions, and often quarrel with my parents? [sent-157, score-0.068]

74 Example of an interaction with PAL (in English) 71 8 Conclusions In this paper, we introduce a simple chatterbot for answering non-obstructive psychological questions. [sent-166, score-0.51]

75 The knowledge behind the chatterbot is learned from the Q&A; pairs derived from an online forum using several extraction strategies. [sent-167, score-0.435]

76 The historical and personal information from the dialogues are also incorporated to output an appropriate answer. [sent-168, score-0.199]

77 , enabling the system to ask questions actively and further improving P-XML to form richer patterns for storing Q&A; knowledge. [sent-171, score-0.241]

78 Another interesting aspect would be to add speech input as well as TTS and to transform PAL into a mobile platform for widespread use. [sent-172, score-0.039]

79 2007AA01Z172 Youth Funds of China social & humanity science (10YJCZH099), and Key Laboratory Opening Funding of China MOE—MS Key Laboratory of Natural Language Processing and Speech (HIT. [sent-174, score-0.031]

80 Research on on-line psychology consultation expert system based on man-machine interaction technique. [sent-182, score-0.17]

81 Research of sentiment model based on HMM and its application in psychological consulting expert system. [sent-185, score-0.132]

82 Improving the performance of question answering with semantically equivalent answer patterns. [sent-189, score-0.235]

83 Language, logic and ontology: Uncovering the structure of commonsense knowledge. [sent-201, score-0.119]


similar papers computed by tfidf model

tfidf for this paper:

wordName wordTfidf (topN-words)

[('pal', 0.622), ('chatterbot', 0.32), ('chatterbots', 0.178), ('questions', 0.157), ('answers', 0.135), ('counseling', 0.126), ('answer', 0.114), ('personal', 0.113), ('deeanna', 0.107), ('merz', 0.107), ('nagel', 0.107), ('pressure', 0.104), ('yvette', 0.094), ('base', 0.09), ('psychological', 0.089), ('logic', 0.088), ('matched', 0.087), ('historical', 0.086), ('harbin', 0.085), ('parents', 0.083), ('dialogue', 0.076), ('col', 0.072), ('clucene', 0.071), ('counselling', 0.071), ('hau', 0.071), ('kosseim', 0.071), ('leila', 0.071), ('shilin', 0.071), ('user', 0.069), ('threshold', 0.068), ('control', 0.068), ('question', 0.068), ('communicate', 0.064), ('lian', 0.063), ('dalmas', 0.063), ('tiphaine', 0.063), ('welcome', 0.063), ('axis', 0.061), ('sites', 0.061), ('china', 0.059), ('online', 0.059), ('walid', 0.058), ('july', 0.056), ('knowledge', 0.056), ('extension', 0.055), ('wanxiang', 0.055), ('answering', 0.053), ('record', 0.052), ('banchs', 0.05), ('tts', 0.05), ('enter', 0.05), ('thresholds', 0.048), ('cong', 0.048), ('baidu', 0.048), ('interaction', 0.048), ('che', 0.045), ('system', 0.044), ('feasibility', 0.044), ('management', 0.043), ('similarity', 0.043), ('expert', 0.043), ('bigger', 0.043), ('rafael', 0.041), ('denotes', 0.041), ('voice', 0.041), ('storing', 0.04), ('response', 0.039), ('input', 0.039), ('index', 0.039), ('suppose', 0.039), ('haizhou', 0.038), ('children', 0.037), ('ding', 0.037), ('indexed', 0.037), ('edition', 0.036), ('psychology', 0.035), ('percent', 0.035), ('shi', 0.034), ('users', 0.034), ('examine', 0.034), ('jing', 0.033), ('basically', 0.033), ('aw', 0.032), ('suggestions', 0.032), ('returned', 0.032), ('troubles', 0.031), ('duplication', 0.031), ('complain', 0.031), ('complaining', 0.031), ('horizon', 0.031), ('listening', 0.031), ('insun', 0.031), ('wangxl', 0.031), ('affairs', 0.031), ('commonsense', 0.031), ('humanity', 0.031), ('personalised', 0.031), ('psychologist', 0.031), ('scheduled', 0.031), ('secretary', 0.031)]

similar papers list:

simIndex simValue paperId paperTitle

same-paper 1 1.0 266 acl-2013-PAL: A Chatterbot System for Answering Domain-specific Questions

Author: Yuanchao Liu ; Ming Liu ; Xiaolong Wang ; Limin Wang ; Jingjing Li

Abstract: In this paper, we propose PAL, a prototype chatterbot for answering non-obstructive psychological domain-specific questions. This system focuses on providing primary suggestions or helping people relieve pressure by extracting knowledge from online forums, based on which the chatterbot system is constructed. The strategies used by PAL, including semantic-extension-based question matching, solution management with personal information consideration, and XML-based knowledge pattern construction, are described and discussed. We also conduct a primary test for the feasibility of our system.

2 0.12499109 107 acl-2013-Deceptive Answer Prediction with User Preference Graph

Author: Fangtao Li ; Yang Gao ; Shuchang Zhou ; Xiance Si ; Decheng Dai

Abstract: In Community question answering (QA) sites, malicious users may provide deceptive answers to promote their products or services. It is important to identify and filter out these deceptive answers. In this paper, we first solve this problem with the traditional supervised learning methods. Two kinds of features, including textual and contextual features, are investigated for this task. We further propose to exploit the user relationships to identify the deceptive answers, based on the hypothesis that similar users will have similar behaviors to post deceptive or authentic answers. To measure the user similarity, we propose a new user preference graph based on the answer preference expressed by users, such as “helpful” voting and “best answer” selection. The user preference graph is incorporated into traditional supervised learning framework with the graph regularization technique. The experiment results demonstrate that the user preference graph can indeed help improve the performance of deceptive answer prediction.

3 0.11245526 291 acl-2013-Question Answering Using Enhanced Lexical Semantic Models

Author: Wen-tau Yih ; Ming-Wei Chang ; Christopher Meek ; Andrzej Pastusiak

Abstract: In this paper, we study the answer sentence selection problem for question answering. Unlike previous work, which primarily leverages syntactic analysis through dependency tree matching, we focus on improving the performance using models of lexical semantic resources. Experiments show that our systems can be consistently and significantly improved with rich lexical semantic information, regardless of the choice of learning algorithms. When evaluated on a benchmark dataset, the MAP and MRR scores are increased by 8 to 10 points, compared to one of our baseline systems using only surface-form matching. Moreover, our best system also outperforms pervious work that makes use of the dependency tree structure by a wide margin.

4 0.10973256 329 acl-2013-Statistical Machine Translation Improves Question Retrieval in Community Question Answering via Matrix Factorization

Author: Guangyou Zhou ; Fang Liu ; Yang Liu ; Shizhu He ; Jun Zhao

Abstract: Community question answering (CQA) has become an increasingly popular research topic. In this paper, we focus on the problem of question retrieval. Question retrieval in CQA can automatically find the most relevant and recent questions that have been solved by other users. However, the word ambiguity and word mismatch problems bring about new challenges for question retrieval in CQA. State-of-the-art approaches address these issues by implicitly expanding the queried questions with additional words or phrases using monolingual translation models. While useful, the effectiveness of these models is highly dependent on the availability of quality parallel monolingual corpora (e.g., question-answer pairs) in the absence of which they are troubled by noise issue. In this work, we propose an alternative way to address the word ambiguity and word mismatch problems by taking advantage of potentially rich semantic information drawn from other languages. Our proposed method employs statistical machine translation to improve question retrieval and enriches the question representation with the translated words from other languages via matrix factorization. Experiments conducted on a real CQA data show that our proposed approach is promising.

5 0.1059404 169 acl-2013-Generating Synthetic Comparable Questions for News Articles

Author: Oleg Rokhlenko ; Idan Szpektor

Abstract: We introduce the novel task of automatically generating questions that are relevant to a text but do not appear in it. One motivating example of its application is for increasing user engagement around news articles by suggesting relevant comparable questions, such as “is Beyonce a better singer than Madonna?”, for the user to answer. We present the first algorithm for the task, which consists of: (a) offline construction of a comparable question template database; (b) ranking of relevant templates to a given article; and (c) instantiation of templates only with entities in the article whose comparison under the template’s relation makes sense. We tested the suggestions generated by our algorithm via a Mechanical Turk experiment, which showed a significant improvement over the strongest baseline of more than 45% in all metrics.

6 0.10285346 272 acl-2013-Paraphrase-Driven Learning for Open Question Answering

7 0.085610956 60 acl-2013-Automatic Coupling of Answer Extraction and Information Retrieval

8 0.083573163 292 acl-2013-Question Classification Transfer

9 0.082808085 290 acl-2013-Question Analysis for Polish Question Answering

10 0.074879035 218 acl-2013-Latent Semantic Tensor Indexing for Community-based Question Answering

11 0.069543287 168 acl-2013-Generating Recommendation Dialogs by Extracting Information from User Reviews

12 0.068973728 241 acl-2013-Minimum Bayes Risk based Answer Re-ranking for Question Answering

13 0.067171678 141 acl-2013-Evaluating a City Exploration Dialogue System with Integrated Question-Answering and Pedestrian Navigation

14 0.065913543 254 acl-2013-Multimodal DBN for Predicting High-Quality Answers in cQA portals

15 0.059269927 230 acl-2013-Lightly Supervised Learning of Procedural Dialog Systems

16 0.051526055 387 acl-2013-Why-Question Answering using Intra- and Inter-Sentential Causal Relations

17 0.050872732 239 acl-2013-Meet EDGAR, a tutoring agent at MONSERRATE

18 0.046603817 121 acl-2013-Discovering User Interactions in Ideological Discussions

19 0.046589617 90 acl-2013-Conditional Random Fields for Responsive Surface Realisation using Global Features

20 0.045227703 78 acl-2013-Categorization of Turkish News Documents with Morphological Analysis


similar papers computed by lsi model

lsi for this paper:

topicId topicWeight

[(0, 0.128), (1, 0.055), (2, 0.003), (3, -0.066), (4, 0.048), (5, 0.02), (6, 0.033), (7, -0.208), (8, 0.09), (9, 0.032), (10, 0.011), (11, -0.012), (12, 0.025), (13, -0.013), (14, 0.024), (15, -0.009), (16, -0.003), (17, -0.022), (18, 0.048), (19, 0.012), (20, -0.015), (21, -0.042), (22, 0.051), (23, -0.061), (24, 0.007), (25, -0.018), (26, 0.024), (27, 0.021), (28, 0.017), (29, -0.021), (30, 0.025), (31, -0.004), (32, 0.018), (33, -0.01), (34, 0.018), (35, 0.051), (36, -0.036), (37, 0.014), (38, 0.013), (39, -0.03), (40, 0.017), (41, 0.0), (42, 0.036), (43, 0.013), (44, -0.018), (45, -0.038), (46, -0.012), (47, -0.017), (48, 0.035), (49, -0.023)]

similar papers list:

simIndex simValue paperId paperTitle

same-paper 1 0.94555855 266 acl-2013-PAL: A Chatterbot System for Answering Domain-specific Questions

Author: Yuanchao Liu ; Ming Liu ; Xiaolong Wang ; Limin Wang ; Jingjing Li

Abstract: In this paper, we propose PAL, a prototype chatterbot for answering non-obstructive psychological domain-specific questions. This system focuses on providing primary suggestions or helping people relieve pressure by extracting knowledge from online forums, based on which the chatterbot system is constructed. The strategies used by PAL, including semantic-extension-based question matching, solution management with personal information consideration, and XML-based knowledge pattern construction, are described and discussed. We also conduct a primary test for the feasibility of our system.

2 0.82713419 107 acl-2013-Deceptive Answer Prediction with User Preference Graph

Author: Fangtao Li ; Yang Gao ; Shuchang Zhou ; Xiance Si ; Decheng Dai

Abstract: In Community question answering (QA) sites, malicious users may provide deceptive answers to promote their products or services. It is important to identify and filter out these deceptive answers. In this paper, we first solve this problem with the traditional supervised learning methods. Two kinds of features, including textual and contextual features, are investigated for this task. We further propose to exploit the user relationships to identify the deceptive answers, based on the hypothesis that similar users will have similar behaviors to post deceptive or authentic answers. To measure the user similarity, we propose a new user preference graph based on the answer preference expressed by users, such as “helpful” voting and “best answer” selection. The user preference graph is incorporated into traditional supervised learning framework with the graph regularization technique. The experiment results demonstrate that the user preference graph can indeed help improve the performance of deceptive answer prediction.

3 0.82293767 241 acl-2013-Minimum Bayes Risk based Answer Re-ranking for Question Answering

Author: Nan Duan

Abstract: This paper presents two minimum Bayes risk (MBR) based Answer Re-ranking (MBRAR) approaches for the question answering (QA) task. The first approach re-ranks single QA system’s outputs by using a traditional MBR model, by measuring correlations between answer candidates; while the second approach reranks the combined outputs of multiple QA systems with heterogenous answer extraction components by using a mixture model-based MBR model. Evaluations are performed on factoid questions selected from two different domains: Jeopardy! and Web, and significant improvements are achieved on all data sets.

4 0.79725015 60 acl-2013-Automatic Coupling of Answer Extraction and Information Retrieval

Author: Xuchen Yao ; Benjamin Van Durme ; Peter Clark

Abstract: Information Retrieval (IR) and Answer Extraction are often designed as isolated or loosely connected components in Question Answering (QA), with repeated overengineering on IR, and not necessarily performance gain for QA. We propose to tightly integrate them by coupling automatically learned features for answer extraction to a shallow-structured IR model. Our method is very quick to implement, and significantly improves IR for QA (measured in Mean Average Precision and Mean Reciprocal Rank) by 10%-20% against an uncoupled retrieval baseline in both document and passage retrieval, which further leads to a downstream 20% improvement in QA F1.

5 0.76807845 218 acl-2013-Latent Semantic Tensor Indexing for Community-based Question Answering

Author: Xipeng Qiu ; Le Tian ; Xuanjing Huang

Abstract: Retrieving similar questions is very important in community-based question answering(CQA) . In this paper, we propose a unified question retrieval model based on latent semantic indexing with tensor analysis, which can capture word associations among different parts of CQA triples simultaneously. Thus, our method can reduce lexical chasm of question retrieval with the help of the information of question content and answer parts. The experimental result shows that our method outperforms the traditional methods.

6 0.75199938 141 acl-2013-Evaluating a City Exploration Dialogue System with Integrated Question-Answering and Pedestrian Navigation

7 0.72120076 292 acl-2013-Question Classification Transfer

8 0.70595235 329 acl-2013-Statistical Machine Translation Improves Question Retrieval in Community Question Answering via Matrix Factorization

9 0.69523931 290 acl-2013-Question Analysis for Polish Question Answering

10 0.69018549 291 acl-2013-Question Answering Using Enhanced Lexical Semantic Models

11 0.67616421 239 acl-2013-Meet EDGAR, a tutoring agent at MONSERRATE

12 0.6626327 272 acl-2013-Paraphrase-Driven Learning for Open Question Answering

13 0.65282315 254 acl-2013-Multimodal DBN for Predicting High-Quality Answers in cQA portals

14 0.61730367 169 acl-2013-Generating Synthetic Comparable Questions for News Articles

15 0.5381211 168 acl-2013-Generating Recommendation Dialogs by Extracting Information from User Reviews

16 0.53496379 387 acl-2013-Why-Question Answering using Intra- and Inter-Sentential Causal Relations

17 0.46974006 158 acl-2013-Feature-Based Selection of Dependency Paths in Ad Hoc Information Retrieval

18 0.44966465 230 acl-2013-Lightly Supervised Learning of Procedural Dialog Systems

19 0.42828095 95 acl-2013-Crawling microblogging services to gather language-classified URLs. Workflow and case study

20 0.42660415 373 acl-2013-Using Conceptual Class Attributes to Characterize Social Media Users


similar papers computed by lda model

lda for this paper:

topicId topicWeight

[(0, 0.04), (6, 0.028), (11, 0.061), (15, 0.021), (24, 0.061), (26, 0.05), (28, 0.011), (35, 0.1), (42, 0.031), (48, 0.049), (64, 0.018), (69, 0.288), (70, 0.063), (88, 0.027), (90, 0.026), (95, 0.063)]

similar papers list:

simIndex simValue paperId paperTitle

same-paper 1 0.76033902 266 acl-2013-PAL: A Chatterbot System for Answering Domain-specific Questions

Author: Yuanchao Liu ; Ming Liu ; Xiaolong Wang ; Limin Wang ; Jingjing Li

Abstract: In this paper, we propose PAL, a prototype chatterbot for answering non-obstructive psychological domain-specific questions. This system focuses on providing primary suggestions or helping people relieve pressure by extracting knowledge from online forums, based on which the chatterbot system is constructed. The strategies used by PAL, including semantic-extension-based question matching, solution management with personal information consideration, and XML-based knowledge pattern construction, are described and discussed. We also conduct a primary test for the feasibility of our system.

2 0.74866438 382 acl-2013-Variational Inference for Structured NLP Models

Author: David Burkett ; Dan Klein

Abstract: unkown-abstract

3 0.52998227 169 acl-2013-Generating Synthetic Comparable Questions for News Articles

Author: Oleg Rokhlenko ; Idan Szpektor

Abstract: We introduce the novel task of automatically generating questions that are relevant to a text but do not appear in it. One motivating example of its application is for increasing user engagement around news articles by suggesting relevant comparable questions, such as “is Beyonce a better singer than Madonna?”, for the user to answer. We present the first algorithm for the task, which consists of: (a) offline construction of a comparable question template database; (b) ranking of relevant templates to a given article; and (c) instantiation of templates only with entities in the article whose comparison under the template’s relation makes sense. We tested the suggestions generated by our algorithm via a Mechanical Turk experiment, which showed a significant improvement over the strongest baseline of more than 45% in all metrics.

4 0.52709681 272 acl-2013-Paraphrase-Driven Learning for Open Question Answering

Author: Anthony Fader ; Luke Zettlemoyer ; Oren Etzioni

Abstract: We study question answering as a machine learning problem, and induce a function that maps open-domain questions to queries over a database of web extractions. Given a large, community-authored, question-paraphrase corpus, we demonstrate that it is possible to learn a semantic lexicon and linear ranking function without manually annotating questions. Our approach automatically generalizes a seed lexicon and includes a scalable, parallelized perceptron parameter estimation scheme. Experiments show that our approach more than quadruples the recall of the seed lexicon, with only an 8% loss in precision.

5 0.52508849 2 acl-2013-A Bayesian Model for Joint Unsupervised Induction of Sentiment, Aspect and Discourse Representations

Author: Angeliki Lazaridou ; Ivan Titov ; Caroline Sporleder

Abstract: We propose a joint model for unsupervised induction of sentiment, aspect and discourse information and show that by incorporating a notion of latent discourse relations in the model, we improve the prediction accuracy for aspect and sentiment polarity on the sub-sentential level. We deviate from the traditional view of discourse, as we induce types of discourse relations and associated discourse cues relevant to the considered opinion analysis task; consequently, the induced discourse relations play the role of opinion and aspect shifters. The quantitative analysis that we conducted indicated that the integration of a discourse model increased the prediction accuracy results with respect to the discourse-agnostic approach and the qualitative analysis suggests that the induced representations encode a meaningful discourse structure.

6 0.5246278 318 acl-2013-Sentiment Relevance

7 0.52248383 341 acl-2013-Text Classification based on the Latent Topics of Important Sentences extracted by the PageRank Algorithm

8 0.5208407 224 acl-2013-Learning to Extract International Relations from Political Context

9 0.51976711 215 acl-2013-Large-scale Semantic Parsing via Schema Matching and Lexicon Extension

10 0.51962239 99 acl-2013-Crowd Prefers the Middle Path: A New IAA Metric for Crowdsourcing Reveals Turker Biases in Query Segmentation

11 0.51828283 185 acl-2013-Identifying Bad Semantic Neighbors for Improving Distributional Thesauri

12 0.51790625 233 acl-2013-Linking Tweets to News: A Framework to Enrich Short Text Data in Social Media

13 0.51756203 159 acl-2013-Filling Knowledge Base Gaps for Distant Supervision of Relation Extraction

14 0.51689202 158 acl-2013-Feature-Based Selection of Dependency Paths in Ad Hoc Information Retrieval

15 0.5168705 291 acl-2013-Question Answering Using Enhanced Lexical Semantic Models

16 0.51672375 85 acl-2013-Combining Intra- and Multi-sentential Rhetorical Parsing for Document-level Discourse Analysis

17 0.51660204 194 acl-2013-Improving Text Simplification Language Modeling Using Unsimplified Text Data

18 0.51633382 17 acl-2013-A Random Walk Approach to Selectional Preferences Based on Preference Ranking and Propagation

19 0.51584899 329 acl-2013-Statistical Machine Translation Improves Question Retrieval in Community Question Answering via Matrix Factorization

20 0.51575786 172 acl-2013-Graph-based Local Coherence Modeling