[From nobody Thu Oct 1 11:00:52 2009 Delivered-To: cjac@colliertech.org Received: by 10.204.66.193 with SMTP id o1cs188963bki; Mon, 21 Sep 2009 19:14:59 -0700 (PDT) Received: by 10.114.251.14 with SMTP id y14mr634593wah.144.1253585698177; Mon, 21 Sep 2009 19:14:58 -0700 (PDT) Return-Path: <cl-announce-bounces@mailman1.u.washington.edu> Received: from mxout1.cac.washington.edu (mxout1.cac.washington.edu [140.142.32.134]) by mx.google.com with ESMTP id 9si1367384pxi.42.2009.09.21.19.14.57; Mon, 21 Sep 2009 19:14:58 -0700 (PDT) Received-SPF: pass (google.com: domain of cl-announce-bounces@mailman1.u.washington.edu designates 140.142.32.134 as permitted sender) client-ip=140.142.32.134; Authentication-Results: mx.google.com; spf=pass (google.com: domain of cl-announce-bounces@mailman1.u.washington.edu designates 140.142.32.134 as permitted sender) smtp.mail=cl-announce-bounces@mailman1.u.washington.edu Received: from mailman1.u.washington.edu (mailman1.u.washington.edu [140.142.17.220]) by mxout1.cac.washington.edu (8.14.3+UW09.06/8.14.3+UW09.05) with ESMTP id n8M2EuCb013416 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO); Mon, 21 Sep 2009 19:14:57 -0700 Received: from mailman1.u.washington.edu (localhost [127.0.0.1]) by mailman1.u.washington.edu (8.14.3+UW09.06/8.14.3+UW09.05) with ESMTP id n8M2EuMP031909; Mon, 21 Sep 2009 19:14:56 -0700 Received: from mxi3.u.washington.edu (mxi3.u.washington.edu [140.142.32.176]) by mailman1.u.washington.edu (8.14.3+UW09.06/8.14.3+UW09.05) with ESMTP id n8M2EtGF031896; Mon, 21 Sep 2009 19:14:55 -0700 Received: from mxout4.cac.washington.edu (mxout4.cac.washington.edu [140.142.33.19]) by mxi3.u.washington.edu (8.14.3+UW09.06/8.14.3+UW09.09) with ESMTP id n8M2EsLH020816 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=OK); Mon, 21 Sep 2009 19:14:54 -0700 Received: from smtp.washington.edu (smtp.washington.edu [140.142.32.139]) by mxout4.cac.washington.edu (8.14.3+UW09.06/8.14.3+UW09.05) with ESMTP id n8M2Er26013426 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=OK); Mon, 21 Sep 2009 19:14:53 -0700 X-Auth-Received: from [192.168.1.102] (c-24-16-110-130.hsd1.wa.comcast.net [24.16.110.130]) (authenticated authid=fxia) by smtp.washington.edu (8.14.3+UW09.06/8.14.3+UW09.05) with ESMTP id n8M2Eq4L024778 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NOT); Mon, 21 Sep 2009 19:14:53 -0700 Message-ID: <4AB8331A.7060703@u.washington.edu> Date: Mon, 21 Sep 2009 19:14:50 -0700 From: Fei Xia <fxia@u.washington.edu> User-Agent: Thunderbird 1.5 (Windows/20051201) MIME-Version: 1.0 To: UW/MS symposium organizers <sympos@microsoft.com>, uw-ai@cs.washington.edu, Linggrads <linggrads@u.washington.edu>, clma-students@u.washington.edu, cl-announce@u.washington.edu Content-Type: text/plain; charset=ISO-8859-1; format=flowed X-PMX-Version: 5.5.8.383112, Antispam-Engine: 2.7.2.376379, Antispam-Data: 2009.9.22.20327 X-Uwash-Spam: Gauge=IIIIIIII, Probability=8%, Report=' BODY_SIZE_4000_4999 0, BODY_SIZE_5000_LESS 0, BODY_SIZE_7000_LESS 0, __CP_URI_IN_BODY 0, __CT 0, __CTE 0, __CT_TEXT_PLAIN 0, __HAS_LIST_HEADER 0, __HAS_LIST_HELP 0, __HAS_LIST_SUBSCRIBE 0, __HAS_LIST_UNSUBSCRIBE 0, __HAS_MSGID 0, __MIME_TEXT_ONLY 0, __MIME_VERSION 0, __MOZILLA_MSGID 0, __SANE_MSGID 0, __TO_MALFORMED_2 0, __URI_NS , __USER_AGENT 0' Cc: Subject: [Cl-announce] the next UW/MS symposium on Oct 9 X-BeenThere: cl-announce@u.washington.edu X-Mailman-Version: 2.1.5 Precedence: list List-Id: "Announcements/discussion re: UW compling lab" <cl-announce.u.washington.edu> List-Unsubscribe: <https://mailman1.u.washington.edu/mailman/listinfo/cl-announce>, <mailto:cl-announce-request@mailman1.u.washington.edu?subject=unsubscribe> List-Archive: <https://mailman1.u.washington.edu/mailman/private/cl-announce> List-Post: <mailto:cl-announce@u.washington.edu> List-Help: <mailto:cl-announce-request@mailman1.u.washington.edu?subject=help> List-Subscribe: <https://mailman1.u.washington.edu/mailman/listinfo/cl-announce>, <mailto:cl-announce-request@mailman1.u.washington.edu?subject=subscribe> Sender: cl-announce-bounces@mailman1.u.washington.edu Errors-To: cl-announce-bounces@mailman1.u.washington.edu Content-Transfer-Encoding: quoted-printable Hi, everyone, The next UW/MS symposium will be at Microsoft on Oct 9. The talks=20 start at 3pm, not 3:30pm. More detail is below. Hope to see you there. -Fei %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Location: Microsoft building 99 Room 1919 How to get there: http://research.microsoft.com/en-us/labs/redmond/visit.aspx Announcing the nineteenth Symposium in Computational Linguistics=20 sponsored by the UW Departments of Linguistics, Electrical Engineering,=20 and Computer Science, Microsoft Research, and UW alumni at Microsoft.=20 Come take advantage of this opportunity to connect with the=20 computational linguistics community at Microsoft and the University of=20 Washington. This is a regular opportunity for computational linguists at=20 the University of Washington and at Microsoft to discuss topics in the=20 field and to connect in a friendly informal atmosphere. This time the symposium consists of three talks from UW summer interns=20 (and their MS mentors), followed by an informal reception. =20 Hitting the Right Paraphrases in Good Time Stanley Kok (CSE) and Chris Brockett (MSR-NLP) =20 We present a random-walk-based approach to extracting paraphrases from=20 bilingual parallel corpora. The corpora are represented as a graph in=20 which a node corresponds to a phrase, and an edge exists between two=20 nodes their corresponding phrases are aligned. We sample random walks=20 to compute the average number of steps it takes to reach a ranking of=20 paraphrases with better ones being "closer" to the phrase of interest. =20 This approach allows "feature" nodes that represent domain knowledge to=20 be easily incorporated into the graph, and incorporates techniques to=20 prevent the graph from growing too large for efficiency. Current=20 state-of-the-art approaches, by contrast, require the graph to be=20 bipartite, are limited to finding paraphrases that are of length two=20 away from a phrase, and do not generally permit easy incorporation of=20 domain knowledge into the graph. Manual evaluation of generated output=20 shows that this approach outperforms state-of-the-art. =20 =20 Toward the Twuring Test: Conversation Modeling using Twitter Alan Ritter (CSE) and Colin Cherry (MSR-NLP) =20 The growing popularity of social media has had an interesting=20 side-effect for language researchers: services such as Twitter have=20 resulted in people having instant-messenger-style conversations using a=20 public medium, where anyone can observe. This creates a unique=20 opportunity to collect, study, and model large-scale conversation data.=20 We present a method for mining conversations from Twitter's public feed.=20 The resulting conversation corpus, which will be made publicly=20 available, has more than 1.3 million conversations, 75 thousand of which=20 have more than 5 turns, providing a rich resource for the study of both=20 Twitter and internet chat. Furthermore, we present several methods that=20 attempt to model the flow of conversation by discovering latent classes=20 over Tweets. We show that a repurposed content model (Barzilay and Lee=20 2004) can discover meaningful dialogue acts, such as "question" and=20 "comment", which indicate not only the role a Tweet plays in its=20 conversation, but also the sorts of Tweets that are likely to follow.=20 This model is improved and extended by employing a Bayesian=20 sampling-based approach, allowing us to model a conversation's topic,=20 and to introduce sparse priors during learning. =20 =20 =20 Joint Inference for Knowledge Extraction from Biomedical Literature Hoifung Poon (CSE) and Lucy Vanderwende (MSR-NLP) Automatically extracting knowledge from online repositories (e.g.,=20 PubMed) holds the promise of dramatically speeding up biomedical=20 research and drug design, and represents an outstanding example for the=20 great vision of knowledge extraction from the Web. After initially=20 focusing on entity recognition and binary interaction for protein, the=20 community has recently shifted their attention towards the more=20 ambitious goal of recognizing complex, nested event structures, which=20 are ubiquitous in the literature. However, the state-of-the-art systems=20 still adopt a pipeline architecture and fail to leverage the relational=20 structures among candidate entities for mutual disambiguation. In this=20 paper, we present the first joint approach for bioevent extraction that=20 obtains state-of-the-art results. Our system is based on Markov logic=20 and jointly predicts events and their arguments. We evaluated it using=20 the BioNLP-09 Shared Task and compared it to the participating systems.=20 Experimental results demonstrate the advantage of our approach. _______________________________________________ Cl-announce mailing list Cl-announce@u.washington.edu https://mailman1.u.washington.edu/mailman/listinfo/cl-announce ]