Google Inc.  and the Linguistic Data         Consortium (LDC) have collaborated to develop new         syntactically-annotated language resources that enable computers         to better understand human language. The project, funded through a gift from         Google in 2010, has resulted in the development of the English Web
          Treebank LDC2012T13 containing over 250,000 words of         weblogs, newsgroups, email, reviews and question-answers         manually annotated for syntactic structure. This resource will         allow language technology researchers to develop and evaluate         the robustness of parsing methods in various new web domains. It         was used in the 2012 shared task on parsing English web text for         the First           Workshop on Syntactic Analysis of Non-Canonical Language           (SANCL) which took place at NAACL-HLT in Montreal on June         8, 2012. The English Web Treebank is available to the research         community through LDC’s
          Catalog.
       
      Natural language processing (NLP) is a field of         computational linguistic research concerned with the         interactions between human language and computers. Parsing is a         discipline within NLP in which computers analyze text and         determine its syntactic structure. While syntactic parsing is         already practically useful, Google funded this effort to help         the research community develop better parsers for web text. The         web texts collected and annotated by LDC provide new, diverse         data for training parsing systems.
       
      Google chose LDC for         this work based on the Consortium’s experience in developing and         creating syntactic annotations, also known as treebanks.         Treebanks are critically important to parsing research since         they provide human-analyzed sentence structures that facilitate         training and testing scenarios in NLP research. This work         extends the existing relationship between LDC and Google.  LDC has published four         other Google-developed data sets in the past six years: English,         Chinese, Japanese and European language n-grams used principally         for language modeling.
       
          

    LDC’s 20th
        Anniversary Workshop is rapidly approaching! The event will take         place on the University of Pennsylvania’s campus on September         6-7, 2012.
             
      Workshop themes
        include: the developments in human language technologies and         associated resources that have brought us to our current state;         the language resources required by the technical approaches         taken and the impact of these resources on HLT progress; the         applications of HLT and resources to other disciplines including         law, medicine, economics, the political sciences and psychology;         the impact of HLTs and related technologies on linguistic         analysis and novel approaches in fields as widespread as         phonetics, semantics, language documentation, sociolinguistics         and dialect geography; and the impact of any of these         developments on the ways in which language resources are         created, shared and exploited and on the specific resources         required.
             
      Please read more here.
       
     
   
     
       

    Applications are
        now being accepted through September 17, 2012, 11:59PM EST for         the Fall 2012 LDC Data Scholarship program! The LDC Data         Scholarship program provides university students with access to         LDC data at no-cost. During previous program cycles, LDC has         awarded no-cost copies of LDC data to over 20 individual         students and student research groups.
       
        This program is open to students pursuing both undergraduate and         graduate studies in an accredited college or university. LDC         Data Scholarships are not restricted to any particular field of         study; however, students must demonstrate a well-developed         research agenda and a bona fide inability to pay. The selection         process is highly competitive.
       
        The application consists of two parts:
       
              (1) Data Use Proposal. Applicants must submit a proposal         describing their intended use of the data. The proposal should         state which data the student plans to use and how the data will         benefit their research project as well as information on the         proposed methodology or algorithm.
           
              Applicants should consult the LDC Corpus Catalog         for a complete list of data distributed by LDC. Due to certain         restrictions, a handful of LDC corpora are restricted to members         of the Consortium. Applicants are advised to select a maximum of         one to two datasets; students may apply for additional datasets         during the following cycle once they have completed processing         of the initial datasets and publish or present work in some         juried venue.
           
              (2) Letter of Support. Applicants must submit one letter         of support from their thesis adviser or department chair. The         letter must confirm that the department or university lacks the         funding to pay the full Non-member Fee for the data and verify         the student's need for data.
                   
        For further information on application materials and program         rules, please visit the LDC Data Scholarship         page.
       
        Students can email their applications to the LDC Data Scholarship program. Decisions will be         sent by email from the same address.
       
        The deadline for the Fall 2012 program cycle is September 17,         2012, 11:59PM EST.
       
   
               

         
      As part
        of our 20th anniversary celebration, the coming newsletters will         include features that provide an overview of the broad range of         LDC’s activities. To begin, we'll examine the Heterogeneous         Audio Visual Internet Collection (HAVIC), one of the many         projects handled by LDC’s Collection/Annotation Group led by         Senior Associate Director Stephanie Strassel.
       
      Under the
        supervision of Senior Research Coordinator Amanda
          Morris, the HAVIC team is developing a large corpus of           unconstrained multimedia data drawn from user-generated videos           on the web and annotated for a variety of features. The HAVIC           corpus has been designed with an eye toward providing           increased challenges for both acoustic and video processing           technologies, focusing on multi-dimensional variation inherent           in user-generated content. Over the past three years the           corpus has provided training, development and test data for           the NIST TRECVID Multimedia Event Detection (MED) Evaluation           Track, whose goal is to assemble core detection technologies           into a system that can search multimedia recordings for           user-defined events based on pre-computed             metadata.
       
      For each
        MED evaluation, LDC and NIST have collaborated to define many         new events, including things like “making a cake” or “assembling         a shelter”. Each event requires an Event Kit, consisting of a         textual description of the event’s properties along with a few         exemplar videos depicting the event. A large team of LDC data         scouts search for videos that contain each event, along with         videos that are only indirectly or superficially related to         defined events plus background videos that are unrelated to any         defined event. After finding suitable content, data scouts label         each video for a variety of features including the presence of         audio, visual or text evidence that a particular event has         occurred. This work is done using LDC’s AScout framework,         consisting of a browser plug-in, a database backend and         processing scripts that together permit data scouts to         efficiently search for videos, annotate the multimedia content,         and initiate download and post-processing of the data. Collected         data is converted to MPEG-4 format, with h.264 video encoding         and AAC audio encoding, and the original video resolution and         audio/video bitrates are retained.
       
      To date,
        LDC has collected and labeled well over 100,000 videos as part         of the HAVIC Project, and the corpus will ultimately comprise         thousands of hours of labeled data. Look for portions of the         corpus to appear among LDC’s future releases.
       
           
           

         
      (1)English
          Web Treebank was developed by the Linguistic Data         Consortium (LDC) with funding through a gift from Google Inc. It         consists of over 250,000 words of English weblogs, newsgroups,         email, reviews and question-answers manually annotated for         syntactic structure and is designed to allow language technology         researchers to develop and evaluate the robustness of parsing         methods in those web domains.
       
      This release
        contains 254,830 word-level tokens and 16,624 sentence-level         tokens of webtext in 1174 files annotated for sentence- and         word-level tokenization, part-of-speech, and syntactic         structure. The data is roughly evenly divided across five         genres: weblogs, newsgroups, email, reviews, and         question-answers. The files were manually annotated following         the sentence-level tokenization guidelines for web text and the         word-level tokenization guidelines developed for English         treebanks in the DARPA GALE project. Only text         from the subject line and message body of posts, articles,         messages and question-answers were collected and annotated.
       
      English Web
        Treebank is distributed via web download.
       
      2012 Subscription
        Members will receive two copies of this data on disc. 2012         Standard Members may request a copy as part of their 16 free         membership corpora. Non-members may license this data by         completing the LDC User Agreement for Non-members. The agreement can         be faxed to +1 215 573 2175 or scanned and emailed to this         address. The first fifty copies of this publication are being         made available at no charge. After the first fifty copies are         distributed, the non-member fee of US$175 applies.
       
           
   
       

   
      (2) GALE
          Phase 2 Arabic Broadcast Conversation Parallel Text Part 2         was developed by LDC. Along with other corpora, the parallel         text in this release comprised training data for Phase 2 of the         DARPA GALE (Global Autonomous Language Exploitation) Program.         This corpus contains Modern Standard Arabic source text and         corresponding English translations selected from broadcast         conversation (BC) data collected by LDC between 2004 and 2007         and transcribed by LDC or under its direction.
       
      GALE Phase
        2 Arabic Broadcast Conversation Parallel Text Part 2 includes 29         source-translation document pairs, comprising 169,488 words of         Arabic source text and its English translation. Data is drawn         from eight distinct Arabic programs broadcast between 2004 and         2007 from Aljazeera, a regional broadcast programmer based in         Doha, Qatar; and Nile TV, an Egyptian broadcaster. The programs         in this release focus on current events topics.
       
      The files
        in this release were transcribed by LDC staff and/or         transcription vendors under contract to LDC in accordance with         the Quick Rich Transcription guidelines         developed by LDC. Transcribers indicated sentence boundaries in         addition to transcribing the text. Data was manually selected         for translation according to several criteria, including         linguistic features, transcription features and topic features.         The transcribed and segmented files were then reformatted into a         human-readable translation format and assigned to translation         vendors. Translators followed LDC's Arabic to English         translation guidelines. Bilingual LDC staff performed quality         control procedures in the completed translations.
       
      GALE Phase
        2 Arabic Broadcast Conversation Parallel Text Part 2 is         distributed via web download.
       
      2012 Subscription
        Members will receive two copies of this data on disc. 2012         Standard Members may request a copy as part of their 16 free         membership corpora. Non-members may license this data for         US$1750.
     
   
   

         
      (3)
        Spanish
          TimeBank 1.0 was developed by researchers at Barcelona Media and consists of         Spanish texts in the AnCora corpus annotated with         temporal and event information according to the TimeML specification language.
       
      Spanish TimeBank
        1.0 contains stand-off annotations for 210 documents with over         75,800 tokens (including punctuation marks) and 68,000 tokens         (excluding punctuation). The source documents are news stories         and fiction from the AnCora corpus.
       
      The AnCora
        corpus is the largest multilayer annotated corpus of Spanish and         Catalan. AnCora contains 400,000 words in Spanish and 275,000         words in Catalan. The AnCora documents are annotated on many         linguistic levels including structure, syntax, dependencies,         semantics and pragmatics. That information is not included in         this release, but it can be mapped to the present annotations.         The corpus is freely available from the Centre de Llenguatge i Computació (CLiC).
       
      Spanish TimeBank
        1.0 is distributed by web download.
       
      2012 Subscription
        Members will receive two copies of this data on disc. 2012         Standard Members may request a copy as part of their 16 free         membership corpora. Non-members may license this data by         completing the LDC User Agreement for Non-members. The agreement         can be faxed to +1 215 573 2175 or scanned and emailed to this         address. The publication is being made available at no charge.
         

In this newsletter: