WP8 2013.01.16
The goal of the meeting is to discuss how to proceed with wp8 before the last deliverable d8.3.
Agenda
The data to be covered in d8.3 • additions of the lexicon entries • transform data to RDF triples • multilingual transformation
Harmonize the painting queries demo and the ontotext demo
• coverage of the query and answer grammar
Book chapter • we are suppose to submit 20 pages by the 31st of March
Evaluations and tracking • time lines
Participants: AR, DD, MM, OC, MD
[1/16/13 1:58:23 PM] Dana Dannélls added Aarne Ranta to this conversation [1/16/13 1:58:29 PM] Dana Dannélls added Mariana Damova to this conversation [1/16/13 1:58:32 PM] Dana Dannélls added Mateva to this conversation [1/16/13 1:58:37 PM] Dana Dannélls: Call started [1/16/13 1:59:16 PM] Olga Caprotti: we are 1 minute early [1/16/13 2:00:09 PM] Mariana Damova: Hi: am am going to join in a bit later ... [1/16/13 2:00:24 PM] Dana Dannélls: ok [1/16/13 2:04:19 PM] Olga Caprotti: DD: added wikipedia data txt to svn, ontotext will need to add this data to the KRI, is it ok? [1/16/13 2:04:55 PM] Mariana Damova: No, we have another better suggestion, we carry a service [1/16/13 2:05:05 PM] Mariana Damova: http://factforge.net [1/16/13 2:05:17 PM] Mariana Damova: http://www.ontotext.com/factforge [1/16/13 2:05:27 PM] Olga Caprotti: MM: we will have to make RDF triples for the new data [1/16/13 2:06:04 PM] Mariana Damova: the FactForge has already triples of DBpedia, which is the RDF version of Wikipedia and 9 more datasets [1/16/13 2:06:22 PM] Mariana Damova: we also have methods to link repositories [1/16/13 2:06:32 PM] Mariana Damova: I was about to send you SPARQL queries [1/16/13 2:06:52 PM] Mariana Damova: whihc are showing what kind of information they can carry [1/16/13 2:06:59 PM] Mariana Damova: they can bring [1/16/13 2:07:01 PM] Mariana Damova: back [1/16/13 2:08:46 PM] Olga Caprotti: OC: does DBpedia have same as wikipedia content? if so, you have RDF already [1/16/13 2:09:12 PM] Olga Caprotti: MM: I'd like demo working by end of january, including data specific samples [1/16/13 2:09:33 PM] Olga Caprotti: AR: we can also then work on optimizing the grammars for that demo [1/16/13 2:10:04 PM] Olga Caprotti: AR: we can translate museum names for this data set [1/16/13 2:10:48 PM] Olga Caprotti: AR. entity strings are now not translated in the grammar [1/16/13 2:11:17 PM] Olga Caprotti: AR, one would like to translate "City Museum" [1/16/13 2:11:39 PM] Olga Caprotti: DD: i offer to translate museum names, painters [1/16/13 2:11:51 PM] Olga Caprotti: AR: translate painting names is hard generally [1/16/13 2:12:28 PM] Olga Caprotti: AR: or transliterated [1/16/13 2:13:04 PM] Olga Caprotti: AR if we have such translations in Wikipedia we can use them [1/16/13 2:13:31 PM] Olga Caprotti: AR: MonaLisa becomes Gioconda in Italian [1/16/13 2:14:01 PM] Olga Caprotti: DD welcome Mariana [1/16/13 2:14:57 PM] Olga Caprotti: DD: tags for languages, a tag for eg PabloPicasso with Hebrew transliteration [1/16/13 2:15:33 PM] Olga Caprotti: DD for the entities we are translating on the RDF on the side [1/16/13 2:16:35 PM] Olga Caprotti: MM: so we take the RDF tags for the language we are using - hope this does not collide with Java [1/16/13 2:17:34 PM] Olga Caprotti: Agenda The data to be covered in d8.3 • additions of the lexicon entries • transform data to RDF triples • multilingual transformation Harmonize the painting queries demo and the ontotext demo • coverage of the query and answer grammar Book chapter • we are suppose to submit 20 pages by the 31st of March Evaluations and tracking • time lines [1/16/13 2:18:57 PM] Olga Caprotti: MD: what kind of data to use beside the goteborg data museum - from other data sets we have triples on artwork not on painting [1/16/13 2:20:09 PM] Mariana Damova: # Cities where paintings of Modigliani are located (PROTON) PREFIX dbpedia: PREFIX ff: PREFIX ptop: PREFIX pext: SELECT DISTINCT ?painting_l ?owner_l ?city_l WHERE { dbpedia:Amedeo_Modigliani pext:isAuthorOf ?painting . ?painting ptop:isOwnedBy ?owner ; ff:preferredLabel ?painting_l. ?owner ff:preferredLabel ?owner_l . ?owner ptop:locatedIn [ a pext:City ; ff:preferredLabel ?city_l ] } [1/16/13 2:20:23 PM] Mariana Damova: here is the query that you can try on http://factforge.net [1/16/13 2:20:25 PM] Olga Caprotti: AR: we must get the grammars done - especially the text generation -- for the evaluation task [1/16/13 2:21:17 PM] Olga Caprotti: AR: who will host the demo? [1/16/13 2:21:30 PM] Olga Caprotti: MD: ontotext if the demo is based on RDF [1/16/13 2:21:46 PM] Olga Caprotti: AR: when did you contact the museum ppl last? [1/16/13 2:21:57 PM] Olga Caprotti: DD: invited to my defence [1/16/13 2:22:33 PM] Olga Caprotti: AR: lets say we have a system that demos architecture and fuctionality for the deliverable, and make sure we expand the data for the flagship [1/16/13 2:24:40 PM] Olga Caprotti: MM: we can make russian and polish grammar, do not know about final quality - what is desirable [1/16/13 2:24:52 PM] Olga Caprotti: AR: full quality is what we want [1/16/13 2:25:20 PM] Olga Caprotti: DD: maybe a native speaker is not necessary to write the grammar. [1/16/13 2:25:31 PM] Olga Caprotti: AR: a native is necessary to evaluate the quality [1/16/13 2:26:04 PM] Olga Caprotti: MM: i was asked to evaluate a bulgarian grammar and i was very happy with it [1/16/13 2:26:48 PM] Aarne Ranta: check coverage of text data set [1/16/13 2:28:22 PM] Olga Caprotti: MD: which infrastructure will you use fo rthe data set? the relational kb or the rdf? [1/16/13 2:28:44 PM] Olga Caprotti: AR: at present to work on grammars we use relational but plan to use RDF in the final demo [1/16/13 2:29:10 PM] Olga Caprotti: MD: including the wikipedia data? [1/16/13 2:29:36 PM] Olga Caprotti: AR: we have some random instances that have to cover all possible cases of the text generation [1/16/13 2:29:49 PM] Olga Caprotti: AR: we will move to more realistic data afterwards [1/16/13 2:30:21 PM] Olga Caprotti: MM: can you send the data to me pls? [1/16/13 2:30:38 PM] Olga Caprotti: DD: i am working on that, will send it in a couple of days [1/16/13 2:32:17 PM] Olga Caprotti: DD: MD, you said we have 2 demos, we need only one. how are we going to harmonize the paintings on the ontotext demo? [1/16/13 2:32:45 PM] Olga Caprotti: AR: we have to implement a SPARQL on the ontotext demo [1/16/13 2:33:14 PM] Olga Caprotti: MD: we need time to invent the model on the transformation to SPARQL [1/16/13 2:33:24 PM] Olga Caprotti: MD: and time to implement it. [1/16/13 2:34:07 PM] Aarne Ranta: after January: finish SPARQL backend of UGOT query language [1/16/13 2:34:30 PM] Aarne Ranta: as the main method of harmonization [1/16/13 2:34:44 PM] Aarne Ranta: who is going to work with this? Maria? [1/16/13 2:35:13 PM] Mariana Damova: I am sorry, i had to hang up, because there was a lot of side noise ... [1/16/13 2:35:29 PM] Olga Caprotti: it is good now that you are gone [1/16/13 2:35:59 PM] Olga Caprotti: it might have been a doppler effect from your loudspeakers [1/16/13 2:36:51 PM] Olga Caprotti: MM: not deployed under the demo yet but we have been working on mapping from NL to SPARQL [1/16/13 2:37:24 PM] Olga Caprotti: MM: we need to generate lots of resources to get this working - created some scripts to do it automatically [1/16/13 2:37:38 PM] Olga Caprotti: MM: now blocks me the lack of RDF to experiment with [1/16/13 2:38:06 PM] Olga Caprotti: AR: can't you generate RDF from the samples currently in the grammar [1/16/13 2:38:37 PM] Olga Caprotti: AR: all the data is in the relational db [1/16/13 2:38:37 PM] Mariana Damova: back again [1/16/13 2:38:53 PM] Olga Caprotti: AR: enough to experiment with the semantics [1/16/13 2:39:36 PM] Olga Caprotti: DD: that example you sent me was based on a record from the goteborg city museum, was good [1/16/13 2:40:26 PM] Olga Caprotti: MM: descriptions are too sparse in current records, my NL is not generated if parts of the records are missing [1/16/13 2:41:10 PM] Olga Caprotti: DD: i added a file called gcm.haskell with more information about the fields that are required [1/16/13 2:41:28 PM] Olga Caprotti: DD: i also sent translations in english of all the fields [1/16/13 2:42:15 PM] Dana Dannélls: GCMDataPainting.hs [1/16/13 2:42:38 PM] Olga Caprotti: MM: will revise the data then - i have not used the new files - i will use these [1/16/13 2:47:11 PM] Olga Caprotti: AR: whether we would have obligatory painter and painting is given [1/16/13 2:48:04 PM] Olga Caprotti: AR: title of painting and name of painters - we could introduce anonymous or unknown, then is no problem also with sparse records [1/16/13 2:49:45 PM] Olga Caprotti: AR: if there is a Picaso then it will be a new painter [1/16/13 2:51:08 PM] Olga Caprotti: MM: the question and answer grammar will be the same? [1/16/13 2:51:11 PM] Olga Caprotti: AR: yes [1/16/13 2:51:28 PM] Olga Caprotti: MM: if we could do the mapping from names to IDs [1/16/13 2:51:43 PM] Olga Caprotti: MM: queries will be easier [1/16/13 2:55:48 PM] Olga Caprotti: MD: we can have normalization done beforehand - [1/16/13 2:56:24 PM] Olga Caprotti: OC: for painter names - we could cross check different languages if records are given in multiple languages (which i guess they are generally not) [1/16/13 2:56:59 PM] Olga Caprotti: DD: about the book chapter, mariana can you please upload the paper to the website? [1/16/13 2:57:11 PM] Olga Caprotti: AR: or use svn please to deal with versions [1/16/13 2:57:24 PM] Olga Caprotti: AR: pls MOLTO SVN under wp8 [1/16/13 2:57:45 PM] Olga Caprotti: MD: mon tue next week for the book chapter [1/16/13 2:58:25 PM] Olga Caprotti: DD: next evaluation and tracking, we need to send texts, then get feedback, then carry out correction. [1/16/13 2:58:55 PM] Olga Caprotti: AR: evaluation will go to molto wp10 deliverable, done early for us to improve based onf eedback [1/16/13 2:59:13 PM] Olga Caprotti: AR: russian and polish will also be improved by the evaluation [1/16/13 3:00:51 PM] Olga Caprotti: MM: about YAQL, i was thinking to map some functors to specific SPARQL constructs. [1/16/13 3:01:14 PM] Olga Caprotti: AR: whatever works, for haskell it was easy because haskell supports same abstractions [1/16/13 3:01:58 PM] Olga Caprotti: MM: how should we proceed wth doing this? [1/16/13 3:02:29 PM] Olga Caprotti: AR: we must do it from YAQL to SPARQL, we do not have to do it in jan, but afterwards [1/16/13 3:02:44 PM] Olga Caprotti: AR: kramisir promised to help [1/16/13 3:03:19 PM] Olga Caprotti: AR: we can book some meetings between you MM and krasimir [1/16/13 3:03:42 PM] Olga Caprotti: AR: I was wondering that YAQL is now under wp8/grammars [1/16/13 3:04:30 PM] Olga Caprotti: AR: its proper place is wp4 [1/16/13 3:04:57 PM] Olga Caprotti: AR: if you read the README file it says to create a symbolic link [1/16/13 3:07:28 PM] Olga Caprotti: AR: if anyone has done any work in this local copy of YAQL, then it should be ported to the wp4 folder [1/16/13 3:09:00 PM] Olga Caprotti: DD: would like to see the mapping text on the ontotext site [1/16/13 3:09:27 PM] Olga Caprotti: MM: hope by the end of the week to be able to create a complete demo
1. DD: added daa
- Printer-friendly version
- Login to post comments
- Slides
What links here
No backlinks found.