LT-based E-science at Språkbanken. Språkbanken kick-off January 2015

Transcription

1 LT-based E-science at Språkbanken Språkbanken kick-off January 2015

2 Definition of e-science E-Science (or escience) is computationally intensive science that is carried out in highly distributed network environments, or science that uses immense data sets that require grid computing; the term sometimes includes technologies that enable distributed collaboration, such as the Access Grid. Most of the research activities into e-science have focused on the development of new computational tools and infrastructures to support scientific discovery.

3 Infrastructures Language Technology SKO2nCjyT3 Digital humanities and social science Historical resources Modern resources Political Medical Historical Korp/Karp front end Methods Korp Karp

4 One example, Lärka Historical resources Application areas Front ends Modern resources Digital humanities and social science Political Medical Historical Korp/Karp front end Methods Korp Karp

5 Swe-Clarin Historical resources Application areas Front ends Modern resources Digital humanities and social science Political Medical Historical Korp/Karp front end Methods Korp Karp

6 Historical resources Application areas Front ends Modern resources Digital humanities and social science Political Medical Historical Korp/Karp front end Methods Korp Karp

7 The SB definition of e-science IT based research methodology With or without large amounts of data Corpus linguistics is a prominent example In the domain of digital humanities and social sciences

8 What has been done at SB? What are we working on? Words (multiwords) Relations Coordinations s Readability Twitter

9 MWE detection to improve parsing quality 1. Use lists as a basis for e.g., idioms, terminology and entities 2. Add reg. exp., pattern matching to find more MWEs 3. Perform Parsing Confirmed intution and previous experiments that prerecognizing MWEs improve parsing (by 16%). Figure from Boleda G. & Evert S.: Multiword Expressions: A pain in the neck of lexical semantics

10 Semantics in Storytelling in Swedish Fiction Relation extraction from Swedish Prose Fiction (SPF) List of relations NEE to extract names and aliases document center approach to link aliases names Extract sentences with min. 2 names. Detect relation Automatic detection would sign. improve coverage of relations

11 Semantics in Storytelling in Swedish Fiction Relation extraction from Swedish Prose Fiction (SPF) List of relations NEE to extract names and aliases document center approach to link aliases names Extract sentences with min. 2 names. Detect relation Automatic detection would sign. improve coverage of relations Relations between 2 males = red, between 2 females = green, otherwise blue.

12 Semantics in Storytelling in Swedish Fiction Relation extraction from Swedish Prose Fiction (SPF) List of relations NEE to extract names and aliases document center approach to link aliases names Extract sentences with min. 2 names. Detect relation Automatic detection would sign. improve coverage of relations Relations between 2 males = red, between 2 females = green, otherwise blue.

13 Swedish Psuedo Coordination (SPC) detection and change Verb pairs where the first is light åka och handla, gå och gifta sig, ringa och berätta Typical properties apply: E.g., both is not possible: jag både satt och läste No paraphrasing: Mona satt och hon läste. Try to classify SPCs from non- SPCs using these features False positives We think non-spc, algorithm guesses SPC. Relaxing drop in P/R Precision and recall for Blogmixen fara, resa, trilla, varda, stog, vända, testa, mejla, maila, kommentera, blogga, googla

14 Modeling SPF used as data set Modeling applied (Mallet) which parts of a document belong to which topic which part of any document belongs to topic i Link original resources to help validate topics

15 Readability of text All paragraphs assigned to topic i that are easy to read. Investigate different readability measures for text. Measures for English Swedish Readability measures are not very reliable when applied directly to Swedish texts.

16 Twitter Analysis around political debates Start with some hashtags, e.g., #pldebatt Find all tweets = core Train classifier to find related tweets Divide into known topics (from debate) CORE 1

17 Twitter Analysis around political debates Start with some hashtags, e.g., #pldebatt Find all tweets = core Train classifier to find related tweets Divide into known topics (from debate) 5 4 CORE

18 Twitter Analysis around political debates Start with some hashtags, e.g., #pldebatt Find all tweets = core Train classifier to find related tweets Divide into known topics (from debate) May October 5 4 CORE

19 Twitter Analysis around political debates Start with some hashtags, e.g., #pldebatt Find all tweets = core Train classifier to find related tweets Divide into known topics (from debate) May 10: 1: Digram jan björklund, attackera, allians, fusklapp frisyr, läcka, åkesson, vinna, slips, ord, siffra, dålig, sverige, analys, prata, tydlig, romson, jobba, önska nöjd T ex October T ex 5 4 CORE

20 Conclusions Future work There are many, many interesting things to do in the field of E-science Workshop on SB related activities for DHSS on April 17th Want to present your work? Come and join us!