1 / 16

Survey Analysis

Survey Analysis. An attempt to develop an Intuition of Semantic Relatedness. Outline. Motivation Survey framework Analysis. Motivation. Semantic Relatedness – broad/subjective concept Given a pair of words – Are they related? If so, to what extent?

nijole
Download Presentation

Survey Analysis

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Survey Analysis An attempt to develop an Intuition of Semantic Relatedness

  2. Outline • Motivation • Survey framework • Analysis

  3. Motivation • Semantic Relatedness – broad/subjective concept • Given a pair of words – • Are they related? • If so, to what extent? • What is the kind of relationship between them? • Answer varies from person to person – depends on his background, culture, work domain etc. • Example: Apple - Computer

  4. Existing Datasets • Rubenstein & Goodenough (1965) – 65 English noun pairs (RG - 65) • Miller and Charles (1991) – subset of RG-65, 30 English noun pairs (MC - 30) • Finkelstein et al. (2002) – 353 word pairs (Fin1-153 and Fin2-200) • Yang and Powers (2006) – 130 verb pairs (YP-130)

  5. Problems with current datasets • Part of speech limitation • Focus on semantic similarity instead of relatedness • Size of dataset usually very small. Constructed manually. Labor intensive. • Only general terms are included. Lack of domain specific terms • Provides no insight into the type of SR

  6. Survey Framework • Was created using 30 word pairs from Miller and Charles (1991) dataset • Participants were asked to rate the relatedness on a scale of 0 – 4, 0 being not related at all and 4 being highly related • They were also asked to specify the kind of relationship • They were made aware of the fact that 2 words may be related in a variety of ways – Synonymy, Antonymy, Frequent association, is a, part of, domain related etc.

  7. Survey Framework • Was conducted among students of IIT Bombay (particularly with a computer science & linguistics background) • 55 students participated in the survey • Was created using Java Servlet and Tomcat container

  8. Screen Shot

  9. Results

  10. Results

  11. Results

  12. Graph

  13. Correlation Coefficient Correlation between MC new and original = 0.91 – quite strong

  14. Graph

  15. Graph

  16. Graph

More Related