Presentation is loading. Please wait.

Presentation is loading. Please wait.

Information Extraction. Extracting Information from Text System : When would you like to meet Peter? User : Let’s see, if I can, I’d like to meet him.

Similar presentations


Presentation on theme: "Information Extraction. Extracting Information from Text System : When would you like to meet Peter? User : Let’s see, if I can, I’d like to meet him."— Presentation transcript:

1 Information Extraction

2 Extracting Information from Text System : When would you like to meet Peter? User : Let’s see, if I can, I’d like to meet him on Tuesday.

3 Template Filling Partner: John Day: Tuesday Location: Mattin Ctr Topic: Coffee Time of Day: Morning I’d like to meet John on Tuesday in the Mattin Center coffee shop. We should discuss policies on coffee consumption in the computer science department. Anytime in the morning would be fine.

4 Template Filling Partner: John Day: Tuesday Location: Mattin Ctr Topic: Coffee Time of Day: Morning I’d like to meet John on Tuesday in the Mattin Center coffee shop. We should discuss policies on coffee consumption in the computer science department. Anytime in the morning would be fine. Or should this be 8am- 12pm?

5 How do we write general rules?  Finite State Machines –(regular expressions)  Extraction from partial parses  Full Parsing

6 Rules with Assigned Semantic Roles Meet with on. On I want to meet. Many patterns fill the same template slots.

7 How can we write these rules?  Manually enumerate them?  manufactures –Any other ways to automatically rewrite this?

8 Semantic Lexicons manufactures Laughter manufactures happiness. How could we avoid this problem?

9 Can we learn them? I’d like to meet John on Tuesday in the Mattin Center coffee shop. We should discuss policies on coffee consumption in the computer science department. Anytime in the morning would be fine. Partner: John Day: Tuesday Location: Mattin Ctr Topic: Coffee Time of Day: Morning

10 Can we learn them? I’d like to meet John on Tuesday in the Mattin Center coffee shop. We should discuss policies on coffee consumption in the computer science department. Anytime in the morning would be fine. Partner: John Day: Tuesday Location: Mattin Ctr Topic: Coffee Time of Day: Morning

11 Can we learn them? I’d like to meet on in the. We should discuss policies on. Anytime in would be fine. Partner: John Day: Tuesday Location: Mattin Ctr Topic: Coffee Time of Day: Morning

12 Generalization I’d like to meet on in the. We should discuss policies on. Anytime in would be fine. meet on in. discuss policies on anytime in How could we learn this?

13 How about if we have templates without text? Person : Mozart Birthyear : 1756 Can we gather text somehow?

14 Web Search to generate patterns Web pages w/“Mozart” “1756” Sentences with “Mozart” “1756” Substrings with “Mozart” “1756”

15 How can we pick good patterns?  Frequent ones may be too general  Infrequent ones not that useful  Want precise, specific ones Use held out templates to evaluate patterns

16 How about pages but no templates?  Have a set of pages marked as either on topic or off topic  Look for all possible patterns  Estimate which patterns are most likely to occur on a marked up page  Manually screen resulting patterns


Download ppt "Information Extraction. Extracting Information from Text System : When would you like to meet Peter? User : Let’s see, if I can, I’d like to meet him."

Similar presentations


Ads by Google