A Knowledge-Rich Approach to Understanding Text about .ppt

1、A Knowledge-Rich Approach to Understanding Text about Aircraft Systems,Peter Clark Lisbeth Duncan Heather Holmback Tom Jenkins John Thompson Boeing Engineering and Information Technology,Overview,Situation: Using a hand-built, airplane knowledge-base (KB) at Boeing (e.g., for concept-based search) G

2、oal: Extend this KB by processing text: from airplane training manuals directly from an aerospace engineer Target: airplane parts, connections, and behaviors,Challenge,Interpretation of text requires background knowledge author trying to convey a model to the reader only some parts of this are expli

3、citly stated in text rest is assumed to be already known to the reader,Approach,Use an aerospace knowledge base (KB): Taken from an earlier project and extended Contains general models about what might be said Contains general knowledge of aircraft and systems Doesnt know details of specific aircraf

4、t Text interpretation = specializing and refining this background knowledge,Constraints on the Scope,Working with restricted sublanguage of English Simple, declarative sentences Regular, “boring” style,“The hydraulic system supplies power to the rudder.” “A ripple damper smooths the pump pressure ou

5、tput.” “The EDP is on the left side of the engine main gearbox.” “The gearbox turns the EDP when the engine turns.”,Constrained domain (aircraft systems) Reduces vocabulary, background KB Domain of mechanical artifacts Simple statements about structure and behavior,KB,Approach,User specifies topic (

6、eg. “777-200-Airplane”),777-200-Airplane,“The hydraulic system supplies power to the rudder.”,Tail,Powered System,Rudder,Vertical Stabilizer,Power Source,Supplying,Fuselage,Wing,Flap,KB,parts,parts,source,sink,script,systems,parts,purpose,connects,agent,recipient,Approach,Then: Text NLP structure,77

7、7-200-Airplane,“The hydraulic system supplies power to the rudder.”,Tail,Powered System,Rudder,Vertical Stabilizer,Power Source,Supplying,Fuselage,Wing,Flap,Rudder,Hydraulic System,Supplying,Power,KB,?,recipient,object,agent,parts,parts,source,sink,script,systems,parts,purpose,connects,agent,recipie

8、nt,Approach,Then: Text NLP structure Match NLP structure with airplane repn.,Stage 1,Stage 2,777-200-Airplane,“The hydraulic system supplies power to the rudder.”,Tail,Powered System,Rudder,Vertical Stabilizer,Supplying,Fuselage,Wing,Flap,Rudder,Hydraulic System,Supplying,Power,KB,Hydraulic System,P

9、ower,Approach,Then: Text NLP structure Match NLP structure with airplane repn.Unify the matching structures.,recipient,object,agent,parts,parts,source,sink,script,systems,parts,purpose,connects,Stage 1,Stage 2,777-200-Airplane,Tail,Powered System,Vertical Stabilizer,Fuselage,Wing,Flap,KB,parts,parts

10、,source,sink,script,systems,parts,purpose,connects,Stage 2,The Knowledge Base,Rudder,Power Source,Supplying,agent,recipient,“The hydraulic system supplies power to the rudder.”,Rudder,Hydraulic System,Supplying,Power,recipient,object,agent,Stage 1,The Knowledge Base,500 concepts, 1500 axioms Mainly

11、airplane parts, behaviors, and connections Axioms converted to prototypes (graph structures),Inference: Building the Initial Airplane Representation,777-200-Airplane,Tail,Airplane,Fuselage,Tail,parts,Vertical Stabilizer,Horizontal Stabilizer,parts,Flight,Takeoff,Cruise,Land,Flight,behavior,Axioms in

12、 KB:,Airplane representation:,KB,777-200-Airplane,“The hydraulic system supplies power to the rudder.”,Tail,Powered System,Vertical Stabilizer,Fuselage,Wing,Flap,Rudder,Hydraulic System,Supplying,Power,KB,recipient,object,agent,parts,parts,source,sink,script,systems,parts,purpose,connects,Stage 1,St

13、age 1: Text to NLP Structure,Rudder,Power Source,Supplying,agent,recipient,Stage 1: Text to NLP Structure,Superficial linguistic variations normalized Add initial semantic interpretation,Commits to: overall parse/syntactic configuration, word senses May underspecify: some semantic relationships (e.g

14、., modifiers),“The hydraulic system supplies power to the rudder.”,777-200-Airplane,“The hydraulic system supplies power to the rudder.”,Tail,Powered System,Vertical Stabilizer,Fuselage,Wing,Flap,Rudder,Hydraulic System,Supplying,Power,KB,recipient,object,agent,parts,parts,source,sink,script,systems

15、,parts,purpose,connects,Stage 1,Rudder,Power Source,Supplying,agent,recipient,777-200-Airplane,Tail,Powered System,Vertical Stabilizer,Fuselage,Wing,Flap,parts,parts,source,sink,script,systems,parts,purpose,connects,Rudder,Power Source,Supplying,agent,recipient,“The hydraulic system supplies power t

16、o the rudder.”,Rudder,Hydraulic System,Supplying,Power,recipient,object,agent,Stage 1,?,Stage 2,KB,Stage 2: Match NLP Structure with Airplane Representation,Stage 2: Match with Airplane Representation,Goal: Integrate text into airplane representation Approach: find matching (subsuming) structure(s),

17、Stage 2: Match with Airplane Representation,Goal: Integrate text into airplane representation Approach: find matching (subsuming) structure(s) If match found, unify the two structures together,Stage 2: Match with Airplane Representation,Goal: Integrate text into airplane representation Approach: fin

18、d matching (subsuming) structure(s) If match found, unify the two structures together,Rudder,Hydraulic System,Supplying,Power,recipient,object,agent,Rudder,Power Source,Supplying,agent,recipient,KB expects:,Rudder,Power Source,Supplying,Power,“The hydraulic system supplies power to the rudder.”,4. ,

19、Matching Problem 1:Linguistic Variation,Use simple transformation rules to modify NLP structure No match transform and re-search for a match,Matching Problem 1: Linguistic Variation,Matching Problem 2: Expression of Implied Facts,Problem: Not all airplane facts are explicit in the KB structures If u

20、ser refers to an implied fact, matcher will not find it,Problem: Not all airplane facts are explicit in the KB structures If user refers to an implied fact, matcher will not find it,Rudder,Pedal,Pilot,Cable,Press,Move,Pull,Rudder Control System,Script,object,agent,behavior,subevents,agent,object,age

21、nt,object,“The pilot presses the pedal.”,Explicit facts:,Matching Problem 2: Expression of Implied Facts,Problem: Not all airplane facts are explicit in the KB structures If user refers to an implied fact, matcher will not find it,Rudder,Pedal,Pilot,Cable,Press,Move,Pull,Rudder Control System,Script

22、,object,agent,behavior,subevents,agent,object,agent,object,“The pilot presses the pedal.” “The pedal pulls a cable.”,Explicit facts:,Matching Problem 2: Expression of Implied Facts,Problem: Not all airplane facts are explicit in the KB structures If user refers to an implied fact, matcher will not f

23、ind it,Rudder,Pedal,Pilot,Cable,Press,Move,Pull,Rudder Control System,Script,object,agent,behavior,subevents,agent,object,agent,object,“The pilot presses the pedal.” “The pedal pulls a cable.” “The cable moves the rudder.”,Explicit facts:,Matching Problem 2: Expression of Implied Facts,Problem: Not

24、all airplane facts are explicit in the KB structures If user refers to an implied fact, matcher will not find it,Rudder,Pedal,Pilot,Cable,Press,Move,Pull,Rudder Control System,Script,object,agent,behavior,subevents,agent,object,agent,object,“The pilot presses the pedal.” “The pedal pulls a cable.” “

25、The cable moves the rudder.”,Explicit facts:,Inference Rule: If A does X, and X causes Y then add A does Y.,Matching Problem 2: Expression of Implied Facts,Problem: Not all airplane facts are explicit in the KB structures If user refers to an implied fact, matcher will not find it,Rudder,Pedal,Pilot

26、,Cable,Press,Move,Pull,Rudder Control System,Script,object,agent,behavior,subevents,agent,object,agent,object,“The pilot presses the pedal.” “The pedal pulls a cable.” “The cable moves the rudder.”,Implied facts:,Move,object,agent,instr- ument.,“The pilot moves the rudder using the pedal.”,Matching

27、Problem 2: Expression of Implied Facts,Problem: Not all airplane facts are explicit in the KB structures If user refers to an implied fact, matcher will not find it,Rudder,Pedal,Pilot,Cable,Press,Move,Pull,Rudder Control System,Script,object,agent,behavior,subevents,agent,object,agent,object,“The pi

28、lot presses the pedal.” “The pedal pulls a cable.” “The cable moves the rudder.”,Implied facts:,Move,object,agent,instr- ument.,“The pilot moves the rudder using the pedal.”,“The pedal moves the rudder using the cable.”,Move,object,instrument,agent,Matching Problem 2: Expression of Implied Facts,Mat

29、ching Problem 2: Expression of Implied Facts,Problem: Not all airplane facts are explicit in the KB structures If user refers to an implied fact, matcher will not find it Approach: Match text against an extended version of the KB Extensions generated on demand,Airplane representation:,Explicit facts

30、,Implied facts,Status and Assumptions,While some components are mature, only complete throughput for a small number of sentences Relies on strong expectations from the KB Everything that might be said is encompassed by KB Assumes stage 1 output is correct Assumes a single model of the airplane is de

31、scribed,Challenges,Inaccurate knowledge “The pump supplies a source of power to the rudder.” Simplifications/multiple models “There are three hydraulic systems in the 777-200.” “The flight control surfaces steer the airplane.” Fluctuating (“fuzzy”) concept boundaries “hydraulic system” includes cont

32、rolled devices (eg rudder)? Ambiguities not locally resolvable “The EDP is attached to the airplanes engine.”,Summary,Goal: extend an airplane KB from text Use of background knowledge can help: constrains possible interpretations provide the surrounding context Presented a simple model of how this can be done: text interpretation = iterative refinement of a representation use of graph matching and unification,