Method and knowledge structures for reasoning about concepts, relations, and rules

• US 7,447,667 B2
• Filed: 12/11/2002
• Issued: 11/04/2008
• Est. Priority Date: 12/11/2002
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Singh, Moninder, Morgenstern, Leora, Riecken, Doug, Mueller, Erik T., Gong, Leiguang
• Primary Examiner(s): STARKS, WILBERT L
• Application Number:
  US10/316,588
• Publication Number:
  US 20040117395A1
• Time in Patent Office: 2,155 Days
• Field of Search: 707/5, 707/101, 707/55, 707/4, 706/52, 706/13, 706/47, 706/55
• US Class Current: 706/47
• CPC Class Codes: G06N 5/04 Inference or reasoning models
• CPC Class Codes: Y10S 707/99935 Query augmenting and refining, e.g. inexact access

Abstract

A system and method for reasoning about concepts, relations and rules having a semantic network comprising at least one node from a predetermined set of node types, at least one link from a predetermined set of link types, and zero or more rules from a predetermined set of rule types, a subset of the rule types being matching rule types, each node and each link being associated with a set of zero or more rules; a network reasoning data structure having a reasoning type database having at least one regular expression, each of the regular expressions being a class of sequences having at least three node types and two link types, wherein the network reasoning data structure further has a context being a set of rules; and a reasoning engine having an activator for activating one or more activated paths in the semantic network, the set of activated paths having a common starting node in the semantic network, wherein the reasoning engine further has a validator for selecting a subset of the activated paths being valid paths, each rule from the set of rule matching types that is associated with one or more path elements on each valid path being matched by one or more rules in the context and wherein the reasoning engine further has a legal inferencer for selecting a subset of the set of valid paths being legal and valid paths, the legal and valid paths matching at least one of the regular expressions.

Claims

1. A computer system comprising:

a computer processor linked to one or more users;

an automated information system which is controlled by the computer processor, wherein the computer processor is adapted to receive the user request for information in one or more information domains supported by the automated information system, wherein the automated information system comprises a semantic network system that uses knowledge-based reasoning of information associated with the one or more information domains to perform an automated task in response to the user request and wherein the computer processor uses the automated information system to generate a response back to each users request in the form of a suggested action for addressing a given need or situation of the user or users,

wherein the semantic network system comprises a memory system for storing computer readable program data and computer-executable program code, and the computer processor for processing the computer-readable program data and computer-executable program code to instantiate the semantic network system;

wherein the computer-readable program data comprises;

a semantic network comprising a logical structure that represents information relating to one or more information domains supported by the automated information system, the semantic network comprising at least one of a plurality of nodes from a predetermined set of node types, at least one of a plurality of links from a predetermined set of link types, and rules from a predetermined set of rule types, a subset of the rule types being matching rule types, each node and each link being associated with a of rules; and

a network reasoning data structure comprising a reasoning type database having at least one of a plurality of regular expressions, each of the regular expressions being a class of sequences having at least three node types and two link types, wherein the network reasoning data structure further comprising a context being a set of rules; and

wherein the computer-executable program code comprises a reasoning engine to process the computer-readable data and generate reasoning results to perform the automated task in response to the user request, wherein the reasoning engine comprises;

an activator for activating one or more activated paths in the semantic network, the set of activated paths having a common starting node in the semantic network, a validator for selecting a subset of the activated paths being valid paths, each rule from the set of rule matching types that is associated with one or more path elements on each valid path being matched by one or more rules in the context; and

a legal inferencer for selecting a subset of the set of valid paths being legal and valid paths, the legal and valid paths matching at least one of the regular expressions, wherein the legal and valid paths are processed for output as reasoning results that are used to perform the automated task in response to the user request.

2. A computer system, as in claim 1, wherein the rules associated with the path elements are sentences in predicate logic and the rules in the context are ground sentences in predicate logic.

3. A computer system as in claim 1, wherein the rules include any one or more of:

a formula on a link stating conditions for a link'"'"'s validity that determines if the path containing the respective link is valid;

a definition formula that provides a definition of a node and states conditions that determines if a path starting with the respective node is valid;

a prerequisite formula stating conditions for a node'"'"'s validity that determines if the path containing the respective node is valid; and

an auxiliary formula providing additional information associated with a node.

4. A computer system as in claim 3, wherein the additional information includes any one or more of a fee for a product, a recommendation, and a weight.

5. A computer system as in claim 1, wherein the path element is a node on the respective valid and legal path.

6. A computer system as in claim 1, wherein the path element is a link on the respective valid and legal path.

7. A computer system as in claim 1, wherein the node types include any one or more of a behavior node, an action node, a situation node, a needs node, and a methods node.

8. A computer system as in claim 7, wherein the link type is a causes link specifying a causal relationship between a behavior or action and a situation.

9. A computer system as in claim 7, wherein the link type is a triggers link relating a situation node and a needs node, specifying that the state of affairs of the situation node triggers the need of the needs node.

10. A computer system as in claim 7, wherein the link type is a served-by link relating a needs node and an action node, specifying that performing the action of the action node will meet the needs of the needs node.

11. A computer system as in claim 7, wherein the link type is a leads-to link relating two situation nodes, specifying that the state of affairs of the first situation node leads to the state of affair of the second situation node.

12. A computer system as in claim 7, wherein the link type is a retroducts link relating a behavior node and a situation node, specifying an inference of the state of affairs of the situation node from the observed actions of the behavior node.

13. A computer system as in claim 7, wherein the link type is a has-method link relating an action node and a method node, specifying that the protocol of the method node gives a method for executing the action of the action node.

14. A computer system as in claim 7, wherein the link type is a subsumes link relating one of two behavior nodes, two situation nodes, two need nodes, two action nodes, or two method nodes;

specifying that the second node is a subclass of the first node.

15. A computer system as in claim 7, wherein the action nodes includes product nodes, the product nodes having information about actions to purchase product or services.

16. A computer system as in claim 15, wherein the method nodes include feature nodes having information about features of a product or service.

17. A computer system as in claim 16, wherein the link types include any one of causes, triggers, served by, subsumes, leads to, retroduct, has method, has feature, and cross-sell.

18. A computer system as in claim 16, wherein the link type is a has-feature link relating a product node and a feature node, specifying that the product of the product node has the feature of the feature node.

19. A computer system as in claim 16, wherein the link type is a cross-sell link relating two product nodes, two feature nodes, or a product and a feature node, specifying that interest on the part of an agent in the first node has a high degree of correlation of interest on the part of this agent in the second node.

20. A computer system as in claim 1, wherein the node type is a behavior node having information about actions of an agent.

21. A computer system as in claim 1, wherein the node type is an action node providing one or more actions that an agent can take.

22. A computer system as in claim 1, wherein the node type is a situation node having information about one or more states of affairs of an agent.

23. A computer system as in claim 1, wherein the node type is a needs node having information about one or more needs of an agent.

24. A computer system as in claim 1, wherein the node type is a method node providing one or more protocols for executing a planned action.

25. A computer system as in claim 1, wherein the link types include any one or more of causes, triggers, served by, subsumes, leads to, retroduct, and has-method.

26. A computer system as in claim 25, wherein the regular expression sequences include any one or more of Situation(ImpliesSituation)*TriggersNeedServedByProduct(SubsumesProduct)*, Need(SubsumesNeed)*ServedByProduct(SubsumesProduct)*, (BehaviorRetroductSituation|BehaviorCausesSituation)(ImpliesSituation)*Triggers NeedServedByProduct(SubsumesProduct)*, Product(IsaProduct)CausesSituation (ImpliesSituation)*, ProductCausesSituation(ImpliesSituation)*TriggersNeed, ProductCausesSituation(ImpliesSituation)*TriggersNeedServedByProduct(SubsumesProduct)*, Product(IsaProduct)*ServesNeed(IsaNeed)*, Product(SubsumesProduct)*, Situation(ImpliesSituation)*TriggersNeedServedByProduct,Need(SubsumesNeed)* ServedByProduct,BehaviorRetroductSituationlBehaviorCausesSituation)(ImpliesSituation) *TriggersNeedServedByProduct, and Product(SubsumesProduct).

27. A computer system as in claim 1, wherein the reasoning type database contains one or more reasoning pairs, each reasoning pair having one of the regular expressions and an associated derived link.

28. A computer system as in claim 1, wherein the reasoning type database contains one or more reasoning pairs, each reasoning pair having one of the regular expressions and an associated derived link and the legal inferencer associating the derived link with the activated path that contains the associated regular expression.

29. A computer system as in claim 28, wherein the associated derived links include any one or more of Recommend, RecommendFromNeed, RecommendFromBehavior, Hypothetical Consequence, InducesNeed, CrossSellFromProductPurchase, PossibleNeedServed, RecommendFromProduct, RecommendProductCategories, RecommendProductCategoriesFromNeed, RecommendProductCategoriesFromBehavior, and ProductCategoriesFromProduct.

30. A computer system as in claim 29, wherein the associated derived link is a Recommends link relating a situation node and an action or product node, specifying that the action of the action node or product of the product node is suitable for the state of affairs of the situation node.

31. A computer system as in claim 30, wherein the associated derived link is a RecommendProductCategoriesFromNeed link relating a need node and a product node, the product node representing a category of products, specifying that all products of the product category of the product node will meet the need of the need node.

32. A computer system as in claim 29, wherein the associated derived link is a Hypothetical Consequence link relating an action or product node and a situation node, specifying that following the action of the action node or purchasing the product of the product node will lead to the state of affairs of the situation node.

33. A computer system as in claim 29, wherein the associated derived link is a RecommendFromNeed link relating a need node and an action or product node, specifying that following the action of the action node or purchasing the product of the product node will meet the need of the need node.

34. A computer system as in claim 29, wherein the associated derived link is a RecommendFromBehavior link relating a behavior node and an action or product node, specifying that the action of the action node or purchasing the product of the product node is suitable for the behavior of the agent specified in the behavior node.

35. A computer system as in claim 29, wherein the associated derived link is an InducesNeed link relating an action or product node and a need node, specifying that following the action of the action node or purchasing product of the product node will lead to the need of the need node.

36. A computer system as in claim 35, wherein the inferential strength is used only if the value of a formula associated with the link is “

True”

.

37. A computer system as in claim 29, wherein the associated derived link is a CrossSellFromPurchase link relating one product node and a second product node, specifying that purchasing the product of the first product node will lead to a situation and trigger a need that can be met by purchasing the product of the second product node.

38. A computer system as in claim 29, wherein the associated derived link is a PossibleNeedServed link relating a product node and a need node, specifying that the need of the need node is one of the needs that can be met by purchasing the product of the product node.

39. A computer system as in claim 29, wherein the associated derived link is a RecommendFromProduct link relating one product node and a second product node, specifying that the second product node is a subclass of the first product node.

40. A computer system as in claim 29, wherein the associated derived link is a RecommendProductCategories link relating a situation node and a product node, the product node representing a category of products, specifying that all products of the product category of the product node are suitable for the state of affairs of the situation node.

41. A computer system as in claim 29, wherein the associated derived link is a RecommendProductCategoriesFromBehavior relating a behavior node and a product node, the product node representing a category of products, specifying that all the products of the product category of the product node are suitable for the agent whose behavior is specified by the behavior node.

42. A computer system as in claim 29 , wherein the derived links representing reasoning types are associated with the regular expressions in the following manner:

for reasoning type Recommend the regular expression is Situation(ImpliesSituation)* TriggersNeedservedByProduct(SubsumesProduct)*,

for reasoning type RecommendFromNeed the regular expression is Need(SubsumesNeed)*ServedByProduct(SubsumesProduct)*,

for reasoning type RecommendFromBehavior the regular expression is (BehaviorRetroductSituation|BehaviorCausesSituation)(ImpliesSituation)* TriggersNeedServedByProduct(SubsumesProduct)*,

for reasoning type Hypothetical Consequences the regular expression is Product(IsaProduct) CausesSituation(ImpliesSituation)*,

for reasoning type InducesNeed the regular expression is ProductCausesSituation(ImpliesSituation)*TriggersNeed,

for reasoning type CrossSellFromProductPurchase the regular expression is ProductCausesSituation(ImpliesSituation)* TriggersNeedServedByProduct(SubsumesProduct)*,

for reasoning type PossibleNeedServed the regular expression is Product(IsaProduct)* ServesNeed(IsaNeed)*,

for reasoning type RecommendFromProduct the regular expression is Product(SubsumesProduct)*,

for reasoning type ProductCategories the regular expression is Situation(ImpliesSituation)*TriggersNeedServedByProduct, for reasoning type ProductCategoriesFromNeed the regular expression is Need(SubsumesNeed)* ServedByProduct,

for reasoning type ProductCategoriesFromBehavior the regular expression is(BehaviorRetroductSituationlBehaviorCausesSituation) (ImpliesSituation)*TriggersNeedServedByProduct, and

for reasoning type ProductCategoriesFromProduct the regular expression is Product(SubsumesProduct).

43. A computer system as in claim 1, wherein one or more of the links in a valid and legal path has a conditional weight, the condition weight being an inferential strength of the link.

44. A computer system as in claim 1, wherein one or more of the links in a valid and legal path has a prior weight, the prior weight being an inferential strength of the link if the value of a formula associated with the link is not known.

45. A computer system as in claim 44, wherein the machine program code is executable to perform a reasoning process comprising the steps of;

receiving data and evaluating the data to determine nodes that are true, said nodes being called starting nodes;

receiving a set of static weights for nodes and formulae on links;

receiving a method of computing dynamic weights at a node or link in the network from the static weights received;

identifying the set of legal paths from the starting nodes;

computing the dynamic weight at each node along any legal path by combining the static weights along all legal paths leading to that node from one or more starting nodes;

selecting a subset of nodes of interest from the set of all nodes on all legal paths; and

outputting the subset of nodes ranked on the basis of their respective computed weights.

46. A computer system as in claim 45, wherein the set of received weights comprise two kinds of weights on formulae on links, and link if no formulae on that link, said kinds of weights comprising a conditional probability, and a prior probability;

and one kind of weights on nodes, said kind of weights comprising prior probabilities.

47. A computer system as in claim 45, wherein the set of received weights are provided by a domain expert.

48. A computer system as in claim 45, wherein the set of received weights are automatically calculated from a database.

49. A computer system as in claim 1, wherein one or more of the nodes has a node weight.

50. A computer system as in claim 49, wherein the machine program code is executable to perform a reasoning process comprising the steps of;

receiving data and evaluating the data determine nodes that are true, said nodes being called starting nodes;

receiving a set of static weights for nodes and formulae on links;

receiving a method of computing dynamic weights at a node or link in the network from the static weights received;

identifying the set of legal paths from the starting nodes;

computing the dynamic weight at each node along any legal path by combining the static weights along all legal paths leading to that node from one or more starting nodes;

selecting a subset of nodes of interest from the set of all nodes on all legal paths; and

outputting the subset of nodes ranked on the basis of their respective computed weights.

51. A computer system as in claim 50, wherein the set of received weights comprise two kinds of weights on formulae on links, and link if no formulae on that link, said kinds of weights comprising a conditional probability, and a prior probability;

and one kind of weights on nodes, said kind of weights comprising prior probabilities.

52. A computer system as in claim 50, wherein the set of received weights are provided by a domain expert.

53. A computer system as in claim 50, wherein the set of received weights are automatically calculated from a database.

54. A computer system as in claim 1 wherein the automated task relates to banking, medical treatment protocols, computer fault diagnosis, or pharmaceuticals applications.

55. The computer system of claim 1, wherein the computer system is adapted to run a business application which suggests a product or service to a consumer to purchase.

56. A computer-implemented method for performing automated reasoning in a semantic network comprising the steps of:

a user providing a request for information to a computer processor linked to the user;

the computer processor receiving the user request for information, wherein the user request for information is information in one or more information domains supported by an automated information system which is controlled by the computer processor;

the computer processor causing the automated information system to perform an automated reasoning process based on the user request using a semantic network that encodes information relating to one or more information domains supported by the automated information system, to generate reasoning results, wherein the semantic network is a logical structure that represents information relating to the one or more information domains supported by the automated information system, wherein the semantic network comprises at least one of the plurality of nodes from a predetermined set of node types, at least one of a plurality of links from a predetermined set of link types, and rules from a predetermined set of rules types, a subset of the rule types being matching rule types, each node and each link being associated with a set of rules; and

performing the automated task using the reasoning results wherein performing an automated reasoning process comprises;

for any node in the semantic network, determining if the formulas defining the node are true in a given context, using a predicate calculus theorem prover, wherein such nodes are labeled starting nodes;

for each starting node, determine legal paths that start with that node, wherein a legal path is defined as a path that corresponds to an instance of the regular expressions included in the permitted reasoning methods, and a path comprises a concatenation of alternating nodes and links, beginning and ending with a node;

traversing legal paths, wherein for each legal path being traversed, the method further comprises;

for each link that has a formula associated with it, checking if the formula is true using theorem prover and continuing traversing if the formula is true, otherwise stop pursuing the path;

for each node that has a prerequisite formula associated with it, checking if the formula is true using the theorem prover and continuing traversing if the formula is true, otherwise stop pursuing the path;

labeling all surviving paths as valid legal paths, wherein each valid legal path is associated with a regular expression;

for each valid legal path, adding a derived link corresponding to the regular expression associated with the path; and

the computer processor responds back to the request from the user in the form of a suggested action to the user for addressing a given need or situation of the user based upon the information obtained from the automated information network.

57. The computer implemented method of claim 56, wherein the method is used for a business application to suggest a product or service to a consumer to purchase.

58. A computer server system comprising;

a computer server linked to one or more users;

an automated recommendation system which runs on the computer server, wherein the computer server is adapted to receive the user request for information, wherein the user request for information is in one or more information domains supported by the automated recommendation system, wherein the automated recommendation system comprises a semantic network system that uses knowledge-based reasoning of information associated with the one or more information domains to formulate a recommendation in response to the user request and wherein the computer server uses the automated recommendation system to generate a response back to each user'"'"'s request in the form of a recommendation for addressing a given need or situation of the user or users, the semantic network system comprising;

a memory system for storing machine-readable program data comprising;

a reasoning path set comprising one or more reasoning paths, each reasoning path being a sequence of reasoning units;

a reasoning type database having one or more regular expressions, each reasoning path corresponding to one of the regular expressions in the reasoning type database;

the memory system further storing machine-executable program code to process the machine-readable program data and user data to formulate user recommendations, wherein the machine-executable program code comprises;

a content transformer transforming each reasoning path to its corresponding display sequence; and

a displayer to process for output one or more display sequences providing user recommendations, each display sequence corresponding to a single reasoning path, each display sequence further comprising a sequence of display units.

59. A computer server system as in claim 58, wherein for each reasoning path being a sequence of N units, the corresponding display sequence comprises a sequence of N display units, the number of display units in the sequence being equal to the total number of reasoning units in the reasoning path, each display unit corresponding to exactly one reasoning unit of the reasoning path, each display unit having a depiction of the corresponding reasoning unit, the sequence of N display units corresponding to a permutation of the basic reasoning units, the content transformer transforming each reasoning unit of each reasoning path to its corresponding display unit.

60. A computer server system as in claim 59, wherein the permutation is order preserving, assigning the first reasoning unit to the first display unit, the second reasoning unit to the second display unit, and the Nth reasoning unit to the Nth display unit.

61. A computer server system as in claim 60, wherein the content transformer includes productions to instantiate a general rule of logic to a particular individual.

62. A computer server system as in claim 59, wherein the permutation is order reversing, assigning the first reasoning unit to the Nth display unit, the second reasoning unit to the N-1rst display unity, and the Nth reasoning unit to the 1rst display element.

63. A computer server system as in claim 58, wherein each reasoning unit is one of;

a basic reasoning element, wherein each basic reasoning element is either a concept or a relation,

an augmented reasoning element, being either an augmented concept or an augmented relation,

an augmented concept being a set comprising a concept and a non-empty set of rules,

an augmented relation being a set comprising a relation and a non-empty set of rules, and

a reasoning pair, comprising at least one of;

a sequence of a concept and a relation,

a sequence of an augmented concept and a relation,

a sequence of a concept and an augmented relation,

a sequence of an augmented concept and an augmented relation.

64. A computer server system as in claim 63, wherein the N reasoning units comprise a first element being one of a concept or an augmented concept and N-1 reasoning pairs.

65. A computer server system as in claim 63, wherein the concepts include any one or more of a behavior concept, an action concept, a situation concept, a needs concept, and a methods concept.

66. A computer server system as in claim 63, wherein the concept element type is a behavior concept, the behavior concept element having information about observed or reported actions of an agent.

67. A computer server system as in claim 63, wherein the concept element type is an action concept, the action concept element providing one or more actions that an agent can take.

68. A computer server system as in claim 63, wherein the concept element type is a situation concept, the situation concept element having information about one or more states of affairs of an agent.

69. A computer server system as in claim 63, wherein the concept element type is a needs concept, the needs concept element having information about one or more needs of an agent.

70. A computer server system as in claim 63, wherein the concept element type is a method concept, the method concept element providing one or more protocols for executing a planned action.

71. A computer server system as in claim 63, wherein the relation element types include any one or more of causes, served by, triggers, subsumes, leads to, retroducts, and has method.

72. A computer server system as in claim 63, wherein the relation element type is a causes relation, the specifying a causal relationship between a behavior or action and a situation.

73. A computer server system as in claim 63, wherein the relation element type is a triggers relation relating a situation concept and a needs concept, specifying that the state of affairs of the situation concept triggers the need of the needs concept.

74. A computer server system as in claim 63, wherein the relation element type is a served-by relation relating a needs concept and an action concept, specifying that performing the action of the action concept will meet the needs of the needs concept.

75. A server system for displaying reasoning paths as in claim 63, wherein the relation element type is a leads-to relation relating two situation) concepts, specifying that the state of affairs of the first situation concept leads to the state of affairs of the second situation concept.

76. A computer server system as in claim 63, wherein the relation element type is a retroducts relation relating a behavior concept and a situation concept, specifying that one can infer the state of affairs of the situation concept from the observed actions of the behavior concept.

77. A computer server system as in claim 63, wherein the relation element type is a has-method relation relating an action concept and a method concept, specifying that the protocol of the method concept gives a method for executing the action of the action concept.

78. A computer server system as in claim 63, wherein the relation element type is a a subsumes relation relating one of two behavior concepts, two situation concepts, two need concepts, two action concepts, or two method concepts, wherein specifying that the second concept is a subclass of the first concept.

79. A computer server system as in claim 63, wherein the action concepts includes product concepts, the product concepts having information about actions to purchase product or services.

80. A computer server system as in claim 63, wherein the method concepts include feature concepts, the feature concepts having information about features of a product or service.

81. A computer server system as in claim 80, wherein the relation type is a cross-sell relation relating two product concepts, two feature concepts, or a product concept and a feature concept, specifying that interest on the part of an agent in the first concept has a high degree of correlation of interest on the part of this agent in the second concept.

82. A computer server system as in claim 63, wherein the relation types include any one of causes, triggers, served by, subsumes, leads to, retroduct, has method, has feature, and cross-sell.

83. A computer server system as in claim 63, wherein the relation type is a has-feature relation relating a product concept and a feature concept, specifying that the product of the product concept has the feature of the feature concept.

84. A computer server system as in claim 63, wherein the concept elements are nodes in a graph, and the relation elements are links in a graph, a reasoning path corresponding to a sequence of alternating nodes and links, starting and ending with a node, in a graph.

85. A computer server system as in claim 63, wherein one or more of the links in the graph has a conditional weight, the conditional weight being the inferential strength of the link.

86. A computer server system as in claim 58, wherein the display unit corresponding to the reasoning unit comprises one or more of text, graphics, video, and voice.

87. A computer server system as in claim 58, wherein each display sequence is transmitted in one transaction.

88. A computer server system as in claim 58, wherein each display sequence is transmitted in several transactions.

89. A computer server system as in claim 58, wherein each unit of a display sequence is transmitted one at a time.

90. A computer server system as in claim 58, wherein a portion of a display sequence is transmitted in response to a request given by the client system, this request corresponding to an action on the part of the user of the system.

91. A computer server system as in claim 58, wherein the content transformer comprises a set of context-free productions, the productions being constructed so that one or more subsequences of each reasoning path in each reasoning path set matches the left-hand side of one or more productions, and the corresponding right-hand side of each such production matches a subsequence of the display sequence in the displayer.

92. A server system as in claim 58, wherein the content transformer comprises a set of context-sensitive productions, the productions being constructed so that one or more subsequences of each reasoning path in each reasoning path set matches the left-hand side of one or more productions, and the corresponding right-hand side of each such production matches a subsequence of the display sequence in the displayer.

93. A server system as in claim 58, wherein the regular expression sequences include any one or more of;

Situation(ImpliesSituation)*TriggersNeedServedByProduct(SubsumesProduct)*, Need(SubsumesNeed)*ServedByProduct(SubsumesProduct)*, BehaviorAbducesSituation|BehaviorCausesSituation)(ImpliesSituation)*TriggersNeedServedByProduct(SubsumesProduct)*, Product(IsaProduct)CausesSituation(ImpliesSituation)*, ProductCausesSituation(ImpliesSituation)*TriggersNeed, ProductCausesSituation(ImpliesSituation)*TriggersNeedServedByProduct(SubsumesProduct)*, Product(IsaProduct)*ServesNeed(lsaNeed)*, Product(SubsumesProduct)*, Situation(ImpliesSituation)*TriggersNeedServedByProduct, Need(SubsumesNeed)*ServedByProduct, BehaviorAbducesSituation,vertline,BehaviorCausesSituation)ImpliesSituation)*TriggersNeedServedByProduct, and Product(SubsumesProduct)*.

94. A computer server system as in claim 58, wherein the reasoning type database contains one or more reasoning pairs, each reasoning pair having one of the regular expressions and an associated derived relation.

95. A computer server system as in claim 94, wherein the associated derived relations include any one or more of;

Recommend, RecommendFromNeed, RecommendFromBehavior, HypotheticalConsequences, InducesNeed, DerivedCrossSellFromProduct, PossibleNeedServed, RecommendFromproduct, ProductCategories, ProductCategoriesFromNeed, ProductCategoriesFromBehavior, ProductCategoriesFromProduct.